Research in Mathematics

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/oama23

Solving equations with real Jordan canonical forms

Zi-Zong Yan & Zhi-Jia Shu

To cite this article: Zi-Zong Yan & Zhi-Jia Shu (2022) Solving equations with real Jordan canonical
forms, Research in Mathematics, 9:1, 2068740, DOI: 10.1080/27684830.2022.2068740

To link to this article: https://doi.org/10.1080/27684830.2022.2068740

© 2022 The Author(s). This open access

article is distributed under a Creative
Commons Attribution (CC-BY) 4.0 license.

Published online: 03 May 2022.

Submit your article to this journal

Article views: 304

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=oama23
RESEARCH IN MATHEMATICS
2022, VOL. 9, NO. 1, 2068740
https://doi.org/10.1080/27684830.2022.2068740

APPLIED & INTERDISCIPLINARY MATHEMATICS | RESEARCH ARTICLE

Solving equations with real Jordan canonical forms

Zi-Zong Yan and Zhi-Jia Shu
School of Information and Mathematics, Yangtze University, Jingzhou, Hubei, China

ABSTRACT ARTICLE HISTORY

A new equivalent version of Gordan’s theorem of the alternative is presented based on the system Received 27 August 2021
of linear inequalities. Such a version has a more intuitive geometric interpretation than Gordan’s Accepted 15 April 2022
theorem. The real Jordan canonical form theorem is utilized in proving the theorem. KEYWORDS
Gordan’s theorem; real
Jordan canonical form;
invariant subspace

1. Introduction investigation. Such an investigation do not make clear

n m�n
Let R denote the set of vectors with size n, and R
denote the collection of m-by-n real matrices. The fol­
lowing theorem examines circumstances under which
a system of linear equations has a positive solution.
Theorem 1 (Gordan’s theorem). Let A 2 R m�n be arbi­
trary. Then, either the system
Ax ¼ 0 and x�0
has a nonzero solution x 2 R n , or the system
AT y > 0
has a solution y 2 R m , but never both.
Theorem 1, due to Gordan (Gordan, 1873), has a long
history and has been reproved numerous times. The
survey by Kjeldsen (Kjeldsen, 2002) attributed the histor­
ical development of the theory of the systems of linear
inequalities (1.1) and (1.2). It was rediscovered by
Stiemke (Stiemke, 1915), representing a large class of
theorems of the alternative that play an important role
in linear and nonlinear programming. Such theorems are
crucial in deriving optimality conditions for wide classes
of extremal problems. For more background information
and applications on the theorems of the alternative and
other relevant matters, we refer to the textbooks (Ciarlet,
1989; Gill et al., 1991; Mccormick, 1983; Osborne, 1985),
the surveys (Dax, 1993; Saunders & Schneider, 1979), and
research papers (Dax & Sreedharan, 1997; Galán, 2017;
Giannessi, 1984; Mangasarian, 1981).
Gordan (Gordan, 1873) and Stiemke (Stiemke, 1915)
proved Theorem 1 by induction independently.
However, they did not give any motivation for this
why the theorem works; see e.g. (Gale, 1960). Recent
proofs are usually based on separation theorems of the
convex sets with a simple geometrical interpretation. In
this note, we present a simple self-contained proof which
is based on elementary arguments in linear algebra.
The paper is organized as follows. In the next section,
we will present a new equivalent form of Theorem 1.
And in Section 3, we will give a simple, elementary proof
(1:1) of the equivalent form by the use of the real Jordan
canonical form. The last section contains conclusions
and some remarks.
(1:2)
2. An equivalent form
Let r denote the rank of the matrix A. Clearly, one has
r � minfm; ng:
For the sake of avoiding trivial cases, we always assume
that r is greater than zero. By the theory of the linear
algebra (Wilkinson & Reinsch, 1971), there is an inver­
tible matrix M 2 R m�m and a permutation matrix P 2
R n�n such that
� �
I B
MA ¼ P; (2:1)
0 0
where B is a r-by-ðn rÞ real matrix, and I denotes
a unit matrix with a suitable size. The left product is
equivalent to performing a series of elementary row
operations on the matrix A. Then the system of equa­
tions in (1.1) can be rewritten as follows:
� �
I B
Px ¼ 0: (2:2)
0 0

