Вы находитесь на странице: 1из 40

Quantum information and computing

lecture 3: Density operators


Jani-Petri Martikainen
Jani-Petri.Martikainen@helsinki.fi
http://www.helsinki.fi/jamartik
Department of Physical Sciences
University of Helsinki
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 1/39
Last weeks exercise 1.1
a) Let
[x) =

1
i
1/2 +i

, [y) =

1 i
1
1 + i

. (1)
Find | x |, x[y), and y[x).
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 2/39
Last weeks exercise 1.1
b) Prove that for general [x) and [y): x[y) = y[x)

Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 3/39


Last weeks exercise 1.1
c) Let
[e
1
) =
1

1
1

, [e
1
) =
1

1
1

(2)
and
[v) =

3
1

c
k
[e
k
). (3)
Find the coefcients c
k
. (Use projection operators
rather than brute force.)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 4/39
Last weeks exercise 1.2
Use the Gram-Schmidt procedure to nd the
orthonormal basis [e
k
) from the (linearly independent)
set of vectors
[a
1
) =

1
2
2

, [a
2
) =

1
1
2

, [a
3
) =

3
0
3

. (4)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 5/39
Last weeks exercise 1.3
a) Suppose 2 2 matrix A has eigenvalues 1, 2 and
corresponding eigenvectors
[e
1
) =
1

1
i

, [e
2
) =
1

1
i

(5)
Find A.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 6/39
Last weeks exercise 1.3
b) Let
A =

5 2 4
2 2 2
4 2 5

(6)
(1) Find the eigenvalues and eigenvectors of A
(2) Find the spectral decomposition of A
(3) Find the inverse of A making use of the spectral
decomposition.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 7/39
Entanglement application:superdense coding
Simple example demonstrating the application of basic
quantum mechanics.
Alice is in possesion of two classical bits of information
and wants to send these bits to Bob. However, she can
only send a single qubit. Is her task possible?
Answer: Yes!
Assume that Alice and Bob share a pair of qubits in an
entangled state
[) =
[00) +[11)

2
(7)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 8/39
Entanglement application:superdense coding
Note: someone else might have prepapared the state
[) and just sending the qubits to Alice and Bob before
hand.
If Alice wishes to transmit 00, she does nothing to her
qubit.
If Alice wishes to transmit 01, she applies phase ip Z to
the her qubit.
If Alice wishes to transmit 10, she applies quantum NOT
gate X to the her qubit.
If Alice wishes to transmit 11, she applies iY to the her
qubit.
iY =
2
4
0 1
1 0
3
5
(8)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 9/39
Entanglement application:superdense coding
The states map according to:
00 : [)
[00) +[11)

2
(9)
01 : [)
[00) [11)

2
(10)
10 : [)
[10) +[01)

2
(11)
11 : [)
[01) [10)

2
(12)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 10/39
Entanglement application:superdense coding
Alice then sends her qubit to Bob
The four states above are an orthonormal basis of the
2-qubit Hilbert space. (known as the Bell basis, Bell
states, and EPR pairs)
Orthogonal states can be distinguished by making an
appropriate quantum measurement
From the measured state Bob can then identify which of
the four alternatives Alice send him.
In some sense, this is delayed communication. The
qubits where entangled and in order for them to get
entangled they must have interacted in the past. The
channel capacity was already waiting as a resource in
the entangled state.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 11/39
Density operator
QM can also be formulated in terms of density operator
or density matrix not just state vector.
Suppose system is at state [
i
) with probability p
i
. We
call p
i
, [
i
) an ensemble of pure states
The density operator is dened through
=

i
p
i
[
i
)
i
[ (13)
Evolution: UU

If the initial state was [


i
) then the probability of the
outcome m is
p(m[i) =
i
[M

m
M
m
[
i
) = Tr(M

m
M
m
[
i
)
i
[) (14)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 12/39
Density operator
For the ensemble the probability of m is
p(m) =

i
p(m[i)p
i
= Tr(M

m
M
m
) (15)
After measurement result m: If initially [
i
) then
[
m
i
) =
M
m
[
i
)

i
[M

m
M
m
[
i
)
(16)
For the ensemble

m
=

i
p(i[m)[
m
i
)
m
i
[ =
M
m
M

m
Tr(M

m
M
m
)
(17)
since p(i[m) = p(m[i)p
i
/p(m)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 13/39
Density operator
Pure state: We know the state exactly and = [)[.
Also, for pure states Tr(
2
) = 1
Otherwise a mixed state. For a mixed state Tr(
2
) < 1
Assume that or record for the result m was lost. We
could now have state
m
with probability p(m), but we
no longer know the value m. Such a system would be
described by a density operator
=

m
p(m)
m
=

m
Tr(M

m
M
m
)
M
m
M

m
Tr(M

m
M
m
)
=

m
M
m
M

m
(18)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 14/39
Density operator: requirements
For an operator to be a density operator:
1. must have a trace equal to one
2. must be a positive operator
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 15/39
Density operator
Postulate 4: If each subsystem is described by a density
matrix
i
the joint state of the total system is

