Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 62

general we cannot convert both complex numbers to real.

However, we can
B
use phase to also reduce dof of by one, for a total phase reduction of
two.

The composite state vector has 4 complex parameters but only one
normalization condition and one phase to ignore, for an equivalent of 6 real
parameters; i.e., dof = 6:

are all complex numbers. We can find an equivalent


vector with one real coefficient and three complex coefficients, say

, thus using phase to reduce the dof by one. That is the best
we can do with phase. We cannot force more than one coefficient be real.

Thus, the state space


SAB is richer than just the set of product states. The

difference is called entanglement. An entangled state is a complete


description of the combined system. However, there are degrees of
entanglement.

Exercise 6.4. Using the matrix expressions (3.20) for and the
d
vector expressions for (2.11) and (2.12) show that Alice’s single-spin
states are

, , ,

, , .

Bob’s states are similar:

We can calculate how Alice’s and Bob’s operators act on the product basis
vectors by remembering that Alice’s s operators act only on the first element and
Bob’s act only on the 2nd. Examples are

By way of explanation, the first equation could be written

51
.

All of the spin operator results are tabulated in the Appendix Tables 1-3, below.
TABLE 1. Up-Down Basis

2-Spin Eigenvectors
|uu⟩ |ud⟩ |du⟩ |dd⟩
z |uu⟩ |ud⟩ - |du⟩ - |dd⟩
x |du⟩ |dd⟩ |uu⟩ |ud⟩
y i |du⟩ i |dd⟩ -i |uu⟩ -i |ud⟩
z |uu⟩ - |ud⟩ |du⟩ - |dd⟩
x |ud⟩ |uu⟩ |dd⟩ |du⟩
y i |ud⟩ -i |uu⟩ i |dd⟩ -i |du⟩

TABLE 2. Right-Left Basis TABLE 3. In-Out Basis

2-Spin Eigenvectors 2-Spin Eigenvectors


|rr⟩ |r l⟩ |l r⟩ |ll⟩ |i i⟩ |i o⟩ |o i⟩ |oo⟩
z |lr⟩ |ll ⟩ |r r⟩ |rl⟩ z |oi⟩ |oo⟩ |i i⟩ |i o⟩
x |rr⟩ |r l⟩ - |l r⟩ - |ll⟩ x i |oi⟩ i |oo⟩ - |i i⟩ - |i o⟩
y |i i⟩ |i o⟩ - |o i⟩ - |oo⟩
y -i |lr⟩ -i |ll ⟩ i |r r⟩ i |rl⟩
z |io⟩ |i i⟩ |oo⟩ |o i⟩
z |rl⟩ |r r⟩ |ll⟩ |lr⟩ x i |io⟩ -i |i i⟩ i |oo⟩ -i |o i⟩
x |rr⟩ - |r l⟩ |lr⟩ - |ll⟩ y |i i⟩ - |io⟩ |o i⟩ - |oo⟩
y -i |rl⟩ i |r r⟩ -i |ll⟩ i |rl⟩

Let be a normalized composite state

and L an observable in
SAB . Then . If is the product

observable that represent Alice’s observable


LA , then

L
Since we know that A operates only on Alice’s part of , we will
usually shorten this expression to

52
. (6.09)

For example, means .

Exercise 6.5 asks us to prove that every prediction about Alice’s half of the
system is the same as it would be in the single-state theory, and similarly for
Bob. Thus, in the 2-state product system the following still holds:

(6.11)

This means that the Spin-Polarization Principle still holds for the product system:
there is some state for which the spin is +1. We claim that this principle does not
sing
hold for the entangled state :

Similarly, , and so (6.11) is not true. ✔

means that +1 and -1 are equally likely for each operator.


Thus, the individual outcomes are completely uncertain even though we know
sing
the exact state of vector .

Exercise 6.9. The operator has eigenvectors


sing , T1 , T2 , and T3
with eigenvalues -3, +1, +1, and +1, respectively. That

53
T1 , T2 , and T3
is, +1 is an eigenvalue with degeneracy 3 which is why are
sing
called triplets and a singlet.

Theorem. Let u = x, y, or z and v = x, y, or z. For

(6.12)
That is,

and

Proof: was shown during the proof of Exercise


6.9. We compute just one mixed case here since the others are similar.

sing
Example. Show represents two particles entangled with opposite spins.

Solution. Suppose that Alice and Bob measure in the direction.

54
.

