Surigao State College of Technology

Surigao City 1

Chapter 1

INTRODUCTION

This chapter presents the background of the study. The

statement of the problem, objectives of the study, significance

of the study, scope and limitation of the study, the methodology,

some fundamental concepts and preliminary results are discussed

in this chapter.

Background of the study

Nonnegative matrices have long been a source of interesting

and challenging mathematical problems. They are real matrices

with all their entries being nonnegative and arise in a number of

important application areas: communications systems, biological

systems, economics, ecology, computer sciences, machine learning,

and many other engineering systems. Inverse eigenvalue problems

constitute an important subclass of inverse problems that arise

in the context of mathematical modeling and parameter

identification [1,5].

Classifying spectra of nonnegative matrices is known as the

nonnegative inverse eigenvalue problem (or NIEP). The real

nonnegative inverse eigenvalue problem (or RNIEP, introduced by

Suleimanova [15], is to determine necessary and sufficient

conditions on the list σ of n real numbers ( λ 1 , λ2 , … , λn ) so that σ

 Surigao State College of Technology
Surigao City 2

is the spectrum of an entry-wise n × n nonnegative matrix A . If

there exists such a nonnegative matrix A with spectrum σ , the

list σ is said to be realizable or we say the matrix A realizes

σ . If we further require that the realizing matrix be symmetric

we call the problem the symmetric nonnegative inverse eigenvalue

problem (or SNIEP).

The RNIEP is solved for n ≤ 4, where the the RNIEP is

equivalent to the NIEP when n=1 and n=2, as the solution applies

when we ask that the eigenvalues be real. The cases n=3, 4 were

solved by Loewy and London [11]. The RNIEP is open for any n ≥ 5,

an evidence of its difficulty.

The nonnegative inverse eigenvalue problem is an open and

difficult problem. A full solution is unlikely in the near

future. A number of partial results are known in the literature

about the problem, most of them in terms of sufficient conditions

[2, 3, 8, 10, 11, 13, 14, 15]. This paper investigates a question

raised by Paparella [13] as to whether any realizable list σ can

be realized by a permutative matrix or a direct sum of

permutative matrices.
 Surigao State College of Technology
Surigao City 3

Statement of the Problem

This paper addresses the question raised by Paparella [13]

as to whether any realizable list σ can be realized by a

permutative matrix or a direct sum of such matrices.

Objectives of the Study

The purpose of the study is to consider the case where the

real list σ can be realized by a permutative matrix or direct sum

of permutative matrices.

Significance of the Study

Nonnegative matrices arise in a number of important

application areas: communications systems, biological systems,

economics, ecology, computer sciences, machine learning, and many

other engineering systems. Inverse eigenvalue problems constitute

an important subclass of inverse problems that arise in the

context of mathematical modeling and parameter identification. A

simple application of such problems is the construction of

Leontief models in economics. This research may serve as a

mathematical tool in studying more characterizations for NIEP

(and RNIEP)—which are currently open problems—and consequently,

be beneficial to both pure and applied sciences.

Scope and Limitation

 Surigao State College of Technology
Surigao City 4

This research focuses on the real nonnegative inverse

eigenvalue problem which can be realized by permutative matrices

or their direct sum.

Methodology

The author started by reading articles related to RNIEP.

Then the author presented definitions of the concepts essential

to the study and further illustrated these definitions by

providing examples. Basic concepts and related results were then

collected and presented in the review of related literature. A

variety of the methods of proof were utilized for the results.

Basic/Preliminary Concepts

This section contains important concepts to better

understand this study.

Definition 1.1 [10] An m× n matrix A is a rectangular array of mn

real or complex number arranged in m horizontal rows and n

vertical columns:

a11 a12 ⋯ a1 n
a
A= 21
⋮
[a22 ⋯ a2 n
⋮ aij ⋮
a m 1 am 2 ⋯ amn
.
]
The ith row of A is

[ ai 1 ai 2 ⋯ a ¿ ] (1 ≤ i≤ m ) ;

The jth column of A is

 Surigao State College of Technology
Surigao City 5

a1 j

[]
a2 j
⋮
amj
( 1 ≤ j≤ n ) .

Definition 1.2 [10] A matrix with the same number of rows as

columns is called square matrix. An n × n matrix A=[ aij ] is called a

diagonal matrix if [ aij ]=0 for i≠ j . For a square matrix [ aij ], the

diagonal entries are the entries a 11 ,a 22 , … , ann, which form the main

diagonal. Thus, for a diagonal matrix, the terms off the main

diagonal are all zero. A scalar matrix is a diagonal matrix whose

diagonal elements are equal. The scalar I n=[ d ij ], where[ d ii ] =1 and

[ d ij ]=0 for i≠ j, is called the n × n identity matrix. For any square

matrix A, define the powers of a matrix, for p a positive

integer, by A P= A ∙ A ∙∙ ∙∙ ∙ A .

Example 1.3 Let

1 0 0 2 0 0 1 0 0

[ ] [ ] [ ]
A= 0 2 0 , B= 0 2 0 ,∧I 3= 0 1 0 .
0 0 3 0 0 2 0 0 1

Then A , B∧I 3 are diagonal matrices; B∧I 3 are scalar matrices; I 3

is the 3 ×3 identity matrix.

