Differential Equation of Connectedness for Massively Interconnected Systems

(Part of PostDoc Dissertation, Moscow 1990, Translation from Russian) Pavel Barseghyan
Abstract Considering the Massively Interconnected system (MIS) as a spatially distributed system of elements and interconnections among them, the differential equation of connectivity is derived. The equation is based on the local balance of densities of interconnections and elements in the space. The capabilities of differential equation of connectedness are validated through checking the behavior of its solutions for different practical examples. The differential equation is capable to reflect any connectivity behavior in the real system of coordinates if the spatial densities of elements and interconnections among them may be considered as continuous functions. In addition the equation can reflect the statistical behavior of the totality of elements and interconnections. One-, two-, and three-dimensional interconnection length distributions are the specific solutions of this equation. I. Introduction. Massively Interconnected Systems Massively Interconnected systems can be characterized by a large number of elements and massive interconnections between them [1, 2]. Such systems are to be found everywhere. Among the examples of MIS are the human society, Internet and World Wide Web, an organization, a human group, a communication system or some semiconductor chip with a millions of elements and interconnections between them. The quantitative description of such systems becomes imperative nowadays because of the variety of potential applications of the latter. In organization science those practical applications are related to the structural analysis and synthesis of the organization and development of mathematical theory of the projects, including quantitative analysis of the behavior of human groups. From the mathematical point of view such systems can be divided into static and dynamic Massively Interconnected Systems. While in static massively interconnected systems the connections between elements are time independent, in dynamic massively interconnected systems they are time dependent. II. Requirements to the Mathematical Models of Massively Interconnected Systems Lets briefly discuss the requirements to the mathematical description of the MIS. 1. Due to the very big number of interconnections and elements within the MIS, the appropriate mathematical description for the connectivity must be a continuous one. 2. Due to the importance of real densities and distances within the MIS, the mathematical description of the latter must be in the real system of coordinates. 3. Mathematical description of such systems must incorporate the density type of parameters (density of interconnections, density of elements, etc.), in other words it must deal with the local or point characteristics of the system. 4. Due to the fact that these densities are functions of the space coordinates, the appropriate mathematical description of MIS must have the form of differential or integral equations. III. Graphical Representations of Massively Interconnected Systems Lets start with simple graphical representations of massively interconnected systems, having a goal to quantitatively describe the spatially distributed elements of the latter and the massive

connections among them. For that purpose it is useful to examine the common behavior for separate elements and their connections in such systems. Each element in the system has chaotic connections with the other elements. Elements of a massively interconnected system and connections of one of those elements are shown in the Fig.1.

Fig.1 One element of a MIS along with its connections with other elements Separating the element out along with its connections we will have the picture presented in Fig.2. The same may be done with the other elements and as a result we will always obtain almost the same picture where an element emits a certain number of connections which are gradually absorbed by the other elements. This chaotic penetration of interconnects into the environment

Fig.2 A separate element along with its connections with some variations is common for all massively interconnected systems. Therefore it may become a focus around which we can concentrate our modeling effort. Another characteristic for such a behavior is that the connection penetration image into the environment is very close to the

well-known diffusion process. Thus for the same modeling purposes it is useful to make parallels between the diffusion and the process of connection penetration into the environment.

IV. Spatial Densities of Interconnections and Elements

The study of an ordinary interconnection image in MIS shows that quantitatively it can be characterized by the spatial densities of elements and connections. If the MIS has a hierarchical structure with different levels of subsystems or elements, it may be described in any level of physical hierarchy. Irrespective to the level of physical hierarchy, the MIS will represent itself as a spatial conglomeration of elements and connections. The spatial densities of elements can be considered as scalar quantities. But the connections cannot be characterized only by the spatial densities, because in each point of the connectivity field they have another important characteristic direction, which must be accounted in connectedness mathematical models.

V. Vector Representation of Connectedness fields as a necessity

Lets consider a simple connectedness field in the form of parallel lines (Fig.3.). In order to reveal the vector behavior of such a connectivity field lets conduct a simple geometrical experiment. Lets consider three line segments in the field which have different angles with the direction of the parallel lines. The goal of the experiment is to find the number of crossings of the line segments with the connectedness field that has a constant density . Quantitatively the vector is equal to the number of crossings with the field of the unit line segment, which is perpendicular to the connections. Obviously the number of crossings with the line segment L will have its maximum value when it is perpendicular to the connections (segment 1 in the Fig.3.). The number of crossings will have a value 0 when line segment is parallel to the connections (segment 3). In order to solve the problem in the general case when the line segment has an arbitrary angle with the connections (or X axis) it is necessary to consider the connectedness field as a vector field. For that it is necessary to split the vector into two components, 21 and 22 . The first one is parallel to the arbitrary line segment and the component 22 is perpendicular to the same segment. Therefore the number of crossings of the line segment with the connectedness field is equal to

m 2 = 22 L = L sin

(1)

These calculations show that in order to account for the directions of connections in the connectedness field the latter has to be considered as a vector field.

