Quantum Algebra-1

A. Pathak
0-0
What are we going to discuss?
It is not in the scope of this 50 minutes lecture to provide all the
mathematical knowledge required for quantum computing. But we will
give some basic ideas which are essential for us to proceed further. In
particular we will discuss the following:
Vector space (only introduction and elementary denitions)
Dirac notation
Inner product and outer product
Gram-Schmidt procedure
Eigen values and eigen operators
Normal, unitary and positive operators
Tensor product
Partial trace
1
Vector Space: Hilbert space is a nite dimensional complex vector space
Recall the vectors in conventional three dimensional space.
In a 3 dimensional coordinate space you need 3 numbers to describe
any vector.
These three numbers are essentially projection along three directions
(or three axes) specied by three orthonormal unit vectors.
In three dimensional space we need a set of three orthonormal unit
vectors

i,

j and

k to describe any other vector

P (in terms of these
unit vectors) as

P = x

i + y

j + z

k.
If we generalize this idea and do not restrict ourself to three
dimension and allow the scaler projections (x, y, z etc.) to be
complex then we obtain a vector space in general. Note: All the
quantum computing literature refers to a nite-dimensional complex
vector space by the name Hilbert Space . We will use H to denote
such a space.
2
Large Vector Space
A vector space consists of a set of vectors (

,

,

.......), together with
a set of scalars (a, b, c.........), which is closed under vector addition and
scaler multiplication. Thus a set of objects constitutes a vector space, if
it obeys following rules:
1. Closure: Addition of two objects of the set gives another object of
the same set (addition of two vectors gives a vector). Therefore a
well dened operation will never take you outside the vector space.
2. Has a zero: for every object

V there exist another object

0 such

V +

0 =

V .
3. Scalar Multiplication : If c is a scaler then c

V is also a vector
4. Inverse: For every

V there exists a

V such that

V + (

V ) =

0
5. Associative: (

V +

W) +

X =

V + (

W +

X)
3
The C
n
Vector Space
In the conventional cartesian coordinate system where the unit vectors,

i,

j and

k forms a complete set of
orthonormal bases. Under this bases we can describe an arbitrary vector P as (x, y, z) or as
_
_
_
_
x
y
z
_
_
_
_
. Thus
the elements of the column matrix is essentially the coecients of the bases.
To describe an arbitrary n-dimensional state/vector we need a complete set of n orthonormal bases and
the state can be described by a set of n complex numbers which are the coecients of the n bases. In
general the coecients of the bases can be complex number and the space spanned by the set of n
orthonormal bases is called C
n
space. A very special case of C
n
space is one in which all the coecients
are real and that space is known as R
n
space.
In C
n
vector space you need n complex numbers to describe any vector of this space. In other words any
object of this set is n tuples of complex numbers. For example
a =
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
z1
z2
z3
.
.
.
.
zn
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
is a vector in C
n
space.
4
Therefore we should have
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
z1
z2
z3
.
.
.
.
zn
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
+
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
z

1
z

2
z

3
.
.
.
.
z

n
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
z1 + z

1
z2 + z

2
z3 + z

3
.
.
.
.
zn + z

n
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
,
and
c
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
z1
z2
z3
.
.
.
.
zn
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
cz1
cz2
cz3
.
.
.
.
czn
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
.
For a discrete quantum system with n possible states, we will be interested in C
n
space.
5
Inner Products
An inner product is a generalization of the dot product. In a vector
space, it is a way to multiply two vectors to yield a complex number
(|, | | C
a
that obeys the following rules:
1. ( , ) is linear in its second argument
(|v
k
,
k
a
k
|w
k
) =
k
a
k
(|v
k
, |w
k
) ,
2.
(|v, |w) = (|w, |v)

,
3.
(|v, |v) 0.
a
Actually is bracket now we can divide it into two parts as | . Thus the
bra-ket is divided in two parts and the left part (i.e |) is known as bra and
the right part (i.e. | is known as ket. Thus the bra and ket in Dirac notation is
essentially division of conventional bracket in to two parts.
6
Bases and Linear Independence:
If there exist a set of vectors {|v1, |v2, ..., |vn} such that any vector |v in the space can be written as a
linear combination of the vectors in the set (i.e. |v =

n
j=1
|vj). Then the set is called spanning set.
A spanning set does not always forms a basis but bases forms a spanning set
a
. Actually, a linearly
independent spanning set forms a basis. The condition of linear independence of a set of vectors
{|v1, |v2, ..., |vn} is mathematically stated as
n