CONTACT Zhi-jia Shu zzyan@yangtzeu.edu.cn;zjshu01@163.com School of Information and Mathematics, Yangtze University, Jingzhou, Hubei, China
© 2022 The Author(s). This open access article is distributed under a Creative Commons Attribution (CC-BY) 4.0 license.
You are free to: Share — copy and redistribute the material in any medium or format. Adapt — remix, transform, and build upon the material for any purpose, even commercially.
The licensor cannot revoke these freedoms as long as you follow the license terms. Under the following terms: Attribution — You must give appropriate credit, provide a link to the
license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. No additional
restrictions You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.
2 Z.-Z. YAN AND Z.-J. SHU
� �
u
If ¼ Px, where u 2 R r and v 2 R n r , then x � 0
v
means that both u � 0 and v � 0 hold. This results in
our main theorem, an equivalent algebraic version of
Gordan’s theorem (Theorem 1).
Theorem 2. (The Main Theorem). Let B 2 R r�n1 be
arbitrary. Then, either the system
Bv � 0 and v�0 (2:3)
n1
has a nonzero solution v 2 R , or the system
BT y > 0 and y>0 (2:4)
r
has a solution y 2 R , but never both.
From this it is simple to prove Gordan’s theorem
(Theorem 1). Our main theorem has a more intuitive
geometric interpretation than Gordan’s theorem in the
sense discussed below but despite this is perhaps simpler
to prove.
In the following statement, we assume without loss of
generality that r ¼ n1 , i.e. the matrix B is square (if not,
make it a square matrix by adding zero rows or columns).
Let R rþþ and R rþ denote the positive and the non­
negative quadrant cones, respectively, that is,
R rþþ ¼ fv 2 R r jv > 0g and
R rþ ¼ fv 2 R r jv � 0g:
If B is a linear operator represented by the matrix B,
then the solvability of the system (2.3) is equivalent to
The basic theorem is that given any r � r matrix with
complex entries, there exists a nonsingular matrix S
such that
1
SBS ¼ J; SB ¼ JS; (3:2)
where J, the Jordan canonical form of B, is block diagonal,
each diagonal matrix being an elementary Jordan block
with λ ¼ λi . Apart from the ordering of the blocks along
the diagonal of J (which can be arbitrary), the Jordan
canonical form is unique, although S is far from unique.
Following we mainly consider the Jordan canonical
form when the matrix B has only real entries. In this
case, all the nonreal eigenvalues must occur in conjugate
pairs. Moreover, if � þ iη is an eigenvector of the real
matrix B associated with the complex eigenvalue
λ ¼ a þ bi, then � iη must be an eigenvector of B
associated with the conjugate eigenvalue �λ ¼ a bi,
where a; b 2 R, �; η 2 R r and i is the imaginary unit,
i.e. i2 ¼ 1. That is
Bð� þ iηÞ ¼ ða þ biÞð� þ iηÞ and
Bð� iηÞ ¼ ða biÞð� iηÞ;
or equivalently,
� �
a b
Bð�; ηÞ ¼ ð�; ηÞ ;
b a
in which the right 2 � 2 real matrix, denoted by Cða; bÞ,
b
has eigenvalues a � bi. If θ ¼ arcsin a2 þb 2 , then

that � �
2 2 cos θ sin θ
Cða; bÞ ¼ ða þ b Þ :
BðR rþ Þ \ ð R rþ Þ�;; (2:5) sin θ cos θ