Fig. 3 Vector Field of connections

VI.

Generic Connectivity Fields Consider the generic connectedness field shown in Fig.4. Each point of this field can be characterized by a vector ( x, y ) that has the direction of the tangent to the connections. The magnitude of this vector

is the density of the connections in the point ( x, y ) .

For each point of the field the connectedness density vector can be split up into the components x and y .

( x, y ) = x ( x, y ) i, + y ( x, y ) j,

(2) (3)

where i and j are the standard basic vectors. The norm of the density vector is

x2 + y2

VII. Number of Connections in the Cross-Sections of the Generic Connectedness Field Lets consider two cross-sections y = y1 and x = x1 in the generic connectedness field (Fig.4). In
order to calculate the number of connections in these cross-sections we can use the same technique of splitting the connections density vectors into x and y components. For instance, the connection density vector ( x, y1 ) in the point A can be split into two vectors -

y ( x, y1 ) j and x ( x, y1 ) i.

Fig.4 Generic connectedness field with its x1 and y1 cross-sections Thus in the point A the connectedness density vector will have the form

x ( x, y ) = Ax ( x A , y1 )i + Ay ( x A , y1 ) j .

Similarly for the point B we can have

x ( x, y ) = Ax ( x A , y1 )i + Ay ( x A , y1 ) j .

Therefore, by knowing the vector field ( x, y ) we can define the densities of connections at any point of the arbitrary cross-sections y = y1 and x = x1 , simply by splitting those connectedness density vectors into their components. Using those density components, we can now estimate the number of connections crossing any segment in the field. So with the aid of the component
y ( x, y1 ) we can calculate the number of connections

m y1 , M

crossing the segment [ y1 ,M].

(4)

m y1,M = y ( x, y1 ) dx .
0

x1

Similarly we can calculate the number of connections m x1 , M , segments [ x1 , M], [0, x1 ], and [0, y1 ] respectively.
y1

m0, x1 and m0, y

(5)

crossing the

m x1 , M = x ( x1 , y )dy
0

m0, x1 = y ( x,0) dx
0

x1

(6)

m0, y1 = x (0, y ) dy .
0

y1

(7)

VIII.

Functional relationship Between Spatial Density of Connections and the Number of terminals of the connectedness field

Using the above results, we can calculate the number of connections that cross the edges of the rectangle [0, y1, M, x1] (Fig.4.) as a sum:

m = m x , M + m0 , x + m0 , y
1

+ m y1, M
y1

(8)
x1

After substitution of the expressions (4), (5), (6), and (7) in (8) we have

m( x1 , y1 ) = x ( x1 , y )dy + y ( x,0)dx + + x (0, y )dy + y ( x, y1 )dx

0

y1

x1

(9)

that represents the number of terminals of the rectangle [0,y1,M,x1]. Replacing x1 and y1 in (9) with the current variables x and y and finding the second order partial derivative of the total number of terminals m( x, y ) , we obtain

2 m( x, y ) x y = + . xy x y

(10)

This expression establishes the functional relationship between the number of terminals and spatial density of connections in a differential form. IX. Total Length of Connections in the Generic Connectivity Fields

It is not difficult to show that for any small area ds = dxdy the total interconnection length can be calculated through the formula

dL = ds = dxdy .

(11)

Integrating (11) for an arbitrary x * y connectedness field we have

L( x, y ) = ( x, y ) dxdy .
0 0

x y

(12)

If the connectedness field has constant density, then from (12) we can obtain a practical simple formula for estimating the total length of connections L=

S,

(13)

Where S is the total area of the connectivity field. Also from (12) it follows that

2 L ( x, y ) = ( x, y ) , xy

(14)

which establishes the functional relationship between the total length of connections and the spatial density of connections in the arbitrary connectedness field.

X.

Differential Equation of Connectedness (Steady State case)

Based on the fact that the number of elements and connections between them for a certain part of the connectedness field are strongly interrelated, lets explore the balance between them in the vicinity of an arbitrary point (x, y). For doing that lets consider a small area ds = dxdy (Fig.5). Connections are incoming into that area from the left and bottom sides of the area ds and are outgoing from the top and right sides of the same area. According to the expression (5), the number of connections min that enters into the small area ds from the left side can be determined as minl = x ( x, y )dy . (15) Similarly, the number of outgoing connections moutr from the right side of the same small area ds can be calculated by the formula moutr = x ( x + dx, y )dy . (16) Thus the difference between the numbers of incoming and outgoing connections in the direction of x axis equals to moutr minl = [ x ( x + dx, y ) x ( x, y )]dy . (17) Similarly we can find the number of incoming connections from the bottom of the small area ds as minb = y ( x, y ) dx . (18) The number of outgoing connections moutt from the top of the same small area ds can be calculated by the formula

moutt = y ( x, y + dy )dx .