Definition 1.4 [10] If A=[aij ] is an m× n matrix, then the tanspose

of A, AT =[aTij ], is the n × m matrix defined by a Tij =a ji . Thus the

transpose of A is obtained from A by interchanging the rows and

 Surigao State College of Technology
Surigao City 6

columns of A . A matrix A with real entries is called symmetric

if AT = A .

1 2 3

[ ]
Example 1.5 The matrix A= 2 3 5 is symmetric.
3 5 6

Definition 1.6 [10] An n × n matrix A is called nonsingular, or

invertible, if there exists an n × n matrix B such that AB=BA=I n ;

such that B is called an inverse of A . Otherwise, an n × n matrix

A is called singular, or noninvertible.

3
Example 1.7 Let A= 2 3 ,andB=
[ ]
2 2
−1
[ ]
2 .
1 −1
Since AB=BA=I n, we

conclude that B is an inverse of A .

Definition 1.8 [10] If we start out with an n × n matrix A=[ aij ] and

then cross out some, but not all, of its rows or columns, we

obtain a submatrix of A . A principal submatrix is a square

submatrix obtained by removing certain rows and columns.

Example 1.9 Let

1 2 −4

[
A= 0 1 1
3 0 3 ]
If we consider the second row and third column, we get the

submatrix

A23= [ 13 20 ]
 Surigao State College of Technology
Surigao City 7

If we have the first row and second column of submatrix A23, then

the principal submatrix is

A12=[ 3 ] .

Definition 1.10 [7] A square matrix P is a permutation matrix if

exactly one entry in each row and column is equal to 1 and all

other entries are 0 .

Example 1.11 The following illustrates how a permutation matrix

produces a permutation of the rows (entries) of a vector: it

sends the first entry to the second position, sends the second

entry to the first position, and leaves the third entry in the

third position:

0 1 0 1 2

[ ][ ] [ ]
1 0 0 2=1
0 0 1 3 3

Definition 1.12 Let A be an n × n matrix. We call A a permutative

matrix if there exist x=[ x 1 , x 2 ,… , x n ]T and n × n permutation matrix

P2 , P3 , … , P n such that

xT

[ ]
(P x)
B= 2
⋮
( Pn x )
T

T
.

Definition 1.13 [12] An n × n matrix A is said to be a reducible

matrix when there exists a permutation matrix P such that

 Surigao State College of Technology
Surigao City 8

PT AP= X Y ,
0 Z[ ] where X and Z are both square.

Otherwise, A is said to be an irreducible matrix.

Example 1.14 Let A= [ 10 11] . Then the reducible matrix is

T
P AP= [10 11] . On the other hand, if we let this matrix A=[ 01 10],
0 1
then the irreducible matrix is P AP=
T
[ ] 1 0
.

Definition 1.15 [10] For an n × n matrix A the determinant of A ,

denoted det ⁡( A) or| A|, is a scalar that is computed as the sum of

all possible products of n entries of A each with its

appropriate sign, with exactly one entry from each row and

exactly one entry from each column.

2 −3 ,| A|= (2 )( 5 )− (−3 ) ( 4 )=22

Example 1.16 Let A=
4 5 [ ]
Definition 1.17 [6] A polynomial p ( x ) =x n+ an−1 x n−1 +…+ a1 x +a0 in which

the coefficient of the highest order term is 1 called a monic

polynomial.

Example 1.18 The polynomial P ( x ) =x5 +2 x 4 −5 x 3 +3 x 2−3 x+1 is monic.

Definition 1.19 If A is an n × n nonnegative matrix with spectrum

σ , then σ said to be realizable and the matrix A is called a

realizing matrix for σ . A list σ =( λ1 , λ 2 , … , λn ) of a spectrum of an

 Surigao State College of Technology
Surigao City 9

n × n nonnegative matrix A that consists of real numbers is called

a realizable list of σ .

Definition 1.20 [12] An m× n matrix A=[aij ] is said to be a

nonnegative matrix whenever each a ij ≥0, and this is denoted by

writing A ≥ 0.

Definition 1.21 [12] Let A be m× n and B be a p ×q matrix. By the

direct sum of A and B, written A ⊕B , we mean the ( m+ p ) ×( n+q)

matrix of the form

[ OA OB ]
where the O ' s represent the zero matrices. The O on the top
right is an m× n matrix, while the O on the bottom left is n × p.

4 3 2 1
Example 1.22 If A=
3 −1
and B= [
5 −1 ]
, then [ ]
4 3 0 0

[ A O =3
O B] 0
0
[ −1 0 0
0 2 1
0 5 −1
.
]
Definition 1.23 [7] The eigenvalues of an n × n matrix A are the

roots of its characteristic polynomial p ( λ )=det( λI −A ). The set of

these roots is called the spectrum of A and is denoted by

σ =( λ :det ( λI − A )=0 ) .

If σ =( λ1 , λ 2 , … , λn ), then det ( A )=λ 1 λ 2 ⋯ λ nand tr ( A )=λ1 + λ 2+ ⋯+ λ n, where the

trace function is the sum of the diagonal entries, that is,

 Surigao State College of Technology
Surigao City 10

n
tr ( A )=∑ aii .
i=1

For the spectrum of a matrix, the spectral radius is defined as

follows:

ρ ( A )=max ¿ λi∨¿ . ¿
λ i ∈σ

Example 1.24 Consider the following matrix.

