EFFECT OF EROs ON DETERMINANTS

Let E be a square matrix:

1) if a multiple of one row of E is added to
another to get a matrix F , then det E det F
(row replacement has no effect on determinant)
# If two rows of E are interchanged to get F ,
then det F =  det E
(each row swap reverses the sign of the
determinant)
3) If one row of E is multiplied by 5 ( !) to
get F , then det F 5 det E
rescaling a row by a factor of 5 also rescales
the determinant by a factor of 5 )
F
E
&+ &,
+ ,
Example: det
& det
.
- .
(3) often used to factor out a number from a
row

Why?

Rephrase in terms of elementary matrices

Theorem

Suppose E is 8 8 and
I is an 8 8 elementary matrix

then

1.

detIE detIdetE

and
2.

detI

"
"
5

if I is: add multiple of row to another

if I is: interchange two rows
if I is: rescale a row by a factor of 5

___________________ WHY?______________________________
Check this first for # # matrices
E

,
is # # and that I is a # # elementary matrix ERO
.

+
-

1. If I represents a row replacement say, 5 *row 1 added to row 2)

IE
detI

"
5

!
+
E

"
5+  -

,
,
5,  .

"

so
detIE 5+,  +.  5+,  ,- +.  , detE =9 this ERO doesn't change
detE
detI detE

2. If I represents a row rescaling (say, a rescaling of row 1)

5
IE
!

!
5+
E

"
-

5,
.

detI 5
so
detIE 5+.  ,- 5 detE
so this ERO multiplies detE by 5
detI detE

3. If I represents a row interchange

IE

!
"

"
E
!
+

.
,

detI  "
so
detIE ,-  +.  "detE
so this ERO changes sign of det E
detI detE

"
So, by direct checking we find that the Theorem is true for # # matrices.
Next argue
Theorem true for # # matrices:
forces the theorem to be true for $ $ matrices;
Then

Theorem true for $ $ matrices:

forces the theorem to be true for % % matrices;

And in general


Theorem true for 8 8 matrices
forces the theorem to be true for 8  " 8  " matrices.
This is called a proof by induction: we argue up the ladder: it's true for # # matrices, and
whenever it's true for one size, 8 8, then it must also be true for the next size up,
8  " 8  " . This means it must be true for all sizes of square matrices.
We didn't do this general argument. But illustrated by showing why, for a row replacement ERO,
true for # # forces true for $ $. The same style argument is how you argue from size
8 8 to size 8  " 8  ".
To illustrate:

Why does the fact that

(**)

detIE detIdetE;

works for # # force it also to be true for $ $ ?

For Row replacement:
detI "

I
E
" ! ! + ,
5 " !
. /
! ! " 1 2

- +
0 .  5+
3 1

,
/  5,
2

IE
-
0  53

expand along a row that's uninvolved with the ERO

,
detI E 1 det
/  5, 0  5-
 2 det

+
.  5+

 3 det

0  5-

+
,
.  5+ /  5,
why? because, without even evaluating the determinants, we know this row
operation does not affect the determint for # # matrices.

, + + ,
1 det
 2 det
 3 det

/ 0
. 0
. /
detE
" detE
detI detE

We then proved in class (see textbook):

If I" I: are 8 8 elementary, and E F are 8 8:

" detI: I" E detI: ... detI" detE

# detI: I" detI: ... detI" (determinant of product = product of
determints, for elementary matrices)
$ detEF detE detF for any two 8 8 matrices E F
and this easily generalizes to:
determinat of product = product of dweterminats for any number of square matrices.
(If you know 3), then 1) and 2) are of course automatically true; but we "inched up" to 3)
by proving 1) and 2) first.

Theorem For an 8 8 matrix E, det EX det E

Proof

True if E is # # (why?
+
just check: is det
-

,
+
det

.
,

?
.

The proof is again by induction: here's why it works for # # forces that is works
for $ $. Why does this make it true when E is $ $ ?

+""
+#"
+$"

+"#
+##
+$#

+"$
+#$
+$$

EX

+""
+"#
a"$

+#"
+##
+#$

+$"
+$#
+$$

First row E first column EX ; use this row & column to compute the determinants
det EX =sum of
down column 1

det E=sum of
(across row 1)
+
+"" ##
+$#

+#$
+$$

+
+"" ##
+
#$

+$#
+$$

+
 +"# #"
+

+#$
+$$

+
 +"# #"
+
#$

+$"
+$$

a
 +"$ #"
+$"

+##
+$#

 +"$

+#"
+##

+$"
+$#

$"

colored pairs have same determinant because they are # # and transposes of each other
So det E det EX when E is $ $.
To show that, then det E det EX must also be true for % % then that it must also be
true for & &, etc. uses exactly the same style calculation.