(19)

Thus the difference between the numbers of incoming and outgoing connections in the direction of y axis equals to

moutt minb = [ y ( x, y + dy ) y ( x, y )]dx .

(20)

So the total difference between the numbers of incoming and outgoing connections for the small area ds can be calculated as the sum of the differences (17) and (20)

m out m in = m outr m inl + m outt m inb

Substituting the expressions (17) and (20) into (21) we will have

(21)

m out min = [ y ( x, y + dy ) y ( x, y )]dx + [ x ( x + dx, y ) x ( x, y )]dy .

(22)

Lets analyze this expression assuming that the density of connections is a continuous function in the small area ds and in its vicinity. Based on that we can represent x ( x + dx, y ) as a power series by dx

x ( x + dx, y ) = x +

Substituting (23) into the (22) and neglecting the higher order terms we have

x dx + x
.

(23)

mout min =
moutt minb =
mout min = (

Similarly representing x ( x, y + dy ) as a power series by dy we have

x dxdy x
dxdy .

(24)

y y

(25)

Substituting the results obtained from (23) and (25) into the expressions (21) or (22) we will have

x y + )dxdy . x y

(26)

As it can be seen from Fig.5, the difference between the numbers of incoming and outgoing connections is formed as a result of the termination of a portion of incoming connections in the small area and generation of a portion of outgoing connections in the same small area. In order to account for that termination and generation process of connections lets introduce the notions of densities of connection sinks s ( x, y) and connection sources n ( x, y ) in the connectivity field. Based on these notions we are able to calculate the same difference
y

mout
y + dy dy y

min x dx x + dx 0 a x Figure 5: Balance between the connections and contacts of elements in the small area ds = dxdy

mout min by the following expression:

mout min = [ n ( x, y ) s ( x, y )]dxdy

Equating the expressions (26) and (27) we will have

(27)

 x y + = n ( x, y ) s ( x, y ) x y

(28)

which represents the above-mentioned balance between the elements and connections in the form point y = 0 can serve as a boundary condition for the equation (28), that is to say - functions

of a differential equation. The functions x ( x, y ) in the point x = 0 and y ( x, y ) in the

x (0, y ) and

y (x,0)

in the left and bottom edges of the connectedness field.

XI. Other Forms of the Connectedness Differential Equation

Because the elements in the connectedness field are the main sinks and sources of the connections, it is evident that the equation (28) contains information about the elements as well, but in an indirect (implicit) form. In order to deal with the spatial densities of elements n( x, y ) lets denote the numbers of the connection sources and sinks for the elements in the vicinity of the point ( x, y ) by p ( x, y ) and q ( x, y ) . It is obvious that

n ( x, y ) = p ( x, y ) n ( x , y )
and

(29) (30)

s ( x, y ) = q ( x, y ) n( x, y ) .
x y + = [ p ( x, y ) q ( x, y )]n( x, y ) . x y

Substituting (29) and (30) into the main equation (28) we can have another form for the latter.
(31)

From mathematical point of view it is convenient to incorporate the components of the density vector of connectedness directly into the right hand side of the equation (28). In order to do that lets introduce two new notions: connection termination intensity ( x, y ) and connection generation intensity ( x, y ) such that ( x, y )dxdy is the probability of the termination of one connection in the area ds , and ( x, y )dxdy is the probability of origination of one connection in the area ds . Based on that we can define the connection sink and source densities by the following formulas:

s ( x, y ) = ( x, y )( x + y )
n ( x, y ) = ( x, y )( x + y ) .

(32)

and

(33)

Substituting (32) and (33) into the main equation (28) we can have a very useful form of the differential equation of connectivity.

x y + = [ ( x, y ) ( x, y )]( x + y ) x y

(34)

This equation with corresponding boundary conditions is able to describe arbitrary massively interconnected systems. Lets apply the equation (34) for connectedness analysis of one- and two-dimensional cases of spatially distributed elements. The main goal here is to describe the spatial connectivity of a stand-alone element. Then the next step is to solve the differential equation for that element and find its connectedness field. If a system incorporates some number of elements, then applying the

solution for one element and the principle of superposition of the connectedness field we can find connectedness characteristics for the arbitrary configuration or shape of connectedness fields. This approach is valid for both homogeneous and heterogeneous connectedness systems. XII. Differential Equation for Separate Elements. One-dimensional case

Lets consider a simple homogeneous connectedness system in the form of an infinite chain of elements positioned along the x axis in the points with integer values. Each element has terminals half of which are connected with the left-hand side elements and another half of the terminals are connected with the elements on the right-hand side (Fig.6.). All the elements in such a system have the same connectedness characteristics (Fig.7) and the system can be described by connectedness equation (34), with

y = 0

where all parameters are the functions of x variable only. In this equation x represents the number of connections length of which exceeds x.