2

n
.
Density operators shine when, a) describing a quantum
system whose state is not known and b) describing
subsystems of a composite quantum system.
Eigenvalues and eigenvectors of the density matrix do
NOT have a special signicance with regard too the
ensemble of quantum states represented by that
density matrix.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 16/39
Density operator
For example, =
3
4
[0)0[ +
1
4
[1)1[ probability of being in
[1) 1/4 and probability of being in [0) 3/4 (???) Not
necessarily!
suppose 1/2 prob. for both
[a) =

3/4[0) +

1/4[1) (19)
and
[b) =

3/4[0)

1/4[1) (20)
The density matrix
= 1/2[a)a[ + 1/2[b)b[ = 3/4[0)0[ + 1/4[1)1[
Different ensembles can give rise to the same density
matrix
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 17/39
Density operator
What class of emsembles gives rise to the same
density matrix?
Vectors (not necessarily normalized) [

i
) generate the
operator

i
)

i
[... connection to usual ensemble
picture of density operators : [

i
) =

p
i
[
i
).
Answer:The sets [

i
) and [

i
) generate the same
density matrix if and only if
[

i
) =

j
u
ij
[

j
), (21)
where u
ij
is a unitary matrix. (Pad which ever set [

i
) or
[

i
) is shorter with additional vectors having probability
0 so that both sets have the same number of elements.)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 18/39
Density operator
Proof: suppose [

i
) =

j
u
ij
[

j
). Then

i
[

i
)

i
[ =

ijk
u
ij
u

ik
[

j
)

k
[ (22)
=

jk

i
u

ki
u
ij

j
)

k
[ =

jk

jk
[

j
)

k
[
=

j
[

j
)

j
[ (23)
so they generate the same operator. Conversely, suppose
A =

i
[

i
)

i
[ =

j
[

j
)

j
[.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 19/39
Density operator
Let A =

k

k
[k)k[ be a decomposition of A into
orthonormal states [k) with
k
> 0. We wish to relate [

i
) to
states [

k) =

k
[k) and similarly [

i
) to states [

k). Let [) be
any vector orthonormal to space spanned by [

k
i
), so
[

k) = 0 for all k. Therefore


0 = [A[) =

i
[

i
)

i
[) =

i
[[

i
)[
2
(24)
and thus [

i
) = 0 for all i. It follows that each [

i
) is a
linear combination [

i
) =

k
c
ik
[

k).
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 20/39
Density operator
Since A =

k
[

k)

k[ =

i
[

i
)

i
[ we see that

k
[

k)

k[ =

kl

i
c
ik
c

il

k)

l[. (25)
Since operators [

k)

l[ are linearly independent

i
c
ik
c

il
=
kl
. This ensures that we may append extra
columns to c to obtain a unitary matrix v such that
[

i
) =

k
v
ik
[

k) where we have appended zero vectors to


the list [

k). Similarly we can nd w, such that


[

i
) =

k
w
ik
[

k). Thus [

i
) =

j
u
ij
[

i
) where u = vw

is
unitary.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 21/39
Reduced density operator
Describe subsystems by using a reduced density operator
Suppose the system is composed of A and B, then the
reduced density operator for system A is

A
Tr
B
(
AB
). (26)
Above the partial trace is dened by
Tr
B
([a
1
)a
2
[ [b
1
)b
2
[) [a
1
)a
2
[Tr([b
1
)b
2
[) (27)
where [a
i
) and [b
i
) are any vectors in the respective
state spaces and Tr([b
1
)b
2
[) = b
2
[b
1
) as usual. Partial
trace must also be linear in its input.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 22/39
Reduced density operator
Reduced density matrix gives the correct measurement statistics (this justies its use
physically)
Take the Bell state (|00 + |11)/

2 (pure state):
=
|0000| + |1100| + |0011| + |1111|
2
(28)
Trace out the second qubit:

