Module 1 1

Vectors Spaces

In this course we work with ‘scalars’ and ‘vectors’. ‘Scalars’ are going to be elements of a chosen

(associative) ring with (identity) K and ‘vectors’ are going to be in general non commutative

and will be equipped with a chosen involution but many significant results are best obtained in a

setting when further restrictions are imposed on K . We provided here a collection of definitions

and examples [of the kind of scalars that we are going to need or think the user may need a bit

later ], remarks, and other tit-bits, on our K.

1. Z, Q, R, C with the usual addition and multiplication are commutative rings. With the

exception of Z, they are fields,

2. fix n > 0 in Z and write a ∼ b iff n divides a − b.then

(i) [ n divides 0 ] a ∼ a so this is reflexive ,

(ii) [ ∵ if n divides a-b then n divides b−a ] a ∼ b ensures b ∼ a so this is a symmetric,and

(iii) [ ∵ n divides a − b and n divides b − c then n divides a − c ] a ∼ b, b ∼ c ensure a ∼ c,

so this is transitive.

Thus we have an equivalence relation on Z; we read a ∼ b as a is equivalent to b modulo

n’. Write [a] := {bεZ/a ∼ b} and define Zn := {[a]/aεZ} ,[a] + [b] := [a + b], [a][b] := [ab].

(i) If a0 ε[a], b0 ε[b] then n divides a−a0 as well as b−b0 ; therefore n divides (a+b)−(a0 +b0 ).

Thus [a + a0 ] = [b + b0 ] if [a] = [a0 ] and [b] = [b0 ]. This means addition above is

well defined. Similarly, if a0 ε[a], say a = nq + a0 , b0 ε[b] , say b = np + b0 then

ab = n2 pq + nqb0 + npa0 + a0 b0 i.e. ab − a0 b0 = n[npq + qb0 + a0 p] i.e. a0 b0 ε[ab]. Then

[ab] = [a0 b0 ] if [a] = [a0 ] and [b] = [b0 ] which means multiplication is well defined.

c Copyright Reserved IIT Delhi

 2
(ii) Clearly [a] + [b] = [a + b] = [b + a] = [b] + [a] and [0] = {nq/qεZ} so that if a0 ε[a],

say a = np + a0 , and nqε[0], we have a0 + nq = a − np + nq = a + n(q − p)ε[a] then

[a] + [0] = [a] = [0] + [a]. Further −a = −np − a0 i.e. [a] + [−a] = [0] which means

[−a] = −[a]. Therefore, under addition as defined, Zn an abelian group [it is clear

that [a] + ([b] + [c]) = [a] + ([b + c]) = [a + (b + c)] = [(a + b) + c] = ([a + b]) + [c] =

([a] + [b]) + [c]].

(iii) If [a] = [1] then a = nq + 1 so that ba = bnq + b hence [ba] = [b] and ab = nqb + b

hence [ab] = [b]. Thus [a] = [a][1] = [1][a] for each [a]εZn . Clearly [a]([b][c]) =

[a][bc] = [a(bc)] = [(ab)c] = [ab][c] = ([a][b])[c] for all [a], [b], [c]εZn . Therefore under

multiplication as defined, Zn is a monoid with [1] as identity.

(iv) We have [a]([b] + [c]) = [a][b + c] = [a(b + c)] = [ab + ac] = [ab] + [ac] = [a][b] + [a][c]

and ([a] + [b])[c] = [a + b][c] = [(a + b)c] = [ac + bc] = [ac] + [bc] = [a][c] + [b][c].Thus

the distributive law hold.

(v) These calculations show that Zn is a ring under the defined operations.

(vi) Since [a][b] = [ab] = [ba] for each [a], [b]εZn , we see that Zn is a commutative ring.

(vii) [a][b] = [1] iff ab = nq + 1. Conversely, if there exist integers c, d with ac = nd + 1,

let c = nq + r,0 ≤ r < n so that nd + 1 = ac = anq + ar, i.e. ar = n(d − aq) + 1

and thus [a][r] = [ar] = [1]. Thus [a]εZn is invertible if f a is relatively prime to n

i.e. gcd(a, n) = 1, i.e. there exist integers c, d with ac − nd = 1; the multiplicative

inverse of [a] is then given by [c] and we have [c] = [a]−1 iff there exists an integer d

such that ac − nd = 1

(viii) Therefore Zn is a field iff n is prime.

f
3. The ring Z occupies a very special position: Z −
→ K defined by f (n) := n1 = 1+. . .+1[1εK]

is the single-morphism for any ring K from Z. [Indeed, f (1) = 1K is must and so is

f (n) = f (1 + . . . + 1) = f (1) + f (1) + . . . + f (1) = 1K + . . . + 1K = 1K so f (n) := n1K is

c Copyright Reserved IIT Delhi

 3
the only possible formula if f is a ring-morphism and it is certainly clear that this formula

supplies a ring -morphism Z → K for any ring K; here we wrote the identity of K as 1K for

extra clarity because 1εZ is also there in the calculation but one usually writes 1K simply

as 1].