Thus, whenever Alice measures +1 spin then Bob measures -1, and vice-versa

Chapter 7. More on Entanglement

Thus far we have defined the tensor product of two vectors, say

A and B
and , as . For n = 2, have 2

components each, so has 4 components: and .

We write and as arrays but we haven’t yet specified

how to write as an array. Moreover, we have not yet defined the tensor
product of two matrices nor how to express the tensor product as an

array. We do these things now for 2-dimensional systems


SA and SB . The
definitions and notation easily extend to n dimensions.

Definition. Let Alice’s and Bob’s systems


SA and SB have respective bases

and . Then
SAB has composite basis

. Let and

be vectors, and and

matrices expressed in terms of the composite basis. We define


A
the tensor product by its action on Alice’s and Bob’s vectors and
B
:

55
. (7.10)

In terms of matrix multiplication, we know how to express

(I)
and

(II)

but not yet how to express nor . We define these now using a
pattern definition.

Definition. (7.6)

(7.7)

This pattern definition can be extended to any size array, larger or smaller. In
A B
particular, replacing matrix L by vector and matrix M by vector gives

. (7.8)

Exercise 7.3 (below) proves that the matrix pattern definition (7.7) is in
agreement with the tensor definition (7.10). It shows that the matrix expressions
for LHS and RHS of (7.10) are equal.

Exercise 7.3. Show in matrix notation that .

56
Solution. Multiplying the right-hand sides of (7.7) and (7.8) yields

. (a)

A LA B M B
Replacing by and by in the pattern definition (7.8) yields

✔ ■

uu
In a 2-spin system, an operator is represented by a 4x4 matrix with rows ,
ud du dd uu ud du dd
, , and and columns , , , and . The next
example shows how to use the tensor pattern definition (7.7) to compute the

57
matrix of a 2-spin operator and then confirms that the answer agrees with the
inner product definition (6.05).

Example. Compute using both the matrix (7.7) and inner-product (6.05)

definitions and confirm that the matrix expression for acts only on Alice’s
vector components while leaving Bob’s vector components alone.

Solution.

(III)
and

. (IV)

Thus,

Recall the inner product definition (6.05) for an element of a linear operator M :

. Letting yields

58
(7.2)

We next confirm that the matrix definition of tensor product enables to act only
on Alice’s vector components and that Bob’s components are unchanged.

Since and , the matrix pattern definition yields

, ,

59
, (7.9)

So

Exercise 7.1 is similar, demonstrating that Alice’s half of the state vector is

unchanged while works on Bob’s half.

In the next example we show that the tensor and inner product definitions agree
for a more complex tensor product.

Exercise 7.2. Compute using the tensor and inner product definitions.
The tensor definition yields

60
.

The inner product definition yields

Definition. Given a bra and a ket , their outer product is the linear
A
operator defined by its action on ket vectors :

. (7.01)

Applying the Bra-ket interchange rule to (7.01) yields

Exchanging and , we generate the outer product action on bra vectors:

. (7.02)

Notation. Suppose . To write the outer product

linear operator as a matrix, set . Then

61
: (7.03)

Thus,

. (7.04)

Observe that the pattern approach used to write tensors in matrix form also

works for outer products. Since ,

. ✔

Example. Confirm that the pattern expression (7.04) and the tensor-product
definition (7.01) for outer product are in agreement.

Solution. Let . Then

and

62

Definition. If is a normalized vector, is called a projection


operator.

A
if is a vector then is a vector
Observe that
A
proportional to . We say projects onto .

Definition. Let L be an operator. The trace of L is

, (7.05)
the sum of the diagonal elements.

Theorem 7.1. Let be a projection operator.


(a) L is Hermitian

(b) is an eigenvector of L with eigenvalue +1

(c) If is an eigenvector of L with eigenvalue 0.

(d) The only eigenvector with unit eigenvalue is .


(e) .
(f) .

(g) (7.11)
(i.e., the sum of all projection operators for a given basis is I )

(h) (7.12)

(the expected value of an operator M prepared in normalized state is


the trace of the product of the projection operator and the operator M.)

Proof.

(a) By 7.03, . So .✔

(b) Since . ✔

63
(c) Since . ✔

(d) If is a vector such that , then

where . That is,

are the same eigenvector because any multiple of an


eigenvector is considered to be the same eigenvector. ✔

(e) since is
normalized. ✔

(f) . ✔

(g) Let be any vector. Then

and similarly . Therefore . ✔

(h) Let . Then . Therefore

Definition. Suppose there are several states and that Alice has

prepared her system in state with probability where . We


define Alice’s density matrix as the operator

(7.12a)
and we say Alice’s system is mixed or the density matrix represents a mixed

state. When the density matrix represents a single state , is simply the
projection operator,

(7.12b)

64
and we say that Alice's system is pure. Notice that “pure” can be considered a
special case of “mixed” where r = 1. When we wish to exclude “pure” we will say
that Alice’s system is entangled. (The book does not make this distinction.)