1
= Tr
2
() =
Tr
2
(|0000|) + Tr
2
(|1100|) + Tr
2
(|0011|) + Tr
2
(|1111|)
2
=
|00|0|0 + |10|1|0 + |01|0|1 + |11|1|1
2
=
|00| + |11|
2
= I/2
This state is a mixed state even though the composite state was pure! Hallmark of
entanglement.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 23/39
Reduced density operator
Why the partial trace?
Partial trace turns out to be a unique operation which
gives rise to the correct description of observable
quantities for subsystems of composite systems.
Suppose M is an observable on A and we have a
measuring device capable of realizing the
measurement of M...let

M be the corresponding
measurement on the composite system AB
If the system is in state [m)[) with [) some arbitrary
state on B and [m) an eigenstate of M, the measuring
device must give the result m with a probability of 1.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 24/39
Reduced density operator
Therefore, if P
m
is the projector onto the m eigenspace
of the observable M, then the corresponding projector
on

M must be P
m
I
B
...we have

M =

m
mP
m
I
B
= M I
B
(29)
Then let us show that the partial trace procedure gives
the correct measurement statistics for observations on
a subsystem.
Physical consistency requires that any prescription
associating a state
A
to system A, must have the
property that measurement averages be the same as
for the whole system i.e.
Tr(M
A
) = Tr(

M
AB
) = Tr((M I
B
)
AB
) (30)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 25/39
Reduced density operator
This equation is certainly satised if
A
= Tr
B
(
AB
).
In fact, partial trace turns out to be a unique function
having this property.
Let f() be any map of density operators on AB to
density operators on A such that
Tr(Mf(
AB
)) = Tr((M I
B
)
AB
) (31)
Let M
i
be orthonormal basis of operators for the space
of Hermitian operators with respect to Hilbert-Schmidt
inner product (see text book page 76) (X, Y ) = Tr(XY ),
then we can expand f(
AB
) in this basis...
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 26/39
Reduced density operator
...
f(
AB
) =

i
M
i
Tr(M
i
f(
AB
))
=

i
M
i
Tr((M I
B
)
AB
) (32)
Therefore, f is uniquely determined by Eq. (30)
Moreover, the partial trace satises Eq. (30), so it is the
unique function having this property.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 27/39
Schmidt decomposition
Suppose [) is a pure state of the composite system
AB. Then there exists orthormal states [i
A
) and [i
B
)
such that
[) =

i
[i
A
)[i
B
), (33)
where
i
0 are known as Schmidt coefcients and
satisfy

i

2
i
= 1
Consequence: the reduced density matrices

A
=

2
i
[i
A
)i
A
[ (34)

B
=

2
i
[i
B
)i
B
[ (35)
have the same eigenvalues.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 28/39
Schmidt decomposition and Purication
Schmidt number is the number of non-zero
i
and in
some sense it quanties the amount of entanglement
between systems A and B.
a state of the composite system is a product state (and
thus not entangled) if and only if it has a Schmidt
number 1.
Purication: suppose we are given a state
A
of a
quantum system A. We can introduce another system
R and a pure state [AR) there, such that

A
= Tr
R
([AR)AR[)
This mathematical procedure is known as purication
and enables us to associate pure states with mixed
states.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 29/39
EPR and Bell inequality
What is actually the difference between quantum
mechanics and the classical world? What makes
quantum mechanics non-classical?
In QM an unobserved particle does not possses
properties that exist independent of observation. For
example, a qubit does not possess denite properties of
spin in the z-direction, and spin in the x-direction each
of which can be revealed by performing the appropriate
measurement.
In the EPR-paper Einstein, Podolsky and Rosen
proposed a thought experiment which they believed
demonstrated the incompleteness of QM as a theory of
Nature.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 30/39
EPR and Bell inequality
Attempt to nd elements of reality which were not
included in QM. Introduced what they claimed was a
sufcient condition for a physical property to be an
element of reality...namely that it is possible to predict
with certainty the value of that property, immediately
before measurement.
Consider, an entangled pair of qubits (spin singlet)
belonging to Alice and Bob
[01) [10)

2
(36)
Measure spin along

v axis i.e.

v

for both spins


Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 31/39
EPR and Bell inequality
No matter what we choose for

v the two
measurements are always opposite to one another.
That, if rst qubit measurement yields +1 the second
will give 1 and vice versa.
Suppose [a) and [b) are the eigenstates of

then
[0) = [a) + [b), [1) = [a) + [b) (37)
and
[01) [10)

