Linear Algebra and Differential Equations
Sartaj Ul Hasan
Department of Mathematics
Indian Institute of Technology Jammu
Jammu, India - 181221
Email: [email protected]
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 0 / 10
Lecture 10
(March 30, 2021)
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 1 / 10
Basis and Dimension
Definition: Let V be a vector space over F. Then a subset B ⊆ V is
a basis for V if
(i) B is a linearly independent and,
(ii) B spans (or generates) V i. e. V = Span B.
Definition: A space V which has a (finite) basis is said to be finite
dimensional.
A space which does not have a finite basis is said to be infinite
dimensional.
Example of Basis: In Rn , consider the vectors (column vectors
written as n-tuples for convenience):
e1 = (1, 0, 0, . . . , 0), e2 = (0, 1, 0, . . . , 0), · · · , en = (0, 0, . . . , 0, 1)
These vectors are linearly independent and Span (or generate) Rn .
Hence, they form a basis for Rn , known as the standard basis. Note
that these vectors change for different n.
The plural of basis is bases.
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 2 / 10
Basis (Conti . . . )
Example of an infinite dimensional space
The space R[t] of polynomials with real coefficients.
Justification: Suppose, by way of contradiction, that R[t] is
finite-dimensional. Then it must have a finite basis, say
B = {p1 (t), p2 (t), . . . , pn (t)}.
Put N = max{deg p1 (t), deg p2 (t), . . . , deg pn (t)}, and let
p(t) = t N+1 . Then we can easily see that p(t) ∈ / Span B. The
contradiction proves the desired result.
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 3 / 10
Basis (Conti . . . )
Alternative Definition for Basis
Recall: A basis for a vector space V is a linearly independent set of
vectors which spans the space V . A space V which has a finite basis
is said to be finite dimensional.
Proposition 11: B = {v1 , v2 , . . . , vn } is a basis of the vector space V
if and only if every vector v ∈ V is uniquely expressible as a linear
combination of the elements of B.
Remark: In some books, the above is used as the definition of a
basis, and then it is shown that a basis is a linearly independent
spanning set.
Proof: Left as an exercise.
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 4 / 10
Fundamental Results
Proposition 12 (Steinitz Exchange Lemma): Suppose v1 , v2 , . . . , vn are
linearly independent vectors in a vector space V , and suppose
V = Span{w1 , w2 , . . . , wm }. Then:
(a) n≤m
(b) {v1 , v2 , . . . , vn , wn+1 , wn+2 , . . . , wm } spans V , after re-ordering the
w ’s if necessary.
Proof: So we have: v1 , v2 , . . . , vn are LI and w1 , w2 , . . . , wm span V i.e
V = Span{w1 , . . . , wm }. Since w1 , w2 , . . . , wm span V , we must have
v1 = c1 w1 + c2 w2 + · · · + cm wm , (1)
for some sacalars ci . If ci = 0 for all i, then v1 = 0, which is not possible
since any set containing the zero vector is LD. Therefore, ci 6= 0 for at
least one i, and re-numbering the wi0 s if necessary, we can assume that
c1 6= 0.
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 5 / 10
Fundamental Results (Conti . . . )
Proof of Proposition 12 (Cont’d)
So we can re-write (1) as:
c1 w1 = v1 − c2 w2 − c2 w2 − · · · − cm wm , (2)
and multiplying by c1−1 , we get:
w1 = c1−1 v1 − c1−1 c2 w2 − · · · − c1−1 cm wm , (3)
or
w1 = d1 v1 + d2 w2 + · · · + dm wm , (4)
where di are scalars. From (4), it follows that:
Span{v1 , w2 , . . . , wm } = Span{w1 , w2 , . . . , wm } = V (5)
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 6 / 10
Fundamental Results (Conti . . . )
Proof of Proposition 12 (Cont’d)
Justification of (5): Suppose x ∈ V , then x ∈ Span{w1 , . . . , wm }, i.e.
x = b1 w1 + b2 w2 + · · · + bm wm . (6)
Substituting for w1 in (6) from (4), we get:
x = b1 (d1 v1 + d2 w2 + · · · + dm wm ) + b2 w2 + · · · + bm wm
= b1 d1 v1 + (b1 d2 + b2 )w2 + · · · + (b1 dm + bm )wm
= h1 v1 + h2 w2 + · · · + hm wm .
Thus x ∈ Span{v1 , w2 , . . . , wm }, which implies that
V ⊆ Span{v1 , w2 , . . . , wm }. Hence V = Span{v1 , w2 , . . . , wm } as claimed.
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 7 / 10
Fundamental Results (Conti . . . )
Proof of Proposition 12 (Cont’d)
So, at the next step, we get that v2 = `1 v1 + `2 w2 + · · · + `m wm for some
scalars `i . We see that at least one of `2 , `3 , . . . , `m is not zero; if all are
zero, then v2 = `1 v1 – contradicting the linear independence of vi0 s. By
re-numbering wj0 s, if necessary, we may assume `2 6= 0. So then:
`2 w2 = −`1 v1 + v2 − `3 w3 − · · · − `m wm , and arguing as before, we get
that:
Span{v1 , v2 , w3 . . . , wm } = Span{v1 , w2 , . . . , wm }
= Span{w1 , w2 , . . . , wm }
=V
Proceeding in this way, we can step-by-step replace w1 by v1 , w2 by
v2 , . . . , etc. The process has to stop after n-th step at most (since there
are only n of the v vectors).
What is the situation when we come to the stop? There are two possible
cases.
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 8 / 10
Fundamental Results (Conti . . . )
Proof of Proposition 12 (Cont’d)
Case 1: n ≤ m. In this case we get the following situation:
v1 , v2 , . . . , vn
↓ ↓ ... ↓
w1 , w2 , . . . , wn , wn+1 , . . . , wm
We have replaced n of the w vectors, with re-numbering if necessary, and
we get V = Span{v1 , v2 , . . . , vn , wn+1 , . . . , wm }. So in case 1, the
proposition is proved.
[If n = m, then the vectors wn+1 , etc are not there in the original spanning
set at all.]
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 9 / 10
Fundamental Results (Conti . . . )
Proof of Proposition 12 (Cont’d)
Case 2: n > m. In this case, we are only able to replace w1 , w2 , . . . , wm
and we are left with the vectors vm+1 , . . . , vn of the original linearly
independent vectors. The situation looks like:
v1 , v2 , . . . , vm , vm+1 , . . . , vn
↓ ↓ ... ↓
w1 , w2 , . . . , wm
i.e. {v1 , v2 , . . . , vm } is now a spanning set for V . But, then
vm+1 ∈ Span{v1 , . . . , vm } or vm+1 = k1 v1 + · · · + km vm for some scalars
ki . But this contradicts linear independence of the vi0 s. Hence, Case 2 can
not happen. Only case 1 can happen, and in this case as we saw before,
the Proposition 12 has been proved.
Sartaj Ul Hasan (IIT Jammu) SMD002U1M 10 / 10
