This article was downloaded by: [Australian National University]

On: 05 January 2015, At: 17:43

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Information and Optimization Sciences

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tios20

Pseudomonotonicity of an affine map and the two

dimensional case
a a
Anna Marchi & Laura Martein
a
Department of Statistics and Applied Mathematics , University of Pisa , Via Cosimo
Rodolfi 10, Pisa , 56214 , Italy
Published online: 18 Jun 2013.

To cite this article: Anna Marchi & Laura Martein (2008) Pseudomonotonicity of an affine map and the two dimensional
case, Journal of Information and Optimization Sciences, 29:3, 407-421, DOI: 10.1080/02522667.2008.10699813

To link to this article: http://dx.doi.org/10.1080/02522667.2008.10699813

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 Pseudomonotonicity of an affine map and the two dimensional case

Anna Marchi ∗
Laura Martein †
Department of Statistics and Applied Mathematics
University of Pisa
Via Cosimo Rodolfi 10
56124 Pisa
Italy

Abstract
Downloaded by [Australian National University] at 17:43 05 January 2015

In this paper the pseudomonotonidty of the affine map F ( x) = Mx + q on the interior

of the positive orthant of <n is studied. A new characterization is suggested involving the
positive and the negative polar of the cone generated by the set W ∗ = { z = Mx + q, x ∈
int <n+ }. The obtained results are applied to the two dimensional case in order to achieve a
complete characterization of pseudomonotonicity in terms of the coefficients of M and q.

Keywords : Pseudomonotonicity, pseudoconvexity.

1. Introduction

Monotonicity and convexity, generalized monotonicity and gener-

alized convexity are closely related since a differentiable function is
convex (generalized convex) if and only if its gradient map is monotone
(generalized monotone). The interest in studying generalized monotonic-
ity is also motivated by its connection with complementarity problems,
variational inequalities and more generally with equilibrium models
([6], [7], [8], [11],[13]). In particular, pseudomonotonicity of an affine map
∗ E-mail: marchiae@ec.unipi.it
† E-mail: lmartein@ec.unipi.it

——————————–
Journal of Information & Optimization Sciences
Vol. 29 (2008), No. 3, pp. 407–421
°c Taru Publications
 408 A. MARCHI AND L. MARTEIN

is of particular interest for its relationships with pseudoconvexity and

with the linear complementarity problem ([2], [7]). This subject has been
studied by several authors (see for instance ([5], [9], [10])) and some
sufficient and/or necessary conditions are given in [5] but, unfortunately,
the obtained results are not very useful in constructing or in testing
pseudomonotonicity in an easily way.
The aim of this paper is to move a first step in this direction; more
precisely we suggest a new approach to the problem which allows us to
find a new characterization of the pseudomonotonicity of an affine map
on the interior of the positive orthant. By means of the obtained results we
are able to give a complete characterization in the two dimensional case in
terms of the elements of the matrix M and of the vector q associated with
the affine map F ( x) = Mx + q.
Downloaded by [Australian National University] at 17:43 05 January 2015

In Section 2 and in Section 3 we establish a new characterization of

pseudomonotonicity of the affine map F ( x) = Mx + q on the interior of
the positive orthant, involving the positive and the negative polar of the
cone generated by the set W ∗ = { z = Mx + q, x ∈ int <n+ }. In Section 4
we characterize the pseudomonotonicity of a bijective linear map Mx in
terms of the elements of the 2 × 2 matrix M; we find as a particular case
some results given in [12]. In Section 5 we extend the obtained results to a
bijective affine map. At last in Section 6 we analyze the case where M is a
singular matrix.

2. Preliminary results

Let H ⊂ <n be a nonempty open convex set. In our approach to

pseudomonotonicity we are interested to characterize the set Z H = { z ∈
<n : ∃ a ∈ H, z T a = 0}. With this aim, let CH be the open cone generated
by H that is CH = {ta, t > 0, a ∈ H }.
The next theorem characterizes Z H by means of the polar cones
+
CH = {α ∈ <n : α T c ≥ 0, for all c ∈ CH }

and
−
CH = {α ∈ <n : α T c ≤ 0, for all c ∈ CH }.