while the solvability of the system (2.4) is equivalent to So the linear operator represented by the matrix Cða; bÞ on
that the real space spanf�; ηg is a superposition of rotation and
stretching. Of course, the rotation angle of the correspond­
B� ðR rþþ Þ \ R rþþ �;; (2:6)
ing adjoint operator, represented by the matrix CT ða; bÞ, is
where B� is the adjoint operator of B, and it can be the opposite of the rotational angle of Cða; bÞ. When θ�0
represented by the matrix BT when B is a real matrix. or π, such two linear operators have no eigenvector on the
2-dimensional real space spanf�; ηg.
Note that the structure of the Jordan blocks corre­
3. Proofs
sponding to the conjugate eigenvalue �λ must be the
The Jordan canonical form is described with reference same as the structure of the Jordan blocks corresponding
to matrices known as elementary Jordan blocks. An to the eigenvalue λ, e.g. see, (Horn & Johnson, 1990). For
elementary Jordan block of size l � l associated with example, if λ is a complex eigenvalue of the real matrix B,
an eigenvalue λ will be denoted by Jl ðλÞ, and its general and if J2 ðλÞ appears in the Jordan canonical form of B
form is adequately illustrated by the definition with a certain multiplicity, J2 ð�λÞ must also appear with
0 1 the same multiplicity. Then, the block matrix
λ 1
0 1
B λ 1 C λ 1 0 0
B .. .. C � �
B
Jl ðλÞ ¼ B . . C: (3:1) J2 ðλÞ 0 B0 λ 0 0C
@
C
A � ¼B @ 0 0 �λ 1 A
C
λ 1 0 J2 ðλÞ
λ 0 0 0 �λ
 RMS: RESEARCH IN MATHEMATICS & STATISTICS 3

is permutation similar to the block matrix (1) If ζ j is an eigenvector of BT associated with the
0 1 Jordan block JrTj ðλj Þ, then
λ 0 1 0
B 0 �λ 0 1 C Wj ðBÞ ¼ spanfζ j ; Bζ j ; B2 ζ j ; � � � ; Brj 1 ζ j g
B C
@ 0 0 λ 0 A:
0 0 0 �λ is an invariant subspace of B, in which B has one
linearly independent real eigenvector on Wj ðBÞ;
If λ ¼ a þ bi and �λ ¼ a bi, then it is also similar to the (2) If �j � iηj are eigenvectors of BT on the complex
matrix space C r associated with the real Jordan block CrTj ðaj ; bj Þ,
� �
Cða; bÞ I where �j ; ηj 2 R r , then
:
0 Cða; bÞ
Wj ðBÞ ¼ spanf�j ; ηj ; B�j ; Bηj ; � � � ; Brj 1 �j ; Brj 1 ηj g
In general, each block pair of conjugate l-by-l Jordan
block is an invariant subspace of B on R r ;
� � (3) R r is a direct sum of the subspaces
Jl ðλÞ 0 Wj ðBÞ; j ¼ 1; 2; � � � ; k, i.e.
0 Jl ð�λÞ
R r ¼ W1 ðBÞ � W2 ðBÞ � � � � � Wk ðBÞ: (3:5)
with nonreal λ ¼ a þ bi is similar to a real 2l-by-2l
block of the form Proof. If rj ¼ 1, then the result is trivial. Otherwise, each
0 1 Jordan block Jrj ðλj Þ and its transpose JrTj ð�λj Þ have only one
Cða; bÞ I
B . C linearly independent real eigenvector erj and e1 , respec­
B
B Cða; bÞ . . C
C: tively, where erj and e1 denote the last column and the
Cl ða; bÞ ¼ B . C
@ . . I A first column of the unit matrix with order rj . Then, it is
Cða; bÞ easy to verify that the above results for the real Jordan
(3:3) canonical form JR hold. Then, the desired results hold
since the similarity invariance of the subspaces Wj ðJÞ.□
These observations lead us to the real Jordan canonical If the real matrix B in Corollary 4 is replaced by its
form, see, (Horn & Johnson, 1990). transpose, then the corresponding result also is true. We
r�r
may obtain another decomposition of R r in the form of
Theorem 3. Each real matrix B 2 R is similar to
the matrix BT
a block diagonal real matrix of the form
R r ¼ W1 ðBT Þ � W2 ðBT Þ � � � � � Wk ðBT Þ: (3:6)
JR ¼ diagðCr1 ða1 ; b1 Þ; n; Crp ðarp ; brp Þ; Jrpþ1 ðλpþ1 Þ; � � � ; Jrk
ðλk ÞÞ; Of course, we have Wj ðBT Þ ¼ Wj ðBÞ for
all j ¼ 1; 2; � � � ; k.
(3:4)
where λj ¼ aj þ bj i is a nonreal eigenvalue of B for It follows from Corollary 4 that if gj and Vλj ðBT Þ
j ¼ 1; 2; � � � ; p, aj ; bj are real, λpþ1 ; � � � ; λk are real eigen­ denote the geometric multiplicity and the eigensubspace
values of B. The real Jordan block Crj ðaj ; bj Þ is of the form of λj associated with the matrix BT , then the null space
(3.3), and the Jordan block Jrj ðλj Þ are exactly the Jordan of ðB λIÞgj is equal to
blocks in (3.2) with λ ¼ λj . spanfVλTj ; BVλj ðBT Þ; � � � ; Bgj 1 Vλj ðBT Þg:

