Professional Documents
Culture Documents
Hilbert Flow Spaces With Operators Over Topological Graphs
Hilbert Flow Spaces With Operators Over Topological Graphs
4(2017), 19-45
Linfan MAO
1. Chinese Academy of Mathematics and System Science, Beijing 100190, P.R.China
2. Academy of Mathematical Combinatorics & Applications (AMCA), Colorado, USA
E-mail: maolinfan@163.com
X +
→
−
LAvu (v, u) = L(v), ∀v ∈ V G .
u∈NG (v)
§1. Introduction
A Banach or Hilbert space is respectively a linear space A over a field R or C equipped with a
complete norm k · k or inner product h · , · i, i.e., for every Cauchy sequence {xn } in A , there
and a topological graph ϕ(G) is an embedding of a graph G with vertex set V (G), edge set
E(G) in a space S , i.e., there is a 1 − 1 continuous mapping ϕ : G → ϕ(G) ⊂ S with
ϕ(p) 6= ϕ(q) if p 6= q for ∀p, q ∈ G, i.e., edges of G only intersect at vertices in S , an embedding
of a topological space to another space. A well-known result on embedding of graphs without
loops and multiple edges in Rn concluded that there always exists an embedding of G that all
edges are straight segments in Rn for n ≥ 3 ([22]) such as those shown in Fig.1.
Fig.1
As we known, the purpose of science is hold on the reality of things in the world. However,
the reality of a thing T is complex and there are no a mathematical subfield applicable unless
a system maybe with contradictions in general. Is such a contradictory system meaningless
to human beings? Certain not because all of these contradictions are the result of human
beings, not the nature of things themselves, particularly on those of contradictory systems in
mathematics. Thus, holding on the reality of things motivates one to turn contradictory systems
to compatible one by a combinatorial notion and establish an envelope theory on mathematics,
i.e., mathematical combinatorics ([9]-[13]). Then, Can we globally characterize the behavior of a
system or a population with elements≥ 2, which maybe contradictory or compatible? The answer
is certainly YES by continuity flows, which needs one to establish an envelope mathematical
theory over topological graphs, i.e., views labeling graphs GL to be mathematical elements
([19]), not only a game object or a combinatorial structure with labels in the following sense.
−
→ −
→
Definition 1.1 A continuity flow G ; L, A is an oriented embedded graph G in a topological
space S associated with a mapping L : v → L(v), (v, u) → L(v, u), 2 end-operators A+ vu :
+
A+
L(v, u) → LAvu (v, u) and A+
uv : L(u, v) → L uv (u, v) on a Banach space B over a field F
L(v)
A+
vu
-
L(v, u) A+
uv
L(u)
v u
Fig.2
Hilbert Flow Spaces with Differentials over Graphs 21
−
→
A+
with L(v, u) = −L(u, v) and A+
vu (−L(v, u)) = −L vu (v, u) for ∀(v, u) ∈ E G holding with
continuity equation −
X + →
LAvu (v, u) = L(v) f or ∀v ∈ V G
u∈NG (v)
L(u1 ) -
L(u1 , v)
-
L(v, u4 )
L(u4 )
u1 u4
A1 A4
L(u2 ) -
L(u2 , v) A2
L(v)
A5
- L(v, u ) 5
L(u5 )
u2 A3 v A6 u5
L(u3 ) -
L(u3 , v)
-
L(v, u6 )
L(u6 )
u3 Fig.3 u6
LA1 (v, u1 ) + LA2 (v, u2 ) + LA3 (v, u3 ) − LA4 (v, u4 ) − LA5 (v, u5 ) − LA6 (v, u6 ) = L(v),
−
→
Clearly, an action flow is an equilibrium state of a continuity flow G ; L, A . We have
shown that Banach or Hilbert space can be extended over topological graphs ([14],[17]), which
can be applied to understanding the reality of things in [15]-[16], and we also shown that
complex flows can be applied to hold on the global stability of biological n-system with n ≥ 3
in [19]. For further discussing continuity flows, we need conceptions following.
Definition 1.2 Let B1 , B2 be Banach spaces over a field F with norms k · k1 and k · k2 ,
respectively. An operator T : B1 → B2 is linear if
for λ, µ ∈ F, and T is said to be continuous at a vector v0 if there always exist such a number
22 Linfan MAO
Definition 1.3 Let B1 , B2 be Banach spaces over a field F with norms k · k1 and k · k2 ,
respectively. An operator T : B1 → B2 is bounded if there is a constant M > 0 such that
kT(v)k2
kT (v)k2 ≤ M kvk1 , i.e., ≤M
kvk1
Theorem 1.4 Let B1 , B2 be Banach spaces over a field F with norms k·k1 and k·k2 , respectively.
Then, a linear operator T : B1 → B2 is continuous if and only if it is bounded, or equivalently,
kT(v)k2
kTk := sup < +∞.
06=v∈B1 kvk1
n− → → − o
Let G 1 , G 2 , · · · be a graph family. The main purpose of this paper is to extend Ba-
nach or Hilbert
n o spaces to Banach or Hilbert continuity flow spaces over topological graphs
−
→ − →
G 1 , G 2 , · · · and establish differentials on continuity flows, which enables one to characterize
their globally change rate constraint on the combinatorial structure. A few well-known results
such as those of Taylor formula, L’Hospital’s rule on limitation are generalized to continuity
flows, and algebraic or differential flow equations are discussed in this paper. All of these
results form the elementary differential theory on continuity flows, which contributes math-
ematical combinatorics and can be used to characterizing the behavior of complex systems,
particularly, the synchronization.
For terminologies and notations not defined in this paper, we follow references [1] for
mechanics, [4] for functionals and linear operators, [22] for topology, [8] combinatorial geometry,
[6]-[7],[25] for Smarandache systems, Smarandache geometries and Smaarandache multispaces
and [2], [20] for biological mathematics.
−
→ −
→
i.e., G i is a subgraph of G for integers 1 ≤ i ≤ n. In this case, these is naturally an embedding
−
→ −
→
ι : Gi → G .