More exactly, denoting with cl S and with ∂S the closure and the boundary
of the set S respectively, we have the following theorems.
 PSEUDOMONOTONICITY OF AN AFFINE MAP 409

Theorem 2.1.
+ − c
Z H = { z ∈ <n : ∃ a ∈ H, z T a = 0} = (CH ∪ CH ). (2.1)

Proof. First of all we recall the following known properties regarding an

open cone V with vertex at the origin:
v ∈ int V if and only if α T v > 0, for all α ∈ V + ; v ∈ cl V if and only if
α T v ≥ 0, for all α ∈ V + .
In order to prove (2.1), let us note that Z H = { z ∈ <n : ∃ c ∈ CH ,
+
zT c= 0}. Let z ∈ Z H and assume z ∈ CH . Since there exists c ∈ CH such
T
that z c = 0, necessarily we have c ∈ ∂CH and this absurd since CH is an
+
open set so that ∂CH = ∅. Consequently z ∈ / CH . In a similar way it can be
− + − c
proven z ∈/ CH and thus z ∈ (CH ∪ CH ) .
Downloaded by [Australian National University] at 17:43 05 January 2015

+ − c + −
Viceversa, let z ∈ (CH ∪ CH ) , that is z ∈
/ CH ∪ CH . Then there
T T
exist c1 ∈ CH , c2 ∈ CH such that z c1 < 0, z c2 > 0. The continuity
of the scalar product and the convexity of CH implies the existence of
c = λ c1 + (1 − λ )c2 , λ ∈ (0, 1) with c ∈ CH and z T c = 0. Consequently
z ∈ Z H and the thesis is achieved. ¤

Theorem 2.2. Let W be an open convex cone with vertex at the origin and set
W ∗ = x0 + W. Then

ZW ⊆ ZW ∗ . (2.2)

Proof. First of all we prove that W = CW ⊆ cl CW ∗ . Let w ∈ W;

we have x0 + nw ∈ x0 + W for every positive integer n. Consequently
1 1
n ( x0 + nw ) = n x0 + w ∈ CW ∗ and thus w is an accumulation point for
+
CW ∗ that is w ∈ cl CW ∗ . It follows (cl CW ∗ )+ = (CW ∗ )+ ⊆ CW , (cl CW ∗ )− =
−
(CW ∗ )− ⊆ CW so that
+ − c + − c
ZW ∗ = (CW ∗ ∪ CW ∗ ) ⊇ (CW ∪ CW ) = ZW . ¤

Consider now the following two special cases:

W = { Mx, x ∈ int <n+ }; W ∗ = { Mx + q, x ∈ int <n+ }

were M is a non singular matrix of order n.

Let us note that W is the open polyhedral cone generated by the

columns of the matrix M while W ∗ = W + q. The following theorem
characterizes the set ZW .
 410 A. MARCHI AND L. MARTEIN

Theorem 2.3. Consider the set ZW . Then

ZW = { z = ( M−1 )T y, y ∈
/ <n+ ∪ <n− }.

Proof. Obviously we have CW = W. It is known that the positive polar

cone of W is given by W + = { z : M T z ≥ 0}. Setting M T z = y we
have W + = { z = ( M T )−1 y = ( M−1 )T y, y ≥ 0} and furthermore
W − = { z = ( M−1 )T y, y ≤ 0}. The thesis follows from Theorem 2.1. ¤

Regarding W ∗ = W + q, we have the following theorem:

+ − c
Theorem 2.4. We have ZW ∗ = (CW ∗ ∪ CW ∗ ) . In particular:

(i) if q ∈ W then ZW ∗ = ZW ;
(ii) if −q ∈ W then ZW ∗ = <n \ {0}.
Downloaded by [Australian National University] at 17:43 05 January 2015

Proof. The first statement follows from Theorem 2.1 setting H = W ∗ .

Statement i) follows by noting that q ∈ W implies CW ∗ = CW .
Statement ii) follows by noting that −q ∈ W implies 0 ∈ int CW ∗ . ¤

3. A new characterization of pseudomonotonicity of a bijective affine

map on int <n+

In this section we will apply the results given in Section 2 in order to

characterize the pseudomonotonicity of an affine map on int <n+ .
We recall that a map F : S ⊆ <n → <n is said to be pseudomonotone
on S if for every x, y ∈ S we have:

( y − x)T F ( x) ≥ 0 implies ( y − x)T F ( y) ≥ 0

or equivalently (see [1])

( y − x)T F ( x) > 0 implies ( y − x)T F ( y) > 0.

A very useful characterization of pseudomonotonicity is given in [10]; in

the case F ( x) = Mx + q where M is a matrix of order n and S = int <n+
such a characterization assumes the form given in the following theorem.

