Cramer’s Rule for Homogeneous Equations

Tanmay Inamdar

Introduction
There is a technique which is used many times for solving a system of homogeneous equations when
there are singly infinite solutions (there is one parameter). In our textbooks, it is referred to as
Cramer’s Rule, although it has nothing to do with the real Cramer’s Rule which is used for solving
n equations with n unknowns, and when the determinant of the LHS is non-zero.

Proof for 3 × 3 matrix :

Let A X = B be a homogeneous system of linear equation in 3 equations and 3 unknowns and let
rank of A be 2. Hence one of the rows is dependent on the other two.

A X=B
    
a1 a2 a3 x1 0
 b1 b2 b3  x2 = 0
 
c1 c2 c3 x3 0
R3 := R3 − (k1 · R1 + k2 · R2 ), so that R3 becomes 0.
    
a1 a2 a3 x1 0
 b1 b2 b3  x2  = 0
0 0 0 x3 0

b1
Now to convert to echelon form, R2 := R2 − · R1
a1
 
a1 a2 a3    
x 0
 b2 a1 − b1 a2 b3 a1 − b1 a3   1   

 x2 = 0
0
 a1 a1
0 0 0 x3 0

Hence, the two equations become,

a1 x1 + a2 x2 + a3 x3 = 0 (1)
(b2 a1 − b1 a2 )x2 + (b3 a1 − b1 a3 )x3 = 0 (2)
Put x3 = t in equation 2.
 
b3 a1 − b1 a3
x2 = −t
b2 a1 − b1 a2
Put the values of x2 and x3 in terms of t in Equation 1.
 a1 x1 = −(a2 x2 + a3 x3 )
 
b1 a3 − b3 a1
= −t a2 + a3
b2 a1 − b1 a2
 
a1 (b3 a2 − b2 a3 )
= −t
b2 a1 − b1 a2
 
b2 a3 − b3 a2
x1 = t
b2 a1 − b1 a2

Hence, all variables are now parameterized. In terms of equal ratios,

x1 −x2
 =  = x3 = t
b2 a3 − b3 a2 b3 a1 − b1 a3
b2 a1 − b1 a2 b2 a1 − b1 a2
t
Dividing all fractions by b2 a1 − b1 a2 . And putting t :=
b2 a1 − b1 a2
x1 −x2 x3
= = =t
b2 a3 − b3 a2 b3 a1 − b1 a3 b2 a1 − b1 a2
That is, in determinant form,
x −x2 x
 1 =  =  3  =t (3)
a3 a2  a1 a3  a1 a2 
     
 b3 b2   b1 b3   b1 b2 

Hence proved.

Generalized proof:
Now we have to prove that the claim is true for all square matrices of order n which have rank
n − 1.
AX=B
    
a1,1 a1,2 · · · a1,n x1 0
 a2,1
 a2,2 · · · a2,n   x2  0
   

 .. .. .. ..   ..  =  .. 
 .
 . . .   .  .
   
an−1,1 an−1,2 · · · an−1,n  xn−1  0
an,1 an,2 · · · an,n xn 0
Now, since the rank of matrix is n − 1, we can make the last row 0 by linear combinations of first
n − 1 rows.
AX=B
    
a1,1 a1,2 · · · a1,n x1 0
 a2,1
 a2,2 · · · a2,n  
  x 2 

0
 
 .. .. . .. . . .
..   ..  =  .. 
   
 . . 
    
an−1,1 an−1,2 · · · an−1,n  xn−1  0
0 0 ··· 0 xn 0
Now, converting back to equation form,

a1,1 x1 + a1,2 x2 + · · · + a1,n xn = 0

a2,1 x1 + a2,2 x2 + · · · + a2,n xn = 0
..
.
an−1,1 x1 + an−1,2 x2 + · · · + an−1,n xn = 0

Shifting terms with xn to RHS and put xn = t,

a1,1 x1 + a1,2 x2 + · · · + a1,n−1 xn−1 = −a1,n t

a2,1 x1 + a2,2 x2 + · · · + a2,n−1 xn−1 = −a2,n t
..
.
an−1,1 x1 + an−1,2 x2 + · · · + an,n−1 xn−1 = −an,n t

Considering RHS as constants (since it is independent of variables x1 , x2 , · · · xn−1 ), it is possible

to solve for x1 , x2 , · · · , xn−1 using Cramer’s Rule.

−t Dx1 −t Dx2 −t Dxn−1

x1 = , x2 = , · · · , xn−1 =
D D D
Where Dx1 , Dx2 , · · · , Dxn are determinants without considering −t from RHS. However, from
each of the Di s, a −t comes outside from ith column.
We can be assured that D is non-zero, since the originial A had order n − 1, so there is at least 1
non-zero minor of order n − 1. If the present D is 0, we could have shifted rows and renamed
variables to match D with the original non-zero minor. Hence,
x1 D x2 D xn−1 D
= = ··· = = −t ( = −xn )
D x1 D x2 Dxn−1

Rearranging, and putting t := −t/D

x1 x2 xn−1 −xn
= = ··· = = =t (4)
D x1 Dx2 Dxn−1 D

Where D is the non-zero minor of order n − 1 Dx1 , Dx2 , · · · , Dxn−1 are determinants Di where ith
column in D is replaced by
 
a1,n
 a2,n 
 
 .. 
 . 
an−1,n
Note that a ‘-’ sign appears in equation 3 with x2 term but in equation 4, it appears with xn
term. The two equations are not contradictory. Equation 3 is an easy-to-remember form of
equation 4, wherein the columns of determinants in denominator are written in ascending order.
If similar representation is needed in equation 4, the number of swaps of columns required will be
1 for xn−1 , 2 for xn−1 and so on, the numerators will have to be multiplied by (−1)n−i for xi .
However, this representation is difficult to remember than equation 4.