A= [−23 10]
Then the characteristic polynomial

p ( λ )=det( λI −A )=( 3−λ ) (−λ ) −1 (−2 ) =λ2−3 λ+ 2.

The eigenvalues are λ 1=1 , λ 2=2. The spectrum is σ =(1 , 2) .

The trace of a matrix is

n
tr ( A )=∑ aii =a11 +a 22=3+0=3
i=1

The spectral radius is ρ ( A )=2 .

Definition 1.25 [4] A loop is an edge (or arc) that joins a

vertex to itself.

Definition 1.26 [4] A circuit { v 1 , v 2 , … , v n , v 1 } (n ≥ 3) whose n vertices vi

are distinct is called a cycle. A cycle of length n is an n-

cycle; a 3-cycle is also called a triangle. A graph of order n

that is a path or a cycle is denoted by Pn or C n, respectively.

Definition 1.27 [4] A graph G consists of finite nonempty set V

of p vertices together with a prescribed set X of q unordered

 Surigao State College of Technology
Surigao City 11

pairs distinct vertices of V . Each pair x= {u , v } of vertex X is a

edge of G, and x is said to join u and v. We write x=uv and say

that u and v are adjacent vertices(sometimes denoted u adj v);

vertex u and edge x are incident with each other, as are v and x .

If two distinct edges x and y are incident with a common vertex,

then they are adjacent edges.

Definition 1.28 [4] A directed graph or digraph is a finite

nonempty set of objects called vertices together with a (possibly

empty)set of ordered pairs of distinct vertices of D called arcs

or directed edges . A digraph has no loops or multiple arcs, see

Figure 1.

Definition 1.29 [11] Let A=(a ij ) be an n × n nonnegative matrix. Let

σ ={ λ 1 , … , λn } be the set of eigenvalues of A and let the kth moment

sk ( σ) of σ defined by

n (1.
k k
sk ( σ ) =∑ λ =tr( A ), k =1,2, …
1
i=1 1)
k
when tr means trace. As A is a nonnegative matrix for k =1,2 , … it

follows that

(1.
sk ( σ ) ≥ 0 , k=1,2 , …
2)
 Surigao State College of Technology
Surigao City 12

Preliminary Results

This section contains preliminary results which are used in this

paper.

5
Theorem 1.1 [14] Let σ = ( λ 1 , λ2 , λ 3 , λ 4 , λ 5 ) and let sk =∑ λki . Suppose
i=1

λ 1 ≥ λ2 ≥ λ 3 ≥ λ 4 ≥ λ5 ≥− λ1.Then σ is a spectrum of a matrix A ∈ R if and

only if the following conditions hold:

1.s1=0 ,

2. s3 ≥0 ,

3. λ 2+ λ 5 ≤ 0.

Theorem 1.2 [10] Let λ 1 , λ2 , λ3 , λ4 and λ 5 be a complex numbers and let

sk =λk1 + λ k2+ λk3 + λ k4 + λ k5fork =1,2,3 , …. Assume s1=0. Then ( λ 1 , λ2 , λ3 , λ4 , λ5 ) is the

spectrum of a nonnegative 5 ×5 matrix if and only if the

following conditions hold:

1.sk ≥ 0 for k =2,3,4,5

2.4 s 4 ≥ s 22 , and

3.12 s 5−5 s 2 s3 +5 s3 √ 4 s 4−s22 ≥ 0.

 Surigao State College of Technology
Surigao City 13

Chapter 2

REVIEW OF RELATED LITERATURE

This chapter presents existing literature related to this study.

Loewy and London (1978) solved the RNIEP for n ∈{3 , 4 }. The

problem for n ≥ 5 remains unsolved. It was shown by Johnson,

Laffey and Loewy in 1996 that the RNIEP and the SNIEP are

equivalent for n ≤ 4 but that they are not for n ≥ 5. Hence for

any realizable list of real eigenvalues σ =( λ1 , λ 2 , … λ n), where n ≤ 4,

there exists also a non-negative symmetric matrix with spectrum

σ. Consequently, the SNIEP is also unsolved for n ≥ 5, though a

solution for n = 5 when the trace of the realizing matrix is zero

was given by Spector [14] in 2011.

In 1996, Johnson et al. [8] proved that RNIEP and the SNIEP

are different. To describe the result of Johnson et al. we need

the following results due to Boyle and Handelman. In 1991 and

1993, Boyle and Handelman published two papers [2,3] in which

they used symbolic dynamics to study the NIEP. The main results

states that there is a realizing nonnegative matrix if enough

zeros are added to a given list of complex numbers.

 Surigao State College of Technology
Surigao City 14

In [13], Paparella considered realizable lists which can be

realized by permutative matrices or their direct sum. He showed

that for n ≤ 4 this is always possible. Moreover, it is shown in

[13] that if σ contains one positive number and is realizable

then it can be realized by a permutative matrix, thus giving a

constructive proof of a well-known theorem of Suleimanova [15].

At the end of [13], the author raises the following problem: “Can

every realizable (real) list be realized by a permutative matrix

or a direct sum of such matrices?” This problem is significant in

the following sense: If the answer is positive, then for any

realizable (real) list there is a realizing matrix which has a

simple and explicit form involving only n numbers. It is our

purpose to answer this question negatively. This will be done in

the next chapter.

 Surigao State College of Technology
Surigao City 15

Chapter 3

RESULTS

Main Results

This chapter presents the result of this study.

