Lecture Notes

Quantum Computing and Quantum Information:

ES 643
Indian Institute of Technology Gandhinagar, 2018

R.R.Puri
Contents

1 Density Operator or Density Matrix 5

1.1 Mixed States . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Properties of Density Matrix . . . . . . . . . . . . . . . . . 7
1.2.1 To Determine whether an Operator is a Density Matrix 11
1.2.2 To Determine whether the Density Matrix Repre-
sents a Pure State . . . . . . . . . . . . . . . . . . . 11
1.3 Density Matrix of a Qubit . . . . . . . . . . . . . . . . . . . 11

3
4 CONTENTS
Chapter 1

Density Operator or Density

Matrix

1.1 Mixed States

In order to understand the concept of density matrix, we introduce first the
concept of a mixed state. To that end, consider the following ensembles of
qubits: (i) ensemble E0 consisting of qubits in the state |0i, (ii) ensemble E1
consisting of qubits in the state |1i,
√ (iii) ensemble Ex consisting of qubits in
the eigenstate |x, 0i = (|0i + |1i)/ 2 of σ̂x corresponding to the eigenvalue
1, and (iv) ensemble Em consisting of qubits half of which are in the state
|0i and half in the state state |1i.
The question is: whereas the state of a qubit in the ensembles E0 , E1
and Ex is |0i, |1i and |x, 0i respectively, what is the state of a qubit drawn
from the ensemble Em ?
To elaborate, note that the measurement of the z-component of any
qubit in the ensemble E0 would give 1 as the outcome and that of any qubit
in the ensemble E1 would give −1 as the outcome. On the other hand, the
measurement of the z-component of the spin of a qubit drawn from Ex
would give 1 as the outcome in half of the large number of measurements
and −1 in the other half. The same is the case with the qubits drawn from
Em . Can we say that the state of the qubits in Em is the same as the state
|x, 0i of the qubits in Ex ? The answer is NO. For, we know that a qubit
drawn from Em is either in the sate |0i or in |1i and hence the outcome of
measurement on a qubit from Em is “known to the qubit beforehand”. On
the other hand, a qubit drawn from Ex is in the superposition of the two
state i.e. it is “simultaneously” in the two states, |0i and |1i. Only after
measurement does it collapse either to |0i or to |1i. The qubit in this case
“does not know beforehand” the result of the measurement or the state to

5
6 CHAPTER 1. DENSITY OPERATOR OR DENSITY MATRIX
which it would collapse when a measurement of its z-component is made.
As an analogy of the ensembles Ex and Em , consider a collection of
normal fair coins i.e. the coins having head on one side and tail on the
other with equal probability of showing head up and tail up on landing on
the ground after tossing. When these coins are tossed up no coin knows
whether it will land with head up or tail up but the number of coins landing
with head up will be same as the number landing with tail up within, of
course, statistical error margin. In this regard the ensemble of normal fair
coins tossed up (i.e. while in air) is akin to the ensemble Ex . The act of
landing on ground is akin to making measurement of the z-component of
the qubit.
Now, consider an ensemble of coins half of which have head on both the
sides and the other half tail on both the sides. This is akin to the ensemble
Em . Pick up a coin at random from the given ensemble to see whether it
has head or tail on it. Within statistical error, we will pick up coins with
head half the time and the coins with tail the other half of the time. This is
akin to making measurement of the z-component of the qubit drawn from
the ensemble Em .
We thus see that the ensemble Em is very much different from the
ensemble Ex . In order to identify the state of Em , we consider the results
of measurement, not only of the z-component of the qubits drawn from
Em , but those of an arbitrary observable Â. The expectation value of the
observable for measurement on the qubits drawn in the state |0i would be
h0|Â|0i and that on the qubits drawn in the state |1i would be h1|Â|1i.
Since the probability that the qubit drawn is in state |0i is the same as
that when it is drawn in the state |1i, the average value of the observable
for measurement on all the qubits will evidently be
1
hÂi = (h0|Â|0i + h1|Â|1i) = Tr(Â{|0ih0| + |1ih1|})
2
= Tr(Âρ̂), (1.1)
where we have used the identity
Tr(P̂ |ψihφ|) = hφ|P̂ |ψi, (1.2)
and introduced the object
1
ρ̂ = (|0ih0| + |1ih1|), (1.3)
2
called the density matrix. The expression (1.1) shows that, knowing ρ̂, we
can determine the expectation value of the measurement of any observable
on the qubits in the ensemble Em . We have thus shown that the state of
Em is characterized by the density matrix given in (1.3).
1.2. PROPERTIES OF DENSITY MATRIX 7
Since it is a “mixture” of different states, we say that the ensemble Em ,
and hence the density matrix (1.3), describe a “mixed state”.
The considerations above may be generalized to an ensemble consisting
of large number N of systems each known to be in one of the n state |ψk i
(k = 1, 2, . . . , n). Let the number of systems in the state |ψk i be Nk . Hence
the probability that a system drwan from the ensemble would be found in
the state |ψk i is pk = Nk /N . As in the example discussed above, consider
the result of measurement of an arbitrary observable on the systems drawn
from the ensemble and show that the state of the ensemble is described by
the density matrix
n
X n
X
ρ̂ = pk |ψk ihψk |, pk = 1. (1.4)
k=1 k=1

