mathematics

Article
Three New Proofs of the Theorem
rank f ( M ) + rank g ( M ) = rank( f , g )( M ) + rank[ f , g ]( M )
Vasile Pop 1,† and Alexandru Negrescu 2, *,†

1 Department of Mathematics, Technical University of Cluj-Napoca, 400114 Cluj-Napoca, Romania;

[Link]@[Link]
2 Department of Mathematical Methods and Models, National University of Science and Technology
Politehnica Bucharest, 060042 Bucharest, Romania
* Correspondence: [Link]@[Link]
† These authors contributed equally to this work.

Abstract: It is well known that in C[ X ], the product of two polynomials is equal to the product of their
greatest common divisor and their least common multiple. In a recent paper, we proved a similar
relation between the ranks of matrix polynomials. More precisely, the sum of the ranks of two matrix
polynomials is equal to the sum of the rank of the greatest common divisor of the polynomials applied
to the respective matrix and the rank of the least common multiple of the polynomials applied to the
respective matrix. In this paper, we present three new proofs for this result. In addition to these, we
present two more applications.

Keywords: rank; Jordan canonical form; matrix polynomials; Sylvester inequality; Frobenius
inequality

MSC: 15A03; 15A20; 15A24

1. Introduction
Citation: Pop, V.; Negrescu, A. Three In an euclidean ring ( A, +, ·), for any elements a, b ∈ A, there exists the greatest com-
New Proofs of the Theorem mon divisor d = ( a, b) ∈ A and there exists the least common multiple m = [ a, b] ∈ A. For
rank f ( M ) + rank g( M) = the rings (Z, +, ·) and (C[ X ], +, ·), the relation a · b = d · m is well known (see [1] [p. 150]).
rank( f , g)( M) + rank[ f , g]( M). By C[ X ], we denote the set of all polynomials in the indeterminate X with complex co-
Mathematics 2024, 12, 360. https:// efficients. In a recent paper (see [2]), we proved a similar relation between the ranks of
[Link]/10.3390/math12030360 matrix polynomials.
Academic Editors: Luca Gemignani
and Xiangmin Jiao
Theorem 1. For any two polynomials f , g ∈ C[ X ] and for any matrix M ∈ Mn (C), the following
relation holds:
Received: 10 November 2023
Revised: 7 January 2024 rank f ( M) + rank g( M ) = rank d( M ) + rank m( M ), (1)
Accepted: 15 January 2024
Published: 23 January 2024 where d := ( f , g) denotes the greatest common divisor and m := [ f , g] denotes the lowest common
multiple of the polynomials f , g.

In [2], the proof is presented using the method of elementary transformations in

Copyright: © 2024 by the authors. block-partitioned matrices. In this paper, we present three more proofs of the previous
Licensee MDPI, Basel, Switzerland. theorem. The first proof uses the reduction of a matrix to Jordan’s canonical form. The
This article is an open access article following two proofs reduce the equality to be proved to the study of the case of equality
distributed under the terms and from Frobenius inequality.
conditions of the Creative Commons The rank of a matrix is a fundamental concept in numerical linear algebra and en-
Attribution (CC BY) license (https:// gineering. For example, in control theory, the rank of a matrix can be used to determine
[Link]/licenses/by/
whether a linear system is controllable or observable.
4.0/).

Mathematics 2024, 12, 360. [Link] [Link]

Mathematics 2024, 12, 360 2 of 6

2. Main Results
First Proof of Theorem 1 (Vasile Pop). For the first proof, we use the reduction of the
matrices which appear in Theorem 1 to Jordan’s canonical form. More precisely, we show
how the relation of this theorem can be reduced to the case in which M is a Jordan cell:

1 0 ... 0 0
 
λ

 0 λ 1 ... 0 0 

 0 0 λ ... 0 0 
Jλ =  .
 
.. .. .. .. .. ..

 . . . . . .

 0 0 0 ... λ 1 
0 0 0 ... 0 λ

For this, we recall some basic facts (see, for example, [3], pp. 164–169; [4], pp. 95–100).