In response to the question raised by Paparella, we will exhibit

a list of five real numbers which is realizable, but cannot be

realized by a permutative matrix or a direct sum of such

matrices.

Let

(3.
8 51 8 51
a= + √ and b= − √
25 50 25 50 1)
and let

(3.
4 21
(
σ = 1 , a , b ,− ,−
5 25 )
2)
For simplicity, and since σ is fixed in this section, we write sk

for sk ( σ ) . We find the following:

5
s1=∑ λ11=¿ 1+
i =1
( 258 + √5051 )+( 258 − √5051 )+( −45 )+( −21
25 )
=0 ¿

5 2 2
8 √ 51 8 √ 51 −4 2 −21 2 3239
s2=∑ λ22=¿ 12 +
i =1
( +
25 50 ) (
+ −
25 50
+
5
+ )( )( )
25
=
1250
¿
 Surigao State College of Technology
Surigao City 16

5 3 3
8 √ 51 8 √ 51 −4 3 −21 3
s3=∑ λ33=¿ 13 +
i=1
( +
25 50 )(+ −
25 50
+
5
+ ) ( )( )
25
=0 ¿

5 4 4
8 √ 51 8 √51 −4 4 −21 4 6107321
i=1
4
s4 =∑ λ =¿ 1 +
4
4
+
25 50 ( + −
25 50 ) (
+
5
+
25
= ) ( ) ( )
3125000
¿

5 5 5
8 √ 51 8 √51 −4 5 −21 5 861
s5=∑ λ55=¿ 15 +
i=1
( +
25 50 )(+ −
25 50
+
5
+
25) ( )( )
=
3125
¿

6107321 3239 2 2
Observe that s1=0, sk ≥ 0 for k =2 ,3 , 4 , 5, 4 s 4 =4
3125000
≥
1250( =s 2, )( )
and

861 3239 6107321 3239 2

12 s 5−5 s 2 s3 +5 s3 √ 4 s 4−s22=12 ( 3125) (
−5
1250 ) √(
( 0 ) +5(0) 4
3125000
− )(
1250 )
≥ 0.

Hence, by Theorem 1.2, σ is realizable. In fact, σ can be

realized even as the spectrum of a 5 ×5 symmetric nonnegative

matrix. This follows from Theorem 1.1 since s1=0 and s3=0, while

λ 2+ λ 5= ( 258 + √5051 )+( −21

25 )
<0. It follows from (3.2) that every realizing

matrix has to be irreducible.

We are left to show that this list cannot be realized by a

permutative matrix or a direct sum of such matrices. Proof for

this is given in the following theorem.

 Surigao State College of Technology
Surigao City 17

Theorem 3.1. The list σ cannot be realized by a permutative

matrix or a direct sum of such matrices.

Proof. Suppose, on the contrary, that there exists a permutative

matrix A with spectrum σ . Then,

x1 x2 x3 x4 x5

[. .
A= . .
. .
. .
.
.
.
.
.
.
.
.
.
. ,
.
.
]
where x i , i=1,2,3,4,5, are nonnegative real numbers. Also, here and

throughout, “ · ” denotes an entry in matrix A which might be 0

or positive. Each row of A consists of a suitable permutation of

these numbers. For instance, we consider

0 1 0 0 0
0
A= 0
1
0
[ ] 0
0
0
0
1
0
0
0
0
0
0
1
0
1 .
0
0

By Definition 1.30,

5
s1 ( σ )=∑ λ11=tr ( A)=0
i =1

and

5
s3 ( σ )=∑ λ31=tr ( A3 )=0
i=1

Note that s1=0implies tr ( A )=0and s3=0implies tr ( A 3 )=0 .

 Surigao State College of Technology
Surigao City 18

It is convenient to think about these equalities in terms of G( A)

, the directed graph corresponding to A ; the vertices of G( A)

are labeled 1 to 5, and there is a (directed) edge from vertexi

to vertex j if and only if a ij >0. We have this directed graph of A

1
3

4 5

Figure 1 Directed graph corresponding to

Notice the matrix A is a permutative matrix; that is, each row

and column has the value of 1 and the rest are 0 . Then it follows

that G( A) contains no loops or 3-cycles. In particular, all main

diagonal entries of A are zero.

It follows that x 1=0. If no other x iis equal to zero, then all the

main diagonal entries of A are equal to x 1, and hence, all off-

diagonal entries of A are positive.

A=[ 0 +¿+ ¿+¿+¿+ ¿ 0 +¿+ ¿+¿+¿+¿ 0 +¿+¿+ ¿+¿+¿ 0 +¿+¿+ ¿+¿+¿ 0 ]

 Surigao State College of Technology
Surigao City 19

Then, clearly, G( A) contains a 3-cycle, which is impossible. So,

at least one additional x i is positive, and since we can apply a

permutation similarity. We can assume without loss of generality

that x 2=0. On the other hand, since A ≠ 0there exists a positive x i

, so without loss of generality (same argument) we may assume

x 3 >0.

We now show that at least one of x 4 , x 5 must be positive. If this

is not the case, then each row of A contains exactly one

positive entry, namely x 3. If there is a column of A containing

no x 3, then A is singular, which is not the case. Otherwise,

there exists a 5 ×5 permutation matrix P such that A=x3 P,

0 1 0 0 0 0 x3 0 0 0
0
A=x3 0
1
0
[ ][
0
0
0
0
1
0
0
0
0
0
0
1
0

0
0
0 0 x3 0 0
1 = 0 0 0 0 x3
x3 0 0 0 0
0 0 0 x3 0
]
which is also impossible. Hence, at least one of x 4 , x 5 must be

positive, which we may assume to be x 4 .