x = [ ( x) ( x)] x , x

(40)

Fig.6. Left and right branches of the connectedness function x (x) Lets separate the element positioned in the point x = 0 along with its connections. In this case connection origination function (x) transforms into a boundary condition x (0) = equation itself has the form

, and the 2

Since by definition

x (x) is the number of connections in the section x that have been

x = ( x) x . x

(41)

originated from element 0, the ratio

x (x) / will be the probability that the length of

2

connections of that element will exceed x. Consequently we can define the connection length distribution function as

F ( x) = 1

The solution of the equation (41) has the form

x x ( x) = Exp[ ( y )]dy , 2 0

2 x

(42)

(43)

which describes the right branch in Fig.6 and is directly related with the connection length distribution function F(x), which can be described as

F ( x) = 1 Exp[ ( y )]dy
0

(44)

Combining the solution (43) with the principle of superposition of the connectivity field, we can calculate connectivity characteristics for the group of N elements (Fig.7). Assume we need to estimate the number of terminals for that group of elements (each one of those has the same connectedness function) with the other neighboring elements.

Fig.7. Group connectedness exploration Due to the symmetric character of the connectedness functions, we can calculate the number of terminals outgoing from the right side of the group only - m r (N ) . Besides, we can represent the number of terminals as the sum (superposition) of the number of terminals for each separate element. The number of terminals for an arbitrary element i will be the value of the function x ( x i ) in the point N , i.e. mi = x ( N i ) . Thus the total number of terminals m(N ) can be calculated as the following sum

m( N ) = 2m r ( N ) = 2 mi = 2 x ( N i ) . (45)
1 1

For the large number of elements we can make a transition to the continuous analogue of the same rule

m( N ) = 2 x ( N x)dx .
0

(46)

Differentiating relationship (46 ) with regard to the total number of elements N we can have
dm( N ) = 2 x ( N ) . dN

(47)

This expression establishes a functional relationship between connectedness characteristics of the system and connectedness characteristics of the element x . Taking into account the relationship (42) between

x (x) and connection length distribution function

dm( N ) = [1 F ( N )] , dN

F (x) we will have

(48)

which establishes a functional relationship between the number of terminals of a homogeneous one-dimensional system and connection length distribution function. XIII. Differential Equation for Separate Elements. Two-dimensional case

Lets analyze a two-dimensional array of the identical elements as a connectedness homogeneous system. Similar to the one-dimensional case we can divide the analysis problem into two stages.

First, to solve the problem with the purpose to find the connectedness field) for a separate element that is positioned in the point (0, 0), then applying the principle of superposition of the connectedness field to find the resulting connectivity field for many elements. For that purpose lets use equation (34). Transforming the connection origination intensity ( x, y ) into a boundary condition, we can describe the 2D wiring field of one element positioned in the point (0, 0) by the equation

x y + = ( x, y )( x + y ) . (49) x y Due to the symmetry of the connectedness field ( x, y ) we can easily solve this equation

making a transition to the polar coordinates. Then we can describe the connectedness field of the element positioned in the arbitrary point ( x1 , y1 ) as ( x x1 , y y1 ) . Applying formulas (4)-(7) to this field we can find the number of terminals for the arbitrary element in the field. Combining solutions for any boundary segment of the field we can estimate the number of terminals for any close contour. I. Differential Equation for the Connectedness Potential

In previous sections we discussed solutions of the connectedness equation for simple homogeneous cases when it is possible to find the solution for one element and make a transition from it to the system level connectedness characteristics. In the general case it is impossible to solve this equation because it contains two unknown functions need to find another general form of connectedness equation that can allow overcoming this problem. The conventional way to do that is to find the so-called connectedness potential, a function that has functional relationships with

x and y . Therefore there is a

= GradP ,

x and y such that

(50) (51)

where

P is the connectedness potential. This simply means, that

P P = y . = x and y x

Taking into account (51) we can obtain the general form of the connectedness equation from (34)

2P 2P P P + 2 = [ ( x, y ) ( x, y )]( + ) . x y x 2 y
1 P 2 P 2 P P P = 2 + 2 [ ( x, y ) ( x, y )]( + ), D t x y x y

(52)

Further generalization of this equation for time dependent connectedness description leads to the following equation (53)

where coefficient D is a constant describing the diffusion of connections into the environment being under study. II. References [1] P. Barseghyan, An Introduction to the Theory of Electronic Packaging and Connectedness. Moscow Aviation Institute Press. Moscow. 1989. 50 pages (in Russian). [2] P. Barseghyan, Mathematical Basics of the Theory of Connectedness and Interconnections. Moscow Aviation Institute Press. Moscow. 1990. 114 Pages (in Russian).