Extending Banach Space on Graph with

−
→
Conservation Laws and G -Flow Solutions of Equations

Linfan MAO

(Chinese Academy of Mathematics and System Science, Beijing 100190, P.R.China)

E-mail: maolinfan@163.com

−
→
Abstract: Let V be a Banach space over a field F . A G -flow is a graph
−
→
G embedded in a topological space S associated with an injective mappings
 
−
→
L : uv → L(uv ) ∈ V such that L(uv ) = −L(v u ) for ∀(u, v) ∈ X G holding
with conservation laws
X  
−
→
L (v u ) = 0 for ∀v ∈ V G ,
u∈NG (v)
 
−
→
where uv denotes the semi-arc of (u, v) ∈ X G , which is an abstract model,
also a mathematical object for things embedded in a topological space, or
matters happened in the world. The main purpose of this paper is to ex-
tend Banach spaces on topological graphs with operator actions and show
all of these extensions are also Banach space with unique correspondence in
elements on linear continuous functionals, which enables one to solve linear
functional equations in such extended space, particularly, solve algebraic, dif-
ferential or integral equations on a topological graph, i.e., find multi-space
solutions for equations, for instance, the Einstein’s gravitational equations. A
generalization of some well-known results in classical mathematics, such as
those of the fundamental theorem in algebra, Hilbert and Schmidt’s result on
−
→
integral equations and the discussion on stability of such G -flow solutions with
applications to controlling of ecologically industrial systems can be also found
in this paper. All of these results in this paper establish the mathematical
foundation for multi-spaces, i.e., mathematical combinatorics.

Key Words: Banach space, topological graph, conservation flow, topological

graph, differential flow, multi-space solution of equation, system control.

AMS(2010): 03A10,05C15,20A05, 34A26,35A01,51A05,51D20,53A35

1
§1. Introduction

−
→ −
→ −
→
Let V be a Banach space over a field F . All graphs G , denoted by (V ( G ), X( G ))
−
→
considered in this paper are strong-connected without loops. A topological graph G
−
→
is an embedding of an oriented graph G in a topological space C . All elements in
−
→ −
→ −
→
V ( G ) or X( G ) are respectively called vertices or arcs of G .
−
→
An arc e = (u, v) ∈ X( G ) can be divided into 2 semi-arcs, i.e., initial semi-arc
uv and end semi-arc v u , such as those shown in Fig.1 following.

u u
v
-
L(uv ) vu v

Fig.1
 
−
→ −
→
All these semi-arcs of a topological graph G are denoted by X 1 G .
2
−
→L −
→ −
→
A vector labeling G on G is a 1 − 1 mapping L : G → V such that L : uv →
−
→ −
→L
L(uv ) ∈ V for ∀uv ∈ X 1 G , such as those shown in Fig.1. For all labelings G
2
−
→
on G , define
−
→L1 − →L2 − →L1 +L2 −
→L − →λL
G +G =G and λ G = G .
−
→
Then, all these vector labelings on G naturally form a vector space.
  Particularly,
−
→ −
→ −
→
a G -flow on G is such a labeling L : uv → V for ∀uv ∈ X 1 G hold with
2

L (uv ) = −L (v u ) and conservation laws

X
L (v u ) = 0
u∈NG (v)

−
→
for ∀v ∈ V ( G ), where 0 is the zero-vector in V . For example, a conservation law
for vertex v in Fig.2 is −L(v u1 ) − L(v u2 ) − L(v u3 ) + L(v u4 ) + L(v u5 ) + L(v u6 ) = 0.
u1
u1 L(v )- -L(vu ) u4
4

u2
u2 L(v )- -L(vu ) u5
5

v
u3
u3 L(v )- -F (vu ) u6
6

Fig.2

2
 −
→ −
→L
Clearly, if V = Z and O = {1}, then the G -flow G is nothing else but the network
−
→ −
→
flow X( G ) → Z on G .
−
→L − →L1 −→L2 −
→ −
→
Let G , G , G be G -flows on a topological graph G and ξ ∈ F a scalar.
−
→L1 − →L2 −
→L −
→
It is clear that G + G and ξ · G are also G -flows, which implies that all
−
→ −
→ −
→0
conservation G -flows on G also form a linear space over F with unit G under
−
→V −
→0 −
→
operations + and·, denoted
 by G , where G is such a G -flow with vector 0 on
−
→ −
→
uv for (u, v) ∈ X G , denoted by O if G is clear by the paragraph..
The flow representation for graphs are first discussed in [5], and then applied to
differential operators in [6], which has shown its important role both in mathematics
and applied sciences. It should be noted that a conservation law naturally deter-
−
→
mines an autonomous systems in the world. We can also find G -flows by solving
conservation equations
X  
−
→
L (v u ) = 0, v∈V G .
u∈NG (v)

−
→
Such a system of equations is non-solvable in general, only with G -flow solutions
−
→
such as those discussions in references [10]-[19]. Thus we can also introduce G
 -flows
by Smarandache multi-system ([21]-[22]). In fact, for any integer m ≥ 1 let Σ; e R e
be a Smarandache multi-system consisting of m mathematical systemsh(Σ1 ; R i 1 ),
L e e
(Σ2 ; R2 ), · · ·, (Σm ; Rm ), different two by two. A topological structure G Σ; R on
 
e R
Σ; e is inherited by
 h i
V GL Σ; e R e = {Σ1 , Σ2 , · · · , Σm },
 h i T
E GL Σ; e R e = {(Σi , Σj ) |Σi Σj 6= ∅, 1 ≤ i 6= j ≤ m} with labeling
T
L : Σi → L (Σi ) = Σi and L : (Σi , Σj ) → L (Σi , Σj ) = Σi Σj

i 1 ≤ i 6= j ≤ m,Ti.e., a topological vertex-edge labeled graph. Clearly,

for hintegers
GL Σ; e Re is a −
→
G -flow if Σi Σj = v ∈ V for integers 1 ≤ i, j ≤ m.
The main purpose of this paper is to establish the theoretical foundation, i.e.,
extending Banach spaces, particularly, extended Hilbert spaces on topological graphs
with operator actions and show all of these extensions are also Banach space with
unique correspondence in elements on linear continuous functionals, which enables
one to solve linear functional equations in such extended space, particularly, solve
algebraic or differential equations on a topological graph, i.e., find multi-space solu-

3
tions for equations, such as those of algebraic equations, the Einstein gravitational
equations and integral equations with applications to controlling of ecologically in-
dustrial systems. All of these discussions provide new viewpoint for mathematical
elements, i.e., mathematical combinatorics.
For terminologies and notations not mentioned in this section, we follow ref-
erences [1] for functional analysis, [3] and [7] for topological graphs, [4] for linear
spaces, [8]-[9], [21]-[22] for Smarandache multi-systems, [3], [20] and [23] for differ-
ential equations.

−
→
§2. G -Flow Spaces

2.1 Existence

Definition 2.1 Let V be a Banach space. A family V of vectors v ∈ V is conser-

vative if
X
v = 0,
v∈V
called a conservative family.

Let V be a Banach space over a field F with a basis {α1 , α2 , · · · , αdimV }. Then,
for v ∈ V there are scalars xv1 , xv2 , · · · , xvdimV ∈ F such that
dimV
X
v= xvi αi .
i=1

Consequently, !
X dimV
X dimV
X X
xvi αi = xvi αi = 0
v∈V i=1 i=1 v∈V
implies that
X
xvi = 0
v∈V
for integers 1 ≤ i ≤ dimV .
Conversely, if
X
xvi = 0, 1 ≤ i ≤ dimV ,
v∈V
define
dimV
X
i
v = xvi αi
i=1

4
 P
and V = {vi , 1 ≤ i ≤ dimV }. Clearly, v = 0, i.e., V is a family of conservation
v∈V
vectors. Whence, if denoted by xvi = (v, αi ) for ∀v ∈ V , we therefore get a condition
on families of conservation in V following.

Theorem 2.2 Let V be a Banach space with a basis {α1 , α2 , · · · , αdimV }. Then, a
vector family V ⊂ V is conservation if and only if
X
(v, αi ) = 0
v∈V

for integers 1 ≤ i ≤ dimV .

For example, let V = {v1 , v2 , v3 , v4 } ⊂ R3 with

v1 = (1, 1, 1), v2 = (−1, 1, 1),

v3 = (1, −1, −1), v4 = (−1, −1, −1)

Then it is a conservation family of vectors in R3 .

Clearly, a conservation flow consists of conservation families. The following
result establishes its inverse.
−
→ −
→L −
→
Theorem 2.3 A G -flow G exists on G if and only if there are conservation
families L(v) in a Banach space V associated an index set V with

L(v) = {L(v u ) ∈ V f or some u ∈ V }

such that L(v u ) = −L(uv ) and

\
L(v) (−L(u)) = L(v u ) or ∅.