2
= ( )
[ab) [ba)

2
(38)
But is a determinant of a unitary matrix and
thus just equal to a phase factor e
i
As if the 2nd qubit knows the result of the measurement
on the 1st
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 32/39
EPR and Bell inequality
Since Alice can predict the measurement result when
Bobs qubit is measured along

v , that is an element of
reality and should be included in a complete physical
theory.
Standard QM does not include any fundamental
element to represent the value of

for all unit


vectors

v .
EPR hoped for a return to a more classical view:
system can be ascribed properties which exist
independently of measurements performed.
Nature experimentally invalidates EPR, while agrees with
QM
Key: Bells inequality
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 33/39
EPR and Bell inequality
Start with a common sense analysis (ala EPR) and
then proceed with QM analysis...nd a difference...let
nature decide.
Perform a measurement outlined in the gure: Charlie
prepares two particles and sends one to Alice and one
to Bob.
Alice can choose to measure two different things which
are physical properties labelled by P
Q
and P
R
. She
doesnt know in advance what measurement she will
conduct...decides by ipping a coin...measurement
outcome is (for simplicity) 1
Alices particle has a value Q for the property P
Q
. This
value is assumed to be an objective property of Alices
particle, which is merely revealed by the measurement.
(similarly R is the value revealed by the measurement
of P
R
)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 34/39
EPR and Bell inequality
1 particle 1 particle

ALICE
BOB
Q=+1
R=+1
S=+1
R=+1
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 35/39
EPR and Bell inequality
Likewise Bob is capable of measuring values S and T
(1) of properties P
S
and P
T
..also he makes the
chooses the measurement randomly
Timing is such that Alice and Bob measure at the same
time (or at least in causally disconnected manner)...so
Alices measurement cannot disturb Bobs
measurement
Look at QS + RS + RT QT = (Q + R)S + (R Q)T
since R, Q = 1 it follows that either (Q + R)S = 0 or
(R Q)T = 0
In either case QS + RS + RT QT = 2
suppose next that p(q, r, s, t) is the prob. that before
measurements the system is in a state with Q = q,
R = r, S = s, and T = t.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 36/39
EPR and Bell inequality
These probabilities may depend on what Charlie does.
Let E() denote the mean value of a quantity...we have
E(QS + RS + RT QT) =

qrst
p(q, r, s, t)(qs + rs + rt qt)

qrst
p(q, r, s, t) 2
= 2 (39)
and
E (QS + RS + RT QT) =
X
qrst
p(q, r, s, t)qs
+
X
qrst
p(q, r, s, t)rs +
X
qrst
p(q, r, s, t)rt

X
qrst
p(q, r, s, t)qt
= E(QS) + E(RS) + E(RT) E(QT) (40)
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 37/39
EPR and Bell inequality
Combine and you have an example of a Bell inequality
E(QS) + E(RS) + E(RT) E(QT) 2 (41)
By repeating the measurements and comparing their
outcomes Alice and Bob can determine the left hand
side
Now lets put QM back in...Let Charlie prepare a
quantum system of two qubits
[01) [10)

2
(42)
Alice and Bob perform measurements Q = Z
1
,
S = (Z
2
X
2
)/

2, R = X
1
, and T = (Z
2
X
2
)/

2
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 38/39
EPR and Bell inequality
Simple calculation (show this...) shows that
QS) = 1/

2, RS) = 1/

2, RT) = 1/

2, and
QT) = 1/

2 and thus
QS) +RS) +RT) QT) = 2

2 (43)
Violates the inequality and is consistent with
experiments! ...some asumptions going into the
derivation of the Bell inequality must have been wrong
However, if Alice and Bob choose (for example),
Q = Z
1
, S = Z
2
, R = X
1
, and T = X
2
then
E(QS) + E(RS) + E(RT) E(QT) = 2 (44)
and there is no violation.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 39/39
EPR and Bell inequality
Questionable assumptions:
1. Physical properties P
Q
, P
R
,P
S
, and P
T
have denite
values which exist independent of observation
(assumption of realism)
2. Alices measurement does not inuence Bobs
measurement (locality)
World is not locally realistic!
entanglement can be a fundamentally new resource
which goes beyond classical resources...How to exploit
it, is the key question of quantum information and
computing.
Department of Physical Sciences, University of Helsinki http://theory.physics.helsinki.fi/kvanttilaskenta/ p. 40/39

Вам также может понравиться