The Student Room Group

Invertible Matrices

Let's suppose I have a matrix BA which is the ordered product of the matrices B and A.

If I claim that BA is invertible, then do I need that B and A are invertible too?

Also, if I have that the inverse of A, A^-1 exists, then can I say the right inverse and left inverse of A also exist?
I mean I know there are plenty of left/right inverses of A, but A^-1 is one of them right, so if it exists, surely I can say A has a left and right inverse?
(edited 3 years ago)
Original post by Takeover Season
Let's suppose I have a matrix BA which is the ordered product of the matrices B and A.

If I claim that BA is invertible, then do I need that B and A are invertible too?

Also, if I have that the inverse of A, A^-1 exists, then can I say the right inverse and left inverse of A also exist?
I mean I know there are plenty of left/right inverses of A, but A^-1 is one of them right, so if it exists, surely I can say A has a left and right inverse?

Think about the determinant of BA and facts you know about the determinant.
Original post by zetamcfc
Think about the determinant of BA and facts you know about the determinant.

|BA| = |B| |A|, if any of |A| or |B| are 0, then |BA| = 0. So, in order for |BA| to be invertible, A and B must both be non-invertible and their inverses must exist. Thank you!!!

What about the right/left inverses part?
Reply 3
Original post by Takeover Season
|BA| = |B| |A|, if any of |A| or |B| are 0, then |BA| = 0. So, in order for |BA| to be invertible, A and B must both be non-invertible and their inverses must exist. Thank you!!!

What about the right/left inverses part?

A is invertible if there's B s.t. AB = BA = I.
Reply 4
Original post by Takeover Season
Let's suppose I have a matrix BA which is the ordered product of the matrices B and A.

If I claim that BA is invertible, then do I need that B and A are invertible too?

Also, if I have that the inverse of A, A^-1 exists, then can I say the right inverse and left inverse of A also exist?
I mean I know there are plenty of left/right inverses of A, but A^-1 is one of them right, so if it exists, surely I can say A has a left and right inverse?

Are A and B necessarily square? You can only define determinants for square matrices.
Original post by S2JN
A is invertible if there's B s.t. AB = BA = I.

Hi, yes that is the definition.

Original post by RichE
Are A and B necessarily square? You can only define determinants for square matrices.

Hi, yes for the first part, A and B are necessarily square. For the second part, no, because left and right inverses also exist for non-square matrices.
Reply 6
Original post by Takeover Season
Hi, yes for the first part, A and B are necessarily square. For the second part, no, because left and right inverses also exist for non-square matrices.

I am aware of this but that was why I was asking. It wasn’t clear that you could assume invertible matrices are necessarily square.
Reply 7
Original post by Takeover Season
Hi, yes that is the definition.

Your whole question is solved by reinstating the definitions.

I suggest you pay attention to the definitions.
Reply 8
Original post by S2JN
Your whole question is solved by reinstating the definitions.

I suggest you pay attention to the definitions.

Not really. For BA to be invertible means there is C such that CBA = I = BAC.

From this it’s clear A has a left inverse and B has a right inverse but you need some result to know that means A and B are invertible.
Original post by RichE
Not really. For BA to be invertible means there is C such that CBA = I = BAC.

From this it’s clear A has a left inverse and B has a right inverse but you need some result to know that means A and B are invertible.


Hi, just to clarify. A may not have an inverse yet it may have a left/right inverse - that I know. Here A has a left inverse, what is the result that means A also has an inverse too? I mean is it just because we are assuming A and B are square matrices, so if it has a left inverse, it also has an inverse = left inverse = right inverse.
Reply 10
Original post by Takeover Season
Hi, just to clarify. A may not have an inverse yet it may have a left/right inverse - that I know. Here A has a left inverse, what is the result that means A also has an inverse too? I mean is it just because we are assuming A and B are square matrices, so if it has a left inverse, it also has an inverse = left inverse = right inverse.

Well that’s the missing bit but how to justify it? It depends what theory you’ve met. You could prove this with determinants, using rank-nullify theorem, knowing row rank equals column rank or using EROs,
Original post by Takeover Season
Let's suppose I have a matrix BA which is the ordered product of the matrices B and A.

If I claim that BA is invertible, then do I need that B and A are invertible too?

Also, if I have that the inverse of A, A^-1 exists, then can I say the right inverse and left inverse of A also exist?
I mean I know there are plenty of left/right inverses of A, but A^-1 is one of them right, so if it exists, surely I can say A has a left and right inverse?

(1) Assuming A, B are square, as someone mentioned: det(BA)=det(B)det(A). Invertible <--> non-zero determinant for square matrices. Hence, if BA is invertible, then det(B) and det(A) must be non-zero - that is, A, B must be invertible too.

(2) Property of square matrices: saying that A has an inverse, or a left inverse, or a right inverse are equivalent. Consider matrix A with inverse B. Then, from definition that BA=AB=I, we see that B is left and right inverse of A. You can also prove the reverse statement i.e. if left and right inverse exists, then inverse exists: let left inverse be B and right inverse C. Then, B=BI=B(AC). By associativity, B(AC)=(BA)C=IC=C.
Reply 12
Original post by matthewleechen
(1) Assuming A, B are square, as someone mentioned: det(BA)=det(B)det(A). Invertible <--> non-zero determinant for square matrices. Hence, if BA is invertible, then det(B) and det(A) must be non-zero - that is, A, B must be invertible too.

(2) Property of square matrices: saying that A has an inverse, or a left inverse, or a right inverse are equivalent. Consider matrix A with inverse B. Then, from definition that BA=AB=I, we see that B is left and right inverse of A. You can also prove the reverse statement i.e. if left and right inverse exists, then inverse exists: let left inverse be B and right inverse C. Then, B=BI=B(AC). By associativity, B(AC)=(BA)C=IC=C.

I know all this. But you said earlier the result came straight from the definition which clearly isn’t the case.

Quick Reply

Latest