Theorem 3.1. F ( x) = Mx + q is pseudomonotone on int <n+ if and only if for

all x ∈ int <n+ and v ∈ <n ,

v T ( Mx + q) = 0 implies v T Mv ≥ 0. (3.3)
 PSEUDOMONOTONICITY OF AN AFFINE MAP 411

Let us note that (3.3) is obviously verified when B = ( M + M T )/2

is positive semidefinite, so that the main problem in characterizing the
pseudomonotonicity of an affine map is related to matrices M for which B
is not positive seimidefinite.
We say that the affine map F is merely pseudomonotone if it is pseu-
domonotone and B is not positive semidefinite.
The following theorem gives a new formulation of pseudomono-
tonicity.
Theorem 3.2. F ( x) = Mx + q is pseudomonotone on int <n+ if and only if

v T Mv ≥ 0 for all v ∈ ZW ∗ . (3.4)

Proof. Taking into account Theorem 2.4, it is sufficient to note that the set
Downloaded by [Australian National University] at 17:43 05 January 2015

of all elements v such that v T ( Mx + q) = 0 is equal to ZW ∗ . ¤

Corollary 3.1. The linear map F ( x) = Mx is pseudomonotone on int <n+ if and
only if

y T ( M −1 ) T y ≥ 0 / <n+ ∪ <n− .
for all y ∈ (3.5)

Proof. We have ZW ∗ = ZW = { z = ( M−1 ) T y, y ∈ / <n+ ∪ <n− } (see

Theorem 2.3). The thesis is achieved taking into account that (3.3) becomes
v T Mv = y T M−1 M( M−1 )T y = y T ( M−1 )T y. ¤

From Theorem 2.2 and Theorem 3.2 we have the following necessary
condition for the affine map F ( x) = Mx + q to be pseudomonotone on
int <n+ .
Theorem 3.3. If the affine map F ( x) = Mx + q is pseudomonotone on int <n+ ,
then the linear map G ( x) = Mx is pseudomonotone on int <n+ .

As an application of the new characterization of pseudomonotonicity

given in Theorem 3.2 and in Corollary 3.1, in the next two sections we
express the pseudomonotonicity of a bijective map on int <2+ in terms of
the elements of the 2 × 2 matrix M in the linear case and in the affine case,
respectively.

4. Pseudomonotonicity of a bijective linear map on int <2+

Consider the linear map F ( x) = Mx on the interior of the positive or-
thant, where M is a 2 × 2 non singular matrix. In such a case, Corollary 3.1
assumes the following simple form:
 412 A. MARCHI AND L. MARTEIN

F ( x) = Mx is pseudomonotone on int <2+ if and only if (4.6) holds:

y T M−1 y ≥ 0 for all y = ( y1 , y2 ) such that y1 y2 < 0 (4.6)

" #
a c
Set M = and ∆ = (b + c)2 − 4ad. By means of (4.6) we are able to
b d
prove the following theorem.

Theorem 4.1. Let M be a nonsingular matrix. The linear map F ( x) = Mx is

pseudomonotone on int <2+ if and only if one of the following conditions holds:
a d
≥ 0, ≥ 0, ∆ ≤ 0 (4.7)
det M det M
a d b+c
≥ 0, ≥ 0, ≥ 0, ∆ > 0. (4.8)
det M det M det M
" #
Downloaded by [Australian National University] at 17:43 05 January 2015

1 d −c
Proof. From (4.6), taking into account that M −1 = det M , F ( x)
−b a
is pseudomonotone if and only if (4.9)holds:
d b+c a
Ψ( y1 , y2 ) = y2 − y1 y2 + y2 ≥ 0, y1 y2 < 0. (4.9)
det M 1 det M det M 2
Let us note that Ψ( y1 , y2 ) ≥ 0 for all y1 , y2 ∈ < if and only if the matrix
associated with the quadratic form is positive semidefinite that is if and
only if (4.7) holds.
y1
Consider now the case ∆ > 0. Setting m = y2 , (4.9) is equivalent to

d b+c a
Ψ(m) = m2 − m+ ≥ 0, for all m < 0. (4.10)
det M det M det M
Assume the pseudomonotonicity of F ( x); the validity of (4.10) implies
lim Ψ(m) ≥ 0 that is detd M ≥ 0 and lim Ψ(m) = deta M ≥ 0. The
m→−∞ m→0−
assumption 4 > 0 together with (4.10) implies that Ψ(m) must have two
b+c
nonnegative roots that is det M ≥ 0. Viceversa if (4.8) holds then Ψ (m ) has
two nonnegative roots so that (4.10) is verified. ¤

Corollary 4.1. Let M be a nonsingular matrix and consider the linear map
F ( x) = Mx. Then (i) and (ii) hold.