−
→
Let V be a linear space over a field F . A vector labeling L : G → V is a mapping with
−
→ −
→
L(v), L(e) ∈ V for ∀v ∈ V ( G ), e ∈ E( G ). Define
−
→L1 −→ 2 −
→ − → L1 [ − → L1 +L2 [ −
→ \− → − → L2
G1 + GL
2 = G1 \ G2 G1 G2 G2 \ G1 (2.1)
and
−
→ −
→
λ · G L = G λ·L (2.2)
−
→ − → 1 − →L2
for ∀λ ∈ F . Clearly, if , and GL , G L1 , G 2 are continuity flows with linear end-operators
−
→ −
→ L1 −
→L2 −
→L
A+ +
vu and Auv for ∀(v, u) ∈ E G , G 1 + G 2 and λ · G are continuity flows also. If we
−
→L −
→Lb b : −→ −
→
consider each continuity flow G i a continuity subflow of G , where L G i = L( G i ) but
−
→ −
→ −
→
b : G \ G i → 0 for integers 1 ≤ i ≤ n, and define O : G → 0, then all continuity flows,
L
−
→ − → →
−
particularly, all complex flows, or graphs G 1 , G 2 , · · · , G n naturally
Dall action flowsEon oriented
−
→ V
form a linear space, denoted by G i, 1 ≤ i ≤ n ; +, · over a field F under operations (2.1)
and (2.2) because it holds with:
(3) An operation “+”, called continuity flow addition, which associates with each pair of
−
→ 1 − →L2 D− → EV −
→ 1 −→L2 D− → EV
continuity flows G L
1 , G 2 in G i , 1 ≤ i ≤ n a continuity flows G L
1 + G 2 in G i , 1 ≤ i ≤ n ,
−
→ L1 −
→L2
called the sum of G 1 and G 2 , in such a way that
−
→ 1 − →L2 − →L2 − →L1
(a) Addition is commutative, G L
1 + G 2 = G 2 + G 1 because of
−
→L1 −→ 2 −
→ → L1 [ −
− → L1 +L2 [ −
→ \− → → L2
−
G1 + GL
2 = G1 − G2 G1 G2 G2 − G1
−
→ → L2 [ −
− → L2 +L1 [ −
→ \− → → L1
−
= G2 − G1 G1 G2 G1 − G2
−
→ 2 − →L1
= GL2 + G1 ;
−
→L1 −→ 2 −→L3 →L1 −
− →L2 − →L3
(b) Addition is associative, G1 + GL
2 + G 3 = G 1 + G 2 + G3 because if we
let → −
− → S− →
Li (x), if x ∈ G i \ G j G k
→ −
− → S− →
Lj (x), if x ∈ G j \ G i G k
→ −
− → S− →
L (x), if x ∈ G k \ G i G j
k
−
+ → T− → − →
Lijk (x) = L+
ij (x), if x ∈ G i G j \ G k (2.3)
−
→ T− → − →
L+
ik (x), if x ∈ G i G k \ G j
−
→ T− → −
→
L+ if x ∈ G j G k \ G i
jk (x),
−
→ T− → T− →
Li (x) + Lj (x) + Lk (x) if x ∈ G i G j G k
24 Linfan MAO
and −
→ − →
if x ∈ G i \ G j
Li (x),
+ −
→ − →
Lij (x) = Lj (x), if x ∈ G j \ G i (2.4)
−
→ T− →
Li (x) + Lj (x), if x ∈ G i G j
G1 + GL
2 + GL
3 = G1 G2 + GL
3 = G 1 G 2 G 3
→L1 − → L23 −
→ [− → 1 − →L3
+
− →L2 −
= G1 + G2 G3 = GL
1 + G2 + G3 ;
−
→ −
→
(c) There is a unique continuity flow O on G hold with O(v, u) = 0 for ∀(v, u) ∈ E G and
−
→ D− → EV −
→ −
→ −
→ D−
→ EV
V G in G i , 1 ≤ i ≤ n , called zero such that G L + O = G L for G L ∈ G i , 1 ≤ i ≤ n ;
−
→ D−
→ EV
(d) For each continuity flow G L ∈ G i , 1 ≤ i ≤ n there is a unique continuity flow
−
→−L −
→L − →−L
G such that G + G = O;
(4) An operation “·, called scalar multiplication, which associates with each scalar k in F
−
→ D−→ EV −
→
and a continuity flow G L in G i , 1 ≤ i ≤ n a continuity flow k · G L in V , called the product
−
→
of k with G L , in such a way that
−
→ −
→ −
→ D−
→ EV
(a) 1 · G L = G L for every G L in G i , 1 ≤ i ≤ n ;
−
→ −
→
(b) (k1 k2 ) · G L = k1 (k2 · G L );
−
→ 1 − →L2 −
→L1 −
→L2
(c) k · ( G L
1 + G2 ) = k · G1 + k · G2 ;
−
→ −
→ −
→
(d) (k1 + k2 ) · G L = k1 · G L + k2 · G L .
D EV D− EV
−
→ →
Usually, we abbreviate G i, 1 ≤ i ≤ n ; +, · to G i , 1 ≤ i ≤ n if these operations
+ and · are clear in the context.
−
→ 1 − →L2 − →L1 −
→ −
→ −
→ − →
By operation (1.1), G L 1 + G 2 6= G 1 if and only if G 1 6 G 2 with L1 : G 1 \ G 2 6→ 0 and
−
→L1 − → 2 − →L2 −
→ −
→ −
→ − →
G1 + GL 2 6= G 2 if and only if G 2 6 G 1 with L2 : G 2 \ G 1 6→ 0, which allows us to introduce
−
→ 1 − →L2 −
→Ln
the conception of linear irreducible. Generally, a continuity flow family { G L 1 , G2 , · · · , Gn }
is linear irreducible if for any integer i,
−
→ [−
→ → [−
− →
G i 6 Gl with Li : G i \ G l 6→ 0, (2.5)
l6=i l6=i
D−
→ EV
G i , 1 ≤ i ≤ n , called basis, i.e.,
D−
→ EV
dim G i , 1 ≤ i ≤ n = n.
−
→ i D−
→ EV
Proof By definition, G L
i , 1 ≤ i ≤ n is a linear generated of G i , 1 ≤ i ≤ n with
−
→
Li : G i → V , i.e.,
D−→ EV
dim G i , 1 ≤ i ≤ n ≤ n.
−
→ i
We only need to show that G L
i , 1 ≤ i ≤ n is linear independent, i.e.,
D−
→ EV
dim G i , 1 ≤ i ≤ n ≥ n,
−
→ − → −
→
then c1 = c2 = · · · = cn = 0. Notice that { G 1 , G 2 , · · · , G n } is linear irreducible. We are easily
−
→ S− → −
→ S− →
know Gi \ G l 6= ∅ and find an element x ∈ E(Gi \ G l ) such that ci Li (x) = 0 for integer
l6=i l6=i
i, 1 ≤ i ≤ n. However, Li (x) 6= 0 by (1.5). We get that ci = 0 for integers 1 ≤ i ≤ n. 2
D−→ EV
A subspace of G i , 1 ≤ i ≤ n is called an A0 -flow space if its elements are all continuity
−
→L −
→
flows G with L(v), v ∈ V G are constant v. The result following is an immediately
conclusion of Theorem 2.1.