Theorem 2. Let M ∈ Mn (C) be a matrix and f ∈ C[ X ] a polynomial. Then,

(i) There is an invertible matrix P ∈ GLn (C) = { A ∈ Mn (C) | det A ̸= 0} such that the matrix

J M = P −1 · M · P

has a diagonal-block form

J M = Jλ1 ⊕ Jλ2 ⊕ · · · ⊕ Jλk ,
where λ1 , λ2 , . . ., λk are the distinct eigenvalues of M and Jλ1 , Jλ2 , . . ., Jλk are the corre-
sponding Jordan cells;
(ii) rank M = rank J M = rank Jλ1 + rank Jλ2 + . . . + rank Jλk ;
(iii) f ( M) = P · f ( J M ) · P−1 ;
(iv) f ( J M ) = f ( Jλ1 ) ⊕ f ( Jλ2 ) ⊕ · · · ⊕ f ( Jλk );
(v) for any Jordan cell Jλ ∈ Mm (C), we have

f ′ (λ) f ′′ (λ) f ( m −2) ( λ ) f ( m −1) ( λ )

 
f (λ) 1! 2! ... ( m −2) ! ( m −1) ! 

f ′ (λ) f ( m −3) ( λ ) f ( m −2) ( λ ) 
···

 0 f (λ) 1! ( m −3) ! ( m −2) ! 


 f ( m −4) ( λ ) f ( m −3) ( λ ) 
f ( Jλ ) = 
 0 0 f (λ) ··· ( m −4) ! ( m −3) !  ;

 .. .. .. .. .. .. 

 . . . . . . 

 f ′ (λ) 
 0 0 0 ··· f (λ) 1!

0 0 0 ··· 0 f (λ)

(vi) for any Jordan cell Jλ ∈ Mm (C), we have

rank f ( Jλ ) = 0

if λ is a root of f with multiplicity at least m (i.e., f (λ) = f ′ (λ) = . . . = f (m−1) (λ) = 0).
For the other cases, we have
rank f ( Jλ ) = m − σ f (λ),
where σ f (λ) is the order of multiplicity of the λ for the polynomial f (i.e., f (λ) = f ′ (λ) =
. . . = f (i) (λ) = 0, but f (i+1) (λ) ̸= 0, with i + 1 = σ f (λ)).

Using Theorem 2 (i)–(iv), proving equality (1) is reduced to proving the follow-
ing relation:

rank f ( Jλ ) + rank g( Jλ ) = rank( f , g)( Jλ ) + rank[ f , g]( Jλ ), (2)

for any Jλ Jordan cell of M.

Mathematics 2024, 12, 360 3 of 6

Let λ be an eigenvalue of the matrix M and Jλ the corresponding Jordan cell. With the
above notations,
( f , g) = d and [ f , g] = m,
we have
σd (λ) = min(σ f (λ), σg (λ)) = σ f (λ) ∧ σg (λ)
and
σm (λ) = max(σ f (λ), σg (λ)) = σ f (λ) ∨ σg (λ).
By Theorem 2 (vi), to prove equality (2), we discuss the following three cases.
(a) If σ f (λ) ≥ k and σg (λ) ≥ k, then σd (λ) ≥ k and σm (λ) ≥ k. Then, relation (2)
becomes 0 + 0 = 0 + 0, which, evidently, is true.
(b) If σ f (λ) ≥ k and σg (λ) = p < k, then σd (λ) = p < k and σm (λ) ≥ k, and relation
(2) becomes p + 0 = p + 0, which, evidently, is true.
(c) If σ f (λ) = p < k and σg (λ) = q < k, then σd (λ) = p ∧ q and σm (λ) = p ∨ q, and
relation (2) becomes p + q = p ∧ q + p ∨ q, which, evidently, is true.

Second Proof of Theorem 1 (Alexandru Negrescu). Consider f 1 , g1 ∈ C[ X ], two polyno-

mials, such that
f = d · f 1 , g = d · g1 , and ( f 1 , g1 ) = 1.
As a consequence, we have
m = d · f 1 · g1 .
We also consider the matrices A, B, C ∈ Mn (C) such that

A = f 1 ( M ), B = d ( M ), and C = g1 ( M ) .

We deduce that
f ( M ) = ( f 1 · d)( M ) = f 1 ( M ) · d( M) = AB,
g( M) = (d · g1 )( M ) = d( M ) · g1 ( M ) = BC,
and
m( M) = ( f 1 · d · g1 )( M) = f 1 ( M) · d( M ) · g1 ( M) = ABC.
Therefore, the equality to be proved is equivalent to

rank( AB) + rank( BC ) = rank( B) + rank( ABC ), (3)

which we recognize as the equality case of Frobenius inequality.

In 2002, Tian and Styan (see [5] (Theorem 1)) proved the following result.

Theorem 3. Let A ∈ Mm,n (C), B ∈ Mn,k (C), and C ∈ Mk,l (C) be given. Then, the following
three statements are equivalent:
 
O AB
(a) rank = rank( AB) + rank( BC );
BC B
(b) rank( ABC ) = rank( AB) + rank( BC ) − rank( B);
(c) there are X and Y such that BCX + YAB = B.