This shows that when x 5 >0 then a 21=0.

Suppose that x 5 >0. Then each row of A contains 3 positive

entries. Hence, a 34+ a35> 0. We now distinguish three cases.

 Surigao State College of Technology
Surigao City 20

 Suppose that a 34 and a 35 are positive. Then, as G(A) contains

no 3-cycles, we must have a 41=a51=0. But then

a 42 , a 43 , a45 , a52 , a 53∧a54 are all positive. So,

x1 x2 x3 x4 x5

.[
A= . 0 . .
. 0 +¿+¿ 0 ¿ ]
. + ¿ 0 ¿ +¿+¿+¿ 0 ¿ ,

where “ + ” denotes throughout a positive entry. But this

is impossible,

because if a 23>0, then a 34 and a 42positive or 2 ↦ 3 ↦4 ↦2is a 3-

cycle in G( A), we have

x1 x2 x3 x4 x5
A=
[ +¿ 0 + ¿ . ¿ . ¿
0 ¿ +¿+ ¿ 0 ¿ +¿+¿ 0 ¿ +¿ 0 ¿+ ¿+¿+¿ 0 ¿
]
which is impossible, where i and j are the vertices of G( A).

While if a 24> 0, then a 45and a 52 must be positive or2 ↦ 4 ↦ 5↦ 2 is

a 3-cycle in G( A), we have

x1 x2 x3 x4 x 5
A=
[ +¿ 0 ¿ . . ¿
+¿+¿ 0 ¿ +¿+ ¿ 0 ¿ +¿ 0 ¿ +¿+ ¿+¿ 0 ¿
]
which is also impossible. Hence, this case is not possible.

 Suppose that a 34> 0 and a 35=0. Hence, a 31and a 32 are positive.

We must have a 53=0 then 1 ↦ 5 ↦ 3 ↦1 is a 3-cycle in G( A), a

contradiction. Likewise, we must have a 43=0. Hence,

a 41 , a 42 , a45 , a51 , a52∧a54are all positive, so

 Surigao State College of Technology
Surigao City 21

x1 x2 x3 x4 x5
A=
[.
+¿+¿ 0
0 .
¿ +¿+¿ 0
.
0 ¿
. +¿ 0 ¿
]
and this leads to a contradiction as follows: At least one

of a 24, a 25 is positive. If a 24 is positive then a 45 and a 52must

be positive or2 ↦ 4 ↦ 5↦ 2 is a 3-cycle in G(A), we have

x1 x2 x3 x4 x5
A=
[ . 0 . ¿ ¿
+¿ 0 ¿ +¿+ ¿ 0 ¿ 0 ¿ +¿+ ¿+¿ 0 ¿ +¿ 0 ¿ ,
]
while if a 25 is positive, then a 54 and a 42 must be positive

or 2 ↦ 5 ↦ 4 ↦ 2 is a 3-cycle in G(A), we have

x1 x2 x3 x4 x5
A=
[ . 0 . . ¿
+¿ 0 ¿ +¿+ ¿ 0 ¿ 0 ¿ +¿+ ¿+¿ 0 ¿ +¿ 0 ¿
]
Hence, this case is impossible.

 Suppose that a 35>0 and a 34=0. Hence, a 31 and a 32 are positive,

and as in (II), we must have a 43=a53=0, implying

a 41 , a 42 , a45 , a51 , a52∧a54are all positive. Hence,

x1 x2 x3 x4 x5
A=
[ .
+¿+¿ 0
0 .
0 +¿+¿+ ¿ 0 0
.
¿
. +¿ 0 ¿
]
and we get a contradiction as in (II), namely either a 24> 0,

in which case we have the 3-cycle 2 ↦ 4 ↦ 5↦ 2 , or a 25>0, in

which case we have the 3-cycle 2 ↦ 5 ↦ 4 ↦ 2.

This shows that assuming x 5 >0leads to a contradiction.

 Surigao State College of Technology
Surigao City 22

This shows that when x 5=0 then a 21>0. Suppose now that x 5=0. Then

each row of A contains 2 positive entries, x 3 and x 4 . We assume

first that a 21>0. Then we must havea 32=a42=0, otherwise G( A)

contains the 3-cycle 2 ↦ 1↦ 3↦ 2 or the 3-cycle 2 ↦ 1↦ 4 ↦ 2, a

contradiction. Then, as A is nonsingular, we must have a 52>0. So

A looks like

x1 x2 x3 x4 x5
A=
[ +¿ 0 . . ¿ ¿
0 ¿ 0 ¿ . ¿ . ¿ . ¿ 0 ¿ .¿ 0 ¿ . ¿ . ¿ +¿ . ¿ . ¿ 0 ¿
]
leading us to consider two cases.

