Escolar Documentos
Profissional Documentos
Cultura Documentos
Lecture 1: 170214
Linear Algebra
Linear space a.k.a (also known as) vector space.
Vectors and Scalars , V
K V V
(, v) 7 v scalar multiplication
V V 7 V
(v1 , v2 ) 7 v1 + v2 adding vectors
axioms :
(v1 + v2 ) = v1 + v2
( + )v = v + v
()v = (v)
zero vector :
1
v+0=v
2
if dim V = dim U < , then V and U are isomorphic.
the main exception: function spaces
(av) = a(v)
3
given V consider
V := F (V ; K ) := {f : V K }
here f is a linear function.
Since K is a linear space.
V is naturally a linear space called the dual space of V .
Example :
Dirac notation : bra h | and ket | i , h | i : scalar
Consider V
the double dual := the set of linear functions on V into K
= the set of K valued linear functions on the set of linear functions of V .
(v) = h | vi
V , vV
given v V there is a unique special element V call it v
v () := (v) = h | vi
essentially one just switches from considering as the function and v as the argument
to viewing v as the function and as its argument.
4
But the situation is different in infinite dimensional spaces.
Example:
let V = F [0, 1] < be the continuous functions on the interval [0,1].
for every function g V one can define an element in V by
R1
g (f ) = 0 g( )f ( )d
thus via g g F (0, 1)
one has a one-to-one map: V V
However V also contains more linear functions which are not of this form.
Example: for any point x one can define a certain linear function (x) V
by (x) (f ) := f (x)
check: (x) (f + g) = f (x) + g(x)
(x) (af ) = af (x)
To deal with such things the physicist Dirac invented a new type of function (not
continuous so not in F [0, 1]) such that
Z 1
(x) (f ) = ( x)f ( )d
0
John von Neumann objected; it took about 20 years for the mathematicians to find
mathematical justification: the theory of distributions.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
V : linear space
K : scalars
L(U , V , ...; W ) := { the set of linear function from U V ... W }
V := L(V ; K ) = L(V ; R) = L(V )
many finite dimensional vector space theorems are not true in general for infinite di-
mensions.
5
We will not worry much about this, and will generally assume finite dimensions without
bothering to always mention it.
********************************************************
M N
& f
f := f
V W
V W
&
K
for A : V W and W
a linear function(A = A) on V , A V
h|Avi = hA |vi adjoint
6
V
%B &A
AB
U W
V
.B -A
(AB) =B A
U W
thus W = L( , W ) = L(W ; K ) = W
Given V
can make V
can make L(V ; V )
= L(V , V ; K )
L(V ; V )
= L(V , V ; K )
L(V ; V )
= L(V , V ; K )
L(V ; V )
= L(V , V ; K )
example: L(V , V ; K )
(v1 , v2 ) is a scalar
(v1 , ) : v2 7 (v1 , v2 )
7
hence (v , ) V
and ( , ) : v1 7 (v1 , ) V
p
Tensors of type q
on V
Tpq V := L(V , V , . . . , V , V , V , . . . , V ; K )
| {z } | {z }
p q
= the set of real valued linear functions on p dual vectors + q vectors. A tensor is just
a multi-linear function, e.g. if
, , V , u, v, V
T(, , , u, v) : (rank 5) real valued, linear in each argument
3
T is a tensor of type 2
T T 32 (V )
Taylor expansions
f = f (0) + ( f )(0)x + 21 ( f )(0)x x + 3!1 ( f )(0)x x x + . . .
f (0) : scalar
0
f (0) : tensor of type 1
0
f (0) : tensor of type 2
1
( f (0))u v
2
= 12 (f )(u, v)