The motivation to prove our main theorem drives us
to investigate the connection between the real Jordan
canonical forms of the real matrix B and its transpose.
A simple observation is that if the real matrix B has a real
Jordan canonical form JR , then its transpose has a real
Jordan canonical form JRT . This yields the following result.
Corollary 4. Let the real Jordan canonical form JR of the
real matrix B be defined by (3.4).
In addition, the first and last terms of Corollary 4 is still
true when B is a complex matrix, only if the transpose
matrix BT is replaced by the conjugate transpose matrix
B� , and the corresponding Jordan block JrTj ðλj Þ is
replaced by J T ð�λj Þ.
rj
Let
Vjþ ðBÞ ¼ spanf�j þ iηj ; Bð�j þ iηj Þ; � � � ; Brj ð�j þ iηj Þg
and
4 Z.-Z. YAN AND Z.-J. SHU
Vj ðBÞ ¼ spanf�j iηj ; Bð�j iηj Þ; � � � ; Brj ð�j
be invariant subspaces of B, where �j � iηj are eigenva­
lues of BT associated with the eigenvalues aj � bj i. Then
�j � iηj 2 Vjþ ðBÞ � Vj ðBÞ such that B on one 2-dimen­
sional real subspace V ¼ spanf�; ηg � Wj ðBÞ rotates
j b
the angles θj ¼ arcsin a2 þb 2 , while the adjoint operator
j j
�
B� on the subspace Vj � Wj BT rotates the
angles θj .
Now we ready to Theorem 2.
Proof of the main theorem. It is easy to show that both
two inequalities (2.5) and (2.6) cannot be true together.
From Corollary 4, we only need prove that at least one
of them holds for each Wj ¼ Wj ðBÞ ¼ Wj ðBT Þ,
j ¼ 1; 2; � � � ; k. The proof requires consideration of two
cases.
When the eigenvalue λj is a real number, the inequal­
ity (2.6) holds on Wj if and only if λj > 0, while the
inequality (2.5) holds on Wj if and only if λj � 0.
When the eigenvalue λj ¼ aj þ bj i is a complex num­
ber, the inequality (2.6) holds on Wj if and only if
b in reading at Yangtze university. His research interest is opti­
00 < arcsin a2 þb
j 0
2 < 90 , while the inequality (2.5) holds
j j
b
on Wj if and only if 900 � arcsin a2 þb
j 0
2 � 180 . The proof
j j
is finished.
Finally, Gordan’s theorem follows from our main
theorem trivially.
4. Concluding remarks
In this note, we have given an alternative proof of Gordan’s
theorem, a proof that is based on the real Jordan canonical
form. Another purely algebraic ways are to apply the
property of the orthogonal or antisymmetric matrix. See,
for example, (Broyden, 1998) and (Tucker, 1956; Vajda,
1961). This approach still don’t make clear why the theo­
rem works. In our proof, a new geometrical interpretation
for a pair of real linear adjoint operators is presented.
Other theorems of the alternative seem to be also
related to the above geometrical interpretation. This
will be our further work.
Acknowledgements
This study was funded by the National Natural Science
Foundation of China (11871118). The author has received
research grants from Yangtze University. The author declares
that he has no conflict of interest. The author is indebted to
the anonymous referees, whose detailed remarks helped in
improving the manuscript.
iηj Þg Disclosure statement
No potential conflict of interest was reported by the author(s).
Funding
Supported by the National Natural Science Foundation of
China (11871118).
Notes on contributors
Zi-Zong Yan is currently
a professor in the school of infor­
mation and mathematics at
Yangtze university. He received
his Ph.D. degree in the school of
mathematics and statistics from
Wuhan university in 2005. His
current research interests are in
the area of optimization theory
and matrix analysis.
Zhi-Jia Shu was born in Hubei,
China, in 1995. He received his
Bachelor’s degree in the school of information and mathe­
matics at Yangtze university, in 2019. Currently, he is a master
mization theory.
PUBLIC INTEREST STATEMENT
This note examines whether a system of linear inequalities is
solvable or a system of linear equations has a positive solution.
In this note, we also investigated the geometrical properties of
real linear operators by the use of its Jordan canonical form.
References
Broyden, C. G. (1998). A simple algebraic proof of Farkas’s
lemma and related theorems. Optimization Methods and
Software, 8(3–4), 185–199. https://doi.org/10.1080/
10556789808805676
Ciarlet, P. G. (1989). Introduction to numerical linear algebra
and optimization. Cambridge University Press.
Dax, A. (1993). The relationship between theorems of the
alternative, least norm problems, steepest descent direc­
tions, and degeneracy: A review. Annals of Operations
Research, 46(1), 11–60 https://doi.org/10.1007/
BF02096256
Dax, A., & Sreedharan, P. (1997). Theorems of the alternative
and duality. Journal of Optimization Theory and
Applications, 94(3), 561–590. https://doi.org/10.1023/
A:1022644832111
Galán, M. R. (2017). A theorem of the alternative with an
arbitrary number of inequalities and quadratic­
programming. Journal of Global Optimization, 69(2),
427–442. https://doi.org/10.1007/s10898-017-0525-x
Gale, D. (1960). The theory of linear economic models.
McGraw-Hill.