We say “Z is an initial object in the category of rings” but this language will not be used

in this course.

(i) The least integer k ≥ 0 such that f (k) = 0 is called the characteristic of K; clearly

f = {km|εZ}.

(ii) (a) Z, Q, R, C have characteristic 0.

(b) Zn has characteristic n. [∵ [0] = {nm|εZ}]

(c) Zn × Zm has characteristic lcm(m, n).

(d) K has characteristic 1 iff f (1) = 0 i.e. 1=0 in K. Since we have agreed that we

have at least two distinct scalars 0 6= 1, no ring of scalars for this course will

have characteristic 1 but it should be noted that not every textbook on linear

algebra takes this position.

4. (i) In connection with 3(ii)(d) above, it is perhaps here we should note that some text-

books on linear algebra accept rings without identity can be enlarged to a ring with

identity so in the good old days one always did it. with modern development, there

are situations [like ‘K-theory’] in which rings, or rather algebras, without identity

play an essential role. But we accept rings only with identity.

(ii) On the other hand, many text books on linear algebra take only commutative rings

as scalars and most of them in fact take only fields. Some accept only division rings.

The basic trouble with these in as follows : Even if one works only with fields as

scalars, one has to works with field as scalars, one has to work with ‘block matrices’ in

c Copyright Reserved IIT Delhi

 4
which are noncommutative so linear algebra with noncommutative rings also comes

into picture.

(iii) Working with only division rings or fields, one can avoid the term ‘module’ and use

only ‘vector spaces’. This is an economy which is only apparent : practically all the

work has to be done anyway. In particular, which one crucial fact : a linear operator
T
V −
→ V [where V is a vector space over field F ] is in fact just a module over the

polynomial ring F[θ], is never mentioned, all the computational work has still to be

done. So all one gains is that a concept has not been given its proper name.

(iv) We note [but will not prove ] that

(i) Every module over a division ring is a free module

(ii)If every module over K-modulo, K must be a division ring. Thus the best result

are certainly developed in the context of vector spaces [ which are just modules over

division rings ] and if one uses the very convenient characterization in terms of de-

terminant [of matrices ] then it is almost essential to work with “finite-dimensional

vector spaces over fields”. However,it becomes an unnecessary journey into surprizes

later when one feels that many things do not work in more matrices are not invert-

ible,that AB = 0 does not mean that A = 0 or B = 0, that AB = BA may not be

true,and so on; but it does not mean that one should never move out of the world

of real numbers.

(v) Some textbooks deal with only complex numbers as scalars; this is to prepare the

student for “Applicable functional analysis ” and hence to many rich and exciting

subjects like quantum mechanics and Wavelets differential equations and so(classical)

computation, one frequently meets with Zn and since there is hardly any saving in

learning efforts by declaring ”by a scalar we mean a real number or a complex

number” we have avoided this declaration.

(vi) Clearly, the ‘choice of reasonable scalars’ is a subjective choice . We hope our option

c Copyright Reserved IIT Delhi

 5
of working with a noncommutative but associative ring with identity equipped with

a preferred conjugation’ prepares the users for a number of situations that may arise

in subsequent mathematical education.

(vii) In all this, let us note that in Zn , we have [n] = [0] so that, for instance, λεZ2 , µεZ2
λ+µ
dose not ensure that 2 εZ2 . Many results will need “let F be a field of characteristic

not equal to 2”. [ There are fields other than Z2 which have characteristic 2;]

5. Let us note that the collection of all vectors in R3 is a ring with cross product as multipli-

cation since a × (b + c) = a × b + a × c. This is not associative , [ a × (b × c) = (a × b) × c

fails] and does not commutative, [ a × b = b × c fails ] and does not have an identity. User

for this course can treat it as an object of curiosity in the beginning but at the advanced

level of linear algebra, non associative multiplication enter into picture [when one deals

with ’Lie Algebra’ with which we will not work].

