Linear Functionals and Dual Spaces (Chapter-3)

Shruti Sharma

October 16, 2016

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

1 / 70

Table of Contents

Linear Functionals
Linear functionals and Subspaces
Annihilators

Double Dual
Hyperspaces

Transpose of a Linear Transformation

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

2 / 70

Linear Functionals

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

3 / 70

Linear Functionals

Definition
If V is a vector space over the field F, a linear transformation f : V F is
called linear functional on V i.e. f is a function from V into F s.t.
f (c + ) = cf () + f ()
for all vectors and in V and all scalars c in F.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

4 / 70

Linear Functionals
Example
Let F be a field and let a1 , ..., an be scalars in F. Define a function f on F n
by:
f (x1 , ..., xn ) = a1 x1 + ... + an xn
Then f is a linear functional because
f (cx1 + y1 , ..., cxn + yn ) = c(a1 x1 + ... + an xn ) + (a1 y1 + ... + an yn )
= cf (x1 , ..., xn ) + f (y1 , ..., yn )
It is the linear dunctional represented by matrix [a1 , ..., an ] relative to the
standard basis for F n and basis {1} for F:
aj = f (j ) j = 1, ..., n
Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

5 / 70

Linear Functionals

Remark
Every linear functional on F n is of this form, for some scalars a1 , ..., an
because:
X


f (x1 , ..., xn ) = f
xj j
j

xj f (j )

aj xj

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

6 / 70

Linear Functionals
Example
Let n be a positive integer and F a field. If A is an n n matrix with
entries in F, the trace of A:
tr (A) = A11 + ... + Ann
Trace function is a linear functional on the matrix space F nn because:
tr (cA + B) =

n
X
(cAii + Bii )
i=1
n
X

=c

i=1

Aii +

n
X

Bii

i=1

= ctr (A) + tr (B)

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

7 / 70

Linear Functionals

Example
Let V be the space of all polynomial functions from the field F into itself.
Let t be an element of F. If we define
Lt (p) = p(t)
then Lt is a linear functional on V.
Evaluation at t is a linear functional on the space of all functions from
F into F.
(cf + g )(t) = cf (t) + g (t)

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

8 / 70

Example
Example
Let [a, b] be a closed interval on the real line and let C ([a, b]) be the space
of continuous real-valued functions [a,b]. Then
Z

g (t)dt

L(g ) =
a

defines a linear functional L on C([a,b]) because:

Z

L(cg + h) =

(cg + h)(t)dt
Z b
Z b
=
cg (t)dt +
h(t)dt
a

= cL(g ) + L(h)
Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

9 / 70

Dual Space
Definition
If V is a vector space, the collection of all linear functionals on V denoted
by L(V,F) forms a vector space in a natural way. This space V is called
dual space of V.
V = L(V , F )
and dim(V )=dim(V)

Recall
Let V be an n-dim vector space over the field F, and let W be an m dim
vector space over F. Then the space L(V,W) is finite dimensional and has
dim mn.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

10 / 70

Dual Space

Let B = {1 , ..., n } be a basis of V. There is a unique linear

functional fi on V for each i s.t.
fi (j ) = ij
and we obtain from B a set of n distinct linear functionals f1 , ..., fn on
V.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

11 / 70

Dual Space
These
Pfunctionals are linearly independent. For, suppose
f = ni=1 ci fi , then
f (j ) =
=

n
X
i=1
n
X

ci fi (j )
ci ij = cj

i=1

If f is the zero functional f (j ) = 0 for each j and hence cj = 0.

f1 , ..., fn are n linearly independent functionals and since V has
dimension n, so it must be that B = {f1 , ..., fn } is a basis for V
which is called Dual Space.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

12 / 70

Dual Space
Theorem
Let V be a finite dimensional vector space over the field F, and let
B = {1 , ..., n } be a basis for V. Then there is a unique dual basis
B = {f1 , ..., fn } for V s.t. fi (j ) = ij . For eacch linear functional f on V
we have
f =

n
X

f (i )fi

i=1

and for each vector in V we have:

=

n
X

fi ()i

i=1

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

13 / 70

Dual Space
Proof.
If =

Pn

i=1 xi i

is a vector in V, then
fj () =
=

n
X
i=1
n
X

fj (i )xi
xi ij = xj

i=1

Suppose f =

Pn

i=1 ci fi ,

then

f (j ) =

n
X

ci fi (j ) = cj

i=1

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

14 / 70

Dual Space

n
X

fi ()i

i=1

provides a nice way of describing what dual basis is.