−
→ −→ − → −
→
Theorem 2.2 Let G , G 1 , G 2 , · · · , G n be oriented graphs embedded in a space S and V
−
→ − → − → −
→
a linear space over a field F. If G v , G v1 1 , G v2 2 , · · · , G vnn are continuity flows with v(v) =
−
→
v, vi (v) = vi ∈ V for ∀v ∈ V G , 1 ≤ i ≤ n, then
D−
→v E
(1) G is an A0 -flow space;
D−
→ → − → E
− −
→ −
→ −
→
(2) G v1 1 , G v2 2 , · · · , G vnn is an A0 -flow space if and only if G 1 = G 2 = · · · = G n or
v1 = v2 = · · · = vn = 0.
−
→ −
→ −
→ −
→ −
→
Proof By definition, G v1 1 + G v2 2 and λ G v are A0 -flows if and only if G 1 = G 1 or
v1 = v2 = 0 by definition. We therefore know this result. 2
2.2 Commutative Rings over Graphs
Furthermore, if V is a commutative ring (R; +, ·), we can extend it over oriented graph family
−
→ − → −
→
{ G 1 , G 2 , · · · , G n } by introducing operation + with (2.1) and operation · following:
−
→L1 −→ 2 −
→ − → L1 [ − → L1 ·L2 [ −
→ \− → − → L2
G1 · GL
2 = G1 \ G2 G1 G2 G2 \ G1 , (2.6)
26 Linfan MAO
where L1 · L2 : x → L1 (x) · L2 (x), and particularly, the scalar product for Rn , n ≥ 2 for
−
→ T− →
x ∈ G 1 G 2.
D ER
−
→
As we shown in Subsection 2.1, G i, 1 ≤ i ≤ n ; + is an Abelian group. We show
D ER
−
→
G i, 1 ≤ i ≤ n ; +, · is a commutative semigroup also.
In fact, define −
→ − →
if x ∈ G i \ G j
Li (x),
−
→ − →
L×
ij (x) = Lj (x), if x ∈ G j \ G i
−
→ T− →
Li (x) · Lj (x), if x ∈ G i G j
→ L2 − → L12 − → L21 −
× ×
−
→ 1− → S− → S− → 2−→L1
Then, we are easily known that G L 1 · G 2 = G 1 G 2 = G 1 G 2 = GL2 ·G1
D ER
−
→ 1 − →L2 −
→
for ∀ G L
1 , G 2 ∈ G i , 1 ≤ i ≤ n ; · by definition (2.6), i.e., it is commutative.
Let → −
− → S− →
Li (x), if x ∈ G i \ G j G k
→ −
− → S− →
Lj (x), if x ∈ G j \ G i G k
→ −
− → S− →
L (x), if x ∈ G k \ G i G j
k
−
→ T− → − →
L×
ijk (x) = Lij (x), if x ∈ G i G j \ G k
−
→ T− → −
→
Lik (x), if x ∈ G i G k \ G j
−
→ T− → − →
Ljk (x), if x ∈ G j G k \ G i
−
→ T− → T− →
Li (x) · Lj (x) · Lk (x) if x ∈ G i G j G k
Then,
−
→L1 −→ 2 −→ 3 −
→ [− → L12 −
×
→ 3 − → [− → [− → L123
×
G1 · GL
2 · GL
3 = G1 G2 · GL 3 = G 1 G 2 G 3
→ 1 −
− → [ −
→ L ×
23 → 1 −
− →L2 −→L3
= GL1 · G2 G3 = GL1 · G2 · G3 .
Thus, −
→L1 −→ 2 −→ 3 −→L1 −
→L2 −→L3
G1 · GL
2 · GL
3 = G1 · G2 · G3
D ER
−
→ − → 1 − →L2 −
→
for ∀ G L , G L
1 , G2 ∈ G i, 1 ≤ i ≤ n ; · , which implies that it is a semigroup.
and −
→L1 −→ 2 −→ 3 −→L1 −→L3 −→L2 −→L3
G1 + GL
2 · GL
3 = G1 · G3 + G2 · G3 (2.8)
Hilbert Flow Spaces with Differentials over Graphs 27
D ER
−
→ − → − → −
→
for ∀ G 3 , G 1 , G 2 ∈ G i, 1 ≤ i ≤ n ; +, · . Clearly,
which is the (2.7). The proof for (2.8) is similar. Thus, we get the following result.
n−
→ 1 − →L2 →Ln o
−
Theorem 2.3 Let (R; +, ·) be a commutative ring and let G L 1 , G2 , · · · , Gn be a linear
−
→ + +
iD → R for integers
irreducible family, Li : G
− E
1 ≤ i ≤ n with linear operators Avu , Auv for
→ −
→ R
∀(v, u) ∈ E G . Then, G i, 1 ≤ i ≤ n ; +, · is a commutative ring.
−
→ 1 − →L2 −
→Ln D−
→ EV
Let { G L
1 , G 2 , · · · , G n } be a basis of G i , 1 ≤ i ≤ n , where V is a Banach space with a
→L D−
− → EV
norm k · k. For ∀ G ∈ G i , 1 ≤ i ≤ n , define
−
X
→L
G
= kL(e)k . (2.9)
→
−
e∈E G
−
→ − → 1 − →L2 D−→ EV
Then, for ∀ G , G L
1 , G 2 ∈ G i, 1 ≤ i ≤ n we are easily know that
−
−
→
→
−
→
(1)
G L
≥ 0 and
G L
= 0 if and only if G L = O;
−
−
→
→
(2)
G ξL
= ξ
G L
for any scalar ξ;
− →L2
→
−
→ 1 −
− L1
→L2
(3)
G L
1 + G 2
≤
G 1
+
G 2
because of
− → 2
X
→L1 −
G1 + GL
2
= kL1 (e)k
→
− → −
e∈E G 1 \ G 2
X X
+ kL1 (e) + L2 (e)k + kL2 (e)k
→
− T→− →
− → −
e∈E G 1 G2 e∈E G 2 \ G 1
X X
≤ kL1 (e)k + kL1 (e)k
→
− → − →
− T→−
e∈E G 1 \ G 2 e∈E G 1 G2
X X
→
−
− 1
→L2
+ kL2 (e)k + kL2 (e)k =
G L
1
+
G2
.