6. If a = a0 + a1 e1 + a2 e2 + a3 e3 is a quaternion then for any r εR we clearly have ra = ar.

Then R ⊆ cenH [ H the division ring of quaternions ]. Conversely, suppose q = (q0 +

q1 e1 + q2 e2 + q3 e3 )ε cenH so that qa = aq for each a εH. Then we have qe1 = e1 q in

particular. Now qe1 = q0 e1 + q1 e21 + q2 e2 e1 + q3 e3 e1

= −q1 + q0 e1 + q3 e2 − q2 e3 , while

e1 q = e1 q0 + e1 q1 e1 + e1 q2 e2 + e1 q3 e3

= q0 e1 + q1 e1 e1 + q2 e1 e2 + q3 e1 e3 [qi εR ⊆ cenH]

= −q1 + q0 e1 − q3 e2 + q2 e3
P3 P3 0
which provides q3 = 0 = q2 [ ∵ i=0 qi ei = i=0 qi ei iff qi = qi0 for each i]

Thus q = q0 + q1 e1 . We also have qe2 = e2 q, i.e. q0 e2 + q1 e1 e2 = e2 q0 + e2 q1 e1 =

q0 e2 + q1 e2 e1 i.e. q0 e2 + q1 e3 = q0 e2 − q1 e3 which means q1 = 0. Thus We have q = q0 εR

i.e. cen(H) ⊆ R. To sum up : cenH = R.

P3 P
[ We wrote a quaternion as i=0 qi ei rather than ei qi but this dose not matter since H

is an algebra over R and R is commutative; ].

c Copyright Reserved IIT Delhi

 6
7. Our rings are rings with a given involutions . We took λ = λ0 − e1 λ1 − e2 λ2 − e3 λ3 for

λ = λ0 +e1 λ1 +e2 λ2 +e3 λ3 εH as the standard involution but λ∗ = λ0 +e1 λ1 −e2 λ2 −e3 λ3 is

also an involution and there could be others [such as ?]. We do not attempt an explanation

of what makes an involution ‘student’. But let us look at some situations.

f
(i) (a) If Z −
→ Z is a ring morphism, we have f (1) = 1 and thus f (m) = f (m.1) = mf (1)

for all mεZ.

f
(b) If Zn −
→ Zn is a ring morphism, we have have f [a] = f (a[1]) = af [1] = a[1] = [a].

f
→ Q is a ring morphism, we have nf ( m
(c) If Q − m
n ) = f (n. n ) = f (m) = f (m.1) =

mf (1) = m = n. m m
n and hence f ( n ) =
m
n for any rεQ, r = m
n, m, nεZ, n > 0
f
(d) Suppose K and L are rings, K −
→ L a ring morphism. Write kerf := {λεK/f (λ) =

0} Then if λ ε kerf and a ε K, we have f (λa) = f (λ)f (a) = 0f (a) = 0 i.e. λaε

ker f ; similarly aλε kerf [ This makes kerf an ideal of K; for the definition

of ‘ideal’ see below] Now if K and L are division rings , 0 6= λ ε kerf , then

1K = λ−1 λ ε kerf so that f (1K ) = 1L and 1 6= 0 in any (division)ring. Therefore

kerf = 0. Hence f is injective [f (λ) = f (µ) means f (λ − µ) = f (λ) − f (µ) = 0

which ensures λ − µ  kerf i.e. λ − µ = 0 i.e. λ = µ ]

To sum up: If K and L are division rings (and in particular, if they are fields),
f
any ring morphism K −
→ L is necessarily injective.

f
(ii) If R −
→ R is a ring morphism, then by above it is injective and hence we have f (r) = r
√ √ √
for rεQ. If xεR, x > 0 then f (x) = f ( x x) = (f ( x))2 so that if a > b and thus

a − b > 0, we get f (a − b) = f (a) − f (b) > 0 i.e.f (a) > f (b). This means f must

be order-preserving. Now let xεR and choose rn , sn εQ such that rn < x < sn with
T∞
n=1 [rn , sn ] = {x}

[This is possible because of the properties of R ]

T∞
Then we have f (rn ) = rn < f (x) < f (sn ) = sn which means f (x)ε n=1 [rn , sn ] =

c Copyright Reserved IIT Delhi

 7
{x} and proves that f (x) = x.

Together with (i),(ii) shows that :

None of the rings Z , Zn , Q , R admits any ring morphism from itself to itself other

than the identity. Since each of them is commutative, morphisms are the same as

anti-morphisms. Therefore
f
No nontrivial involution, isomorphisms are the same as antiisomorphism . If C −
→ C is

an isomorphism such that f (x) = x for xεR, then f (x + iy) = f (x) + f (i)f (y) with

(f (i))2 = f (i)f (i) = f (i2 ) = f (−1) = −1 so that f (i) = ±i.To sum up:

The only nontrivial involution on C with f (x) = x for xεR is x + iy 7→ x − iy.

(iii) If d 6= 0 is an integer which is square free [ in the sense that its prime factorization
√ √ √ √
has no square ], we write Z[ d] := {a + b d|a, bεZ} and Q[ d] := {a + b d|a, bεQ}

Then under obvious multiplication and addition, they are commutative rings and
√ √ √
Q[ d] is actually a field ; both carry the nontrivial involution a + b d 7→ a − b d.