 Suppose that a 31>0 . Then, a 43=0, or else we have the 3-cycle

1 ↦ 4 ↦ 3↦ 1. We conclude that a 41 and a 45 are positive (two

positive entries in every row). Not allowing 3-cycles

implies a 51=a24=a34=0. Hence, a 35>0, implying a 23=0, and hence

a 21, a 25>0.

x1 x2 x3 x4 x5
A=
[ +¿ 0 0 0 + ¿+¿ 0 ¿
0 ¿ +¿+ ¿ 0 ¿ 0 ¿ 0 ¿ + ¿0 ¿ +¿ . ¿ . ¿ 0 ¿
]
It follows that each of the second, third and fourth rows

has the pattern ¿, where one of the positive entries is x 3

and the other is x 4. This means that two of these rows are

identical, so A is singular, a contradiction.

 Suppose that a 31=0 then this implies when a 21>0 . Then A looks

like
 Surigao State College of Technology
Surigao City 23

x1 x2 x3 x 4 x 5
A=
[ +¿ 0 0 0 +¿0 ¿ ]
0 ¿ . ¿ . ¿ . ¿ 0 ¿ . ¿ 0 ¿ . ¿ . ¿ +¿ . ¿ . ¿ 0 ¿ ,

and we must have (two positive entries in each row) a 34,

a 35>0 . Continuing as in previous arguments we conclude

a 41=a51=0, a 43, a 45> 0, a 53=0 and a 54> 0, we have

x1 x2 x3 x 4 x 5
A=
[ +¿ 0 0 0 +¿0 ¿
0 ¿ +¿+¿ 0 ¿ 0 ¿ +¿ 0 ¿ +¿ 0 ¿ +¿ 0 ¿+ ¿ 0 ¿ ,
]
This leads to a contradiction, as we get the 3-cycle

3 ↦5 ↦4 ↦ 3.

This shows that when x 5=0 we must have a 21=0.

Then (two positive entries in every row) at least one of a 23, a 24

must be positive, and by symmetry(between the indices 3 and 4) we

may assume a 23>0. We consider two cases.

 Suppose that a 24> 0.. Then a 25=0. If a 51=a52=0, then a 53

and a 54 must be positive, implying that each of the

first, second and fifth rows has the pattern ¿, where

one of the positive entries isx 3and the other is x 4.

Hence, A has two equal rows, so is singular, a

contradiction.

x1 x2 x 3 x4 x5
A=
[ 0 0 ¿ +¿ . ¿
.¿ . ¿ . ¿ . ¿ . ¿ 0 ¿ . ¿ 0 ¿ 0 ¿ +¿+¿ 0 ¿
]
 Surigao State College of Technology
Surigao City 24

By the symmetry of the indices 1 and 2 we may

therefore assume that a 51>0. Since no 3-cycle is

allowed we conclude that a 35=a45=0, leading to a

contradiction as the last column of A is zero.

x1 x2 x3 x4 x5
A=
[ 0 0 ¿ +¿ . ¿
.¿ 0 ¿ . ¿ . ¿ . ¿ 0 ¿ 0 ¿ 0 ¿ 0 ¿ +¿+ ¿ 0 ¿
]
 Suppose that a 24=0. Then a 25>0 . We claim that a 31+a 32>0 .

Otherwise a 34, a 35>0, and as no 3-cycles are allowed we

must have a 41=a51=0, so the first column of A is zero,

a contradiction. This allows us to assume that a 31>0,

so A looks like

x 1 x 2 x3 x4 x5
A=
[ 0 0 +¿ 0 + ¿ 0 ¿ ]
.¿ 0 ¿ 0 ¿ . ¿ 0 ¿ 0 ¿ . ¿ 0 ¿ . ¿ . ¿ . ¿ 0 ¿ ,

implying (no 3-cycles) a 43=0.

We claim that a 41> 0. Otherwise (two positive entries in

each row)a 42,a 45> 0, implying (no 3-cycles) a 51=a54=0, and so

a 52, a 53>0. This forces a 34=a35=0and a 32>0. Hence, G( A)

contains the 3-cycle 2 ↦ 5 ↦ 3 ↦2, a contradiction.

Therefore,a 41> 0, so A looks like

x 1 x 2 x3 x4 x5
A=
[ 0 0 +¿ 0 + ¿+¿ . ¿ ]
.¿ . ¿+ ¿ . ¿ 0 ¿ 0 ¿ . ¿ . ¿ . ¿ . ¿ . ¿ 0 ¿ ,

implying a 34=0.
 Surigao State College of Technology
Surigao City 25

We claim that a 32>0. Otherwise we must have a 35>0, and so

a 51=a52=0, while a 53, a 54> 0. As no column of A can be zero

we must have a 42> 0, but then we have the 3-cycle 2 ↦ 5 ↦ 4 ↦ 2

, a contradiction. Hence,a 32>0, so a 35=a53=0. If a 54=0, then

we must have a 51, a 52>0, implying a 45=0. But now we have

that A[345], the principal submatrix of A based on indices

3 , 4 ,5 is zero, so A is singular, which is impossible.

Hence,a 54> 0, implying a 42=0, a 45> 0, a 51=0, a 52>0. We finally

get to the only (+, 0) pattern that A can take, namely,

A=[ 00 0 + ¿+¿ 0
0 ¿ ]
+¿+¿+ ¿ 0 ¿ 0 ¿ 0 ¿ + ¿ 0 ¿ 0 ¿ 0 ¿ + ¿ 0 ¿ +¿ 0 ¿ +¿ 0 ¿ ,

so the undirected graph corresponding to G( A) is just the

5-cycle, and in G( A) if i↦ j , then j ↦ i .