→
− → − →
− T→−
e∈E G 2 \ G 1 e∈E G 1 G2
for kL1 (e) + L2 (e)k ≤ kL1 (e)k + kL2 (e)k in Banach space V . Therefore, k · k is also a norm
28 Linfan MAO
D−
→ EV
on G i, 1 ≤ i ≤ n .
−
→ 1 − →L2 D−→ EV
Furthermore, if V is a Hilbert space with an inner product h·, ·i, for ∀ G L
1 , G 2 ∈ G i, 1 ≤ i ≤ n ,
define
D−
→L1 −→ 2E X
G1 , GL
2 = hL1 (e), L1 (e)i
→
− → −
e∈E G 1 \ G 2
X X
+ hL1 (e), L2 (e)i + hL2 (e), L2 (e)i . (2.10)
→
− T→− →
− → −
e∈E G 1 G2 e∈E G 2 \ G 1
−
→ D−
→ EV
(1) For ∀ G L ∈ G i , 1 ≤ i ≤ n ,
D−
→L −→ E X
G , GL = hL(e), L(e)i ≥ 0
→
−
e∈E G
D−
→L −→ E −
→
and G , G L = 0 if and only if G L = O.
→ −
− → D−
→ EV
(2) For ∀ G L1 , G L2 ∈ G i , 1 ≤ i ≤ n ,
D−
→L1 −→ 2 E D−
→ 2 − →L1 E
G1 , GL
2 = GL2 , G1
because of
D−
→L1 −→ 2E X X
G1 , GL
2 = hL1 (e), L1 (e)i + hL1 (e), L2 (e)i
→
− → − →
− T→−
e∈E G 1 \ G 2 e∈E G 1 G2
X
+ hL2 (e), L2 (e)i
→
− → −
e∈E G 2 \ G 1
X X
= hL1 (e), L1 (e)i + hL2 (e), L1 (e)i
→
− → − →
− T→−
e∈E G 1 \ G 2 e∈E G 1 G2
X D−
→ 2 − →L1 E
+ hL2 (e), L2 (e)i = G L
2 , G1
→
− → −
e∈E G 2 \ G 1
−
→ → − 1 → − L 2 D−
→ EV
(3) For G L , G L
1 , G 2 ∈ G i, 1 ≤ i ≤ n and λ, µ ∈ F , there is
D −
→ 1 −
→L2 →− LE D−
→ 1 −→L E D−
→ 2 −→L E
λGL1 + µG2 , G = λ GL1 ,G + µ GL2 ,G
Hilbert Flow Spaces with Differentials over Graphs 29
because of
D −
→ 1 − →L E D−
→L2 − → 1 − → 2 − → E
λGL1 + µG2 , G = G λL
1 + G µL
2 , GL
−
→ − → λL1 [ −
→ \− → λL1 +µL2 [ −
→ − → µL2 −
→L
= G1 \ G2 G1 G2 G2 \ G1 ,G .
−
→ S−→
Define L1λ 2µ : G 1 G 2 → V by
−
→ − →
if x ∈ G 1 \ G 2
λL1 (x),
−
→ − →
L1λ 2µ (x) = µL2 (x), if x ∈ G 2 \ G 1
−
→ T− →
λL1 (x) + µL2 (x), if x ∈ G 2 G 1
and
D−
→ 1 −→L E D−→ 2 − →L E
λ GL1 ,G + µ GL 2 ,G
X X
= hλL1 (e), λL1 (e)i + hλL1 (e), L(e)i
→
− → − →
− T→−
e∈E G 1 \ G e∈E G 1 G
X X
+ hL(e), L(e)i + hµL2 (e), µL2 (e)i
→
− →− →
− → −
e∈E G \ G 1 e∈E G 2 \ G
X X
+ hµL2 (e), L(e)i + hL(e), L(e)i .
→
− T→− →
− →−
e∈E G 2 G e∈E G \ G 2
Notice that
X
L1λ 2µ (e), L1λ 2µ (e)
→ − →
− S→ −
e∈E G1 G2 \G
X X
= hλL1 (e), λL1 (e)i + hµL2 (e), µL2 (e)i
→
− → − →
− → −
e∈E G 1 \ G e∈E G 2 \ G
X
+ L1λ 2µ (e), L(e)
→ − T→
− S→ −
e∈E G1 G2 G
30 Linfan MAO
X X
= hλL1 (e), L(e)i + hµL2 (e), L(e)i
→
− T→− →
− T→−
e∈E G 1 G e∈E G 2 G
X
+ hL(e), L(e)i
→
− →−
e∈E G \ G 2
X X
= hL(e), L(e)i + hL(e), L(e)i .
→
− →− →
− →−
e∈E G \ G 1 e∈E G \ G 2
−
→
Thus, G V is an inner space
−
→ 1 → − L2 −
→Ln D−
→ EV
If { G L
1 , G 2 , · · · , G n } is a basis of space G i , 1 ≤ i ≤ n , we are easily find the exact
formula on L by L1 .L2 , · · · , Ln . In fact, let
−
→L −
→ 1 −
→L2 −
→Ln
G = λ1 G L
1 + λ2 G 2 + · · · + λn G n ,
i
!
i
\ −
→ [ −
→ X
b:
L G kl \ Gs → λkl Lkl
l=1 s6=kl ,··· ,ki l=1
for integers 1 ≤ i ≤ n. Then, we are easily knowing that Lb is nothing else but the labeling L
−
→
on G by operation (2.1), a generation of (2.3) and (2.4) with
−
n
X X
Xi
→L
G
=
λkl Lkl (e)
, (2.11)
−
i=1 e∈E →
Gi l=1
* +
D−
→L −→ ′E n
X X i
X i
X
G , G′L = λkl L1kl (e), λ′ks L2ks , (2.12)
−
i=1 e∈E →
Gi l=1 s=1
i
!
−
→ ′ −
→ 1 \ [
′−
→L2 ′ −
→Ln −
→ −
→ −
→
where G′L = λ′1 G L
1 + λ2 G 2 + · · · + λn G n and G i = G kl \ Gs .
l=1 s6=kl ,··· ,ki
−
→ − → −
→
We therefore extend the Banach or Hilbert space V over graphs G 1 , G 2 , · · · , G n following.
−
→ − → −
→
Theorem 2.4 Let G 1 , G 2 , · · · , G n be oriented graphs embedded in a space S and V a Banach
D−→ EV
space over a field F . Then G i , 1 ≤ i ≤ n with linear operators A+ +
vu , Auv for ∀(v, u) ∈
−→ D −
→ EV
E G is a Banach space, and furthermore, if V is a Hilbert space, G i , 1 ≤ i ≤ n is a
Hilbert space too.