Like the state vector provides a quantum theoretic description of an isolated

system, the density matrix describes the state of a mixture of particles in
different states.
Note that the states |ψk i in (1.4) need not be orthogonal. For example,
we can have a mixture consisting of qubits each in one of the states |0i and
|x, 0i with fractions p0 and px so that the state of the ensemble is described
by

ρ̂ = p0 |0ih0| + px |x, 0ihx, 0|. (1.5)

The states |0i and |x, 0i are not orthogonal. In general we can have a
mixture of qubits each of which is in one of the n states |θk , φk ; 0i (k =
1, 2, . . . , n). Such an ensemble is characterized by the density matrix
n
X
ρ̂ = pk |θk , φk ; 0ihθk , φk ; 0|. (1.6)
k=1

In the following we enumerate some properties of the density operator. For

the sake of convenience we replace ρ̂A → ρ̂ .

1.2 Properties of Density Matrix

By starting with (1.4) as defining the density matrix, we enumerate be-
low some its important properties. We assume that the system lies in
n-dimensional Hilbert space so that each of the |ψk i in (1.4) is a column
with n rows and hence ρ̂ is an n × n matrix. The matrix representing ρ̂
may be constructed by taking the matrix elements of (1.4) in some basis.
For example, if |e1 i, |e2 i, . . . , |en i is some orthonormal basis of the space in
8 CHAPTER 1. DENSITY OPERATOR OR DENSITY MATRIX
which the system lies then (we will indicate the matrix representing ρ̂ also
by ρ̂)
 
he1 |ρ̂|e1 i he1 |ρ̂|e2 i · ·he1 |ρ̂|en i
 he2 |ρ̂|e1 i he2 |ρ̂|e2 i · ·he2 |ρ̂|en i 
 
ρ̂ = 
 · · · · 
 (1.7)
 · · · · 
hen |ρ̂|e1 i hen |ρ̂|e2 i · ·hen |ρ̂|en i
1. Take the hermitian conjugate of (1.4) to get
n
X
ρ̂† = pk |ψk ihψk | = ρ̂. (1.8)
k=1

This shows that ρ̂ is hermitian.

2. The trace of ρ̂ is given by
n
X n
X
Tr(ρ̂) = pk Tr(|ψk ihψk |) = pk = 1. (1.9)
i=k i=k

where we have invoked (1.2) (with P̂ = 1) to write Tr(|ψk ihψk |) =

hψk |ψk i = 1.
3. Take the matrix element of (1.4) in an arbitrary state |ψi to get
n
X n
X
hψ|ρ̂|ψi = pk hψ|ψk ihψk |ψi = pk |hψ|ψk i|2 . (1.10)
k=1 k=1

The last line is due to the property hψ|ψk i = hψk |ψi∗ of the scalar
product. Since the right hand side of the equation above is positive,
it follows that hψ|ρ̂|ψi ≥ 0. Now, a matrix  is called positive, and is
indicated by  ≥ 0, if hψ|Â|ψi ≥ 0 for all |ψi. Hence ρ̂ is a positive
matrix i.e.
ρ̂ ≥ 0. (1.11)

4. Since ρ̂ is hermitian, it admits a complete set of orthonormal vectors

{|λi i} corresponding to real eigenvalues {λi }:
ρ̂|λi i = λi |λi i, i = 1, 2, . . . , n, (1.12)

hλi |λj i = δij . (1.13)

Consequently, the density matrix admits spectral decomposition
n
X
ρ̂ = λi |λi ihλi |. (1.14)
i=1
1.2. PROPERTIES OF DENSITY MATRIX 9
5. On taking the matrix element of (1.11) in the eigenstate |λk i of ρ̂, we
have hλk |ρ̂|λk i ≥ 0. On using (1.12) this implies

λk ≥ 0, k = 1, 2, . . . , n. (1.15)

This shows that the eigenvalues of ρ̂ are positive.