Giannessi, F. (1984). Theorems of the alternative and optim­
ality conditions. Journal of Optimization Theory and
Applications, 42(3), 331–365. https://doi.org/10.1007/
BF00935321
Gill, P. E., Murry, W., & Wright, M. H. (1991). Numerical
linear algebra and optimization. Vol. 1 Addison-Wesley.
Gordan, P. (1873). Über die auflösung linearer gleichungen
mit reelen coefficienten. Mathematische Annalen, 6(1),
23–28. https://doi.org/10.1007/BF01442864
Horn, R. A., & Johnson, C. R. (1990). Matrix analysis.
Cambridge University Press.
Kjeldsen, T. H. (2002). Different motivations and goals in the
historical development of the theory of systems of linear
inequalities. Archive for History of Exact Sciences, 56(6),
469–538. https://doi.org/10.1007/s004070200057
Mangasarian, O. L. (1981). A stable theorem of the alternative:
An extension of the Gordan theorem. Linear Algebra and
Its Applications, 41, 209–223. https://doi.org/10.1016/0024-
3795(81)90100-2
RMS: RESEARCH IN MATHEMATICS & STATISTICS 5
Mccormick, G. P. (1983). Nonlinear programming. John
Wiley.
Osborne, M. R. (1985). Finite algorithms in optimization and
data analysis. John Wiley & Sons.
Saunders, B. D., & Schneider, H. (1979). Applications of the
Gordan-Stiemke theorem in combinatorial matrix theory.
SIAM Review, 21(4), 528–541. https://doi.org/10.1137/
1021094
Stiemke, E. (1915). Über positive Lösungen homogener line­
arer Gleichungen. Mathematische Annalen, 76(2–3),
340–342. https://doi.org/10.1007/BF01458147
Tucker, A. W. (1956). Dual systems of homogeneous linear
relations. In H. W. Kuhn & A. W. Tucker (Eds.), Linear
Inequalities and Related Systems (pp. 3–18). Ann Math
Stud. Princeton Univ. Press.
Vajda, S. Mathematical programming, Addison Wesley: 1961.
Wilkinson, J. H., & Reinsch, C. (1971). Linear algebra,
handbook for automatic computation (Vol. 2). Springer-
Verlag.