Hilbert Flow Spaces with Differentials over Graphs 31
D−→ EV
Proof We have shown, G i , 1 ≤ i ≤ n is a linear normed space or inner space if V is a
linear normed space or inner space, and for the later, let
−
rD− → E
→L
→L −
G
= G , GL
−
→ D−
→ EV D−→ EV
for G L ∈ G i , 1 ≤ i ≤ n . Then G i , 1 ≤ i ≤ n is a normed space and furthermore, it is
a Hilbert space if it is complete. Thus, we are only need to show that any Cauchy sequence is
D−
→ EV
converges in G i , 1 ≤ i ≤ n .
n−
→ no D−
→ EV
In fact, let H Ln be a Cauchy sequence in G i , 1 ≤ i ≤ n , i.e., for any number ε > 0,
there always exists an integer N (ε) such that
− → m
→Ln −
H n − H L m
< ε
−
→ n −
S → −
→ n
if n, m ≥ N (ε). Let G V be the continuity flow space on G = G i . We embed each H L
n to
i=1
−
→b − →
a G L ∈ G V by letting
Ln (e), if e ∈ E (Hn )
b n (e) =
L −→ − →
0, if e ∈ E G \ H n .
Then
− →b
X X
→Lb n −
G − G Lm
= kLn (e)k + kLn (e) − Lm (e)k
→
− → − →
− T→−
e∈E G n \ G m e∈E G n Gm
X
− →Lm
→ n −
+ k−Lm (e)k =
H L
n − H m
≤ ε.
→
− −
→
e∈E G m \ G n
n−
→Lb n o −
→
Thus, G is a Cauchy sequence also in G V . By definition,
− →
b b m (e)
→ −
Ln (e) − L
≤
G Ln − G Lm
< ε,
−
→
i.e., {Ln (e)} is a Cauchy sequence for ∀e ∈ E G , which is converges on in V by definition.
Let
b
L(e) b n (e)
= lim L
n→∞
−
→ −
→b −
→b −
→b
for ∀e ∈ E G . Then it is clear that lim G Ln = G L , which implies that { G Ln }, i.e.,
n→∞
n−→Ln o −
→b − → D−
→ EV
Hn is converges to G L ∈ G V , an element in G i , 1 ≤ i ≤ n b
because of L(e) ∈ V for
−
→ −
→ n
S −→
∀e ∈ E G and G = G i. 2
i=1
32 Linfan MAO
Theorem 2.4 enables one to establish differentials and generalizes results in classical calculus in
D−
→ EV
space G i , 1 ≤ i ≤ n . Let L be kth differentiable to t on a domain D ⊂ R, where k ≥ 1.
Define
−
→ Zt Rt
dGL −
→ dL −
→L −
→0 Ldt
= G dt and G dt = G .
dt
0
D−
→ EV
Then, we are easily to generalize Taylor formula in G i, 1 ≤ i ≤ n following.
−
→ D−→ ER×R n
−
→L − → t − t0 −
→L′ (t0 ) k
(t − t0 ) −→L(k) (t0 ) →
−k −
G = G L(t0 ) + G + ···+ G + o (t − t0 ) G , (3.1)
1! k!
→
−
for ∀t0 ∈ D, where o (t − t0 )−k G denotes such an infinitesimal term L b of L that
b u)
L(v, −
→
lim k
=0 f or ∀(v, u) ∈ E G .
t→t0 (t − t0 )
−
→LC
Particularly, if L(v, u) =
−f (t)c
vu , where cvu is a constant, denoted by f (t) G with LC :
→
(v, u) → cvu for ∀(v, u) ∈ E G and
(t − t0 ) ′ (t − t0 )2 ′′ (t − t0 )k (k)
f (t) = f (t0 ) + f (t0 ) + f (t0 ) + · · · + f (t0 ) + o (t − t0 )k ,
1! 2! k!
then
−
→ −
→
f (t) G LC = f (t) · G LC .
−
→
Proof Notice that L(v, u) has kth order derivative to t on D for ∀(v, u) ∈ E G . By
applying Taylor formula on t0 , we know that
b u)
L(v,
lim t =0
t→t0 (t − t0 )
Hilbert Flow Spaces with Differentials over Graphs 33
−
→
for ∀(v, u) ∈ E G . By operations (2.1) and (2.2),
−
→L1 −
→ −
→ −
→ −
→ →−
G + G L2 = G L1 +L2 and λ G L = G λ L
−
→
because A+ +
vu , Auv are linear for ∀(v, u) ∈ E G . We therefore get
−
→L − → (t − t0 ) −
→L′ (t0 )
k
(t − t0 ) −→L(k) (t0 ) →
−
G = G L(t0 ) + G + ···+ G + o (t − t0 )−k G
1! k!
→
−k − b of L, i.e.,
for t0 ∈ D, where o (t − t0 ) G is an infinitesimal term L
b u)
L(v,
lim t = 0
t→t0 (t − t0 )
−
→
for ∀(v, u) ∈ E G . Calculation also shows that
k
−
→ −
→ → f (t0 )+ (t−t
− 0 ) f ′ (t )···+ (t−t0 ) f (k) (t )+o (t−t )k
0 0 ( 0 ) cvu
f (t) G LC (v,u) = G f (t)LC (v,u) = G
1! k!
→ f ′ (t0 )(t − t0 )
− → f ”(t0 )(t − t0 )2
− −
→
= f (t0 )cvu G + cvu G + cvu G
1! 2!
f (k) (t0 ) (t − t0 )k −→ →
k −
+···+ cvu G + o (t − t0 ) G
k! !
(t − t0 ) ′
k
(t − t0 ) (k) −
→
k
= f (t0 ) + f (t0 ) · · · + f (t0 ) + o (t − t0 ) cvu G
1! k!
−
→ −
→
= f (t)cvu G = f (t) · G LC (v,u) ,
i.e.,
−
→ −
→
f (t) G LC = f (t) · G LC .
Theorem 3.2 Let f (t) be a k differentiable function to t on a domain D ⊂ R with 0 ∈ D and
−
→ −
→ −
→
f (0 G) = f (0) G . If A+ +
vu , Auv are linear for ∀(v, u) ∈ E G , then
−
→ −→
f (t) G = f t G . (3.2)
Notice that
−
→ f ′ (0) f ′′ (0) 2 f (k) (0) k k
−→
f (t) G = f (0) + t+ t + ···+ t +o t G
1! 2! k!