Proof Notice that

X
L(v u ) = 0
u∈NG (v)
 
−
→
for ∀v ∈ V G implies
X
L(v u ) = − L(v w ).
w∈NG (v)\{u}

Whence, if there is an index set V associated conservation families L(v) with

L(v) = {L(v u ) ∈ V for some u ∈ V }

5
 T
for ∀v ∈ V such that L(v u ) = −L(uv ) and L(v) (−L(u)) = L(v u ) or ∅, define a
−
→
topological graph G by
    [
−
→ −
→
V G = V and X G = {(v, u)|L(v u ) ∈ L(v)}
v∈V

−
→L −
→
with an orientation v → u on its each arcs. Then, it is clear that G is a G -flow
by definition.
−
→L −
→
Conversely, if G is a G -flow, let
 
−
→
L(v) = {L(v u ) ∈ V for ∀(v, u) ∈ X G }
   
−
→ −
→
for ∀v ∈ V G . Then, it is also clear that L(v), v ∈ V G are conservation
 
−
→
families associated with an index set V = V G such that L(v, u) = −L(u, v) and
  
\  L(v u ) if (v, u) ∈ X −→
G
L(v) (−L(u)) =  
 ∅ −
→
if (v, u) 6∈ X G

by definition. 

Theorems 2.2 and 2.3 enables one to get the following result.
−
→ −
→
Corollary 2.4 There are always existing G -flows on a topological
   graph   Gwith
−
→  −
→   −
→ 
weights λv for v ∈ V , particularly, λe αi on ∀e ∈ X G if X G  ≥ V G +1.
 
−
→
Proof Let e = (u, v) ∈ X G . By Theorems 2.2 and 2.3, for an integer
−
→
1 ≤ i ≤ dimV , such a G -flow exists if and only if the system of linear equations
X  
−
→
λ(v,u) = 0, v ∈ V G

→
−
u∈V G

     
 −
→   −
→ 
is solvable. However, if X G  ≥ V G  + 1, such a system is indeed solvable
by theory of linear equations. 
−
→
2.2 G -Flow Spaces

Define L
−
→ X
G = kL(uv )k
−
→
(u,v)∈X G

6
 −
→L − →V
for ∀ G ∈ G , where kL(uv )k is the norm of F (uv ) in V . Then
L L
−
→ −
→ −
→L − →0
(1) G ≥ 0 and G = 0 if and only if G = G = O.
ξL L
−
→ −
→
(2) G = ξ G for any scalar ξ.
L L L
−
→ 1 − →L2 − → 1 − → 2
(3) G + G ≤ G + G because of
L X
−
→ 1 −
→L2
G + G = kL1 (uv ) + L2 (uv )k
−
→
(u,v)∈X G
X X L L
v v → 1 −
− → 2
≤ kL1 (u )k + kL2 (u )k = G + G .

→
− 
→
−
(u,v)∈X G (u,v)∈X G

−
→V
Whence, k · k is a norm on linear space G .
Furthermore, if V is an inner space with inner product h·, ·i, define
D L E X
−
→ 1 → − L2
G ,G = hL1 (uv ), L2 (uv )i .
 
−
→
(u,v)∈X G

Then we know that

D L LE P D L LE
−
→ → − v v −
→ −→
(4) G , G =  
hL(u ), L(u )i ≥ 0 and G ,G = 0 if and only
−
→
(u,v)∈X G

−
→ −
→L
if L(uv ) = 0 for ∀(u, v) ∈ X( G ), i.e., G = O.
D L1 E D L2 E
−
→ − →L2 −
→ − →L1 −
→L1 → − L2 − →V
(5) G , G = G ,G for ∀ G , G ∈ G because of
D E X X
−
→L1 →
− L2
G ,G = hL1 (uv ), L2 (uv )i = hL2 (uv ), L1 (uv )i
−
→ −
→
(u,v)∈X G (u,v)∈X G
X D E
−
→L2 −
→L1
= hL2 (uv ), L1 (uv )i = G ,G

→
−
(u,v)∈X G

−
→L −→L1 −→L2 −
→V
(6) For G , G , G ∈ G , there is
D E
−
→L1 −
→L2 −→L
λG + µG , G
D L1 E D L2 E
−
→ → −L −
→ → −L
= λ G ,G +µ G ,G

7
because of
D E D λL1 E
−
→L1 −
→L2 − →L −
→ −
→µL2 − →L
λG + µG , G = G +G ,G
X
= hλL1 (uv ) + µL2 (uv ), L(uv )i
−
→
(u,v)∈X( G )
X X
= hλL1 (uv ), L(uv )i + hµL2 (uv ), L(uv )i
−
→ −
→
(u,v)∈X( G ) (u,v)∈X( G )
D λL E D µL E
−
→ 1 − →L −
→ 2 − →L
= G ,G + G ,G
D L1 E D L2 E
−
→ − →L −
→ − →L
= λ G ,G +µ G ,G .

−
→V
Thus, G is an inner space also and as the usual, let
L rD L L E
−
→ −
→ − →
G = G ,G

−
→L −
→V
for G ∈ G . Then it is a normed space. Furthermore, we know the following
result.
−
→ − →V
Theorem 2.5 For any topological graph G , G is a Banach space, and furthermore,
−
→V
if V is a Hilbert space, G is a Hilbert space also.
−
→V
Proof As shown in the previous, G is a linear normed space or inner space if
V is an inner space. We show nthat ito is also complete, i.e., any Cauchy sequence in
−
→V −
→Ln −
→V
G is converges. In fact, let G be a Cauchy sequence in G . Thus for any
number ε > 0, there always exists an integer N(ε) such that
L
−→ n − →Lm
G − G <ε

if n, m ≥ N(ε). By definition,
L
−
→ n −→Lm
kLn (uv ) − Lm (uv )k ≤ G − G < ε
 
v −
→
i.e., {Ln (u )} is also a Cauchy sequence for ∀(u, v) ∈ X G , which is converges on
in V by definition.  
−
→
Let L(uv ) = lim Ln (uv ) for ∀(u, v) ∈ X G . Then it is clear that
n→∞

−
→Ln −
→L
lim G = G .
n→∞

8
 −
→L −→V
However, we are needed to show G ∈ G . By definition,
X
Ln (uv ) = 0
v∈NG (u)
 
−
→
for ∀u ∈ V G and integers n ≥ 1. Let n → ∞ on its both sides. Then
 
X X
lim  Ln (uv ) = lim Ln (uv )
n→∞ n→∞
v∈NG (u) v∈NG (u)
X
= L(uv ) = 0.
v∈NG (u)

−
→L −→V
Thus, G ∈ G . 
−
→ −
→L1 −
→L2
D Similarly,
E two conservation G -flows G and G are said to be orthogonal if
−
→L1 → − L2
G ,G = 0. The following result characterizes those of orthogonal pairs of
−
→
conservation G -flows.
−
→L1 → − L2 − →V −
→L1 −
→L2
Theorem 2.6 let G , G ∈ G . Then  G is orthogonal to G if and only if
−
→
hL1 (uv ), L2 (uv )i = 0 for ∀(u, v) ∈ X G .
 
−
→
Proof Clearly, if hL1 (uv ), L2 (uv )i = 0 for ∀(u, v) ∈ X G , then,
D E X
−
→L1 −
→L2
G ,G = hL1 (uv ), L2 (uv )i = 0,

→
−
(u,v)∈X G

−
→L1 −
→L2
i.e., G is orthogonal to G .
−
→L1 −
→L2
Conversely, if G is indeed orthogonal to G , then
D L1 E X
−
→ → − L2
G ,G = hL1 (uv ), L2 (uv )i = 0
−
→ 
(u,v)∈X G

 
−
→
by definition. We therefore know that hL1 (uv ), L2 (uv )i = 0 for ∀(u, v) ∈ X G
because of hL1 (uv ), L2 (uv )i ≥ 0. 

Theorem 2.7 Let V be a Hilbert space with an orthogonal decomposition V =

V ⊕ V⊥ for a closed subspace V ⊂ V . Then there is a decomposition
−
→V e ⊕V
e ⊥,
G =V

9
where,
n L    o
e = −
V
→ 1 −→V  −
→
G ∈ G L1 : X G → V
n L2    o
e −
→ −
→V  −
→
V = G ∈ G  L2 : X G → V⊥ ,
⊥

−
→L − →V
i.e., for ∀ G ∈ G , there is a uniquely decomposition
−
→L − →L1 −
→L2
G =G +G
   
−
→ −
→
with L1 : X G → V and L2 : X G → V⊥ .
 
−
→
Proof By definition, L(uv ) ∈ V for ∀(u, v) ∈ X G . Thus, there is a decom-
position
L(uv ) = L1 (uv ) + L2 (uv )
v v ⊥
with uniquely h L iL1 (u ) ∈ V but L2 (u ) ∈ V .
h L idetermined  
−
→ 1 −
→ 2 −
→ −
→
Let G and G be two labeled graphs on G with L1 : X 1 G → V
  h L1 i h L2 i D 2
E
−
→ −
→ −
→ −
→
and L2 : X 1 G → V⊥ . We need to show that G , G ∈ G , V . In fact,
2

the conservation laws show that

X X
L(uv ) = 0, i.e., (L1 (uv ) + L2 (uv )) = 0
v∈NG (u) v∈NG (u)
 
−
→
for ∀u ∈ V G . Consequently,
X X
L1 (uv ) + L2 (uv ) = 0.
v∈NG (u) v∈NG (u)

Whence,
* +
X X X X
0 = L1 (uv ) + L2 (uv ), L1 (uv ) + L2 (uv )
v∈NG (u) v∈NG (u) v∈NG (u) v∈NG (u)
* + * +
X X X X
= L1 (uv ), L1 (uv ) + L2 (uv ), L2 (uv )
v∈NG (u) v∈NG (u) v∈NG (u) v∈NG (u)
* + * +
X X X X
+ L1 (uv ), L2 (uv ) + L2 (uv ), L1 (uv )
v∈NG (u) v∈NG (u) v∈NG (u) v∈NG (u)
* + * +
X X X X
= L1 (uv ), L1 (uv ) + L2 (uv ), L2 (uv )
v∈NG (u) v∈NG (u) v∈NG (u) v∈NG (u)

10
 X X
+ hL1 (uv ), L2 (uv )i + hL2 (uv ), L1 (uv )i
v∈NG (u) v∈NG (u)
* + * +
X X X X
= L1 (uv ), L1 (uv ) + L2 (uv ), L2 (uv ) .
v∈NG (u) v∈NG (u) v∈NG (u) v∈NG (u)

Notice that
* + * +
X X X X
L1 (uv ), L1 (uv ) ≥ 0, L2 (uv ), L2 (uv ) ≥ 0.
v∈NG (u) v∈NG (u) v∈NG (u) v∈NG (u)

We therefore get that

* + * +
X X X X
L1 (uv ), L1 (uv ) = 0, L2 (uv ), L2 (uv ) = 0,
v∈NG (u) v∈NG (u) v∈NG (u) v∈NG (u)

i.e.,
X X
L1 (uv ) = 0 and L2 (uv ) = 0.
v∈NG (u) v∈NG (u)
h Li h Li
−
→ 1 −
→ 2 −
→V
Thus, G , G ∈ G . This completes the proof. 