If B = {1 , ..., n } is an ordered basis for V and B = {f1 , ..., fn } is
the dual basis, then fi is the function which assigns to each vector
in V the i th coordinate of relative to the ordered basis B.
fi can also be seen as the coordinate functions for B.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

15 / 70

Dual space

If f V and we let f (i ) = ai then when

= x1 1 + ... + xn n
we have
f () = x1 a1 + ... + xn an

(1)

If ordered basis B for V and every vector (here ) in V is described as

n-tuple of coordinates (x1 , ..., xn ) relative to B, then every linear
functional on V has the form eq.1.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

16 / 70

Dual Space

Example
Let V be the vector space of all polynomial functions from R into R which
have degree less than or equal to 2. Let t1 , t2 , t3 be any distinct real no.
and let
Li (p) = p(ti )
Then L1 , L2 , L3 are linear functionals on V. These functionals are linearly
independent; for suppose
L = c1 L1 + c2 L2 + c3 L3

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

17 / 70

Dual Space
Example (Contd.)
If L=0, i.e. if L(p)=0 for each p V , then applying L to the
particular polynomial functions 1, x, x 2 , we obtain
c1 + c2 + c3 = 0
t1 c1 + t2 c2 + t3 c3 = 0
t12 c1 + t22 c2 + t32 c3 = 0
It follows that c1 = c2 = c3 = 0 as the matrix

1 1 1
t1 t2 t3
t12 t22 t32
is invertible when t1 , t2 , t3 are distinct.
Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

18 / 70

Dual Space

Example (Contd.)
Now, Li are independent and since V has dimension 3, these
functionals form a basis for V . What is the basis for V of which this
is the dual?
Such basis {p1 , p2 , p3 } must satisfy:
Li (pj ) = pj (ti ) = ij

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

19 / 70

Dual Space
Example (Contd.)
These polynomials can be seen to be:
(x t2 )(x t3 )
(t1 t2 )(t1 t3 )
(x t1 )(x t3 )
p2 (x) =
(t2 t1 )(t2 t3 )
(x t1 )(x t2 )
p3 (x) =
(t3 t1 )(t3 t2 )
p1 (x) =

The basis for V is interesting as we can express every p in V as:

p = p(t1 )p1 + p(t2 )p2 + p(t3 )p3

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

20 / 70

Dual Space

Example (Contd.)
Thus, if c1 , c2 , c3 are any real no., there is exactly one polynomial
function p over R which has degree at most 2 and satisfies
p(tj ) = cj j = 1, 2, 3
This polynomial function is
p = c1 p1 + c2 p2 + c3 p3

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

21 / 70

Linear functionals and Subspaces

If f is non-zero linear functional:

Rank(f ) = 1 (as the range of f is non zero subspace of the scalar field
of dimension 1).
Nullity (f ) = dim(V ) 1 ( from rank-nullity theorem).

In a vector space of dimension n, a subspace of dimension n-1 is

called a hyperspace (or hyperplanes or subspaces of co-dimension 1).
Every hyperspace is the null space of a linear functional.

Remark
Each d-dimensional subspace of n-dim space is the intersection of the null
spaces of (n-d) linear functionals.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

22 / 70

Annihilators

Definition
If V is a vector space over the field F and S is a subset of V, the
annihilator of S is the set S 0 of linear functionals f on V s.t. f () = 0 for
every S. i.e.
S 0 = {f : f () = 0, S V , f V }

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

23 / 70

Annihilators

Remark
S 0 is a subspace of V whether S is a subspace of V or not. For
suppose: Let f1 , f2 be distinct annihilators on space V s.t.
f1 () = f2 () = 0, then
(c1 f1 + c2 f2 )() = c1 f1 () + c2 f2 () = 0
If S = {0}, then S 0 = V .
If S = V , then S 0 = {0}.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

24 / 70

Annihilators

Theorem
Let V be a finite dim vector space over the field F, and let W be a
subspace of V. Then
dim(W ) + dim(W 0 ) = dim(V )

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

25 / 70

Annihilators

Proof:
Let dim(W)=k and 1 , ..., k be a basis for W.
Choose vectors k+1 , ..., n in V s.t. 1 , ..., n is a basis for V.
Let {f1 , ..., fn } be the basis for V which is dual basis for V. Claim is
k+1 , ..., n is a basis for annihilator W 0 .
Certainly fi W 0 for i k + 1 and j k as:
fi (j ) = 0
It follows fi () = 0 for i k + 1 whenever is a linear combination of
1 , ..., k

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

26 / 70

Annihilators
The functionals fk+1 , ..., fn are independent, and claim is they span
W 0:
Suppose f V . Now,
f =

n
X

f (i )fi

i=1

so that if f W 0 we have f (i ) = 0 for i k and

f =

n
X

f (i )fi

i=k+1

Thus, if dim(W)=k and dim(V)=n then dim(W 0 )=n-k. Q.E.D.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

27 / 70

Annihilators

Corollary (1)
If W is a k-dim subspace of an n dim vector space V, then W is the
intersection of (n-k) hyperspaces in V.