−
→f (0)+ f ′1!(0) t+ f ′′2!(0) t2 +···+ f (k)k!(0) tk +o(tk )
= G
→ f ′ (0)t −
− → f (k) (0)tk − → −
→
= f (0) G + G + ··· + G + o tk G
1! k!
and by definition,
−→ −
→ f ′ (0) − → f ′′ (0) −
→2
f tG = f 0G + tG + tG
1! 2!
f (k) (0) −→k −
→ k
+···+ tG + o tG
k!
−
→ f ′ (0) −
→ f ′′ (0) 2 −
→ f (k) (0) k −
→ −
→
= f 0G + tG + t G + ···+ t G + o tk G
1! 2! k!
− →i −
→i −
→ −
→ −
→
because of t G = G t = ti G for any integer 1 ≤ i ≤ k. Notice that f (0 G) = f (0) G . We
therefore get that −
−
→ →
f (t) G = f t G . 2
−→
Theorem 3.2 enables one easily getting Taylor expansion formulas by f t G . For example,
let f (t) = et . Then
−
→ →
−
et G = et G (3.3)
′
by Theorem 3.5. Notice that (et ) = et and e0 = 1. We know that
→
− −
→ − → t− → t2 −
→ tk −
→
et G = et G = G + G + G + · · · + G + · · · (3.4)
1! 2! k!
and
→
− →
− −
→t − →s − →t s − → t+s →
−
et G · es G = G e · G e = G e ·e = G e = e(t+s) G , (3.5)
where t and s are variables, and similarly, for a real number α if |t| < 1,
−
→ →α −
− → αt −→ α(α − 1) · · · (α − n + 1)tn −
→
G + tG =G+ G + ···+ G + ··· (3.6)
1! n!
3.2 Limitation
−
→ − → 1 D−
→ EV
Definition 3.3 Let G L , G L 1 ∈ G i , 1 ≤ i ≤ n with L, L1 dependent on a variable t ∈
−→
[a, b] ⊂ (−∞, +∞) and linear continuous end-operators A+ vu for ∀(v, u) ∈ E G . For t0 ∈
[a, b]
and any number ε > 0, if there is always a number δ(ε) such that if |t − t0 | ≤ δ(ε)
−→L1 − →L
−
→L1 −
→L
then
G 1 − G
< ε, then, G 1 is said to be converged to G as t → t0 , denoted by
−
→ 1 − → −
→ −
→
lim G L 1 = G L . Particularly, if G L is a continuity flow with a constant L(v) for ∀v ∈ V G
t→t0
−
→ 1 −
→
and t0 = +∞, G L 1 is said to be G -synchronized.
Hilbert Flow Spaces with Differentials over Graphs 35
Applying Theorem 1.4,we know that there are positive constants cvu ∈ R such that
−
→
kA+ +
vu k ≤ cvu for ∀(v, u) ∈ E G .
−
→ −
→
and kL(v, u)k ≥ 0 for (v, u) ∈ E G and E G 1 . Let
cmax
G1 G = max c+
vu , max c+
vu .
(v,u)∈E(G1 ) (v,u)∈E(G1 )
− →L
− →
→ 1 −
→ −
If
G L
1 − G
< ε, we easily get that kL1 (v, u)k < cmax
G1 G ε for (v, u) ∈ E G 1 \ G ,
−
→ T− →
k(L1 − L)(v, u)k < cmax
G1 G ε for (v, u) ∈ E G 1 G and k − L(v, u)k < cmax G1 G ε for (v, u) ∈
−
→ − →
E G \ G1 .
−
→ − →
Conversely, if kL1 (v, u)k < ε for (v, u) ∈ E G 1 \ G , k(L1 − L)(v, u)k < ε for (v, u) ∈
−
→ T− → −
→ − →
E G 1 G and k − L(v, u)k < ε for (v, u) ∈ E G \ G 1 , we easily know that
− →
X
′+
X
′+
→L1 −
A vu
A vu A+
G 1 − G L
=
L1 (v, u)
+
L1 − Lvuvu (v, u)
u∈NG1 \G (v) u∈NG1 T G (v)
X
A+
+
−L vu (v, u)
u∈NG\G1 (v)
X X
≤ c+
vu kL1 (v, u)k + c+
vu k (L1 − L) (v, u)k
u∈NG1 \G (v) u∈NG1 T
G (v)
X
+ c+
vu k − L(v, u)k
u∈NG\G1 (v)
− → − → max − → max − → max
→ − → \ − → − → [ −
< G 1 \ G cmax
G1 G ε + G 1 G c G1 G ε + G \ G c
1 G1 G ε = G 1 G cG1 G ε.
−
→ 1 − →L
Thus, we get an equivalent condition for lim G L
1 = G following.
t→t0
−
→ 1 − →L
Theorem 3.4 lim G L 1 = G if and only if for any number ε > 0 there is always a number δ(ε)
t→t0 −
→ − →
such that if |t − t0 | ≤ δ(ε) then kL1 (v, u)k < ε for (v, u) ∈ E G 1 \ G , k(L1 − L)(v, u)k < ε
−→ T− → −
→ − → −
→ 1 − →L
for (v, u) ∈ E G 1 G and k − L(v, u)k < ε for (v, u) ∈ E G \ G 1 ,i.e., G L 1 − G is an
−
→L1 − →L
infinitesimal or lim G 1 − G = O.
t→t0
36 Linfan MAO
−
→ −
→ −
→
If lim G L , lim G1 L1 and lim G2 L2 exist, the formulas following are clearly true by defi-
t→t0 t→t0 t→t0
nition:
−
→ →
− −
→ −
→
lim G1 L1 + G2 L2 = lim G1 L1 + lim G2 L2 ,
t→t0 t→t0 t→t0
−
→ →
− −
→ −
→
lim G1 L1 · G2 L2 = lim G1 L1 · lim G2 L2 ,
t→t0 t→t0 t→t0
−
→ − → −→ −
→ −
→ −
→ −
→
lim G L · G1 L1 + G2 L2 = lim G L · lim G1 L1 + lim G L · lim G2 L2 ,
t→t0 t→t0 t→t0 t→t0 t→t0
−→L1 − →L2 −→L −→L1 −
→L −
→ L2 −
→
lim G1 + G2 ·G = lim G1 · lim G + lim G2 · lim G L
t→t0 t→t0 t→t0 t→t0 t→t0
−
→
and furthermore, if lim G2 L2 6= O, then
t→t0
−
→
→ !
− − lim G1 L1
G1 L1 → L1 −
→L−1 t→t0
lim −
→ = lim G1 · G2 2 = → .