(i) F ( x) is merely pseudomonotone on int <2+ if and only if (4.11) or (4.12)

holds.

det M > 0, a ≥ 0, d ≥ 0, b + c ≥ 0, ∆ > 0 (4.11)

 PSEUDOMONOTONICITY OF AN AFFINE MAP 413

det M < 0, a ≤ 0, d ≤ 0, b + c ≤ 0. (4.12)

(ii) F ( x) is pseudomonotone (not merely) on int <2+ if and only if (4.13) holds.
a ≥ 0, d ≥ 0, ∆ ≤ 0. (4.13)

Proof. It is sufficient to note that det M < 0 implies ∆ > 0. In fact

(b + c)2 − 4ad ≤ 0, ad − bc < 0 implies (b + c)2 ≤ 4ad < 4bc that is
(b − c)2 < 0 and this is absurd. ¤

Example 4.1. The following matrices are associated with merely pseu-
domonotone linear maps:
" #
0 α
M= , αβ < 0, α + β ≥ 0, γ ≥ 0
β γ
" #
Downloaded by [Australian National University] at 17:43 05 January 2015

0 α
M= , αβ > 0, α + β ≤ 0, γ ≤ 0
β γ
" # " #
1 2 −1 −2
M= ;M = .
4 3 −1 −1

5. Pseudomonotonicity of a bijective affine map on int <2+

Consider the affine map F ( x) = Mx + q where M is a 2 × 2 non

singular matrix; by means of Theorem 3.2 we are able to characterize the
pseudomonotonicity of F ( x) on int <2+ .
First of all we must characterize the set ZW ∗ . Since M is nonsingular,
q is a linear combination of the columns m(1) , m(2) of M, that is there exist
α , β ∈ < such that q = α m(1) + βm(2) . In what follows we will use the
following notations:

• (m(1) )⊥ , (m(2) )⊥ are the orthogonal space of m(1) , m(2) respectively;

• M1 is the matrix having q, m(1) as columns;
• M2 is the matrix having q, m(2) as columns.

We have the following theorem.

Theorem 5.1. It results:

(i) If α ≥ 0, β ≥ 0 then ZW ∗ = ZW .
(ii) If α < 0, β < 0 then ZW ∗ = <2 .
 414 A. MARCHI AND L. MARTEIN

(iii) If α < 0, β = 0 then ZW ∗ = <2 \ (m(1) )⊥ .

(iv) If α = 0, β < 0 then ZW ∗ = <2 \ (m(2) )⊥ .
(v) If α > 0, β < 0 then ZW ∗ = {v = ( M2−1 )T y, y1 y2 < 0}.
(vi) If α < 0, β > 0 then ZW ∗ = {v = ( M1−1 )T y, y1 y2 < 0}.

Proof. (i) It follows by noting that W ∗ = q + W ⊂ W.

(ii) It follows by noting that 0 ∈ int W ∗ , so that CW ∗ = ZW ∗ = <2 .
(iii) we have CW ∗ = { z = t( x1 +α )m(1) + tx2 m(2) , t > 0, x1 > 0, x2 > 0, α < 0}.
Set λ = t( x1 + α ) and µ = tx2 ; choosing ² > 0 such that x1 = −α ± ² > 0
and t = ²z , z > 0, we obtain CW ∗ = {λ m(1) + µ m(2) , λ ∈ <, µ > 0}. It
follows ZW ∗ = ((CW ∗ )+ ∪ (CW ∗ )− )c = <2 \ (m(1) )⊥ .
Downloaded by [Australian National University] at 17:43 05 January 2015

(iv) The proof is similar to the one given in (iii).

(v) We have CW ∗ = { z = t( x1 + α )m(1) + t( x2 + β)m(2) , t > 0, x1 > 0, x2 >
0, α > 0, β < 0}. Set λ = t x1α+α > 0 and µ = t α x2 − α
β x1
> 0; by simple
calculations we obtain CW = {αλ m + (µ + βλ )m = λ q + µ m(2) , λ > 0,
∗ ( 1 ) ( 2 )

µ > 0}. In other words, CW ∗ is the open convex cone generated by the
columns of the matrix M2 , so that the thesis follows from Theorem 2.3
applied to the matrix M2 .
(vi) The proof is similar to the one given in (v). ¤
" #
a c
As in the previous section set M = and ∆ = (b + c)2 − 4ad.
b d
Now we are able to characterize the pseudomonotonicity of an affine map
on the interior of the positive orthant.