−
→
2.3 Solvable G -Flow Spaces
  
−
→L −
→ −
→
Let G be a G -flow. If for ∀v ∈ V G , all flows L (v u ) , u ∈ NG+ (v) \ u+
0 are
determined by equations


Fv L (v u ) ; L (w v ) , w ∈ NG− (v) = 0
−
→
unless L(v u0 ), then G -flow is called
 solvable,
 and L(v u0 ) the co-flow at vertex v.
−
→ +
Such a G -flow is linear if each L v u , u+ ∈ NG+ (v) \ {u+
0 } is determined by

 + X v

L vu = au− L u− ,
−
u− ∈NG (v)
   +
−
→
with scalars au− ∈ F for ∀v ∈ V G unless L v u0 , and is ordinary or partial
differential if L (v u ) is determined by ordinary differential equations
  
d + v

Lv ;1 ≤ i ≤ n L(v u ); L u− , u− ∈ NG− (v) = 0
dxi
or   
∂ u+ −v

− −
Lv ; 1 ≤ i ≤ n L(v ); L u , u ∈ NG (v) = 0
∂xi

11
  +      
u0 −
→ d ∂
unless L v for ∀v ∈ V G , where, Lv ; 1 ≤ i ≤ n , Lv ;1 ≤ i ≤ n
dxi ∂xi
denote an ordinary or partial differential operators,
 respectively.  
+ v +
Notice that a linear equation au+ L v u + L (u− ) = 0 is solvable for L v u
if and only if au+ 6= 0, such as those shown in Fig.3 following.

v1 -
(0, 1) v2
(1, 0)

(0, 1) ? 6(1, 0) (1, 0) ? 6(0, 1)


(1, 0)
v4
-
(0, 1) v3

Fig.3
−
→
We consequently know the following result on linear solvable G -flows.
−
→ −
→ −
→L
Theorem 2.8 For a strong-connected graph G , there exist linear G -flows G , not
−
→
all flows being zero on G .
−
→
Proof Notice that G is strong-connected. There must be a decomposition
m1
! m2
!
−
→ [ → [ [−
− →
G= Ci Ti ,
i=1 i=1

−
→ − → −
→
where C i , T j are respectively directed circuit or path in G with m 1≥ 1, m2 ≥ 0.
−
→ u k
For an integer 1 ≤ k ≤ m1 , let C k = uk1 uk2 · · · uksk and L uki i+1 = vk , where
−
→ j j j
i + 1 ≡ (mods). Similarly, for integers 1 ≤ j ≤ m2 , if T j = w  1 w2 · · · wt , let

−
→
F wjt , wj+1
t
= 0. Clearly, the conservation law hold at ∀v ∈ V G by definition,
 k 
u
and each flow L uki i+1 , i + 1 ≡ (mods) is linear determined by

 uk
  k
 X X  k X m2 X  k
k ui
L uki i+1 = L ui−1 +0× L v ui
+ L v ui
j6=i v∈N − (uk ) j=1 v∈N − (uk )
C i j T j i

m2
X X
= vk + 0 + 0 = vk .
j=1 v∈NT− (uki )
j

−
→L −
→ −
→
Thus, G is a linear solvable G -flow and not all flows being zero on G . 

12
 
f(x, t)
v1 - v2
x

f(x, t) ? 6x x ? 6f(x, t)
x
v4
- v3
f(x, t)

Fig.4
−
→
All G -flows constructed in Theorem 2.8 can be also replaced by vectors de-
pendent on the time t, i.e., v(t). According to the theory of ordinary differ-
ential equations ([23]), the differential equation ẋ = f(x,  t) with x|t=t0 =x0 is
dx1 dx2 dxn
always solvable in a neighborhood of x0 , where, ẋ = , ,···, and
dt dt dt 
−
→
f(x, t) = (f1 (x, t), f2 (x, t), · · · , fn (x, t)). Thus, if there is a vertex v ∈ V G with
−
→ →−′
ρ− (v) ≥ 2 and ρ+ (v) ≥ 2, i.e., v is in at least 2 directed circuits C , C , let flows on
−
→ −
→′
C and C be respectively x and f(x, t). Similar to the proof of Theorem 2.8, we
−
→
know the conservation laws hold for vertices in G , and there are indeed flows on
−
→
G determined by ordinary differential equations. We therefore know the following
result.
−
→
Theorem
  2.9 Let G be a strong-connected graph. If there exists a vertex v ∈
−
→ −
→
V G with ρ− (v) ≥ 2 and ρ+ (v) ≥ 2 then there exist ordinary differential G -flows
−
→L −
→
G , not all flows being zero on G .
−
→ −
→
For example, the G -flow shown in Fig.4 is an ordinary differential G -flow in a
vector space if ẋ = f(x, t) is solvable with x|t=t0 = x0 .
−
→
Similarly, we know the result on partial differential G -flows. By the theory of
partial differential equations of first order ([3]), let x = (x1 , x2 , · · · , xn ). If


 xi = xi (t, s1 , s2 , · · · , sn−1)

u = u(t, s1 , s2 , · · · , sn−1 )


 p = p (t, s , s , · · · , s ), i = 1, 2, · · · , n
i i 1 2 n−1

13
is a solution of system
dx1 dx2 dxn du
= = ··· = = P
n
Fp1 Fp2 Fpn
pi Fpi
i=1
dp1 dpn
= − = ··· = − = dt
Fx1 + p1 Fu Fxn + pn Fu

with initial values



 xi0 = xi0 (s1 , s2 , · · · , sn−1 )

u0 = u0 (s1 , s2 , · · · , sn−1 )


 p = p (s , s , · · · , s ), i = 1, 2, · · · , n
i0 i0 1 2 n−1

such that 

 F (x10 , x20 , · · · , xn0 , u, p10 , p20 , · · · , pn0 ) = 0
n
∂u0 X ∂xi0

 ∂s − pi0 = 0, j = 1, 2, · · · , n − 1,
j i=0
∂s j

then it is the solution of partial differential equation

F (x1 , x2 , · · · , xn , u, p1, p2 , · · · , pn ) = 0

∂u ∂F
with the same initial values, where pi = and Fpi = for integers 1 ≤ i ≤ n.
∂xi ∂pi
For partial differential equations of second order, the Cauchy problems on heat
or wave equations
X ∂2un X ∂2u n
∂u ∂2u
= a2 2
, = a2
∂t i=1
∂xi ∂t2 i=1
∂x2i
with initial values 
∂u 
u|t=0 = ϕ (x) , = ψ (x)
∂t t=0
are solvable. For example, we know closed formulae ([3])
Z +∞
1 (x1 −y1 )2 +···+(xn −yn )2
u (x, t) = n e− 4t ϕ(y1 , · · · , yn )dy1 · · · dyn
(4πt) 2 −∞

for heat equations, and

 
Z Z
∂  1  1
u (x1 , x2 , x3 , t) =  ϕdS  + ψdS
∂t 4πa2 t 4πa2 t
M
Sat M
Sat

14
 M
for wave equations in n = 3, where Sat denotes the sphere centered at M (x1 , x2 , x3 )
−
→
with radius at. Similar to Theorem 2.9, we get a result on partial G -flows following.
−
→
Theorem
  2.10 Let G be a strong-connected graph. If there exists a vertex v ∈
−
→ −
→
V G with ρ− (v) ≥ 2 and ρ+ (v) ≥ 2 then there exist partial differential G -flows
−
→L −
→
G , not all flows being zero on G .