Proof.
W is exactly the set of vectors s.t. fi () = 0, i = k + 1, ..., n.
In case k=n-1, W is the null space of fn .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

28 / 70

Annihilators
Corollary (2)
If W1 and W2 are subspaces of a finite dim vector space then W1 = W2 iff
W10 = W20 .

Proof.
If W1 = W2 , then of course W10 = W20 .
If W1 6= W2 , then one of the two subspaces contains a vector which is
not in the other.
Let W2 but
/ W1 .
There is a linear functional f s.t. f () = 0 for all W1 but f () 6= 0.
Then f W10 but not in W20 and hence W10 6= W20 .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

29 / 70

Annihilators

Corollary 1 says if we select some ordered basis for the space, each
k-dim subspace can be described by specifying (n-k) homogeneous
linear conditions on the coordinates relative to that basis.
From the point of linear functionals, suppose we have a system of
linear equations:
A11 x1 + . . . + A1n xn = 0
..
.
Am1 x1 + . . . + Amn xn = 0
for which we wish to find the solutions.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

30 / 70

Annihilators

If we let fi , i = 1, ..., m be the linear functional on F n defined by:

fi (x1 , ..., xn ) = Ai1 x1 + . . . + Ain xn
then we are seeking the subspace of F n for all s.t.
fi () = 0 i = 1, ..., m
In other words, we are seeking the subspace annihilated by f1 , ..., fm .
Row-reduction of the coefficient matrix provides us with a systematic
method of finding this subspace.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

31 / 70

Annihilators

The n-tuple (Ai1 , ..., Ain ) gives the coordinates of the linear functional
fi relative to the basis which is dual to the standard basis for F n .
Row space of the coefficient matrix may also be regarded as the space
of linear functionals spanned by f1 , ..., fm .
Solution space is the subspace annihilated by this space of functionals.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

32 / 70

Dual point of view

Suppose we are given m vectors in F n
i = (Ai1 , ..., Ain ) i = 1, ..., m
and we wish to find the annihilator of the subspace spanned by these
vectors.
A typical linear functional on F n has the form:
fi (x1 , ..., xn ) = c1 x1 + . . . + cn xn
The condition that f is an annihilator is that:
n
X

Aij cj = 0 i = 1, ..., m

j=1

i.e. (c1 , ..., cn ) is a solution of the system AX=0.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

33 / 70

Double Dual

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

34 / 70

Double Dual

Dual space of V V .
If V , then induces a linear functional L on V defined by:
L (f ) = f () f V
L is linear as:
L (cf + g ) = (cf + g )()
= (cf )() + g ()
= cf () + g ()
= cL (f ) + L (g )

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

35 / 70

Double Dual

If V is finite dim and 6= 0, then L 6= 0, in other words, there exists

a linear functional f s.t. f () 6= 0.
Choose an ordered basis B = {1 , ..., n } for V s.t. 1 = and let f
be the linear functional which assigns to each vector in V its first
coordinate in the ordered basis B.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

36 / 70

Double Dual

Theorem
Let V be a finite dim vector space over the field F. For each vector V
define:
L (f ) = f () f V
The mapping 7 L is then an isomorphism of V onto V .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

37 / 70

Double Dual
Proof.
For each , the function L is linear. Suppose and are in V and c
is in F, and let = c + . Then for each f in V :
L (f ) = f ()
= f (c + )
= cf () + f ()
= cL (f ) + L (f )
This shows that the mapping 7 L is a linear transformation from
V to V .
Transformation is non-singular; for L = 0 iff = 0.
dim(V)=dim(V )=dim(V) which means transformation is onto.

Transformation is invertible and thus an isomorphism of V onto V .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

38 / 70

Double Dual

Corollary
Let V be a finite dim vector space over the field F. If L is a linear functional
on the dual space V of V, then there is a unique vector in V s.t.
L(f ) = f ()
for every f in V .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

39 / 70

Double Dual
Corollary
Let V be a finite dim vector space over the field F. Each basis for V is
the dual of some basis for V.

