Transforms and Inverses

As stated on the previous page, AT-1 = A-1T. And a demonstration is in order.

The new equations

The original equations will not suffice, because we need an invertible equation set. For this demonstration we'll use these:

x+z=b1
y-z=b2
x-z=b3

In matrix notation this is:

Ax = b

Where the matrix A is:

101
01-1
10-1

Inverting this new matrix

Augment the matrix with identity and begin the row operations:

101
01-1
10-1
100
010
001

Subtract row 1 from row 3.

101
01-1
00-2
100
010
-101

Divide through row 3 by -2.

101
01-1
001
100
010
1/20-1/2

Add row 3 to row 2.

101
010
001
100
1/21-1/2
1/20-1/2

Subtract row 3 from row 1.

100
010
001
1/201/2
1/21-1/2
1/20-1/2

The inverse is:

1/201/2
1/21-1/2
1/20-1/2

Proving the inverse is correct

1/201/2
1/21-1/2
1/20-1/2
*
101
01-1
10-1
=
100
010
001

This inverse is apparently the correct one. Now let's generate this solution from the transpose of the original matrix.

Inverting the transpose

The transpose of the matrix A at the top of this page is (AT):

101
010
1-1-1

Augment and begin the row operations.

101
010
1-1-1
100
010
001

Subtract row 1 from row 3.

101
010
0-1-2
100
010
-101

Add row 2 to row 3.

101
010
00-2
100
010
-111

Divide through row 3 by -2.

101
010
001
100
010
1/2-1/2-1/2

Subtract row 3 from row 1.

100
010
001
1/21/21/2
010
1/2-1/2-1/2

The inverse of AT (designated AT-1) is:

1/21/21/2
010
1/2-1/2-1/2

The transpose of this matrix (AT-1T) is:

1/201/2
1/21-1/2
1/20-1/2

And this is the result calculated above for the stated problem. We demonstrated the transpose of the inverse of the transpose is the inverse of the matrix.

In the next section we'll use column operations instead of doing the row operations on the transposed matrix.