Theorem 5.2. Let M be a nonsingular matrix and consider the affine map
F ( x) = Mx + q. Set (α , β) T = M−1 q. Then F ( x) is merely pseudomonotone
on int <2+ if and only if one of the following cases is verified.

(i) α ≥ 0, β ≥ 0 and one of the following conditions holds:

det M > 0, a ≥ 0, d ≥ 0, b + c ≥ 0, ∆ > 0 (5.14)

det M < 0, a ≤ 0, d ≤ 0, b + c ≤ 0, ∆ > 0 (5.15)

(ii) α > 0, β < 0 and one of the following conditions holds:

β a
det M > 0, a > 0, d = 0, b + c > 0, − ≤ (5.16)
α b+c
 PSEUDOMONOTONICITY OF AN AFFINE MAP 415
√
β b+c− ∆
det M > 0, a > 0, d > 0, b + c > 0, ∆ > 0, − ≤ (5.17)
α 2d
β a
det M < 0, a < 0, d = 0, b + c < 0, − ≤ (5.18)
α b+c
√
β b+c+ ∆
det M < 0, a < 0, d < 0, b + c < 0, ∆ > 0, − ≤ (5.19)
α 2d
(iii) α < 0, β > 0 and one of the following conditions holds:
α d
det M > 0, a = 0, d > 0, b + c > 0, − ≤ (5.20)
β b+c
√
α b+c− ∆
det M > 0, a > 0, d > 0, b + c > 0, ∆ > 0, − ≤ (5.21)
β 2a
α d
det M < 0, a = 0, d < 0, b + c < 0, − ≤ (5.22)
β b+c
√
Downloaded by [Australian National University] at 17:43 05 January 2015

α b+c+ ∆
det M < 0, a < 0, d < 0, b + c < 0, ∆ > 0, − ≤ (5.23)
β 2a

Furthermore, F ( x) is pseudomonotone (not merely) on int <2+ if and only if

(5.24) holds.

a ≥ 0, d ≥ 0, ∆ ≤ 0. (5.24)

Proof. From Theorem 3.2 the affine map F ( x) is pseudomonotone on

int <2+ if and only if v T Mv ≥ 0 for all v ∈ ZW ∗ .

• If α ≥ 0, β ≥ 0, from (i) of Theorem 5.1 we have ZW ∗ = ZW so that

the pseudomonotonicity of F ( x) reduces to the pseudomonotonicity
of the linear map G ( x) = Mx which is equivalent to (5.14), (5.15) (see
Corollary 4.1).
• If α < 0, β < 0, from (ii) of Theorem 5.1 we have ZW ∗ = <2 so that
T
F ( x) is pseudomonotone if and only if M+2M is positive semidefinite,
that is if and only if (5.24) holds.
• If α < 0, β = 0 or α = 0, β < 0, from (iii), (iv) of Theorem 5.1 ZW ∗ is
the union of two open halfspaces so that F ( x) is pseudomonotone if
T
and only if M+2M is positive semidefinite, that is if and only if (5.24)
holds.
• If α > 0, β < 0, taking into account (v) of Theorem 5.1, F ( x) is
pseudomonotone if and only if (5.25) holds:
y T M2−1 M( M2−1 )T y ≥ 0, for all y = ( y1 , y2 )T with y1 y2 < 0. (5.25)
 416 A. MARCHI AND L. MARTEIN

Consequently (see (4.6)), the pseudomonotonicity of F ( x) is equiva-

lent to the pseudomonotonicity of the linear map H ( x) = M2∗ x where
M2∗ = [ M2−1 M( M2−1 )T ]−1 = M2T M−1 ( M2 ).
" #
α a + βc c
Setting M2 = , by simple calculations we have
α b + βd d
" #
∗ aα 2 + (b + c)αβ + dβ2 α b + βd
M2 = .
α c + βd d
From Corollary 4.1, taking into account that det M2∗ = α 2 det M and
∆∗ = ((b + c)α + 2βd)2 − 4d( aα 2 + (b + c)αβ + dβ2 )
= α 2 [(b + c)2 − 4ad]
= α 2 ∆,
Downloaded by [Australian National University] at 17:43 05 January 2015