j=1
aj|vj = 0 if and only if all aj = 0.
In other words if there does not exist any linear combination of the set of vectors {|v1, |v2, ..., |vn} which
adds to zero nontrivially then it forms a set of linearly independent vectors.
The elements of the basis set are linearly independent of each other so they satisfy the following
condition of orthogonality
b
vi|vj = 0 for all i = j.
a
A set of vectors that spans the space is also called complete.
b
In general any two vectors |v and |w are orthogonal to each other if they satisfy v|w = 0.
7
Qubit: A vector in 2 dimensional space
If we have a vector in a particular direction we can easily obtain an unit vector in the same direction by
dividing the vector by its norm. Since in the direction of|v.
unit vector =
|v
_
v|v
, (1)
where
normof |v = ||v| =
_
v|v. (2)
It is usual practice to use a basis set whose elements are of unit magnitude and orthogonal to each other.
Such a basis set is known as orthonormal basis set and the elements of the set satisfy
vi|vj = ij,
where ij is Kronecker delta function:
In two dimensional quantum space we use a basis {|0 and|1} to describe an arbitrary state (qubit)
| = |0 + |1 and |0 and |1 satisfy
0|0 = 1|1 = 1 and 0|1 = 1|0 = 0. (3)
Thus {|0, |1} forms an orthonormal basis in 2 dimensional quantum vector space.
Problem:Assume that (4) is satised and then show that the
_
|00+|11

2
,
|00|11

2
,
|01+|10

2
,
|01|10

2
_
forms a
complete set of orthonormal basis in C
4
space.
8
C
2
space: the space spanned by a single qubit
To describe a two level quantum system or a qubit we need a
two-dimensional complex Hilbert space, which we realize as C
2
- the set
of all column vectors
a =
_
_
_
_
a
1
a
2
_
_
_
_
where a
1
and a
2
are the complex numbers, and with scalar product (a,b)
dened by

(a,

b) = a
1
b
1
+ a
2
b
2
(4)
where the bar stands for the complex conjugation. This particular
Hilbert space is of immense importance because it is the space spanned
by the set of all possible single qubit. In the matrix notation we write a
qubit |as
| = |0 + |1 =
_
_
_
_

_
_
_
_
(5)
9
Outer Products
Let |v be a vector in the vector space V and |w be a vector in the
vector space W. Then the outer product of |w and |v (which is denoted
as |wv| in Dirac notation) is a linear map from V into W dened by
|wv|(|v

) = |wv|v

. (6)
10
Linear Operators
An operator A is said to be linear if and only if it satises the following
properties:
A[f(x) + g(x)] = Af(x) + Ag(x) (7)
and
A[cf(x)] = cAf(x) (8)
where f and g are arbitrary functions and c is an arbitrary constant.
Examples:
d
dx
is a linear operator since,
d
dx
[f(x) + g(x)] =
d
dx
f(x) +
d
dx
g(x)
and
d
dx
[cf(x)] = c
d
dx
f(x).
But ( )
2
is not a linear operators since,
[f(x) + g(x)]
2
= [f(x)]
2
+ [g(x)]
2
.
Similarly, we can show that x
2
,
d
2
dx
2
etc. are linear operators and

, log()
are nonlinear operators. It is an outstanding curiosity to note that the
operators occurring in quantum mechanics are essentially linear.
11
Pauli Matrices
Following set of four useful matrices which acts on a 2-dimensional
vector space (or on single qubit) are known as Pauli matrices:

0
= I =
_

_
1 0
0 1
_

_
, (9)

1
=
x
= X =
_

_
0 1
1 0
_

_
, (10)

2
=
y
= Y =
_

_
0 i
i 0
_

_
(11)
and

3
=
z
= Z =
_

_
1 0
0 1
_

_
. (12)
The identity matrix is not always included in the set of Pauli
matrices but we have included it to make a complete set of
orthogonal basis. Thus any 2 2 complex Hermitian matrices can be
expressed in terms of the the Pauli matrices.
12
Properties of Pauli matrices
All single-qubit quantum gates are 2 2 unitary matrices. The Pauli matrices are some of the most
important single-qubit operations, for example, 1 represents the NOT gate. As this matrices are very
useful in quantum computing here we will note some of the properties of these matrices.
For i, j = 1, 2, 3 Pauli matrices satisfy

