0

Here is Prob. 15, Sec. 2.2, in the book Linear Algebra With Applications by Steven J. Leon and Lisette de Pillis, tenth edition:

Let $A$ and $B$ be $n \times n$ matrices. Prove that if $AB = I$, then $BA = I$. ...

Although this question has several answers here on Math Stack Exchange, I feel that my answer below is more elementary since it uses minimal background knowledge.

My Attempt:

Since $AB = I$, where $I$ is the $n \times n$ identity matrix, we have $$ \det (A) \det(B) = \det (AB) = \det(I) = 1, \tag{1} $$ and hence $$ \det ( B) \neq 0, $$ which shows that $B$ is a non-singular matrix.

Therefore we have $$ B = BI = B(AB) = (BA)B, $$ which implies that $$ B - (BA)B = O, $$ where $O$ is the $n \times n$ zero matrix. The preceding relation can be rewritten as $$ IB - (BA)B = O, $$ and hence $$ (I - BA)B = O, $$ and since $B$ is non-singular, we have $$ (I - BA)BB^{-1} = OB^{-1}, $$ that is, $$ (I - BA)I = O, $$ or in other words, $$ I - BA = O, $$ and hence $$ I = BA, $$ which is the same as $$ BA = I, $$ as required.

Alternatively, from (1) above we can conclude that $$ \det(A) \neq 0 \qquad \mbox{ and } \qquad \det(B) \neq 0, $$ which shows that both $A$ and $B$ are non-singular matrices. Then we obtain $$ A \left( A^{-1} - B \right) = AA^{-1} - AB = I- I = O, $$ and hence $$ A \left( A^{-1} - B \right) = O, $$ which implies $$ A^{-1}A \left( A^{-1} - B \right) = A^{-1} O, $$ or in other words, $$ I \left( A^{-1} - B \right) = O, $$ and hence $$ A^{-1} - B = O, $$ which implies $$ A^{-1} = B, $$ and so $$ BA = A^{-1} A = I, $$ as required.

Is this approach correct and clear enough? If so, then is this proof not simpler than that requiring a knowledge of vector spaces and linear transformations?

  • 1
    This is correct but it can be made shorter: First use the $\det$ argument to prove $B$ is nonsingular. Then take $AB = I$ . Multiply the left side by $B$ to get: $B;AB = B$ . Multiply to the right side by $B^{-1}$ to get:

    $BAB ;B^{-1} = B; B^{-1}$. Reduce with $B; B^{-1} = I$ to get $BA = I$

    – user3257842 May 26 '23 at 12:14
  • This site is not a proof checking machine. Please say, where exactly you have a doubt on your solution, given the well known standard solution. – Dietrich Burde May 26 '23 at 12:16
  • @DietrichBurde I'm sorry if you've found reading my post unagreeable, but I thought I should seek the opinion of the Math Stack Exchange community as to whether my proposed proof is more elementary than those already given. – Saaqib Mahmood May 26 '23 at 12:22
  • I am sorry if you think that I found your post unagreeable. This is not the case. But I find the standard proofs really very nice and elementary, and I think you first should compare them with your proof. – Dietrich Burde May 26 '23 at 12:31
  • @DietrichBurde thank you for the advice. I did review a couple of proofs of this same result here on Math Stack Exchange before posting mine. Since the textbook I've referred to discusses determinants before it discusses vector spaces and linear transformations, I thought I should post a proof using only the machinary available to a student learning linear algebra from this particular textbook or other with the same order of presentation of topics as this one. – Saaqib Mahmood May 26 '23 at 12:53
  • The solution by Davidac897 at the duplicate assumes even less than determinants. It just uses linear independent vectors. I do not see why your proof is less elementary. I find determinants already more complicated than the definition of a vector space and linearly independent vectors. – Dietrich Burde May 26 '23 at 12:55
  • @DietrichBurde Steven J. Leon discusses determinants before he even defines what vector spaces are. – Saaqib Mahmood Jun 03 '23 at 13:10

0 Answers0