−
→
§3. Operators on G -Flow Spaces

3.1 Linear Continuous Operators

Definition 3.1 Let T ∈ O be an operator on Banach space V over a field F . An

−
→V −
→V
operator T : G → G is bounded if
 L  L
−
→ −
→
T G ≤ ξ G
−
→L − →V
for ∀ G ∈ G with a constant ξ ∈ [0, ∞) and furthermore, is a contractor if
 L  L  L
−→ 1 −
→ 2 −
→ 1 − →L2
T G −T G ≤ ξ G − G )
−
→L1 →− L1 −
→V
for ∀ G , G ∈ G with ξ ∈ [0, 1).
−
→V −
→V
Theorem 3.2 Let T : G → G be a contractor. Then there is a uniquely
−
→L −→V
conservation G-flow G ∈ G such that
 L
−
→ −
→L
T G =G .
n Ln o
−
→L0 −
→V −
→
Proof Let G ∈ G be a G-flow. Define a sequence G by
 L
−
→L1 −
→ 0
G =T G ,
 L  L
−
→ 2
L −
→ 1 −
→ 0
G =T G = T2 G ,
................................ ,
 Ln−1   L0 
−
→Ln −
→ −
→
G =T G = Tn G ,
................................ .
n Ln o
−
→ −
→V
We prove G is a Cauchy sequence in G . Notice that T is a contractor.
For any integer m ≥ 1, we know that
L  L   Lm−1 
−
→ m+1 − →Lm −
→ m −
→
G − G = T G −T G

15
 L  L   L 
−
→ m − →Lm−1 −
→ m−1 −
→ m−2
≤ ξ G − G = T G −T G
L
−
→ m−1 −
→Lm−2 −→L1 −
→L0
≤ ξ2 G −G ≤ · · · ≤ ξm G − G .

Applying the triangle inequality, for integers m ≥ n we therefore get that

L
−
→ m − →Ln
G − G
L L
−→ m − →Lm−1 −→ n−1 − →Ln
≤ G − G + + · · · + G −G

−
→L1 − →L0
≤ ξ m + ξ m−1 + · · · + ξ n−1 × G − G
ξ n−1 − ξ m −→L1 − →L0

= × G − G
1−ξ
ξ n−1 L
−→ 1 − →L0
≤ × G − G for 0 < ξ < 1.
1−ξ
L n Ln o
−
→ m − →Ln −
→
Consequently, G − G → 0 if m → ∞, n → ∞. So the sequence G
−
→L
is a Cauchy sequence and converges to G . Similar to the proof of Theorem 2.5, we
−
→L − →V
know it is a G-flow, i.e., G ∈ G . Notice that
L  L  L L  L 
−→ −
→ −→ −
→Lm − → m −
→
G − T G ≤ G − G + G − T G
L L
−→ −
→Lm −
→ m−1 − →L
≤ G − G + ξ G − G .
L  L 
−
→ −
→
Let m → ∞. For 0 < ξ < 1, we therefore get that G − T G = 0, i.e.,
 L
−
→ −
→L
T G =G .
−
→L′ −
→V
For the uniqueness,
 L′  if there is an another conservation G-flow G ∈ G
−
→ −
→L
holding with T G = G , by
L  L  L′  L
−
→ −
→L′ −
→ −
→ −
→ −
→L′
G − G = T G − T G ≤ ξ G − G ,
−
→L −→L′
it can be only happened in the case of G = G for 0 < ξ < 1. 
−
→V −
→V
Definition 3.3 An operator T : G → G is linear if
   L  L
−
→L1 −
→L2 −
→ 1 −
→ 2
T λG + µG = λT G + µT G
−
→L1 −→L2 −
→V −
→ −
→L0
for ∀ G , G ∈ G and λ, µ ∈ F , and is continuous at a G -flow G if there
always exist a number δ(ε) for ∀ǫ > 0 such that
 L  L0  L
−
→ −
→ −
→ −
→L0
T G − T G < ε if G − G < δ(ε).

16
 The following result reveals the relation between conceptions of linear continu-
ous with that of linear bounded.
−
→V −
→V
Theorem 3.4 An operator T : G → G is linear continuous if and only if it is
bounded.

Proof If T is bounded, then

 L  L0   L   L 
−
→ −
→ −
→ −
→L0 −
→ −
→L0
T G − T G = T G − G ≤ ξ G − G

−
→L − →L0 −
→V
for an constant ξ ∈ [0, ∞) and ∀ G , G ∈ G . Whence, if
L ε
−→ −
→L0
G − G < δ(ε) with δ(ε) = , ξ = 6 0,
ξ
then there must be  L 
−
→ −
→L0
T G − G < ε,
−
→V
i.e., T is linear continuous on G . However, this is obvious for ξ = 0. n Ln o
−
→
Now if T is linear continuous but unbounded, there exists a sequence G
−
→V
in G such that L L
−
→ n −
→ n
G ≥ n G .

Let
−
→L∗n 1
G = L ×− →Ln
→ n G .
−
n G
L∗ 1  L∗ 
−
→ n −
→ n
Then G = → 0, i.e., T G → 0 if n → ∞. However, by definition
n
 
 L∗  −
→Ln
−
→ n G
T G = T  
→Ln
−
n G
 L  L
−
→ n −
→ n
T G n G
= L ≥ L = 1,
−→ n −
→ n
n G n G

a contradiction. Thus, T must be bounded. 

The following result is a generalization of the representation theorem of Fréchet

−
→V −
→ −
→V
and Riesz on linear continuous functionals, i.e., T : G → C on G -flow space G ,
where C is the complex field.

17
 −
→V
Theorem 3.5 Let T : G → C be a linear continuous functional. Then there is a
−
→Lb −
→V
unique G ∈ G such that
 L   L Lb 
−
→ −
→ − →
T G = G ,G

−
→L − →V
for ∀ G ∈ G .
−
→V
Proof Define a closed subset of G by
n L   L o
−
→ −
→V  −
→
N (T) = G ∈ G  T G =0
 L
−
→V −
→
for the linear continuous functional T. If N (T) = G , i.e., T G = 0 for
−
→L − →V −
→Lb
∀ G ∈ G , choose G = O. We then easily obtain the identity
 L   L Lb 
−
→ −
→ → −
T G = G ,G .

−
→V
Whence, we assume that N (T) 6= G . In this case, there is an orthogonal decom-
position
−
→V
G = N (T) ⊕ N ⊥ (T)
with N (T) 6= {O} and N ⊥ (T) 6= {O}.
−
→ −
→L0 −
→L0
Choose a G -flow G ∈ N ⊥ (T) with G 6= O and define
  L  L0   L0  L
−
→L∗ −
→ −
→ −
→ −
→
G = T G G − T G G
−
→L − →V
for ∀ G ∈ G . Calculation shows that
 L∗    L   L    L   L 
−
→ −
→ −
→ 0 −
→ 0 −
→
T G = T G T G − T G T G = 0,
−
→L∗
i.e., G ∈ N (T). We therefore get that
D L∗ E
−
→ − →L0
0 = G ,G
D  L  L   L  L L E
−
→ −
→ 0 −
→ 0 − → → − 0
= T G G − T G G ,G
 L D L E  L D L L E
−
→ −
→ 0 →− L0 −
→ 0 − → → − 0
= T G G ,G −T G G ,G .
D L E L 2
−
→ 0 −
→L0 −
→ 0
Notice that G , G = G 6= 0. We find that
 L *  L +
−
→ 0 −
→ 0
 L T G D L L0 E T G
−
→ −
→ −→ −
→L −
→L0
T G = G ,G = G , G .
−→L0 2 −→L0 2
G G

18
 Let  L0 
−
→
−
→Lb T G −
→L0 −→λL0
G = G = G ,
−→L0 2
G
 L
−
→ 0
T G  L   L Lb 
−
→ −
→ − →
where λ = 2 . We consequently get that T G = G ,G .
−→ 0
L
G
 L  
−
→Lb ′ −
→V −
→ −
→L − →Lb ′
Now if there is another G ∈ G such that T G = G ,G for
 
−
→L − →V −
→L − →Lb −→Lb ′
∀ G ∈ G , there must be G , G − G = 0 by definition. Particularly, let
−
→L −→Lb −
→Lb ′
G = G − G . We know that
 
− →Lb ′
→ − −
→Lb −
→Lb ′ −
→Lb −
→Lb ′
b
L

G − G = G − G , G − G = 0,

−
→Lb −
→Lb ′ −
→Lb −
→Lb ′
which implies that G − G = O, i.e., G = G . 

3.2 Differential and Integral Operators

Let V be Hilbert space consisting of measurable functions f (x1 , x2 , · · · , xn ) on a set

∆ = {x = (x1 , x2 , · · · , xn ) ∈ Rn |ai ≤ xi ≤ bi , 1 ≤ i ≤ n} ,

i.e., the functional space L2 [∆], with inner product

Z
hf (x) , g (x)i = f (x)g(x)dx for f (x), g(x) ∈ L2 [∆]
∆
−
→V −
→ −
→
and G its G -extension on a topological graph G . The differential operator and
integral operators
n
X Z Z
∂
D= ai and ,
i=1
∂xi ∆ ∆

−
→V
on G are respectively defined by
−
→L −→DL(uv )
DG = G

and
Z Z R
−
→L −
→L[y] −
→ ∆
K(x,y)L(uv )[y]dy
G = K(x, y) G dy = G ,
∆ ∆
Z Z R
−
→L −
→L[y] −
→ ∆ K(x,y)L(uv )[y]dy
G = K(x, y) G dy = G
∆ ∆

19
   ∂ai
−
→
for ∀(u, v) ∈ X G , where ai , ∈ C0 (∆) for integers 1 ≤ i, j ≤ n and K(x, y) :
∂xj
∆ × ∆ → C ∈ L2 (∆ × ∆, C) with
Z
K(x, y)dxdy < ∞.
∆×∆

Z Z
Such integral operators are usually called adjoint for = by K (x, y) =
∆ ∆
−
→L1 −→L2 − →V
K (x, y). Clearly, for G , G ∈ G and λ, µ ∈ F ,
   λL (uv )+µL (uv ) 
−
→L1 (uv ) −
→L2 (uv ) −
→ 1 2 −
→D(λL1 (uv )+µL2 (uv ))
D λG + µG =D G =G
−
→D(λL1 (uv ))+D(µL2 (uv )) −
→D(λL1 (uv )) −→D(µL2 (uv ))
=G =G +G
 (λL (uv ))   L (uv )   
−
→ 1 −
→(µL2 (uv )) −
→ 1 −
→L2 (uv )
=D G +G = λD G + D µG
 
−
→
for ∀(u, v) ∈ X G , i.e.,
 
−
→L1 −
→L2 −
→L1 −
→L2
D λG + µG = λD G + µD G .