1
3

4 5

2
Figure 2 Undirected graph corresponding to A
 Surigao State College of Technology
Surigao City 26

We find it convenient to change the notation, and we replace x 3

, x 4 by x , y . Recall that the spectral radius of σ is 1, and on

the other hand each row sum of A is x + y , hence we must have

x + y=1, so y=1−x. We also find it convenient to perform a

permutation similarity of our matrix so that in G( A) we have

1 ↔2, 2 ↔3, 3 ↔ 4, 4 ↔5, and 5 ↔1. Hence, the first row of A is

(0 , x ,0 , 0 , y=1−x), and every other row has each of x, y one time,

so altogether we have 16 possible matrices to check. We will

show that none will give the desired spectrum.

Denote by q (t) the monic polynomial in the indeterminate t whose

roots are the elements of σ , that is,

q ( t )=(t−1)( t−a)(t−b)(t+ 4 /5)(t +21/45),

where a and b are defined by (3.2). It follows that

q ( t )=t 5− ( 3239 3
t +
21919
t−
861
2500 ) ( 62500 ) 15625
.

−3239
Then, the coefficient oft 3 in q ( t )= . We call the 16
2500

possible matrices A1 , A 2 ,… A16 . For each of those matrices, we

will compute the characteristic polynomial, which we call q i ( t ).

For each matrix Ai , if its spectrum is σ , the coefficient of t 3

−3239
in q i ( t ) must be . This will be used to determine x . We
2500

then show that the computed x does not satisfy other

 Surigao State College of Technology
Surigao City 27

conditions forced by equality of q (t ) and q i ( t ). To describe

uniquely each matrix Ai it suffices to indicate the location

of x in each of its rows except the first one (where, as

indicated, the x is located in the second place). In the

following list, the row vectorvi consists of 4 components,

where in the ith component we give the location of x in the

(i +1)-th row of Ai .

 v1 =( 1,4,3,4 ) , v 2=( 1,4,3,1 ) , v 3=( 1,4,5,4 ) , v 4=( 1,4,5,1 ) ,

 v5 =( 1 ,2 , 3 , 4 ) , v 6 =( 1, 2 , 3 ,1 ) , v 7= (1 , 2 ,5 , 4 ) , v 8=( 1 , 2 ,5 , 1 ) ,

 v 9=( 3 , 4 , 3 , 4 ) , v 10=( 3 , 4 , 3 , 1 ) , v11 =( 3 , 4 ,5 , 4 ) , v 12=( 3 , 4 , 5 , 1 ) ,

 v13= ( 3 ,2 , 3 , 4 ) , v 14 =( 3 ,2 , 3 ,1 ) , v 15=( 13 , 2 ,5 , 4 ) , v 16=( 3 , 2 ,5 , 1 ) ,

so, for example,

0 x 0 0 1−x
x
A1= 0
0
1−x
[
0
1−x
0
0
1−x 0
0
x
0
x

x
0
0
0 1−x
0
]
and

0 x 0 0 1−x
1−x
A12= 0
0
x
[0
1−x
0
0
x
0
1−x
0
0
x
0
1−x
0
0
x
0
]
Computing the characteristic polynomials of the matrices

Ai=1,2, … , 16, yields four different polynomials, namely:

 Surigao State College of Technology
Surigao City 28

 q 1 ( t )=t 5−( 3 x 2−3 x+ 2 ) t 3−( −x 4 +2 x 3−4 x 2+ 3 x −1 ) t−x 4 + 2 x 3−x 2,

 q 4 ( t ) =t 5 −(−x 2−x+ 1 ) t 3−( −3 x 4 −6 x 3+ 5 x 2−2 x ) t+3 x 4 −6 x3 + 4 x 2−x ,

 q 6 ( t )=t 5−( −x2 + x +1 ) t 3− (−x 4 +2 x3 −x ) t−x 4 +2 x 3−x 2,

 q 12 ( t )=t 5 +5 x (−1+ x ) t 3−(−x 5 4 +10 x 3−5 x2 ) t−5 x 4 +10 x 3−10 x2 +5 x−1,

and we also have

 q 1 ( t )=q 2 ( t )=q 3 ( t )=q7 ( t ) =q15 ( t ) ,

 q 4 ( t ) =q5 ( t )=q 10 ( t )=q 11 ( t )=q16 (t ),

 q 6 ( t )=q8 ( t ) =q9 ( t )=q 13 ( t )=q14 ( t ).

Consider first the matrix A1. If its spectrum is equal to σ , then

the coefficients of t 3 in q ( t )and q 1 ( t )must coincide, so x has to

satisfy

−3239
−( 3 x 2−3 x+ 2 )= ,
2500

1 38
which yields the two solutions x 1, x 2, given by x 1 ,= − √ ,
2 50

1 38
x 2 ,= + √ . Substituting either one of those numbers into A1 we
2 50

get, in both cases, the characteristic polynomial

t 5−(3239/2500)t 3 +(2192069/6250000)t−344569/6250000 ,

which is not equal to q ( t ), so the spectrum of A1 cannot be σ .

Consider now A 4. A similar argument leads to the equation

 Surigao State College of Technology
Surigao City 29

−( x 2−x +1 )=¿−3239/2500,

which has no real roots. Hence, the spectrum of A 4 cannot be σ .