Proof.
Let B = {f1 , ..., fn } be a basis for V . There is a basis {L1 , ..., Ln }
for V s.t. Li (fj ) = ij .
So, for each i there is a vector i in V s.t.
Li (f ) = f (i )
for every f in V , i.e. s.t. Li = Li .
It follows that {1 , ..., n } is a basis for V and that B is the dual of
this basis.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

40 / 70

Double Dual

If E is a subset of V , then annihilator E 0 is a subset of V .

If we choose to identify V and V as in prev theorem, then E 0 is a
subspace of V, i.e. L (f ) = f () = 0 the set of all in V s.t.
f () = 0 for all f in E.
Each subspace W is determined by its annihilator W 0 i.e. W 0 is the
set of linear functionals f s.t. f (W ) = 0. How???
W is the subspace annihilated by all f in W 0 , i.e. the intersection of
the null spaces of all fs in W 0 .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

41 / 70

Double Dual
Theorem
If S is any subset of a finite dim vector space V, then (S 0 )0 is the
subspace spanned by S.

Proof.
Let W be the subspace spanned by S. Clearly, W 0 = S 0 . Claim is
W = W 00 .
Since
dim(W ) + dim(W 0 ) = dim(V )
dim(W 0 ) + dim(W 00 ) = dim(V )
Since, dim(V ) = dim(V ) we get dim(W ) = dim(W 00 ). Since, W is
a subspace of W 00 , we see W = W 00 .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

42 / 70

Hyperspaces

Definition
If V is a vector space, a hyperspace in V is a maximal proper subspace of
V.

Theorem
If f is a non-zero linear functional on the vector space V, then the null
space of f is a hyperspace in V. Conversely, every hyperspace in V is the
null space of a (not unique) non-zero functional on V.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

43 / 70

Hyperspaces
Proof: Let f 6= 0 V and Nf is its null space. Let V which is not in
Nf i.e. a vector s.t. f () 6= 0.
Claim is that every vector in V is in the subspace spanned by Nf
and i.e. = + c Nf , c F .
Let be in V. Define
c=

f ()
f ()

which makes sense as f () 6= 0.

Then vector = c is in Nf as:
f () = f ( c) = f () cf () = 0
So, is in the subspace spanned by Nf and .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

44 / 70

Hyperspaces

Let N be a hyperspace in V.
Fix some vector which is not in N.
Since N is maximal proper subspace, the subspace spanned by N and
is the entire space V.
So each vector in V has the form = + c where N and
c F . Here, vector and scalar c is uniquely determined by . For if
suppose = 0 + c 0 then (c 0 c) = 0 .
If c 0 c 6= 0 then would be in N; hence, c 0 = c and 0 = .

Q.E.D.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

45 / 70

Hyperspaces
Remark
If V , there is a unique scalar c s.t. c is in N. Call scalar g ().
Then g is a linear functional on V and N is the null space of g. i.e.
f ( g ()) = 0
f ()
g () =
f ()
For pa + b V and p scalar in F,
f (pa + b
f ()
pf (a) + f (b)
=
= pg (a) + g (b)
f ()

g (pa + b) =

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

46 / 70

Hyperspaces

Lemma
If f and g are linear functionals on a vector space V, then g is a scalar
multiple of f i.e. g=cf iff the null space of g contains the null space of f,
i.e. iff f () = 0 implies g () = 0. Equivalently, if g=cf then Nf Ng .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

47 / 70

Hyperspaces
Proof.
If f=0 then g=0 and thus g is trivially a scalar multiple of f.
Suppose f 6= 0, so that null space Nf is a hyperspace in V. Choosing
V with f () 6= 0 and let
c=

g ()
f ()

Linear functional h = g cf is 0 on Nf , since both f and g are 0

there and h() = g () cf () = 0.
Thus, h is 0 on the subspace spanned by Nf and and that subspace
is V. So, h=0 i.e. g = cf .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

48 / 70

Hyperspaces
Lemma
If f and g are linear functionals on a vector space V s.t. null space of f is
equal to null space of g then g is a scalar multiple of f i.e. g=cf.