Thus, the equality to be demonstrated is reduced to showing that there are two
matrices, X, Y ∈ Mn (C), such that

g ( M ) · X + Y · f ( M ) = d ( M ),

which is an immediate consequence of the following theorem (see [1] (Theorem 3.31,
p. 135)).
Mathematics 2024, 12, 360 4 of 6

Theorem 4. If K is a field, f , g ∈ K[ X ], and d is their greatest common divisor, then there are
φ1 , φ2 ∈ K[ X ] such that
g · φ1 + φ2 · f = d.

Indeed, we can choose X = φ1 ( M ) and Y = φ2 ( M ), whereby the proof is finished.

Third Proof of Theorem 1 (Mihai Opincariu). As put forward in the previous proof, the
equality to be demonstrated is equivalent to

rank( AB) + rank( BC ) = rank( B) + rank( ABC ),

where the notations are those of the second proof (A = f 1 ( M ), B = d( M), C = g1 ( M)).
To prove relation (3), we use Sylvester’s Theorem (see [6] (Theorem 2.6)), which
presents the equality case in Sylvester’s inequality.

Theorem 5. For any matrices A, B ∈ Mn (C), the following relation holds:

rank B − rank( AB) = dim(Ker A ∩ Im B). (4)

If, in relation (4), we replace B with BC, we obtain

rank( BC ) − rank( ABC ) = dim(Ker A ∩ Im( BC )). (5)

Subtracting relation (5) from (4), we obtain

rank B + rank( ABC ) = rank( AB) + rank( BC ) + dim(Ker A ∩ Im B) − dim(Ker A ∩ Im( BC )). (6)
Since Im( BC ) ⊂ Im B, we obtain the well-known Frobenius inequality

rank( AB) + rank( BC ) ≤ rank B + rank( ABC ), (7)

and by (6), it follows that equality case (3) is true if and only if

Ker A ∩ Im B = Ker A ∩ Im( BC ).

So, taking into account the notations at the beginning of this proof, it remains to show
the relation

Ker( f 1 ( M)) ∩ Im(d( M)) = Ker( f 1 ( M)) ∩ Im(d( M ) g1 ( M )).

It is enough to proof the inclusion “⊂”, since the other is obvious.

We have:

x ∈ Ker( f 1 ( M )) ∩ Im(d( M )) iff f 1 ( M ) · x = 0 and x = d( M) · y.

According to Theorem 4, since ( f 1 , g1 ) = 1, there exist two polynomials, f 2 , g2 ∈ C[ X ],

such that
f 2 · f 1 + g2 · g1 = 1,
hence
f 2 ( M ) · f 1 ( M) + g2 ( M ) · g1 ( M ) = In .
By multiplying in the last relation by x and taking into account that x ∈ Ker( f 1 ( M )),
we obtain
g2 ( M) · g1 ( M) · x = x.
Since x = d( M ) · y, we have

g2 ( M ) · g1 ( M ) · d ( M ) · y = x
Mathematics 2024, 12, 360 5 of 6

equivalent to
d( M ) · g1 ( M )( g2 ( M) · y) = x,
or
x = d( M) · g1 ( M ) · z, where z = g2 ( M) · y,
that is,
x ∈ Im(d( M) · g1 ( M ))
and the proof is finished.

Besides the applications highlighted in [2], Section 3, we present two others. The first
of these is closely related to a problem given in the 2004 edition of the Undergraduate
Mathematics Competition at the Mechanics and Mathematics Faculty of Taras Shevchenko
National University of Kyiv, Ukraine (see [7], p. 27).

Proposition 1. Let A, B, C, and D be n × n real matrices such that

A T = BCD, B T = CDA, C T = DAB, and D T = ABC.

We denote S = ABCD. Then,

rank S + rank(S − In ) + rank(S + In ) = 2n.

Proof. First, we show that S is a tripotent matrix. Indeed, we have

S3 = ( ABCD )( ABCD )( ABCD ) = ( ABC )( DAB)(CDA)( BCD )

= D T C T B T ( BCD ) = ( BCD ) T ( BCD ) = ( A T ) T ( BCD ) = ABCD = S.

If we consider in Theorem 1 the polynomials f
= x and g = x2 − 1,
for which d = 1

and m = x3 − x, we obtain that rank S + rank S2 − In = n + rank S3 − S ; thereby,
taking
into account that the matrix S is tripotent, we deduce that rank S + rank S2 − In = n.
If we consider in Theorem 1 the polynomials f = x − 1 and g = x + 1, for which
d = 1 and m
= x2 − 1, we obtain that rank(S − In ) + rank(S + In ) = n + rank S2 − In , so

rank S2 − In = rank(S − In ) + rank(S + In ) − n and the conclusion follows.

