Row Rank of a Matrix Equals its Column Rank
S. Kumaresan
[email protected]
Let F be a field. If you wish, you may take F = R, the field of real numbers in the sequel.
Let F m×n be the set of matrices of type m × n with values in F .
We consider F 1×n as the n-dimensional vector space Frow
n
.
We consider F m×1 as the m-dimensional vector space Fcol
m
consisting of all column vectors
y1
.
( y1 , . . . , ym ) := .. with yi ∈ F for 1 ≤ i ≤ m.
t
ym
Let A = (ai j ) ∈ F m×n be an m × n matrix over a field F . We denote by Ai , the i-throw of A:
Ai := (ai1 , . . . , ain ). The j-th column of A is denoted by A0j and is given by (a1 j , . . . , am j ) t .
n
We usually consider the row-vectors Ai as elements of the n-dimensional vector space Frow
consisting of all row vectors (x 1 , . . . , x n ) with x i ∈ F for 1 ≤ i ≤ n. We may consider Frow as the
n
n-dimensional vector space F 1×n consisting of matrices of type 1 × n with entries in F .
The row rank of the matrix A is the number of elements in a maximal linearly independent
n
subset of {Ai : 1 ≤ i ≤ m}. Let Wr ⊂ Frow be the vector subspace spanned by the vectors Ai ,
1 ≤ i ≤ m. The subspace Wr is known as the row-subspace of the matrix A. Then the row rank of
A is dim Wr .
Similar considerations apply to the column vectors A0j . The column rank of the matrix A is the
number of elements in a maximal linearly independent subset of {A0j : 1 ≤ j ≤ n}. The subspace
Wc spanned by {A0j : 1 ≤ j ≤ n} is known as the column space of the matrix A. The the column
rank of A is nothing other than dim Wc .
The result of the tile says that the column and row ranks of a matrix are equal.
We give a simple proof of this result which is also visually appealing. The proof depends on
the following two observations which you might have learned earlier in a course on linear algebra.
Observation 1. Let V be a vector space over F . Let S := {v1 , . . . , vk } ⊂ V . Let W be thePvector
subspace spanned by S. That is, W consists of all finite linear combinations of the from i ci vi ,
ci ∈ F . Then dim V ≤ qk.
This follows from the well-known result that there exists a subset B ⊂ S which is a basis of V .
(Recall that nay basis is a minimal spanning set!)
Observation 2. Let the notation be as in Observation 1. Let Z ≤ V be a vector subspace of V such
that V ⊂ W . Then dim Z ≤ dim W ≤ k.
1
With the preliminaries over, we are ready to state and prove the theorem.
Theorem 3. The row rank and the column rank of a matrix A are equal.
Proof. Let r be the row rank of A.
Let B1 , . . ., B r be a set of linearly independent rows of A. Let us write Bi = (bi1 , . . . , bi j , . . . , bin ).
Then any i-th row of A is a linear combination of B’s. We write these linear combinations explicitly.
(a11 , . . . , a1n ) = c11 (b11 , . . . , b1 j , . . . , b1n ) + · · · + c1r (b r1 , . . . , b r j , . . . , b r n )
..
.
(ai1 , . . . , ain ) = ci1 (b11 , . . . , b1 j , . . . , b1n ) + · · · + cir (b r1 , . . . , b r j , . . . , b r n )
..
.
(am1 , . . . , amn ) = cm1 (b11 , . . . , b1 j , . . . , b1n ) + · · · + cmr (b r1 , . . . , b r j , . . . , bn j )
Let us read these vertically and write the j-th column of this array of equations.1 We get
a1 j c11 c1r
a2 j c21 c2r
. = b + ··· +
br j , for 1 ≤ j ≤ n
.. 1j
... ...
am j c m1 c mr
= b1 j C1 + · · · + b r j C r , say.
That is, the columns are linear combinations of Ck , 1 ≤ k ≤ r. Hence the maximum number of
linearly independent columns is at most r, the row rank of A. Thus the column rank of A is less
than or equal to the row rank of A.
m
We have made use of the Observations. Let W be the vector subspace of Frow spanned
by the the column vectors C1 , . . . , C r . Then W ≤ Wc is a vector subspace of W . By
Observation 1, dim W ≤ r and by Observation 2, dim Wc ≤ r.
Starting with columns, we may prove that the row rank of A is less than or equal ot the column
rank of A. Or, we observe that the column rank (respectively, row rank) of AT is the row rank
(respectively, column rank) of A. Thus we get
row-rank of A = column rank of AT ≤ row rank of AT = column rank of A.
Hence both the ranks are equal.
The proof above is the same as the one in my book “Linear Algebra–A Geometric Approach”.
Since my knowledge of LATEX is better now, it is typeset better to bring our the visual appeal.
This article may be downloaded from https://4dspace.mtts.org.in/ea
1
If you remove the brackets of the terms on the left side and stack the rows, you get the matrix A.