6. On taking the trace of the equation (1.14) and noting that Tr(ρ̂) = 1
we get
n
X
λi = 1. (1.16)
i=1

Since, as shown in (1.15), the λ0i s are non negative, it follows from the
equation above that, in order that their sum be unity, each eigenvalue
must be less than one i.e. the eigenvalues of ρ̂ satisfy the condition

0 ≤ λi ≤ 1, i = 1, 2, . . . , n. (1.17)

7. When the systems A is isolsated then it is described by a state vec-

tor, say, |ψA i and hX̂A i = hψA |X̂A |ψA i = Tr(X̂A |ψA ihψA |). It then
follows that the density matrix of an isolated system in state |ΨA i is
given by

ρ̂ = |ΨA ihΨA |, for isolated system in state |ΨA i. (1.18)

The density matrix expressible in the form above is said to represent

a pure state. The ρ̂ is said to represent a mixed state if it can not be
expressed in the form (1.18). The meaning of the term “mixed state”
is elaborated further in the Section 1.1 below.
8. The Eq.(1.18) shows that ρ̂|ΨA i = |ΨA i. In other words, unity is an
eigenvalue of the density operator for a pure state. Since the sum of
the eigenvalues of the density matrix is unity, it follows that all the
other eigenvalues of the density matrix representing a pure state are
zero.
9. With ρ̂ given by (1.14) we know that ρ̂2 is expressible as
n
X
ρ̂2 = λ2i |λi ihλi |. (1.19)
i=1

Since 0 ≤ λi ≤ 1, we have λ2i ≤ λi and hence the last equation implies

n
X n
X
Tr(ρ̂2 ) = λ2i ≤ λi = 1. (1.20)
i=1 i=1
10 CHAPTER 1. DENSITY OPERATOR OR DENSITY MATRIX
In other words
Tr(ρ̂2 ) ≤ 1. (1.21)
The equality would hold when one of the eigenvalues is unity and all
the other eigenvalues are zero. In general (1.19) implies
n
X n
X
hψ|ρ̂2 |ψi = λ2i |hψ|λi i|2 ≤ λi |hψ|λi i|2 = hψ|ρ̂|ψi, (1.22)
i=1 i=1

where the inequality is due to the fact that λ2i ≤ λi and |hψ|λi i|2 ≥ 0.
The result above implies that hψ|ρ̂ − ρ̂2 |ψi ≥ 0 for all |ψi i.e.
ρ̂2 ≤ ρ̂. (1.23)
The equality would hold when ρ̂ represents a pure state as in that case
it is given by ρ̂ = |ΨA ihΨA | which implies ρ̂2 = |ΨA ihΨA |ΨA ihΨA | =
|ΨA ihΨA | = ρ̂.
10. Let the density matrix of a system at time t = 0 be given by
n
X
ρ̂(0) = ρij |ai ihaj |. (1.24)
i,j=1

Let the system evolve under the hamiltonian Ĥ so that the state |ai i
at time t is given by |ai (t)i. Hence, the density matrix at time t would
be
Xn
ρ̂(t) = ρij |ai (t)ihaj (t)|. (1.25)
i,j=1

The states evolve according to the Schrod̈inger equation

d|ai (t)i dhai (t)|
ih̄ = Ĥ|ai (t)i, ih̄ = −hai (t)|Ĥ. (1.26)
dt dt
From (1.25) it follows that
n    
dρ̂(t) X d|ai (t)i dhaj (t)|
ih̄ = ih̄ ρij haj (t)| + |ai (t)i
dt dt dt
i,j=1
n
X h i
= ρij Ĥ|ai (t)ihaj (t)| − |ai (t)ihaj (t)|Ĥ
i,j=1
h i
= Ĥ, ρ̂(t) . (1.27)

This equation, known as Liouville-von Neumann equation, governs

the time evolution of the density matrix.
1.3. DENSITY MATRIX OF A QUBIT 11
1.2.1 To Determine whether an Operator is a Density Ma-
trix
The expression (1.4) shows that ρ̂ is an operator on the state space of the
system. However, not every operator can qualify to be called a density
matrix. From the discussion above, we say that an operator  is a density
matrix if it possesses the following three properties: (i) Â is hermitian i.e.
† = Â, (ii) Tr(Â) = 1, (iii) the eigenvalues of  lie between zero and one.

1.2.2 To Determine whether the Density Matrix Represents

a Pure State
In order to determine whether the given density matrix ρ̂ represents a pure
state, find ρ̂2 and then find its trace. The given ρ̂ would represent a pure
state if Tr(ρ̂2 ) = 1 else it represents a mixed state.