−
t→t0 G2 L2 t→t0 lim G2 L2
t→t0
−→ −
→
Theorem 3.5(L’Hospital’s rule) If lim G1 L1 = O, lim G2 L2 = O and L1 , L2 are differentiable
t→t0 t→t
− →
0
→ −
respect to t with lim L′1 (v, u) = 0 for (v, u) ∈ E G 1 \ G 2 , lim L′2 (v, u) 6= 0 for (v, u) ∈
− t→t0 t→t0
→ T− → ′
−
→ − →
E G 1 G 2 and lim L2 (v, u) = 0 for (v, u) ∈ E G 2 \ G 1 , then,
t→t0
−
→ ′
→ !
− lim G1 L1
G1 L1 t→t0
lim −
→ = → ′.
−
t→t0 G2 L2 lim G2 L2
t→t0
−
→ − → L1 − → L1 ·L2 −
→ \−
−1
→ − → L2
= lim G1 \ G2 G1 G2 G2 \ G1
t→t0
− → L1 ·L2
→ \−
−1 −
→ \−→ L−1
L1
= lim G1 G2 = lim G 1 G2 2
t→t0 t→t0
lim L′ 1
0 t→t
− → t→t
→ \− lim
L1
−1
− → t→t
→ \− lim L′
−1
L 2
= G1 G2 0 2 = G1 G2 0
−
→ − → lim L1 −
′
→ \− → t→t
lim L′ 1 · lim L′−1
2
−
→ − → lim L2
′
= G 1 \ G 2 t→t0 G1 G2 0 t→t0
G 2 \ G 1 t→t0
−
→L′
lim L ′
lim L ′−1 lim G 1 1
−
→t→t0 1 −
→t→t 2 t→t0
= G1 · G2 0 = →L′ .
−
lim G 2 2
t→t0
−
→ −
→
Corollary 3.6 If lim G L1 = O, lim G L2 = O and L1 , L2 are differentiable respect to t with
t→t0 −t→t
→
0
−
→ ′
→L1 !
− lim G L1
G t→t0
lim →L2 =
− → ′.
−
t→t0 G lim G L2
t→t0
−
→L −→ t − t0 −
→L′ (t0 ) (t − t0 )k −
→L(k) (t0 ) →
−
G = G L(t0 ) + G + ··· + G + o (t − t0 )−k G ,
1! k!
(k−1) (k−1)
if L1 (t0 ) = L′1 (t0 ) = · · · = L1 (t0 ) = 0 and L2 (t0 ) = L′2 (t0 ) = · · · = L2 (t0 ) = 0 but
(k)
L2 (t0 ) 6= 0, then
−
→L1 k
(t − t0 ) −→L(k) →
(t ) −k −
G1 = G 1 1 0 + o (t − t0 ) G 1 ,
k!
−
→L2
k
(t − t0 ) −→L(k) →
−
(t )
G2 = G 2 2 0 + o (t − t0 )−k G 2 .
k!
We are easily know the following result.
−
→ −
→ (k−1)
Theorem 3.7 If lim G1 L1 = O, lim G2 L2 = O and L1 (t0 ) = L′1 (t0 ) = · · · = L1 (t0 ) = 0
t→t0 t→t0
(k−1) (k)
and L2 (t0 ) = L′2 (t0 ) = · · · = L2 (t0 ) = 0 but L2 (t0 ) 6= 0, then
−
→L(k) (t )
−
→L1 lim G 1 1 0
G1 t→t0
lim →L = .
t→t0 −
G2 2 −
→L(k) (t )
lim G 2 2 0
t→t0
−
→ −
→ −
→
Example 3.8 Let G 1 = G 2 = C n , A+ +
vi vi+1 = 1, Avi vi−1 = 2 and
f1 + 2i−1 − 1 F (x) n!
fi = +
2i−1 (2n + 1)et
?f
fn
6
2
fi
f3
vn vi+1 vi v3
Fig.4
38 Linfan MAO
−
→ −
→
A continuity
−flow G L is in fact an operator L : G → B determined by L(v, u) ∈ B for
→
∀(v, u) ∈ E G . Generally, let
L11 L12 ··· L1n
L L22 ··· L2n
21
[L]m×n =
··· ··· ··· ···
Lm1 Lm2 · · · Lmn
−
→
with Lij : G → B for 1 ≤ i ≤ m, 1 ≤ j ≤ n, called operator matrix. Particularly, if for integers
−
→
1 ≤ i ≤ m, 1 ≤ j ≤ n, Lij : G → R, we can also determine its rank as the usual, labeled the
→
− −
→
edge (v, u) by Rank[L]m×n for ∀(v, u) ∈ E G and get a labeled graph G Rank[L] . Then we
get a result following.
where
L11 L12 ··· L1n L11 L12 ··· L1n L1
L L22 ··· L2n ··· L2
21 L21 L22 L2n
[L] = and L = .
··· ··· ··· ··· ··· ··· ··· ···
Ln1 Ln2 · · · Lnn Ln1 Ln2 · · · Lnn Ln
Hilbert Flow Spaces with Differentials over Graphs 39
−
→
Proof Clearly, if (4.1) is solvable, then for ∀(v, u) ∈ E G , the linear equations
x1 L11 (v, u) + x2 L12 (v, u) + · · · + xn Ln1 (v, u0 = L1 (v, u)
x L (v, u) + x L (v, u) + · · · + x L (v, u0 = L (v, u)
1 21 2 22 n 21 2
························
x1 Ln1 (v, u) + x2 Ln2 (v, u) + · · · + xn Lnn (v, u0 = Ln (v, u)
x1 L11 (v, u) + x2 L12 (v, u) + · · · + xn Ln1 (v, u0 = L1 (v, u)
x L (v, u) + x L (v, u) + · · · + x L (v, u0 = L (v, u)
1 21 2 22 n 21 2
························
x1 Ln1 (v, u) + x2 Ln2 (v, u) + · · · + xn Lnn (v, u0 = Ln (v, u)
Q −
→Lλ −
→
vu
→
n solutions G , and particularly, if L s (v, u) 6
= 0 for ∀(v, u) ∈ E G , there are
−
(v,u)∈E G
→
−
E G −
→
s solutions G Lλ of equation (4.3).