F ( x) is merely pseudomonotone if and only if (5.26) or (5.27) holds:

det M > 0, ∆ > 0, d ≥ 0, (b + c)α + 2βd ≥ 0,
aα 2 + (b + c)αβ + dβ2 ≥ 0 (5.26)
det M < 0, d ≤ 0, (b + c)α + 2βd ≤ 0,
aα 2 + (b + c)αβ + dβ2 ≤ 0. (5.27)
Consider the case det M > 0. Taking into account the necessary
pseudomonotonicity condition stated in Theorem 3.3, F ( x) is merely
pseudomonotone if and only if the conditions det M > 0, a ≥ 0,
d ≥ 0, b + c ≥ 0, ∆ > 0 are verified and furthermore the following
two inequalities hold:
β
2d + b + c ≥ 0, (5.28)
α
µ ¶2
β β
d + (b + c) + a ≥ 0 . (5.29)
α α
Let us note that the case a = 0 cannot occur. In fact if a = 0 and d = 0,
the conditions ∆ > 0 and b + c ≥ 0 imply b + c > 0 so that (5.29) is
not verified since β < 0; if a = 0 and d > 0, (5.29) is verified for αβ ≥ 0
β −(b+c)
(and this is a contradiction) or for α ≤ d while (5.28) is verified
β −(b+c) −(b+c)
for α ≥ 2d > d and this is a contradiction.
Consider the case a > 0 and d = 0; since ∆ > 0 we have b + c > 0
so that (5.28) is satisfied. Condition (5.29) is verified if and only if
− αβ ≤ b+a c , that is if and only if (5.16) holds.
 PSEUDOMONOTONICITY OF AN AFFINE MAP 417
√
−(b+c)− ∆
When a > 0 and d > 0, we have γ1 < γ2 < 0 where γ1 = 2d ,
√
−(b+c)+ ∆ β −(b+c)
γ2 = 2d . Condition (5.28) is verified if and only if α ≥ 2d
while (5.29) is verified if and only if αβ ≥ γ2 or αβ ≤ γ1 . Consequently
(5.28) and (5.29) hold if and only if αβ ≥ γ2 that is if and only if (5.17)
holds.
Consider now the case det M < 0. Let us note that (5.27) is equivalent
to (5.26) changing a, d, b + c with − a, −d, −(b + c), respectively. Ap-
plying these changes to (5.16) and (5.17) we obtain (5.18) and (5.19),
respectively.
• At last, consider the case α < 0, β > 0. Taking into account (vi) of The-
−1 −1 T
orem 5.1, F ( x) is pseudomonotone if and only if y T M" 1 M( M1 )# y ≥
α a + βc a
0, for all y = ( y1 , y2 )T with y1 y2 < 0, where M1 = .
Downloaded by [Australian National University] at 17:43 05 January 2015

α b + βd b

By simple calculations we have

" #
∗ T −1 aα 2 + (b + c)αβ + dβ2 α a + βc
M1 = M1 M M1 = .
α a + βb a

It follows that M1∗ can be obtained from M2∗ changing a, α with d, β,

respectively. Applying these changes in (ii) we obtain (iii).
The proof is complete. ¤
6. The singular case

In this section we will characterize

" # the pseudomonotonicity of the
a c
affine map F ( x) = Mx + q, M = , rank ( M) = 1, on the interior of
b d
the positive orthant.
Set x T = ( x1 , x2 ), q T = (q1 , q2 ) and denote by m(1) , m(2) the first
and the second column of M, respectively. Since rank ( M) = 1 at least
one column is not the null vector. Without loss of generality assume the
existence of α ∈ < such that m(2) = α m(1) .
Two cases arise: q is proportional to m(1) , that is there exists β ∈ <
such that q = βm(1) , or q is not proportional to m(1) , that is rank [q,m(1) ] = 2.
In the first case we have the following theorem:

Theorem 6.1. Consider the affine map F ( x) = Mx + q, q = βm(1) , β ∈ <.

Then F ( x) is pseudomonotone on int <2+ if and only if one of the following
 418 A. MARCHI AND L. MARTEIN

conditions holds:

(i) α ≥ 0, β ≥ 0;
(ii) a > 0, b = α a.