1.3 Density Matrix of a Qubit

We construct the density matrix of a qubit by finding the general form of
an operator on the state space of a qubit having the properties listed in
1.2.1
The first property that the desired operator must have is that of her-
miticity. Recall that a hermitian operator  acting on the state space of a
qubit has the form

 = a0 I + bx σ̂x + by σ̂y + bz σ̂z , (1.28)

where a0 and the b0i s are real. Next, it should have unit trace. To that end,
note that Tr(I) = 2 and

Tr(σ̂µ ) = 0, µ = x, y, z. (1.29)

Hence Tr(Â) = 2a0 . Consequently, a hermitian operator of unit trace is

1
 = (I + ax σ̂x + ay σ̂y + az σ̂z ), (1.30)
2
where we have defined bµ = aµ /2 for later convenience. The third property
that  should possess is that its eigenvalues must lie between 0 and 1. To
find the eigenvalues of Â, rewrite it in the form
1 1
 = (I + a · σ̂) = (I + |a|n · σ̂) , (1.31)
2 2
where a = ax ex + ay ey + az ez = |a|n where n is the unit vector in the
direction of a. The expression above shows that the eigenstates of  are
12 CHAPTER 1. DENSITY OPERATOR OR DENSITY MATRIX
the same as those of n · σ̂. Now, recall that the eigenstates of n · σ̂ are
|n, 0i and |n, 1i corresponding respectively to the eigenvalues ±1. Hence
1 1
Â|n, 0i = (1 + |a|) |n, 0i, Â|n, 1i = (1 − |a|) |n, 1i. (1.32)
2 2
Thus the eigenvalues of  are (1/2)(1 ± |a|). These eigenvalues would lie
between 0 and 1 if |a| ≤ 1. In other words, the general form of the density
matrix of a qubit is
1
ρ̂ = (I + |a|n · σ̂) , |a| ≤ 1. (1.33)
2
The form of ρ̂ above leads to the following important interpretation of a.
To see that, we find hσ̂x i = Tr(σ̂x ρ̂) by multiplying the equation above by
σ̂x and by taking the trace of the resulting equation to obtain
1
hσ̂x i ≡ Tr(σ̂x ρ̂) = Tr σ̂x + |a|{nx σ̂x2 + ny σ̂x σ̂y + ny σ̂x σ̂z } .


(1.34)
2
Now, (i) σ̂x2 = I and hence Tr(σ̂x2 ) = 2, (ii) σ̂x σ̂y = iσ̂z and hence Tr(σ̂x σ̂y ) =
iTr(σ̂z ) = 0, (iii) σ̂x σ̂z = −iσ̂y and hence Tr(σ̂x σ̂z ) = −iTr(σ̂y ) = 0, Thus
(1.34) yields

hσ̂x i = nx |a| ≡ ax . (1.35)

In the same manner it can be verified that

hσ̂y i = ny |a| ≡ ay , hσ̂z i = nz |a| ≡ az . (1.36)

We thus see that a = hσ̂i i.e. a is the Bloch vector of the qubit. We can
thus rewrite (1.33) as
1
ρ̂ = (I + hσ̂i · σ̂) , |hσ̂i| ≤ 1. (1.37)
2
(You are expected to remember this equation). The condition
|hσ̂i| ≤ 1 means that the length of the Bloch vector of a qubit in mixed
state is less than unity. The case |hσ̂i| = 1 corresponds to the qubit in pure
state. In explicit form, the condition |hσ̂i| ≤ 1 reads

hσ̂x i2 + hσ̂y i2 + hσ̂z i2 ≤ 1. (1.38)

The Eq.(1.37) shows that the density matrix is characterized by the vector
hσ̂i of length less than unity. We can visualize this vector as the one from
the centre of a ”ball” of unit radius to a point inside it. The vectors from
the centre to the points on the surface of the ball represent pure states.
1.3. DENSITY MATRIX OF A QUBIT 13
The said representation of the density matrix is also referred to as Bloch
sphere representation with the understanding that, it is not only the radial
vectors of the sphere, but all the vectors joining the centre of the sphere
with the points inside it represent the density matrix.
Note that, in the case of a pure state it is only two angular coordinates
(θ, φ) which caracterize the state of the qubit but in the case of a mixed
state, apart from two angles, the length of the Bloch vector is an additional
parameter that characterizes the state of the qubit.
Verify that the density matrix (1.37) may be recast in the form

ρ̂ = p|0ih0| + µ|0ih1| + µ∗ |1ih0| + (1 − p)|1ih1|, (1.39)

where p is real. Show that the condition (1.38) then reads

|µ|2 ≤ p(1 − p). (1.40)