Proof By the fundamental theorem of algebra, we know there are s roots λvu vu vu
1 , λ2 , · · · , λs
−
→
for the equation (4.3). Whence, Lλ G is a solution of equation (4.2) because of
−
→s − → −→s−1 − → −
→0 −→
λ G · G Ls + λ G · G Ls−1 + · · · + λ G · G L0
−
→ s −
→ s−1 −
→ 0 −
→ s s−1
= G λ Ls + G λ Ls−1 + · · · + G λ L0 = G λ Ls +λ Ls−1 +···+L0
and
λs Ls + λs−1 Ls−1 + · · · + L0 : (v, u) → αvu s vu
s λ + αs−1 λ
s−1
+ · · · + αvu
0 = 0,
−
→
for ∀(v, u) ∈ E G , i.e.,
−
→s −→ −
→s−1 −→ −
→0 −→ −
→
λ G · G Ls + λ G · G Ls−1 + · · · + λ G · G L0 = 0 G = O.
→
−
Count the number of different Lλ for (v, u) ∈ E G . It is nothing else but just nvu .
Q
Therefore, the number of solutions of equation (4.3) is
nvu .
→
−
2
(v,u)∈E G
−
→L −→ →
−
G = G Lβ · etLα G ,
−
→
where Lα : (v, u) → αvu , Lβ : (v, u) → βvu are constants for ∀(v, u) ∈ E G .
i.e.,
−
→
dGL −→ −
→
= G Lα · G L
dt
−
→ −
→ →
−
if G L = G Lβ · etLα G . This completes the proof. 2
−
→
Theorem 4.3 can be generalized to the case of L : (v, u) → Rn , n ≥ 2 for ∀(v, u) ∈ E G .
−
→L −→ →
−
G = G Lβ · etLα G ,
where Lα : (v, u) → α1vu , α2vu , · · · , αnvu
1
, Lβ : (v, u) → βvu 2
, βvu n
, · · · , βvu with constants
i i −
→
αvu , βvu , 1 ≤ i ≤ n for ∀(v, u) ∈ E G .
− →
for (v, u) ∈ E G , where hmi (t) is a polynomial of degree≤ mi −1 on t, m1 +m2 +· · ·+ms = n
and λvu vu vu
1 , λ2 , · · · , λs are the distinct roots of characteristic equation
αvu n vu
n λ + αn−1 λ
n−1
+ · · · + αvu
0 = 0
42 Linfan MAO
−
→
with αvu
n 6
= 0 for (v, u) ∈ E G .
−
→
Proof Clearly, the equation (4.8) on an edge (v, u) ∈ E G is
i.e.,
−
→(λn ·Lαn +λn−1 ·Lαn−1 +···+Lα0 )·L
G = O,
−
→
which implies that for ∀(v, u) ∈ E G ,
λn αvu
n +λ
n−1 vu
αn−1 + · · · + α0 = 0 (4.10)
or
L(v, u) = 0.
Let λvu vu vu vu vu vu
1 , λ2 , · · · , λs be the distinct roots with respective multiplicities m1 , m2 , · · · , ms
of equation (4.8). We know the general solution of (4.9) is
s
X vu
L(v, u) = hi (t)eλi t
i=1
with hmi (t) a polynomial of degree≤ mi −1 on t by the theory of ordinary differential equations.
−
→
Therefore, the general solution of (4.8) is G Lλ with
( s
)
X
λvu
i t
Lλ : (v, u) → 0, hi (t)e
i=1
−
→
for (v, u) ∈ E G . 2
Hilbert Flow Spaces with Differentials over Graphs 43
−
→ −
→
The difference of a complex flow G L with that of a continuity flow G L is the labeling L on a
vertex is L(v) = ẋv or xv . Notice that
d X + X d A+
LAvu (v, u) = L vu (v, u)
dt dt
u∈NG (v) u∈NG (v)
−
→ −
→ −
→
for ∀v ∈ V G . There must be relations between complex flows G L and continuity flows G L .
We get a general result following.
Z t Z t
+ + + +
Theorem 5.1 If end-operators Avu , Auv are linear with , Avu = , Auv = 0 and
0 0
d + d + −
→ −
→L
,A = ,A = 0 for ∀(v, u) ∈ E G , and particularly, A+ vu = 1V , then G ∈
dt vu dt uv
D−→ ER×R n −
→
G i, 1 ≤ i ≤ n is a continuity flow with a constant L(v) for ∀v ∈ V G if and only if
Z t −
−
→L →
G dt is such a continuity flow with a constant one each vertex v, v ∈ V G .
0
Z t
d +
Proof Notice that if A+
vu = 1V , there always is , A+
vu = 0 and , A = 0, and by
0 dt vu
definition, we know that
Z t Z t Z t
, A+
vu =0 ⇔ ◦A+ +
vu = Avu ◦ ,
0 0 0
d + d d
,A =0 ⇔ ◦ A+ +
vu = Avu ◦ .
dt vu dt dt
−
→ −
→
If G L is a continuity flow with a constant L(v) for ∀v ∈ V G , i.e.,
X +
−
→
LAvu (v, u) = v for ∀v ∈ V G ,
u∈NG (v)
X Z t Z t
= A+
vu L(v, u)dt = vdt
u∈NG (v) 0 0
− Z t Z t
→ −
→L
for ∀v ∈ V G with a constant vector vdt, i.e., G dt is a continuity flow with a
−
→
0 0
constant flow on each vertex v, v ∈ V G .
Z t
−
→L
Conversely, if G dt is a continuity flow with a constant flow on each vertex v, v ∈
0
44 Linfan MAO
−
→
V G , i.e.,
Z −
X t
→
A+
vu ◦ L(v, u)dt = v for ∀v ∈ V G ,
u∈NG (v) 0
then Z
t
−
→L
d G dt
−
→L 0
G =
dt
−
→
is such a continuity flow with a constant flow on vertices in G because of
!
P A+
d L vu (v, u) Z
X t
u∈NG (v) d
= ◦ A+
vu ◦ L(v, u)dt
dt dt 0
u∈NG (v)
X Z t X
d + dv
= A+
vu ◦ ◦ L(v, u)dt = L(v, u)Avu =
dt 0 dt
u∈NG (v) u∈NG (v)
dv −
→
with a constant flow
dt
on vertex v, v ∈ V G . This completes the proof. 2
− Z t
+ + → +
If all end-operators Avu and Auv are constant for ∀(v, u) ∈ E G , the conditions , Avu =
Z t 0
d d +
, A+
uv = 0 and , A+
vu = , Auv = 0 are clearly true. We immediately get a conclu-
0 dt dt
sion by Theorem 5.1 following.
−→
Corollary 5.2 For ∀(v, u) ∈ E G , if end-operators A+ +
vu and Auv are constant cvu , cuv for
−
→ −
→ D−
→ E R×R n
References