Similarly, we know also that

Z   Z Z
−
→L1 −
→L2 −
→L1 −
→L2
λG + µG =λ G +µ G ,
∆ ∆ ∆
Z   Z Z
−
→L1 −
→L2 −
→L1 −
→L2
λG + µG =λ G +µ G .
∆ ∆ ∆
Z Z
−
→V
Thus, operators D, and are al linear on G .
∆ ∆
−
→L
For example, let f (t) = t, g(t) = et , K(t, τ ) = t2 + τ 2 for ∆ = [0, 1] and let G
−
→
be the G -flow shown on the left in Fig.6. Then, we know that Df = 1, Dg = et ,
Z 1 Z 1 Z 1

t2 1
K(t, τ )f (τ )dτ = K(t, τ )f (τ )dτ = t2 + τ 2 τ dτ = + = a(t),
0 0 0 2 4
Z 1 Z 1 Z 1


K(t, τ )g(τ )dτ = K(t, τ )g(τ )dτ = t2 + τ 2 eτ dτ
0 0 0
2
= (e − 1)t + e − 2 = b(t)
Z Z
−
→L −
→L −
→L
and the actions D G , G and G are shown on the right in Fig.5.
[0,1] [0,1]

20
 -et
-t -
et
-1
= et ? D t
}= et ?1
6 } t 1 e
t
t e
6
-t 6 -1 
et et

-et -
-
b(t)
-t R R a(t)
?a(t)
}= e ?t 6b(t) =} b(t)
t t , [0,1]
t 6e
[0,1] a(t)

- -
t a(t)

et b(t)
Fig.5
−
→V
Furthermore, we know that both of them are injections on G .
Z
−
→V −
→V −
→V −
→V
Theorem 3.6 D : G → G and : G →G .
∆
Z
−
→L − →V −
→L −
→L −→V
Proof For ∀ G ∈ G , we are needed to show that D G and G ∈ G ,
∆
i.e., the conservation laws
X X Z
v
DL(u ) = 0 and L(uv ) = 0
v∈NG (u) v∈NG (u) ∆

 
−
→ −
→L(uv ) −→V
hold with ∀v ∈ V G . However, because of G ∈ G , there must be
X  
v −
→
L(u ) = 0 for ∀v ∈ V G ,
v∈NG (u)

we immediately know that

 
X X
0 = D L(uv ) = DL(uv )
v∈NG (u) v∈NG (u)

and  
Z X X Z
0=  L(u ) =v
L(uv )
∆ v∈NG (u) v∈NG (u) ∆
 
−
→
for ∀v ∈ V G . 

21
 −
→
§4. G -Flow Solutions of Equations

As we mentioned, all G-solutions of non-solvable systems on algebraic, ordinary or

−
→
partial differential equations determined in [13]-[19] are in fact G -flows. We show
−
→
there are also G -flow solutions for solvable equations in this section.

4.1 Linear Equations

Let V be a field (F ; +, ·). We can further define

−
→L1 −
→L2 −
→L1 ·L2
G ◦G =G
 
v v v −
→
with L1 · L2 (u ) = L1 (u ) · L2 (u ) for ∀(u, v) ∈ X G . Then it can be verified
 F 
−
→F −
→ c isomorphic to F if the
easily that G is also a field G ; +, ◦ with a subfield F
conservation laws is not emphasized, where
n L  o
c −
→ −
→F v −
→
F = G ∈ G |L (u ) is constant in F for ∀(u, v) ∈ X G .
   
 →
−
E G   F  →
−
−
→F   −
→  nE G 
Clearly, G ≃ F . Thus  G  = p if |F | = pn , where p is a
−
→
prime number. For this F -extension on G , the linear equation
−
→L
aX = G
−
→a−1 L −
→F
is uniquely solvable for X = G in G if 0 6= a ∈ F . Particularly,
 if one views
−
→L −
→
an element b ∈ F as b = G if L(uv ) = b for (u, v) ∈ X G and 0 6= a ∈ F , then
an algebraic equation
ax = b
−
→F −
→a−1 L
in F also is an equation in G with a solution x = G such as those shown in
−
→ − →
Fig.6 for G = C 4 , a = 3, b = 5 following.

-3
5

5
3 6 ?53

5
3

Fig.6

22
 Let [Lij ]m×n be a matrix with entries Lij : uv → V . Denoted by [Lij ]m×n (uv )
−
→
the matrix [Lij (uv )]m×n . Then, a general result on G -flow solutions of linear systems
is known following.

n
Theorem 4.1 A linear system (LESm ) of equations

 −
→L1
 a11 X1 + a12 X2 + · · · + a1n Xn = G



 a X +a X +···+a X = − →L2
21 1 22 2 2n n G n
(LESm )

 ......................................



 a X +a X +···+a X = − →Lm
m1 1 m2 2 mn n G

→Li −
− →V
with aij ∈ C and G ∈ G for integers 1 ≤ i ≤ n and 1 ≤ j ≤ m is solvable for
−
→V
Xi ∈ G , 1 ≤ i ≤ m if and only if

rank [aij ]m×n = rank [aij ]+ v

m×(n+1) (u )

−
→
for ∀(u, v) ∈ G , where
 
a11 a12 · · · a1n L1
 
 a21 a22 · · · a2n L2 
[aij ]+
m×(n+1) =

.

 ... ... .. ... ... 
am1 am2 · · · amn Lm

 
−
→Lx −
→
Proof Let Xi = G i with Lxi (uv ) ∈ V on (u, v) ∈ X G for integers
 
−
→ n
1 ≤ i ≤ n. For ∀(u, v) ∈ X G , the system (LESm ) appears as a common linear
system


 a11 Lx1 (uv ) + a12 Lx2 (uv ) + · · · + a1n Lxn (uv ) = L1 (uv )


 a L (uv ) + a L (uv ) + · · · + a L (uv ) = L (uv )
21 x1 22 x2 2n xn 2

 ........................................................



am1 Lx1 (uv ) + am2 Lx2 (uv ) + · · · + amn Lxn (uv ) = Lm (uv )

By linear algebra, such a system is solvable if and only if ([4])

rank [aij ]m×n = rank [aij ]+ v

m×(n+1) (u )

−
→
for ∀(u, v) ∈ G .

23
 Labeling the
 semi-arc
 uv respectively by solutions Lx1 (uv ), Lx2 (uv ), · · · , Lxn (uv )
−
→ −
→Lx − →Lx −
→Lxn
for ∀(u, v) ∈ X G , we get labeled graphs G 1 , G 2 , · · · , G . We prove that
−
→Lx1 →− Lx −
→Lxn − →V
G , G 2, · · ·, G ∈G .
Let rank [aij ]m×n = r. Similar to that of linear algebra, we are easily know that

 Pm
−
→Li

 X j1 = c 1i G + c1,r+1 Xjr+1 + · · · c1n Xjn




i=1
 Xj = P c2i −
m
 →Li
2 G + c2,r+1 Xjr+1 + · · · c2n Xjn
i=1 ,



 .........................................

 Pm

 −
→Li
 Xj r = cri G + cr,r+1 Xjr+1 + · · · crn Xjn
i=1

−
→Lx −
→Lx −
→V
where {j1 , · · · , jn } = {1, · · · , n}. Whence, if G jr+1 , · · · , G jn ∈ G , then
X m
X X
v
Lxk (u ) = cki Li (uv )
v∈NG (u) v∈NG (u) i=1
X X
+ c2,r+1 Lxjr+1 (uv ) + · · · + c2n Lxjn (uv )
v∈NG (u) v∈NG (u)
 
m
X X
= cki  Li (uv )
i=1 v∈NG (u)
X X
+c2,r+1 Lxjr+1 (uv ) + · · · + c2n Lxjn (uv ) = 0
v∈NG (u) v∈NG (u)

n −
→V
Whence, the system (LESm ) is solvable in G . 

The following result is an immediate conclusion of Theorem 4.1.

Corollary 4.2 A linear system of equations



 a11 x1 + a12 x2 + · · · + a1n xn = b1


 a x +a x +···+a x = b
21 1 22 2 2n n 2

 ..................................



am1 x1 + am2 x2 + · · · + amn xn = bm

with aij , bj ∈ F for integers 1 ≤ i ≤ n, 1 ≤ j ≤ m holding with

rank [aij ]m×n = rank [aij ]+

m×(n+1)

−
→ −
→
has G -flow solutions on infinitely many topological graphs G .

24
 Let the operator D and ∆ ⊂ Rn be the same as in Subsection 3.2. We consider
−
→V
differential equations in G following.
−
→V
Theorem 4.3 For ∀GL ∈ G , the Cauchy problem on differential equation

DGX = GL
−
→X| 0 −
→L0
is uniquely solvable prescribed with G xn =xn = G .
 
