University:
Massachusetts Institute of Technology
Course:
18.701 | Algebra I
Academic year:

2021
Views:

320

Pages:

1
Author:

ejaculis4cq

Row Rank = Column Rank This is in remorse for the mess I made at the end of class on Oct 1. The column rank of an m × n matrix A is the dimension of the subspace of F m spanned by the columns of A. Similarly, the row rank is the dimension of the subspace of the space F n of row vectors spanned by the rows of A. Theorem. The row rank and the column rank of a matrix A are equal. proof. We have seen that there exist an invertible m × m matrix Q and an invertible n × n matrix P such that A1 = Q−1 AP has the block form � � I 0 A1 = 0 0 where I is an r × r identity matrix for some r, and the rest of the matrix is zero. For this matrix, it is obvious that row rank = column rank = r. The strategy is to reduce an arbitrary matrix to this form. We can write Q−1 = Ek · · · E2 E1 and P = E1� E2� · · · E�� for some elementary m × m matrices Ei and n × n matrices Ej� . So A1 is obtained from A by a sequence of row and column operations. (It doesn’t matter whether one does the row operations before the column operations, or mixes them together: The associative law for matrix multiplication shows that E(AE � ) = (EA)E � , i.e., that row operations commute with column operations.) This being so, it suﬃces to show that the row ranks and column ranks of a matrix A are equal to those of a matrix of the form EA, and also to those of a matrix of the form AE � . We’ll treat the case of a row operation EA. The column operation AE � can be analyzed in a similar way, or one can use the transpose to change row operations to column operations. We denote the matrix EA by A� . Let the columns of A be C1 , ..., Cn and let those of A� be C1� , ..., Cn� . Then Cj� = ECj . Therefore any linear relation among the columns of A gives us a linear relation among the columns of A� : If C1 x1 + · · · + Cn xn = 0 then E(C1 x1 + · · · + Cn xn ) = C1� x1 + · · · + Cn� xn = 0. So if j1 , ..., jr are distinct indices between 1 and n, and if the set {Cj�1 , ..., Cj�r } is independent, the set {Cj1 , ..., Cjr } must also be independent. This shows that column rank(A� ) � column rank(A). Because the inverse of an elementary matrix is elementary and A = E −1 A� , we can also conclude that column rank(A) � column rank(A� ). The column ranks of the two matrices are equal. Next, let the rows of A be R1 , ..., Rn and let those of A� be R1� , ..., Rn� , and let’s suppose that E is an elementary matrix of the ﬁrst type, that adds a · row k to row i. So Rj� = Rj for j �= i and Ri� = Ri + aRk . Then any linear combination of the rows Rj� is also a linear comination of the rows Rj . Therefore Span{Rj� } � Span{Rj }, and so row rank(A� ) � row rank(A). And because the inverse of E is elementary, we obtain the other inequality. Elementary matrices of the other types are treated easily, so the row ranks of the two matrices are equal. �

Description

Related Documents

Free up your schedule!

Take 5 seconds to unlock