6

MTL101
Lecture 5 (Solving a system of linear equations)
(20) Observation: If (A|B) ∼ (A0 |B 0 ) (row equivalent) then the following system have exactly same
set of solutions: AX = B and A0 X = B 0 .
Reasoning: If ρ = ρk ◦ ρk−1 ◦ · · · ◦ ρ1 is a composition of elementary row operations then
ρ(AX) = ρ(B) ⇔ ρ(A)X = ρ(B) ⇔ A0 X = B 0 .
Suppose A ∈ Mm×n (F ) is a row reduced echelone matrix having r nonzero rows (and m − r
zero rows). Further assume that the leading nonzero coefficient of the i-th nonzero row occurs
in the ki -th column. Call the xk1 , xk2 , . . . , xkr to be dependent unknowns and the remaining
unknowns as independent or free unknowns.
The system has no solution if any one of the last m − r rows of B is nonzero because such a
row gives an inconsistent equation 0 = br+i for some i ≥ 1. If the last m − r rows are zero
rows then the system has a solution. In this case, assign arbitrary values to the independent
unknowns and then write the dependent unknowns in terms of arbitrary values assigned to the
independent unknowns.
(21) Thus, AX = B has a solution if and only if rank(A) = rank(A|B). Further, AX = B has a
unique solution if and only if rank(A) = rank(A|B) = n, the number of unknowns (so that
n − r = 0 is the number of free unknowns).
(22) Now we discuss
 the three system
  of equations
  in the first lectures
 using the method.

1 1 1 3 x1 1 1 1 | 3
In a), A = 1 2 3 , B = 6, X = x2  and (A|B) = 1 2 3 | 6 .
0 1 2 1 x3 0 1 2 | 1
Apply elementary row operations
 on (A|B)
 to find the row reduced
 echelon form
 of the first
1 1 1 | 3 1 0 −1 | 0
R →R −R R1 →R1 −R2 ,R3 →R3 −R2
block. (A|B) 2 ∼2 1 0 1 2 | 3 ∼ 0 1 2 | 3  .
0 1 2 | 1 0 0 0 | −2
Since the third row of the first block is zero but the third entry of the second column is nonzero
the system  has no solution.
      
1 1 1 3 x1 1 1 1 | 3
In b), A = 1 2 3 , B = 6, X = x2  and(A|B) = 1 2 3 | 6 .
1 1 2 4 x3 1 1 2 | 4
   
1 1 1 | 3 1 0 −1 | 0
R2 →R2 −R1 ,R3 →R3 −R1
In this case, (A|B) = ∼ 0 1 2 | 3 R1 →R −R
∼1 2 0 1 2 | 3
0 0 1 | 1 0 0 1 | 1
 
1 0 0 | 1
R1 →R1 +R3 ,R2 →R2 −2R3
∼ 0 1 0 | 1 . Thus, r = 3 = n = m, ki = i for each i = 1, 2, 3
0 0 1 | 1
so that thereis no independent
 unknown.
  In factthe solution is x1= x2 = x3 = 1(unique).
1 1 1 3 x1 1 1 1 | 3
R →R −R
In c), A = 1 2 3 , B = 6 , X = x2 and (A|B) = 1 2 3 | 6 2 ∼2 1
      
0 1 2 3 x3 0 1 2 | 3
   
1 1 1 | 3 1 0 −1 | 0
0 1 2 | 3 R1 →R1 −R2∼,R3 →R3 −R2 0 1 2 | 3 .
0 1 2 | 3 0 0 0 | 0
In this case, n = m = 3, r = 2, k1 = 1, k2 = 2. x3 is the independent unknown so assign any
arbitrary value to it, say x3 = λ. Then x1 − λ = 0 and x2 + 2λ = 3. Thus the general solution
of the system is (λ, 3 − 2λ, λ).
(23) Denote by ei , (1 ≤ i ≤ n) the n-tuple with 1 in the i-th component and 0 in the other
n
P
components. Then (a1 , a2 , . . . , an ) = ai ei . In fact any n-tuple is a unique linear combination
i=1
of ei ’s (recall that two n-tuples are equal if and only if each component of the first tuple is equal
to the corresponding component of the second tuple). Suppose f1 = e1 +e2 +e3 , f2 = e2 +e3 , f3 =
e3 ∈ R3 . Then (x, y, z) = af1 + bf2 + cf3 for some a, b, c ∈ R. In fact, a = x, b = y − x, c = z − y.
Further, choices for a, b, c are unique (prove this).
 7

More generally, the set {e1 , e2 , . . . , en } can be replaced by some other set X = {f1 , f2 , . . . , fn }
of n elements from Rn so that every element of Rn can be written as a linear combination of
tuples belonging to X.
Note that that not any choice of n vectors fi will do. We need a set of n linearly independent
set of vectors. X is said to be linear independent if a linear combination of vectors belonging
to X is zero only when each coefficient is zero. Further the number of elements in X cannot
be changed if we want every vector of Rn is represented as a linear combination of elements of
X in a unique way.
Explain geometrically when n = 1, 2, 3.