We note that the coefficients of t 3 in q 4 ( t ) and q 6 ( t ) are the same

polynomial in x , implying that the spectrum of A6 cannot be σ as

well. Finally, we consider A12. A similar argument leads to the

equation

−5 x+ 5 x 2=¿ −3239/2500

which has no real roots, so the spectrum of A12cannot be σ . As

these are the only distinct cases we have shown that σ cannot be

realized as the spectrum of a permutative matrix. Since any

nonnegative matrix with spectrum σ must be irreducible, this

completes the proof.

Chapter 4

SUMMARY AND RECOMMENDATIONS

Summary

 This study is all about the real nonnegative inverse

eigenvalue problem which has a list σ =( λ1 , λ 2 , … , λn ), consisting

of real numbers the spectrum of a n x n nonnegative matrix.

 Surigao State College of Technology
Surigao City 30

 The purpose of this study is to show that the given list

σ =( λ1 , λ 2 , … , λn ) cannot be realized by a permutative matrix or a

direct sum of such matrices.

 The method use to the main theorem is by proof of

contraction which assume every entries of a permutative

matrix to create a cycle which the 3-cycle is impossible

since we have a 5 ×5 permutation matrix. The other method

that use is by denote the q ( t ) as Monic polynomial which has

the indeterminate t whose roots are the elements of σ .

 The result of this study is clearly shown that the spectrum

cannot be realized by a permutative matrix or a direct sum

of such matrices.

Recommendation

This paper recommends further research to:

 solve the n≥5 for RNIEP, since it is open and still

unsolved.

 to show that the realizable list σ =( λ1 , λ 2 , … , λn ), can be

realized by a permutative matrix.

REFERENCES
 Surigao State College of Technology
Surigao City 31

1. Berman, A. & Plemmons R. J. (1994) Nonnegative Matrices in the

Mathematical Sciences. In: Classics in Applied

Mathematics 9, Society for Industrial and Applied

Mathematics (SIAM), Philadelphia, PA.

2. Boyle, M. & Handelman, D. (1991) The spectra of nonnegative

matrices via symbolic dynamics. Annals of Mathematics,

133 (2), 249–316.

3. Boyle, M. & Handelman, D. (1993). Algebraic shift equivalence

and primitive matrices. Transactions of the American

Mathematical Society 336(1), 121-149.

4. Chartrand, G., and Lesniak, L. (2005). Graphs and digraphs.

Chapman Hall/CRC.

5. Chu, M. T. & Golub, G. H. (2005) Inverse eigenvalue problems:

theory, algorithms and applications. Oxford University

Press, New York.

6. Dummit, D. S. and Foote, R. M. (1998) Abstract Algebra, 2nd

ed. Englewood Cliffs, NJ: Prentice-Hall.

7. Horn, R. A. & Johnson, C. R. (1987). Matrix Analysis.

Cambridge, England: Cambridge University Press.

8. Johnson, C., Laffey, T.J. & Loewy, R. (1996). The real and the

symmetric nonnegative inverse eigenvalue problems are

 Surigao State College of Technology
Surigao City 32

different. Proceedings of the American Mathematical

Society, 124(12), 3647–3651.

9. Kolman, B., and Hill, D. R. (2008). Elementary Linear Algebra

With Applications 9 th

Edition. Pearson Education Inc.

10. Laffey, T.J. & Meehan, A. (1999). A characterization of trace

zero nonnegative 5 ×5 matrices. Linear Algebra and its

Applications, 302/303:295–302.

11. Loewy, R. & London, D. (1978). A note on an inverse problem

for nonnegative matrices. Linear and Multilinear

Algebra, 6, 83–90.

12. Meyer, C. D. (2000). Matrix Analysis and Applied Linear

Algebra. Philadelphia, PA: SIAM.

13. Paparella, P. (2016). Realizing Suleimanova-type spectra via

permutative matrices. Electronic Journal of Linear

Algebra, 31, 306–312.

14. Spector, O. (2011). A characterization of trace zero

symmetric nonnegative 5 ×5 matrices. Linear Algebra and

its Applications, 434, 1000–1017.

15. Suleimanova, H.R. (1949). Stochastic matrices with real

characteristic numbers. Doklady Akademii Nauk SSSR

(N.S.), 66, 343–345.

Surigao State College of Technology
Surigao City 33

Appendix A
Surigao State College of Technology
Surigao City 34

Main Source Journal Article

 Surigao State College of Technology
Surigao City 35

CURRICULUM VITAE

John Kennard T.Seville

Puerto Princesa Basilisa,Dinagat Islands.
Contact No. 09382558583

PERSONAL DATA

Age: 23 years old

Sex: Male
Date of Birth: February 25, 1998
Place of Birth: Tarong Madridejos,Cebu
Religion: Roman Catholic
Civil Status: Single
Height: 5’4
Weight: 48kgs.
Father’s Name: Juan T. Seville
Mother’s Name: Raquel T. Seville

EDUCATIONAL ATTAINMENT

Tertiary: Surigao State College of Technology 2014-present

Narcisostreet, Surigao City
BACHELOR OF SCIENCE IN MATHEMATICS

Secondary: Puerto Princesa-Moleta National High School 2014-2015

Puerto Princesa Basilisa, Dinagat Islands

Elementary: Tarong Elementary School 2010-2011

Tarong,Madridejos, Cebu
Surigao State College of Technology
Surigao City 36