−
→
Proof For ∀(u, v) ∈ X G , denoted by F (uv ) the flow on the semi-arc uv .
Then the differential equation DGX = GL transforms into a linear partial differential
equation
n
X ∂F (uv )
ai = L (uv )
i=1
∂xi
on the semi-arc uv . By assumption, ai ∈ C0 (∆) and L (uv ) ∈ L2 [∆], which
implies that there is a uniquely solution F (uv ) with initial value L0 (uv ) by the
characteristic theory of partial differential equation of first order. In fact, let
φi (x1 , x2 , · · · , xn , F ) , 1 ≤ i ≤ n be the n independent first integrals of its char-
acteristic equations. Then


F (uv ) = F ′ (uv ) − L0 x′1 , x′2 · · · , x′n−1 ∈ L2 [∆],

where, x′1 , x′2 , · · · , x′n−1 and F ′ are determined by system of equations



 φ1 (x1 , x2 , · · · , xn−1 , x0n , F ) = φ1


 φ (x , x , · · · , x , x0 , F ) = φ
2 1 2 n−1 n 2

 ...............................



φn (x1 , x2 , · · · , xn−1 , x0n , F ) = φn

Clearly,
 
X X X
D F (uv ) = DF (uv ) = L(uv ) = 0.
v∈NG (u) v∈NG (u) v∈NG (u)

Notice that 

X  X
v 
F (u ) = L0 (uv ) = 0.
v∈NG (u)  v∈NG (u)
xn =x0n

25
We therefore know that
X
F (uv ) = 0.
v∈NG (u)

−
→X −→F −→V
Thus, we get a uniquely solution G = G ∈ G for the equation

DGX = GL
−
→X| 0 −
→L0
prescribed with initial data G xn =xn = G . 

We know that the Cauchy problem on heat equation

X ∂2u n
∂u
= c2
∂t i=1
∂x2i

is solvable in Rn ×R if u(x, t0 ) = ϕ(x) is continuous and bounded in Rn , c a non-zero

−
→L − →V
constant in R. For G ∈ G in Subsection 3.2, if we define
−
→L −
→L
∂G −
→ ∂L ∂G −
→ ∂L
= G ∂t and = G ∂xi , 1 ≤ i ≤ n,
∂t ∂xi
−
→V
then we can also consider the Cauchy problem in G , i.e.,
X ∂2X n
∂X
= c2
∂t i=1
∂x2i

with initial values X|t=t0 , and know the result following.

−
→L′ −→V
Theorem 4.4 For ∀ G ∈ G and a non-zero constant c in R, the Cauchy problems
on differential equations
X ∂2X n
∂X
= c2
∂t i=1
∂x2i
−
→L′ −→V −
→V
with initial value X|t=t0 = G ∈ G is solvable in G if L′ (uv ) is continuous and
−
→
bounded in Rn for ∀(u, v) ∈ X G .
 
−
→
Proof For (u, v) ∈ X G , the Cauchy problem on the semi-arc uv appears as

X ∂2u n
∂u
= c2
∂t i=1
∂x2i

26
 −
→F
with initial value u|t=0 = L′ (uv ) (x) if X = G . According to the theory of partial
differential equations, we know that
Z +∞
1 (x1 −y1 )2 +···+(xn −yn )2
v
F (u ) (x, t) = n e− 4t L′ (uv ) (y1 , · · · , yn )dy1 · · · dyn .
(4πt) −∞
2

 
−
→
Labeling the semi-arc uv by F (uv ) (x, t) for ∀(u, v) ∈ X G , we get a labeled
−
→F −
→ −
→F − →V
graph G on G . We prove G ∈ G .  
−
→L′ − →V −
→
By assumption, G ∈ G , i.e., for ∀u ∈ V G ,
X
L′ (uv ) (x) = 0,
v∈NG (u)

we know that
X
F (uv ) (x, t)
v∈NG (u)
X Z +∞
1 (x1 −y1 )2 +···+(xn −yn )2
= n e− 4t L′ (uv ) (y1 , · · · , yn )dy1 · · · dyn
v∈NG (u)
(4πt) 2 −∞
 
Z +∞ X
1 (x −y )2 +···+(xn −yn )2
− 1 1 
= n e 4t L′ (uv ) (y1 , · · · , yn ) dy1 · · · dyn
(4πt) 2 −∞ v∈NG (u)
Z +∞
1 (x1 −y1 )2 +···+(xn −yn )2
= n e− 4t (0) dy1 · · · dyn = 0
(4πt) 2 −∞
 
−
→ −
→F − →V
for ∀u ∈ V G . Therefore, G ∈ G and

X ∂2X n
∂X
= c2
∂t i=1
∂x2i

−
→L′ −
→V −
→V
with initial value X|t=t0 = G ∈ G is solvable in G . 

Similarly, we can also get a result on Cauchy problem on 3-dimensional wave

−
→V
equation in G following.
−
→L′ −→V
Theorem 4.5 For ∀ G ∈ G and a non-zero constant c in R, the Cauchy problems
on differential equations
 
∂2X ∂2X ∂2X ∂2X
= c2 + +
∂t2 ∂x21 ∂x22 ∂x23

27
 −
→L′ −→V −
→V ′ v
with initial value X|t=t0 = G ∈ G is solvable in G if L (u ) is continuous and
−
→
bounded in Rn for ∀(u, v) ∈ X G .
∗
For an integral kernel K(x, y), the two subspaces N , N ⊂ L2 [∆] are deter-
mined by
 Z 
2
N = φ(x) ∈ L [∆]|
K (x, y) φ(y)dy = φ(x) ,
∆
 Z 
∗ 2
N = ϕ(x) ∈ L [∆]| K (x, y)ϕ(y)dy = ϕ(x) .
∆

Then we know the result following.

−
→V
Theorem 4.6 For ∀GL ∈ G , if dimN = 0, then the integral equation
Z
−
→X −
→X
G − G = GL
∆

−
→V
is solvable in G with V = L2 [∆] if and only if
D L L′ E
−
→ → − −
→L′
G ,G = 0, ∀ G ∈ N ∗ .

 
−
→
Proof For ∀(u, v) ∈ X G
Z D E
−
→X −
→X −
→L →
− L′ −
→L′ ∗
G − G = GL and G ,G = 0, ∀G ∈ N
∆
v
on the semi-arc u respectively appear as
Z
F (x) − K (x, y) F (y) dy = L (uv ) [x]
∆
v
if X (u ) = F (x) and
Z
−
→ L′
L (uv ) [x]L′ (uv ) [x]dx = 0 for ∀ G ∈ N ∗ .
∆

Applying Hilbert and Schmidt’s theorem ([20]) on integral equation, we know

the integral equation
Z
F (x) − K (x, y) F (y) dy = L (uv ) [x]
∆
2
is solvable in L [∆] if and only if
Z
L (uv ) [x]L′ (uv ) [x]dx = 0
∆

28
 −
→L′
for ∀ G ∈ N ∗ . Thus, there are functions F (x) ∈ L2 [∆] hold for the integral
equation Z
F (x) − K (x, y) F (y) dy = L (uv ) [x]
∆
 
−
→
for ∀(u, v) ∈ X G in this case.
 
−
→
For ∀u ∈ V G , it is clear that

X  Z  X
v
F (u ) [x] − K(x, y)F (u ) [x] =v
L (uv ) [x] = 0,
v∈NG (u) ∆ v∈NG (u)

which implies that,

 
Z X X
K(x, y)  F (uv ) [x] = F (uv ) [x].
∆ v∈NG (u) v∈NG (u)

Thus,
X
F (uv ) [x] ∈ N .
v∈NG (u)

However, if dimN = 0, there must be

X
F (uv ) [x] = 0
v∈NG (u)

 
−
→ −
→F −→V
for ∀u ∈ V G , i.e., G ∈ G . Whence, if dimN = 0, the integral equation
Z
−
→X −
→X
G − G = GL
∆

−
→V
is solvable in G with V = L2 [∆] if and only if
D L L′ E
−
→ → − −
→L′
G ,G = 0, ∀ G ∈ N ∗ .

This completes the proof. 

Theorem 4.7 Let the integral kernel K(x, y) : ∆ × ∆ → C ∈ L2 (∆ × ∆) be given

with Z
|K(x, y)|2dxdy > 0, dimN = 0 and K(x, y) = K(x, y)
∆×∆

29
 o y) ∈ ∆ × ∆. Then there is a finite or countably infinite system
n allL (x,
for almost
−
→ −
→ i
G -flows G ⊂ L2 (∆, C) with associate real numbers {λi }i=1,2,··· ⊂ R such
i=1,2,···
that the integral equations
Z
−
→Li [y] −
→Li [x]
K(x, y) G dy = λi G
∆

hold with integers i = 1, 2, · · ·, and furthermore,

|λ1 | ≥ |λ2 | ≥ · · · ≥ 0 and lim λi = 0.

i→∞

Proof Notice that the integral equations

Z
−
→Li [y] −
→Li [x]
K(x, y) G dy = λi G
∆

is appeared as Z
K(x, y)Li (uv ) [y]dy = λi Li (uv ) [x]
∆
 
−
→
on (u, v) ∈ X G . By the spectral theorem of Hilbert and Schmidt ([20]), there
is indeed a finite or countably system of functions {Li (uv ) [x]}i=1,2,··· hold with this
integral equation, and furthermore,

|λ1 | ≥ |λ2 | ≥ · · · ≥ 0 with lim λi = 0.

i→∞

Similar to the proof of Theorem 4.5, if dimN = 0, we know that

X
Li (uv ) [x] = 0
v∈NG (u)

 
−
→ −
→Li −
→V
for ∀u ∈ V G , i.e., G ∈ G for integers i = 1, 2, · · ·. 