Proof.
Let N be null space and be vector outside H, then span(N,)=V.
All x V can be written as x = n + t with n N, then f () and
g () uniquely determine function f and :
f (x)
f (n + t)
f ()
=
=
=c
g (x)
g (n + t)
g ()
Hence, f=cg.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

49 / 70

Hyperspaces
Theorem
Let g, f1 , ..., fr be linear functionals on a vector space V with respective
null space N, N1 , ..., Nr . Then g is a linear combination of f1 , ..., fr iff N
contains the intersection N1 ... Nr .
Proof:
If g = c1 f1 + ... + cr fr and fi () = 0 for each i, then clearly g () = 0.
So, N contains N1 ... Nr .
Converse is proved by induction. From prev lemma, r=1 case is
handled i.e. if g and f are linear functionals with null space Ng and
Nf , then g is scalar multiple of f iff Ng contains Nf .
Suppose we know the result is true for r=k-1, and let f1 , ..., fk be
linear functionals with null spaces N1 , ..., Nk s.t. N1 ... Nk is
contained in N.
Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

50 / 70

Hyperspaces
0
0
Let g 0 , f10 , ..., fk1
to the subspace Nk . Then g 0 , f10 , ..., fk1
are linear
functionals on the vector space Nk .
If Nk and fi 0 () = 0 i=1,...k-1, then N1 ... Nk and so
g 0 () = 0.
By induction, there are scalars ci s.t.
0
g 0 = c1 f10 + ... + ck1 fk1

Now let
h=g

k1
X

ci fi

i=1

Then h is a linear functional on V and thus h() = 0 Nk .

By prev. lemma, h is a scalar multiple of fk . If h = ck fk , then
g=

k
X

ci fi

i=1

Q.E.D.
Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

51 / 70

Transpose of a Linear Transformation

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

52 / 70

Transpose of a Linear Transformation

Suppose we have two vector spaces over the field F, V and W and a
linear transformation T : V W . Then T induces a linear
transformation from W into V as follows:
Suppose g is a linear functional on W and let
f () = g (T )
for each V . (T W ).
So, this eq. defines a function from V into F, namely the composition
of T (a function from V to W) with g (a function from W to F).
Since, T and g are linear so is f. (Recall composition of two linear
functions is linear).

So, T provides us a rule T t which associates with each linear

functional g on W a linear functional f = T t g on V.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

53 / 70

Transpose of a Linear Transformation

T t is a linear transformation from W into V ; for if g1 and g2 are in

W and c is a scalar:
[T t (cg1 + g2 )]() = (cg1 + g2 )(T )
= cg1 (T ) + g2 (T )
= c(T t g1 )() + (T t g2 )()

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

54 / 70

Transpose of a Linear Transformation

Theorem
Let V and W be vector spaces over the field F. For each linear
transformation T : V W , there is a unique linear transformation
T t : W V s.t.
(T t g )() = g (T )
for every g in W and in V. T t is called Transpose or adjoint of T.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

55 / 70

Transpose of a Linear Transformation

Theorem
Let V and W be vector spaces over the field F and let T is linear
transformation T : V W . The null space of T t is the annihilator of the
range of T. If V and W are finite dimensional, then
(1) rank(T t )=rank(T)
(2) the range of T t is the annihilator of the null space of T.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

56 / 70

Transpose of a Linear Transformation

Proof: Suppose dim(V)=n and dim(W)=m.
If g is in W , then
(T t g )() = g (T )
for each V .
Null space of T t means g (T ) = 0 for every in V which is precisely
annihilator of range of T.

(1) Let r be the rank of T. Dimension of range of annihilator of range of

T is m-r (From slide 25).
By the first statement of this theorem, nullity of T t must be (m-r).
But since, T t is linear transformation, from rank nullity theorem, rank
of T t is m-(m-r)=r.
So, T and T t have same rank.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

57 / 70

Transpose of Linear Transformation

(2) Let N be the null space of T. Every functional in the range of T t is

the annihilator of N; for:
Suppose f = T t g for some g W ; then if N:
f () = (T t g )() = g (T ) = g (0) = 0
Now range of T t is a subspace of the space N 0 and
dim(N 0 ) = n dim(N) = rank(T ) = rank(T t )
so that the range of T t must be exactly N 0 .