Proof. It results v T ( Mx + q) = ( x1 + α x2 + β)v T m(1) . If for every ( x1 , x2 )

∈ int <2+ it results x1 + α x2 + β 6= 0, then vT ( Mx + q) = 0 ⇔ vT m(1) = 0;
on the other hand v T m(1) = 0 implies v T Mv = 0 and thus F ( x) is
pseudomonotone. Let us note that x1 + α x2 + β 6= 0, for all ( x1 , x2 ) ∈
int <2+ if and only if α ≥ 0, β ≥ 0, so that condition i) is verified.
If there exists ( x1 , x2 ) ∈ int <2+ such that x1 + α x2 + β = 0 then
it results v T ( Mx + q) = 0, for all v ∈ <2 , so that v T Mv ≥ 0 if and
T
only if B = M+2M is positive semidefinite. Let us note that B is positive
semidefinite if and only if a ≥ 0, α b ≥ 0 and det B = −(α a − b)2 ≥ 0 that
Downloaded by [Australian National University] at 17:43 05 January 2015

is taking into account that m(1) 6= 0, if and only if a > 0, b = α a, so that

condition (ii) holds. ¤

Corollary 6.1. Consider the affine map F ( x) = Mx + q, q = βm(1) , β ∈ <.

Then:

(i) F ( x) is merely pseudomonotone if and only if

" # Ã !
a αa βa
M= , q= , α ≥ 0, β ≥ 0, b 6= α a (6.30)
b αb βb

(ii) F ( x) is pseudomonotone (not merely) if and only if

" #
1 α
M=a , a > 0, α ∈ <. (6.31)
α α2

Proof. We have v T ( Mx + q) = ( x1 + α x2 + β)v T m(1) .

If α < 0 it is possible to choose x∗2 > 0 large enough such that
x∗1 = −α x∗2 − β > 0; it follows that ( x∗1 + α x∗2 + β)v T m(1) = 0 is verified
for all v ∈ <2 . Consequently, in the case α < 0, F ( x) is pseudomonotone
T
on int <2+ if and only if the matrix M+2M is positive semidefinite (in such
a case it is easy to prove that it assumes the form (6.31)).
If α ≥ 0, β < 0 it is possible to choose x∗2 > 0 close to zero such
that x∗1 = −α x∗2 − β > 0; it follows that ( x∗1 + α x∗2 + β)v T m(1) = 0 is
verified for all v ∈ <2 . Consequently, also in the case α ≥ 0, β < 0, F ( x)
T
is pseudomonotone on int <2+ if and only if the matrix M+2M is positive
semidefinite.
 PSEUDOMONOTONICITY OF AN AFFINE MAP 419

In the case α ≥ 0, β ≥ 0, we have v T ( Mx + q) = ( x1 + α x2 +

β)v T m(1) = 0, x1 > 0, x2 > 0 if and only if v T m(1) = 0, so that v T Mv = 0
and F ( x) is pseudomonotone.
The proof is complete. ¤

Corollary 6.2. Consider the linear map F ( x) = Mx. Then:

(i) F ( x) is merely pseudomonotone if and only if

" #
a αa
M= , α ≥ 0, b 6= α a (6.32)
b αb
(ii) F ( x) is pseudomonotone (not merely) if and only if
" #
1 α
M=a , a > 0, α ∈ <. (6.33)
α α2
Downloaded by [Australian National University] at 17:43 05 January 2015

The following theorem characterizes the pseudomonotonicity of the

affine map F ( x) when rank [q, m(1) ] = 2.

Theorem 6.2. Consider the affine map F ( x) = Mx + q with rank[q, m(1) ] = 2.

Then:

(i) F ( x) is merely pseudomonotone if and only if one of the following conditions

holds:
" #
a αa
M= , α ≥ 0, bq1 − aq2 > 0, q2 ≤ α q1 , b < α a (6.34)
b αb
" #
a αa
M= , α ≥ 0, bq1 − aq2 < 0, q2 > α q1 , b > α a (6.35)
b αb
(ii) F ( x) is pseudomonotone (not merely) if and only if (6.33) holds.

Proof. Let CW ∗ be the cone generated by W ∗ = { Mx + q, x ∈ int <2+ }. It

is easy to prove that:

• if α < 0 then CW ∗ is the half-plane containing q and having r = { x =

tm(1) , t ∈ <} as supporting line;
• if α ≥ 0 then CW ∗ is the open convex cone generated by the vectors m(1)
and q.