4.2 Non-linear Equations

−
→ −
→
If G is chosen with a special structure, we can get a general result on G -solutions
of equations, including non-linear equations following.
−
→
Theorem 4.8 If the topological graph G can be decomposed into circuits
l
→ [−
− →
G= Ci
i=1

30
  
v −
→
such that L(u ) = Li (x) for ∀(u, v) ∈ X C i , 1 ≤ i ≤ l and the Cauchy problem
(
Fi (x, u, ux1 , · · · , uxn , ux1 x2 , · · ·) = 0
u|x0 = Li (x)

is solvable in a Hilbert space V on domain ∆ ⊂ Rn for integers 1 ≤ i ≤ l, then the

Cauchy problem (
Fi (x, X, Xx1 , · · · , Xxn , Xx1 x2 , · · ·) = 0
−
→L
X|x0 = G
 
−
→ −
→V
such that L (uv ) = Li (x) for ∀(u, v) ∈ X C i is solvable for X ∈ G .
 
−
→Lu(x) v −
→
Proof Let X = G with Lu(x) (u ) = u(x) for (u, v) ∈ X G . Notice that
the Cauchy problem
(
Fi (x, X, Xx1 , · · · , Xxn , Xx1 x2 , · · ·) = 0
X|x0 = GL

then appears as (
Fi (x, u, ux1 , · · · , uxn , ux1 x2 , · · ·) = 0
u|x0 = Li (x)
 
−
→
on the semi-arc uv for (u, v) ∈ X G , which is solvable by assumption. Whence,
there exists solution u (uv ) (x) holding with
(
Fi (x, u, ux1 , · · · , uxn , ux1 x2 , · · ·) = 0
u|x0 = Li (x)
 
−
→Lu(x) −
→ −
→
Let G be a labeling on G with u (uv ) (x) on uv for ∀(u, v) ∈ X G . We
−
→Lu(x) − →V
show that G ∈ G . Notice that
l
→ [−
− →
G= Ci
i=1

−
→
and all flows on C i is the same, i.e., the solution u (uv ) (x). Clearly, it is holden
−
→
with conservation on each vertex in C i for integers 1 ≤ i ≤ l. We therefore know
that  
X −
→
Lx0 (uv ) = 0, u∈V G .
v∈NG (u)

31
 −
→Lu(x) −→V
Thus, G ∈ G . This completes the proof. 
−
→
There are many interesting conclusions on G -flow solutions of equations by
Theorem 4.8. For example, if Fi is nothing else but polynomials of degree n in one
variable x, we get a conclusion following, which generalizes the fundamental theorem
in algebra.
−
→
Corollary 4.9(Generalized Fundamental Theorem in Algebra) If G can be decom-
posed into circuits
l
→ [−
− →
G= Ci
i=1
 
−
→
and Li (uv ) = ai ∈ C for ∀(u, v) ∈ X C i and integers 1 ≤ i ≤ l, then the
polynomial
−
→L1 −
→L2 −
→Ln −
→Ln+1
F (X) = G ◦ X n + G ◦ X n−1 + · · · + G ◦ X + G
−
→C −
→L1
always has roots, i.e., X0 ∈ G such that F (X0 ) = O if G 6= O and n ≥ 1.
Particularly, an algebraic equation

a1 xn + a2 xn−1 + · · · + an x + an+1 = 0
−
→ −
→C
with a1 6= 0 has infinite many G -flow solutions in G on those topological graphs
l
−
→ → [−
− →
G with G = C i.
i=1
−
→
Notice that Theorem 4.8 enables one to get G -flow solutions both on those
linear and non-linear equations in physics. For example, we know the spherical
solution  rg  2 1
ds2 = f (t) 1 − dt − dr 2 − r 2 (dθ2 + sin2 θdφ2 )
r 1 − rrg
for the Einstein’s gravitational equations ([9])
1
Rµν − Rg µν = −8πGT µν
2
with Rµν = Rαµαν = gαβ Rαµβν , R = gµν Rµν , G = 6.673×10−8cm3 /gs2, κ = 8πG/c4 =
−
→
2.08 × 10−48 cm−1 · g −1 · s2. By Theorem 4.8, we get their G -flow solutions following.

Corollary 4.10 The Einstein’s gravitational equations

1
Rµν − Rg µν = −8πGT µν ,
2
32
 −
→ −
→C
has infinite many G -flow solutions in G , particularly on those topological graphs
l
→ [−
− →
G= C i with spherical solutions of the equations on their arcs.
i=1
−
→ −
→ −
→
For example, let G = C 4 . We are easily find C 4 -flow solution of Einstein’s
gravitational equations,such as those shown in Fig.7.
v1
-S1 v2

S4 6 ?S2

v4
y v3
S3
Fig.7

where, each Si is a spherical solution

 rs  2 1
ds2 = f (t) 1 − dt − dr 2 − r 2 (dθ2 + sin2 θdφ2 )
r 1 − rrs
of Einstein’s gravitational equations for integers 1 ≤ i ≤ 4.
As a by-product, Theorems 4.5-4.6 can be also generalized on those topological
graphs with circuit-decomposition following.

Corollary 4.11 Let the integral kernel K(x, y) : ∆ × ∆ → C ∈ L2 (∆ × ∆) be given

with Z
|K(x, y)|2 dxdy > 0, K(x, y) = K(x, y)
∆×∆

for almost all (x, y) ∈ ∆ × ∆, and

l
→L [ −
− →
G = Ci
i=1
 
−
→
such that L(uv ) = L[i] (x) for ∀(u, v) ∈ X C i and integers 1 ≤ i ≤ l. Then, the
integral equation Z
−
→X −
→X
G − G = GL
∆
−
→V
is solvable in G with V = L2 [∆] if and only if
D L L′ E
−
→ → − −
→L′
G ,G = 0, ∀ G ∈ N ∗ .

33
Corollary 4.12 Let the integral kernel K(x, y) : ∆ × ∆ → C ∈ L2 (∆ × ∆) be given
with Z
|K(x, y)|2 dxdy > 0, K(x, y) = K(x, y)
∆×∆

for almost all (x, y) ∈ ∆ × ∆, and

l
→L [ −
− →
G = Ci
i=1
 
v −
→
such that L(u ) = L[i] (x) for ∀(u, v) ∈ X C i and integers 1 ≤ i ≤ l. Then,
n Lo
−
→ −
→ i
there is a finite or countably infinite system G -flows G ⊂ L2 (∆, C) with
i=1,2,···
associate real numbers {λi }i=1,2,··· ⊂ R such that the integral equations
Z
−
→Li [y] −
→Li [x]
K(x, y) G dy = λi G
∆

hold with integers i = 1, 2, · · ·, and furthermore,

|λ1 | ≥ |λ2 | ≥ · · · ≥ 0 and lim λi = 0.

i→∞

§5. Applications to System Control

−
→
5.1 Stability of G -Flow Solutions
−
→Lu(x) −
→Lu (x)
Let X = G and X2 = G 1 be respectively solutions of

F (x, Xx1 , · · · , Xxn , Xx1 x2 , · · ·) = 0

−
→L −
→L1 − →V
on the initial values X|x0 = G or X|x0 = G in G with V = L2 [∆], the Hilbert
−
→
space. The G -flow solution X is said to be stable if there exists a number δ(ε) for
any number ε > 0 such that
L
−
→ u (x) −
→Lu(x)
kX1 − X2 k = G 1 − G <ε

if L
−
→ 1 −→L
G − G ≤ δ(ε).
By definition, L
−→ 1 −→L X
G − G = kL1 (uv ) − L (uv )k
−
→
(u,v)∈X G

34
and L
−
→ u1 (x) −
→Lu(x) X
G −G = ku1 (uv ) (x) − u (uv ) (x)k .
−
→
(u,v)∈X G
−
→
Clearly, if these G -flow solutions X are stable, then
X
ku1 (uv ) (x) − u (uv ) (x)k ≤ ku1 (uv ) (x) − u (uv ) (x)k < ε

→
−
(u,v)∈X G

if
X
kL1 (uv ) − L (uv )k ≤ kL1 (uv ) − L (uv )k ≤ δ(ε),
−
→
(u,v)∈X G
 
−
→
i.e., u (uv ) (x) is stable on uv for (u, v) ∈ X G .
 
−
→
Conversely, if u (uv ) (x) is stable on uv for (u, v) ∈ X G , i.e., for any number
 
−
→
ε/ε G > 0 there always is a number δ(ε) (uv ) such that

ε
ku1 (uv ) (x) − u (uv ) (x)k <  
−
→
ε G

if
kL1 (uv ) − L (uv )k ≤ δ(ε) (uv ) ,

then there must be

X   ε
−
→
ku1 (uv ) (x) − u (uv ) (x)k < ε G × − 
→ =ε

→
− ε G
(u,v)∈X G

if
δ(ε)
kL1 (uv ) − L (uv )k ≤  ,
−
→
ε G
 
−
→ −
→
where ε G is the number of arcs of G and
n  o
−
→
δ(ε) = min δ(ε) (uv ) |(u, v) ∈ X G .

Whence, we get the result following.