The user is requested to verify this.

(iv) For a prime number p, Zp is a finite field with exactly p element ; this has been

introduced in 2(viii) page 2 above. We supply the following facts without proof; the

proof are not exactly obvious and are properly speaking part of an algebra course

which pays some attention to finite fields.

(a) A finite field has exactly pr elements, p a prime, r > 0 an integer.

(b) There is exactly one finite field with pr elements

(c) For the finite field with pr elements, the characteristic is p. Further, λ 7→ λp is

a ring morphism and (λ + µ)p = λp + µp holds.

(d) Thus on the field F with p2 elements, x 7→ x := xp is an involution.

(e) Let us examine the field F with p2 elements for p = 2. Quite clearly, this cannot

be Z4 [in fact the field with pr elements, r > 1, will never be Zq , q = pr ] because

Zn is a field iff p is a prime [2(viii), page 2 above ]. verify that it carries addition

c Copyright Reserved IIT Delhi

 8
and multiplication given by the tables
+ 0 1 a b . 0 1 a b

0 0 1 a b 0 0 0 0 0

1 1 0 b a and 1 0 1 a b

a a b 0 1 a o a b 1

b b a 1 0 b 0 b 1 a

Its multiplicative fragment {1, a, b} must be an abelian group since it is a field.

Verify that this is same as the additive fragment of Z3 . There are no finite division

rings other than finite fields.

  
 
 α β 
 

8. Consider the collection K := 
 
 /α, βεC .
 
 −β α
 

       
 α β   γ δ   α+γ β+δ   α+γ β+δ 
Since 

+
 
=
 
=
 
,

−β α −δ γ −β − δ α + γ −(β + δ) α + γ
      
 α β   γ δ   αγ − βδ αδ + βγ   αγ − βδ αδ + βγ 
  = = ,
      
−β α −δ γ −βγ − αδ −βδ + α γ −(αδ + βγ) αγ − βδ
We find that K is closed under the usual matrix addition and multiplication and thus obvi-

ously a ring [the distributive law follows

 from
 the usual matrix situation M at2 (C)] which
 1 0 
is a sub ring of M at2 (C) because 

 = Id2 ε K.

0 1
    
 0 1   0 i  i 0 
(i)We write J := 

,K = 
 
,

 = I and note

−1 0 i 0 0 −i
    
 0 1  0 i   i 0 
JK = 



=
 
 = I,

−1 0 i 0 0 −i
    
 0 i   0 1   −i 0 
KJ =   =  = −I,
    
i 0 −1 0 0 i

c Copyright Reserved IIT Delhi

      9

 0 i  i 0   0 1 
KI = 



=
 
=J

i 0 0 −i −1 0
    
 i 0   0 i   0 −1 
IK =   =  = −J
    
0 −i i 0 1 0
    
 i 0  0 1   0 i 
IJ = 



=
 
=K

0 −i −1 0 i 0
    
 0 1   i 0   0 −i 
JI =   =  = −K
    
−1 0 0 −i −i 0
    
 0 1   0 1   −1 0 
J2 = 



=
 
 = −Id2

−1 0 −1 0 0 −1
    
 0 i   0 i   −1 0 
K2 = 



=
 
 = Id2

i 0 i 0 0 −1
    
 i 0   i 0   −1 0 
I2 = 



=
 
 = Id2 If α = a + bi, β = c + di, a, b, c, dεR

0 −i 0 −i 0 −1
we can write a quaternion uniquely as q = aId2 +bI+cJ +dK and conversely a matrix of the

sort aId2 +bI +cJ +dK would stand for the quaternion This can be done by correspondence

e0 ↔ 1d2 , e1 ↔ I, e2 ↔ J, e3 ↔ K The calculation above show that the multiplication

table for quaternions corresponds to the matrix multiplication performed above; since the

addition obviously does the mapping f (e0 ) = Id2 , f (e1 ) = I, f (e2 ) = J, f (e3 ) = K estab-

lishes an isomorphism of the division algebra of the quaternions 

onto the 
matrix algebra

 α 0 
introduced just now. So the two are the same. Further,f (α) = 

 = aId2 + bI

0 α
 
 α 0 
for α = a + biC embeds the field C into H via a 7→ a + bi = α = 



0 α
 
 a 0 
Similarly, with aεR being identified to 

 we can show that the matrices a + bj =

0 a

c Copyright Reserved IIT Delhi

  10

 a −b 
  form a field under multiplication and addition which is isomorphic to C via
 
b a
 
 a −b 
a + bi ↔ 

.

b a

c Copyright Reserved IIT Delhi