In the first case, that is α < 0, ZW ∗ = ((m(1) )⊥ )c (see Theorem 2.1); when
α ≥ 0, we have ZW ∗ = {v = ( A−1 )T y, y ∈ / <2+ ∪ <2− } where A is the
( 1 )
matrix having m and q as columns (see Theorem 2.3).
 420 A. MARCHI AND L. MARTEIN

From Theorem 3.2, F ( x) is pseudomonotone if and only v T Mv ≥ 0,

for all v ∈ ZW ∗ . Consequently:

• if α < 0 we have vT Mv ≥ 0, for all v ∈ ((m(1) )⊥ )c , that is vT Mv ≥ 0,

T
for all v ∈ <2 , so that F ( x) is pseudomonotone if and only if M+2M is
positive semidefinite;
• if α ≥ 0, F ( x) is pseudomonotone if and only if
y T A−1 M( A−1 )T y ≥ 0 for all y = ( y1 , y2 )T , y1 y2 < 0 (6.36)
By simple calculations we find
" #
−1 −1 T 1 0 0
A M( A ) = (6.37)
bq1 − aq2 b − α a −q2 + α q1
so that (6.36) becomes
Downloaded by [Australian National University] at 17:43 05 January 2015

−q2 + α q1 2 (α a − b)
y − y1 y2 ≥ 0 for all y1 y2 < 0. (6.38)
bq1 − aq2 2 bq1 − aq2

Following the same lines given in the proof of Theorem 4.1, the thesis is
achieved. ¤

References

[1] M. Avriel, W. E. Diewert, S. Schaible and I. Zang, Generalized

Concavity, Plenum Publishing Corporation, New York, 1988.
[2] A. Cambini, L. Martein and S. Schaible, Pseudoconvexity, pseu-
domonotonicity and the generalized Charnes-Cooper transforma-
tion, Pacific Journal of Optimization, Vol. 1 (2), pp. 1–10, 2005.
[3] J. P. Crouzeix and J. A. Ferland, Criteria for differentiable generalized
monotone maps, Mathematical Programming, Vol. 75, pp. 399–406,
1996.
[4] J. P. Crouzeix and S. Schaible, Generalized monotone affine maps,
SIAM J. Matrix Anal. Appli., Vol. 17, pp. 992–997, 1996.
[5] J. P. Crouzeix, Characterizations of generalized convexity and mono-
tonicity – a survey, in Generalized Convexity, Generalized Monotonicity,
J. P. Crouzeix, J. E. Martinez-Legaz and M. Voile (editors), Kluwer
Academic Publisher, Dordrecht, pp. 237–256, 1998.
[6] J. P. Crouzeix, A. Hassouni, A. Lahlou and S. Schaible, Positive
subdefinite matrices, generalized monotonicity and linear comple-
mentarity problems, SIAM J. Matrix Anal. Appli., Vol. 22, pp. 66–85,
2000.
 PSEUDOMONOTONICITY OF AN AFFINE MAP 421

[7] M. S. Gowda, Affine pseudomonotone mapping and the linear

complementarity problem, SIAM J. Matrix Anal. Appl., Vol. 11,
pp. 373–380, 1990.
[8] N. Hadjisavvas and S. Schaible, Quasimonotonicity and pseu-
domonotonicity in variational inequalities and equilibrium prob-
lems, in Generalized Convexity, Generalized Monotonicity, J.P. Crouzeix,
J.E. Martinez-Legaz and M. Voile (editors), Kluwer Academic Pub-
lisher, Dordrecht, pp. 257–275, 1998.
[9] N. Hadjisavvas and S. Schaible , Generalized monotone maps, Hand-
book of Generalized Convexity and Generalized Monotonicity, pp. 388–
420, Springer, 2005.
[10] S. Karamardian, S. Schaible and J. P. Crouzeix, Characterizations
of generalized monotone maps, Journal of Optimization Theory and
Application, Vol. 76 (3), pp. 399–413, 1993.
Downloaded by [Australian National University] at 17:43 05 January 2015

[11] I. Konnov, Generalized monotone equilibrium problems and varia-

tional inequalities, Handbook of Generalized Convexity and Generalized
Monotonicity, pp. 559–618, Springer, 2005.
[12] R. Pini and S. Schaible, Invariance properties of generalized mono-
tonicity, Optimization, Vol. 28, pp. 212–222, 1994.
[13] J. Yao and O. Chadli, Pseudomonotone complementarity problems
and variational inequalities, Handbook of Generalized Convexity and
Generalized Monotonicity, pp. 501–558, Springer, 2005.

Received February, 2006; Revised July, 2006