35
 −
→
Theorem 5.1 Let V be the Hilbert space L2 [∆]. The G -flow solution X of equation
(
F (x, X, Xx1 , · · · , Xxn , Xx1 x2 , · · ·) = 0
−→L
X|x0 = G

−
→V
in G is stable if and only if the solution u(x) (uv ) of equation
(
F (x, u, ux1 , · · · , uxn , ux1 x2 , · · ·) = 0
−
→L
u|x0 = G
 
v −
→
is stable on the semi-arc u for ∀(u, v) ∈ X G .
−
→
This conclusion enables one to find stable G -flow solutions of equations. For ex-
ample, we know that the stability of trivial solution y = 0 of an ordinary differential
equation
dy
= [A]y
dx
with constant coefficients, is dependent on the number γ = max{Reλ : λ ∈ σ[A]}
([23]), i.e., it is stable if and only if γ < 0, or γ = 0 but m′ (λ) = m(λ) for all
eigenvalues λ with Reλ = 0, where σ[A] is the set of eigenvalue of the matrix [A],
m(λ) the multiplicity and m′ (λ) the dimension of corresponding eigenspace of λ.

Corollary 5.2 Let [A] be a matrix with all eigenvalues λ < 0, or γ = 0 but
m′ (λ) = m(λ) for all eigenvalues λ with Reλ = 0. Then the solution X = O of
differential equation
dX
= [A]X
dx
−
→V −
→ −
→
is stable in G , where G is such a topological graph that there are G -flows hold
with the equation.
−
→
For example, the G -flow shown in Fig.8 following
v1
b(x) - v2

j  a(x)b(x) ? ?a(x)
a(x)
a(x)
6 6 b(x)

v46
 v6
b(x) 3

Fig.8

36
 −
→
is a G -flow solution of the differential equation
d2 X dX
2
+5 + 6X = 0
dx dx
with f (x) = C1 e−2x + C1′ e−3x and g(x) = C2 e−2x + C2′ e−3x , where C1 , C1′ and C2 , C2′
are constants.
Similarly, applying the stability of solutions of wave equations, heat equations
and elliptic equations, the conclusion following is known by Theorem 5.1.
−
→
Corollary 5.3 Let V be the Hilbert space L2 [∆]. Then, the G -flow solutions X of
equations following
 2  2 2


 ∂ X 2 ∂ X ∂ X −
→L
 − c 2
+ 2
= G
∂t2 ∂x1 ∂x2
,

 →Lφ(x1 ,x2 ) ∂X 
− −
→Lϕ(x1 ,x2 ) −
→Lµ(t,x1 ,x2 )
 X|t0 = G , =G , X|∂∆ = G
∂t t0
 2 
 ∂ X − c2 ∂X = − →L 
 ∂2X ∂2X ∂2X
G + + =O
∂t2 ∂x1 and ∂x21 ∂x22 ∂x23
 −
→Lφ(x1 ,x2 )  X|∂∆ = −
 →Lg(x1 ,x2 ,x3 )
X|t0 = G G

−
→V −
→ −
→
are stable in G , where G is such a topological graph that there are G -flows hold
with these equations.

5.2 Industrial System Control

An industrial system with raw materials M1 , M2 , · · · , Mn , products (including by-

products) P1 , P2 , · · · , Pm but w1 , w2 , · · · , ws wastes after a produce process, such as
those shown in Fig.9 following,

M1
x1 y1 - P1

M2
x2 - ?F (x) y2 - P2
6
Mn
xn ym - Pm
?
w1
?
w2
?
ws

Fig.9

37
i.e., an input-output system, where,

(y1 , y2, · · · , ym ) = F (x1 , x2 , · · · , xn )

determined by differential equations, called the production function and constrained

with the conservation law of matter, i.e.,
m
X s
X n
X
yi + wi = xi .
i=1 i=1 i=1

Notice that such an industrial system is an opened system in general, which

can be transferred into a closed one by letting the nature as an additional cell, i.e.,
all materials comes from and all wastes resolves by the nature, a classical one on
human beings with the nature. However, the resolvability of nature is very limited.
Such a classical system finally resulted in the environmental pollution accompanied
with the developed production of human beings.
Different from those of classical industrial systems, an ecologically industrial
system is a recycling system ([24]), i.e., all outputs of one of its subsystem, including
products, by-products provide the inputs of other subsystems and all wastes are
−
→
disposed harmless to the nature. Clearly, such a system is nothing else but a G -flow
−
→
because it is holding with conservation laws on each vertex in a topological graph G ,
−
→
where G is determined by the technological process for products, wastes disposal
−
→V
and recycle, and can be characterized by differential equations in Banach space G .
−
→Lu v v
Whence, we  determine such a system by G with Lu : u → u (u ) (t, x) for
 can
−
→
(u, v) ∈ X G , or ordinary differential equations


 →L0 dk X −
− →L1 dk−1X −
→Lu(t,x)
 G ◦ k + G ◦ k−1 + · · · + G =O
dt dt 
Lh0 (x) dX 
k−1 
 −
→  −
→ Lh1 (x) dX −
→Lh (x)

 X|t=t0 = G ,  =G , · · · , k−1  = G k−1
dt t=t0 dt t=t0

for an integer k ≥ 1, or a partial differential equation


 −→L0 ∂X − →L1 ∂X →Ln ∂X
− −
→Lu(t,x)
G ◦ +G ◦ +···+ G ◦ =G
∂t ∂x1 ∂xn
 −
→Lu(x)
X|t=t0 = G

−
→Li
and characterize its stability by Theorem 5.1, where, the coefficients G , i ≥ 0 are
determined by the technological process of production.

38
References

[1] John B.Conway, A Course in Functional Analysis, Springer-Verlag New York,Inc.,

1990.
[2] J.L.Gross and T.W.Tucker, Topological Graph Theory, John Wiley & Sons,
1987.
[3] F.John, Partial Differential Equations (4th edition), Springer-Verlag New York,Inc.,
1982.
[4] Kenneth Hoffman and Ray Kunze, Linear Algebra(Second Edition), Prentice-
Hall, Inc. Englewood Cliffs, New Jersey, 1971.
[5] Linfan Mao, The simple flow representation for graph and its application, Re-
ported at The First International Symposium on Combinatorial Optimization
of China, Tianjing, 1988.
[6] Linfan Mao, Analytic graph and its applications, Reported at The 6th Confer-
ence on Graph Theory and Its Applications of China, Qingdao, 1989.
[7] Linfan Mao, Automorphism Groups of Maps, Surfaces and Smarandache Ge-
ometries, First edition published by American Research Press in 2005, Second
edition is as a Graduate Textbook in Mathematics, Published by The Education
Publisher Inc., USA, 2011.
[8] Linfan Mao, Smarandache Multi-Space Theory(Second edition), First edition
published by Hexis, Phoenix in 2006, Second edition is as a Graduate Textbook
in Mathematics, Published by The Education Publisher Inc., USA, 2011.
[9] Linfan Mao, Combinatorial Geometry with Applications to Field Theory, First
edition published by InfoQuest in 2005, Second edition is as a Graduate Text-
book in Mathematics, Published by The Education Publisher Inc., USA, 2011.
[10] Linfan Mao, Non-solvable spaces of linear equation systems. International J.
Math. Combin., Vol.2 (2012), 9-23.
[11] Linfan Mao, Graph structure of manifolds listing, International J.Contemp.
Math. Sciences, Vol.5, 2011, No.2,71-85.
[12] Linfan Mao, A generalization of Seifert-Van Kampen theorem for fundamental
groups, Far East Journal of Math.Sciences, Vol.61 No.2 (2012), 141-160.
[13] Linfan Mao, Global stability of non-solvable ordinary differential equations with
applications, International J.Math. Combin., Vol.1 (2013), 1-37.
[14] Linfan Mao, Non-solvable equation systems with graphs embedded in Rn , In-

39
 ternational J.Math. Combin., Vol.2 (2013), 8-23.
[15] Linfan Mao, Geometry on GL -system of homogenous polynomials, International
J.Contemp. Math. Sciences (Accepted, 2014).
[16] Linfan Mao, A topological model for ecologically industrial systems, Interna-
tional J.Math. Combin., Vol.1(2014), 109-117.
[17] Linfan Mao, Cauchy problem on non-solvable system of first order partial differ-
ential equations with applications, Methods and Applications of Analysis (Ac-
cepted).
[18] Linfan Mao, Mathematics on non-mathematics – a combinatorial contribution,
International J.Math.Combin., Vol.3, 2014, 1-34.
[19] Linfan Mao, Geometry on non-solvable equations – A review on contradictory
systems, Reported at the International Conference on Geometry and Its Appli-
cations, Jardpour University, October 16-18, 2014, Kolkata, India.
[20] F.Sauvigny, Partial Differential Equations (Vol.1 and 2), Springer-Verlag Berlin
Heidelberg, 2006.
[21] F.Smarandache, Paradoxist Geometry, State Archives from Valcea, Rm. Valcea,
Romania, 1969, and in Paradoxist Mathematics, Collected Papers (Vol. II),
Kishinev University Press, Kishinev, 5-28, 1997.
[22] F.Smarandache, Multi-space and multi-structure, in Neutrosophy. Neutrosophic
Logic, Set, Probability and Statistics, American Research Press, 1998.
[23] Wolfgang Walter, Ordinary Differential Equations, Springer-Verlag New York,Inc.,
1998.
[234] Zengjun Yuan and Jun Bi, Industrial Ecology (in Chinese), Science Press, Bei-
jing, 2010.

40