The following result was proposed by Bogdan Sebacher in the 2019 edition of the
Traian Lalescu National Mathematics Contest for University Students, Romania (see [8]).

Proposition 2. Let a ∈ C, with a ̸= 0, n ≥ 2, and A ∈ Mn (C). Then,

 
rank aA − A2 = rank A + rank( aIn − A) − n.

Proof. If we consider in Theorem 1 the polynomials f = x and g = a − x, for which
d = 1

and m = ax − x2 , we obtain that rank A + rank( aIn − A) = rank In + rank aA − A2 , and
the conclusion follows.

In addition to those presented in [2], we offer two more consequences of Theorem 1,

which can be easily shown by the method of mathematical induction.
 
Corollary 1. Let f 1 , f 2 , . . . , f m ∈ K[ X ] be polynomials and A ∈ Mn (K). Let f i1 , f i2 , . . . , f i p
denote the greatest common divisor of polynomials f i1 , f i2 , . . . , f i p (the maximum-degree monic
h i
polynomial that divides all polynomials f i1 , f i2 , . . . , f i p ) and f i1 , f i2 , . . . , f i p denote the lowest
common multiple of the polynomials f i1 , f i2 , . . . , f i p (the minimum-degree monic polynomial that is
a multiple of all polynomials f i1 , f i2 , . . . , f i p ). Then,
Mathematics 2024, 12, 360 6 of 6

m
∑ ∑



rank[ f 1 , f 2 , . . . , f m ]( A) = rank f i1 ( A) − rank f i1 , f i2 ( A) +
i1 =1 i1 < i2

∑


+ rank f i1 , f i2 , f i3 ( A) + . . . +
i1 < i2 < i3

+(−1)m−1 rank( f 1 , f 2 , . . . , f m )( A)

and
m
∑ ∑
   
rank( f 1 , f 2 , . . . , f m )( A) = rank f i1 ( A) − rank f i1 , f i2 ( A) +
i1 =1 i1 < i2

∑
 
+ rank f i1 , f i2 , f i3 ( A) + . . . +
i1 < i2 < i3

+(−1)m−1 rank[ f 1 , f 2 , . . . , f m ]( A).

3. Conclusions
This paper provides three new proofs that the sum of the ranks of two matrix polyno-
mials is equal to the sum of the rank of the greatest common divisor of the polynomials
applied to the respective matrix and the rank of the least common multiple of the poly-
nomials applied to the respective matrix. In addition to these, two more applications
are provided.

Author Contributions: Conceptualization, V.P. and A.N.; methodology, V.P. and A.N.; formal analysis,
V.P. and A.N.; investigation, V.P. and A.N.; writing—original draft preparation, V.P. and A.N.; writing—
review and editing, V.P. and A.N.; supervision, V.P. and A.N.; funding acquisition, V.P. All authors
have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: Data sharing is not applicable to this article.
Acknowledgments: We thank Mihai Opincariu, “Avram Iancu” National College, Brad, Hunedoara,
Romania, for providing us with the third proof.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Rotman, J.J. Advanced Modern Algebra; Prentice Hall: Saddle River, NJ, USA, 2003.
2. Pop, V. Relations between ranks of matrix polynomials. J. Algebr. Number Theory Adv. Appl. 2021, 24, 35–41. [CrossRef]
3. Horn, R.A.; Johnson, C.R. Matrix Analysis, 2nd ed.; Cambridge University Press: New York, NY, USA, 2013.
4. Gantmacher, F.R. The Theory of Matrices; Chelsea Publishing Company: New York, NY, USA, 1960; Volume 1.
5. Tian, Y.; Styan, G.P.H. When does rank( ABC ) = rank( AB) + rank( BC ) − rank( B) hold? Int. J. Math. Educ. Sci. Technol. 2002, 33,
127–137.
6. Zhang, F. Matrix Theory. Basic Results and Techniques; Springer: New York, NY, USA, 2011.
7. Brayman, V.; Kukush, A. Undergraduate Mathematics Competitions (1995–2016). Taras Shevchenko National University of Kyiv, 2nd ed.;
Springer: New York, NY, USA, 2017.
8. Băeţica, C.; Mincu, G.; Pop, V.; Todea, C.C. Traian Lalescu national mathematics contest for university students, 2019 edition. Gaz.
Mat. Ser. 2019, 3–4, 22–35.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.