2
i
= I, (13)
ij = ji for i = j, (14)
det(1) = 1, (15)
Tr(i) = 0 (16)
and
[i, j] = i
ijk

k
(17)
a
where
ijk
is Levi-Civita symbol
b
a
[A,B]=AB-BA denes commutation of A and B. If [A,B]=0, then we say A and B commutes.
b
The value of Levi-Civita symbol is 1 if (i, j, k) is an even permutation of (1,2,3), -1 for odd permutations and 0 for all other cases, to
be more precise

ijk
=
_

_
1 if (i, j, k) is (1, 2, 3) or (2, 3, 1) or (3, 1, 2)
1 if (i, j, k) or (1, 3, 2) or (3, 2, 1) or (2, 1, 3)
0 otherwise
13
Gram-Schmidt procedure:
Suppose we start with a non orthonormal basis (|e
1
, |e
2
, .....|e
n
).
Gram-Schmidt procedure provides us a simple prescription to generate
an orthonormal basis (|e

1
, |e

2
, .....|e

n
) from the non orthonormal basis
(|e
1
, |e
2
, .....|e
n
). The prescription is as follows:
1. Normalize the rst basis vector:
|e

1
=
|e
1

||e
1
||
.
2. Find the projection of the second vector along the rst and subtract
it o:
|e
2
e

1
|e
2
|e

1
.
This is orthogonal to |e

1
. Normalize it to obtain |e

3. Subtract from |e
3
, its projection along |e

1
and |e

2
:
|e
3
e

1
|e
3
|e

1
e

2
|e
3
|e

and normalize the resultant vector to get |e

3
. Continue the process
till we obtain |e

n
.
14
Example
Let us construct an orthonormal basis from the non orthonormal set {|e1, |e2:|e1 =
_
2
1
_
and |e2 =
_
1
3
_
}
Step 1: Normalize the rst vector
|e

1
=
_
2
1
_
_
_
2 1
_
_
2
1
_
=
1

5
_
2
1
_
.
Step 2.1: construct the second vector orthogonal to the normalized rst vector
|e2 e

1
|e2|e

1
=
_
1
3
_

_
.4 .2
_
_
1
3
__
2
1
_
=
_
1
3
_

_
2
1
_
=
_
1
2
_
Step 2.2: Normalize the vector constructed in step 2.1
|e

2
=
_
1
2
_
_
_
1 2
_
_
1
2
_
=
1

5
_
1
2
_
.
Thus we have the required orthonormal set of basis as
_
1

5
_
2
1
_
,
1

5
_
1
2
__
.
Problem: Use Gram-Schmidt procedure to orthonormalize the basis |e1 = (1 i)

i + 2

j +i

k , |e2 = (2i)

i + 3

j + 2

k
and |e3 = (1 i)

i +

j + (1 + i)

k.
15
Eigenvalues and Eigenvectors
If an operator is operated on a vector (state) and yield a scaler
multiplied by the vector (state), then the equation is called the eigen
value equation. The operator is called the eigen operator, the vector
(state) is called eigen state (eigen vector) and the scaler is called eigen
value. Therefore, if we have
A|v = |v, (18)
then A is the eigen operator, |v is the eigen vector (eigen state) and
is the eigen value. For example, if we consider
d
dx
as the eigen operator
then exp(nx) is a valid eigen state having eigen value n but sin(x) is not
an eigen function of the operator
d
dx
. Problem:
1. Find out the eigenvalues of the Pauli matrices [hints: nd roots to
the equation (
i

i
= 0)].
16
Hermitian Operators
To dene Hermitian operator rst we have to dene adjoint operators.
A

is the adjoint operator of the operator A if A and A

satises,
(A

|v, |w) = (|v, A|w) (19)

= A, A

= (A

)
T
, (AB)

= B

. (20)
Now we can loosely dene Hermitian operator as an operator which
satises,
A = A

. (21)
This is essentially the denition of a self adjoint operator. In a strict
sense all the self adjoint operators dened by (21) is Hermitian but all
Hermitian operators are not self adjoint. But the Hermitian operators
which are not self adjoint are not very important in the context of
quantum computing.
17
Normal, unitary and positive operators:
Following denitions are also important for the understanding of the
text:
1. Normal operator:any operator that satises A

A = AA

is known as
normal operator
2. Unitary operator: If the operator A satises A

A = AA

= I, then A is
called unitary. So it is easy to note that the unitary operators are
essentially normal and an unitary operator A must satisfy A
1
= A

