Professional Documents
Culture Documents
LMao Conservationflow PDF
LMao Conservationflow PDF
−
→
Conservation Laws and G -Flow Solutions of Equations
Linfan MAO
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.
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
−
→
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
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
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
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 Gwith
−
→ −
→ −
→
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
−
→ −
→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.
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
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)
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)
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
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
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 −∞
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
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
.
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.
−
→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
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 .
∆ = {x = (x1 , x2 , · · · , xn ) ∈ Rn |ai ≤ xi ≤ bi , 1 ≤ i ≤ n} ,
−
→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 .
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)
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
-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
−
→
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 )
−
→
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 .
−
→ −
→
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 [∆],
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 .
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 .
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) .
∆
−
→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 ∗ .
∆
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)
Thus,
X
F (uv ) [x] ∈ N .
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 ∗ .
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
∆
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,
−
→ −
→Li −
→V
for ∀u ∈ V G , i.e., G ∈ G for integers i = 1, 2, · · ·.
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)
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.
S4 6 ?S2
v4
y v3
S3
Fig.7
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)
∆×∆
−
→
5.1 Stability of G -Flow Solutions
−
→Lu(x) −
→Lu (x)
Let X = G and X2 = G 1 be respectively solutions of
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 ) ,
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 .
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.
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,
−
→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
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