Q.E.D.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

58 / 70

Representation of transpose of linear transformation by

matrix

Theorem
Let V and W be finite dim vector spaces over the field F. Let B be an
ordered basis for V with dual basis B , and let B 0 be an ordered basis for
W with dual basis B 0 . Let T be a linear transformation T : V W ; let
A be the matrix of T relative to B and B 0 and let B be the matrix of T t
relative to B 0 , B. Then Bij = Aji .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

59 / 70

Representation of transpose of linear transformation by

matrix
Proof: Let
B = {1 , ..., n }
B = {f1 , ..., fn }

B 0 = {1 , ..., m }
B 0 = {g1 , ..., gm }

By definition
T j =
T t gj =

m
X
i=1
n
X

Aij i j = 1, ..., n
Bij fi j = 1, ..., m

i=1

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

60 / 70

Representation of transpose of linear transformation by

matrix
On the other hand,
(T t gj )(i ) = gj (T i )
m
X
= gj (
Aki k )
k=1

=
=

m
X
k=1
m
X

Aki gj (k )
Aki jk

k=1

= Aji

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

61 / 70

Representation of transpose of linear transformation by

matrix

For any linear functional f on V

f =

n
X

f (i )fi

i=1

Since, f = T t gj and using the fact that (T t gj )(i ) = Aji , we have

T t gj =

n
X

Aji fi

i=1

which immediately follows: Bij = Aji . Q.E.D.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

62 / 70

Representation of transpose of linear transformation by

matrix

Definition
If A is an m n matrix over the field , the transpose of A is the m n
matrix At defined by Atij = Aji .

Remark
If T is a linear transformation T : V W , the matrix of which in some
pair of bases is A, then the transpose transformation T t : W V is
represented in the dual pair of bases by the transpose of matrix At .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

63 / 70

Representation of transpose of linear transformation by

matrix
Theorem
Let A be any m n matrix over the field F. Then the row rank of A is
equal to column rank of A.
Proof: Let B be the standard basis for F n and B 0 the standard basis for
F m . Let T : F n F m s.t. the matrix of T relative to the pair B, B 0 is A
i.e.
T (x1 , ..., xn ) = (y1 , ..., ym )
n
X
yi =
Aij xj
j=1

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

64 / 70

Representation of transpose of linear transformation by

matrix

Column rank of A is the rank of transformation T, as the range of T

consists of all m-tuples which are linear combinations of the column
vectors of A.
Relative to the dual basis B 0 , B , the transpose mapping T t is
represented by the matrix At .
Since columns of At are rows of A, we see by the same reasoning that
the row rank of A is equal to the rank of T t .

Since, T and T t have same rank, thus row rank of A is equal to the
column rank of A.
Q.E.D.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

65 / 70

Representation of transpose of linear transformation by

matrix
Example
Let V be an n-dim vector space over the field F, and let T be a linear
operator on space V. Suppose B = {1 , ..., n } is an ordered basis for V.
The matrix of T in the ordered basis B is defined to be the n n
matrix A s.t.
T j =

n
X

Aij i

i=1

i.e. Aij is the i th coordinate of the vector T j in the ordered basis B.

P
If {f1 , ..., fn } is the dual basis of B and since, = ni=1 fi ()i this
can be simply stated as:
Aij = fi (T j )
Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

(2)
October 16, 2016

66 / 70

Representation of transpose of linear transformation by

matrix

Example (Contd.)
What happens when we change basis?
Suppose,
B 0 = {10 , ..., n0 }
is another basis for B with dual {f10 , ..., fn0 }.
If B is the matrix of T in the ordered basis B 0 , then from eq2.
Bij = fi 0 (T j0 )

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

67 / 70

Representation of transpose of linear transformation by

matrix
Example (Contd.)
Let U be the invertible linear operator s.t. Uj = j0 . Then transpose
of U is :
U t fi 0 (j ) = fi 0 (U(j )) = fi 0 (j0 ) = ij
Hence,
U t fi 0 = fi
Since, U is invertible, so is U t and (U t )1 = (U 1 )t . Thus,
fi 0 = (U 1 )t fi , i=1,...,n.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

68 / 70

Representation of transpose of linear transformation by

matrix
Example (Contd.)
Therefore,
Bij = fi 0 (T j0 )
= [(U 1 )t fi ](T j0 )
= fi (U 1 (T j0 ))
= fi (U 1 TUj )
So, fi (U 1 TUj ) is the i,j entry of the matrix of U 1 TU in the
ordered basis B which is also the i,j entry of the matrix of T in the
ordered basis B 0 .

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

69 / 70

Representation of transpose of linear transformation by

matrix

Example (Contd.)
In other words,
[T ]B0 = [U 1 TU]B
= [U 1 ]B [T ]B [U]B
= [U]1
B [T ]B [U]B
which is precisely the change-of-basis formula.

Shruti Sharma

Linear Functionals and Dual Spaces (Chapter-3)

October 16, 2016

70 / 70