.
3. Positive Operators: operator B is called positive operator if it
satises, (|v, B|v) 0 for all |v in the vector space V . That means a
positive operator does not have any negative eigen value. But a
positive operator can have eigen value 0. Now if we exclude this
possibility and demand that all the eigenvalues B are positive then
the B is called positive denite operator. Thus a positive denite
operator B satises, (|v, B|v) > 0 for all |v in the vector space V .
18
Tensor Products:
A tensor product space is a larger vector space formed from two smaller ones simply by combining elements
from each in all possible ways that preserve both linearity and scalar multiplication. If V is a vector space of
dimension n and W is a vector space of dimension m. Then V W is a vector space of dimension nm. The idea
of tensor product can be claried from the following examples: Consider Pauli matrices X =
_
0 1
1 0
_
and
Y =
_
0 i
i 0
_
tensor product of these matrices is
X Y =
_
0.Y 1.Y
1.Y 0.Y
_
=
_

_
0 0 0 i
0 0 i 0
0 i 0 0
i 0 0 0
_

_
.
Similarly we have
|0 |1 = |01 =
_
1
0
_

_
0
1
_
=
_
_
_
_
_
_
0
1
0
0
_
_
_
_
_
_
.
The fact that tensor product preserves linearity and scalar multiplication can be mathematically stated as
z(|v |w) = (|v z|w) = (z|v |w) scalar multiplication
and
|v (|w1 + |w2) = |v |w1 + |v |w2)
(|w1 + |w2) |v = |w1 |v + |w2 |v
_
linearity.
19
Trace and Partial trace:
The trace
a
of a n n square matrix A is dened as the sum of the elements on the main diagonal. Thus
Tr(A) =
n

j
Ajj. (22)
Example:
The trace operation has some interesting properties, let us list few of them here:
1.
Tr(A
T
) = TrA.
Since the diagonal elements of a square matrix does not change on transposition therefore
2.
Tr(A) = Tr(AB) = Tr(BA)
3. The trace is a linear map i.e
Tr(A + B) = Tr(A) + Tr(B)
and
Tr(rA) = r Tr(A).
a
The use of the term trace arises from a German word spur which is anonymous to English word spoor.
20
Partial Trace
Partial trace is a generalized version of trace. It is easy to understand if
we consider state |
AB
H
A

H
B
. Even if the states are entangled the
state of the rst qubit can in general be described by a density operator

A
on H
A
. Popularly,
A
is called reduced density operator. The
mathematical operation that calculates the reduced density operator is
the partial trace. The reduced density operator
A
can be dened in
terms of the density operator of the composite system as

A
Tr
B
(
AB
), (23)
where Tr
B
is the partial trace over system B, which is dened as
Tr
B
(|a
1
a
2
| |b
1
b
2
|) |a
1
a
2
|Tr (|b
1
b
2
|) . (24)
Using the cyclic property of the trace
Tr (|b
1
b
2
|) = Tr (b
2
|b
1
) = b
2
|b
1
(25)
we can simplify (24) as
Tr
B
(|a
1
a
2
| |b
1
b
2
|) |a
1
a
2
|b
2
|b
1
. (26)
This operation (Tr
B
) is often called tracing out system B.
21
Example
Trace out the second qubit of the two qubit entangled state
| =
1

2
(|00 +|11) .The density matrix for this state is
= || =
1
2
(|0000| +|0011| +|1100| +|1111|) . (27)
Now we can compute the partial trace as

A
= Tr
B
()
=
1
2
Tr
B
(|0000| +|0011| +|1100| +|1111|)
=
1
2
(|00|Tr(|00|) +|01|Tr(|01|) +|10|Tr(|10|) +|11|Tr(|11|))
=
1
2
(|00|0|0 +|01|1|0 +|10|0|1 +|11|1|1)
=
1
2
(|00| +|11|) .
(28)
Problem: Trace out the rst qubit of all four Bell states.
22
What happens in quantum mechanics?
In quantum mechanics there exists an operator corresponding to each dynamical observable of classical
physics. Examples: Momentum and energy are classical observable, in quantum mechanics ih
d
dx
represents
momentum operator pxalong X direction, ih
d
dt
represents energy.
These operators may be represented by matrices and these operators satisfy eigen value equations of the form
Aop| = | (29)
where, Aop is the eigen operator, | is the eigen-state and is the eigen value. As it is expected the eigen
values are discrete and this means all values of the observable are not possible. The meaning of the word
quantum is discrete and this discrete nature of the quantum mechanics intrinsically lies in the inherent
operator algebra.
Again as a measurement of the observable can only yield one of the eigen values of the operator that
represents the particular observable and we cannot measure anything imaginary, so the eigenvalues of
the meaningful physical operators will be real. And this requirements make it essential that the
quantum mechanical operators has to be Hermitian. Other physical conditions makes it unitary too.
Therefore, most of the time we will deal with Hermitian unitary operators.
23