Theorem 7.2. Let LA be any observable of Alice's system and suppose her
system is mixed.
(a) Show that Alice’s expectation is

. (7.13)

(b) Let be a basis for the Hilbert space of Alice’s states, and denote
LA
and in this basis. Show that can be expressed

(7.14)

Proof. Had Alice prepared her system in a single state then she would have

computed . So, for a mixed state, she computes

In the basis , . That is, the element of

in row a ' and column a " is . The diagonal element in row a ' and

column a ' is Thus, is the sum of the diagonal

elements of . Finally, ■

Note: The matrix is a sum of r terms.

65
(7.12c)

Thus each term in (7.14) is a sum of r terms:

(7.12d)

Definition. A pure composite system is one in which there is a state vector

(or, equivalently, a wave function ). If there is a distribution of possible


starting states, we say that AB is mixed. If AB is pure we say that Alice knows

the wave function (and also the composite system state ). Because

exists we can define a density matrix :

. (7.17)

Exercise. Suppose Alice and Bob have single-spin systems and Alice selects an
observable LA of her subsystem A. LA has no effect on Bob’s subsystem B. Find
the operator L in the composite system AB that represents LA. If AB is pure, find
LA
Alice's expected outcome .

Solution. We use the basis . An element of LA in row a' and column

a is where a’ and a range over u and d. Thus,

Similarly, an element of L in row a’b’ and column ab is . Thus,


any operator on the composite system has the form

66
Keep in mind that the 1st and 3rd subscripts pertain to Alice and the 2nd and 4th
pertain to Bob.

If L is the operator on the composite system that represents LA, we must have

In addition, the half of the elements, like , that represent a change in Bob's
state must equal zero and the remaining elements that leave Bob’s states

unchanged must equal the corresponding LA element. For example,

equates to . We represent this mathematically as

(7.18)

where on the RHS captures Alice’s observations and captures whether


or not Bob changes states, “0” if he does, “1” if not.

We replace the elements of L using (7.18), carefully managing the subscripts


because they are in a different order on each side of (7.18):

To confirm that L represents LA, we set

and we use the pattern definition for matrix representation of tensor product to
get

. ✔

67
Now, suppose AB is pure. Then there is a unit composite state vector . Let

be a set of normalized eigenvectors of L. Then forms an


orthonormal basis for AB by the Fundamental Theorem and also by Exercise 3.1.

Hence, we can express as

So

and

L
We are now set up to compute Alice’s expected outcome. To compute ,
recall

. By letting we get

(7.15)

(7.19)

. (7.19b)

✔ (7.21)
This is the same as equation (7.14) in which  represented Alice's density
S
matrix for a mixed system A . Thus, even though AB is pure, A is described
as a mixed state. We will quantify just how "mixed" A is shortly.

68
LA
To generate the other expression, (7.13), for , we denote Alice's density

matrix as . Then

(7.20b)
and

Noticing that this expression is the sum of the diagonal of , as shown in

(7.20b), we obtain . ✔ ■

Susskind says that in order to calculate Alice’s density matrix she may have
had need of the composite wave function, but once she has she can forget
where it came from and use it compute anything about her observations. I
believe this is because the diagonal of contains the probabilities of all of
Alice’s possible states. But I find that sometimes one has to be very careful with
the notation involved. (See “Caution” below). Thus, I have formalized Susskind’s
comment in next theorem. The last section of the proof to the Corollary to
Theorem 7.5, coming soon, is an example of when one must be careful with the
notation.

Theorem 7.3. Let LA be an observable of Alice’s system


SA having orthonormal

eigenvectors and associated eigenvalues . The collection

constitutes an orthonormal basis for


SA . Let LB be an observable of Bob’s

system
SB having orthonormal eigenvectors . Suppose the composite
S
system AB is prepared in the state . Let be Alice’s
density matrix per equation (7.17). After a measurement LA is performed, the
a
probability of being in state is

(7.22)

Solution. From Principle 4, . So

69
. ■

Note. is a diagonal element of . So the diagonal elements are probabilities.

Caution. The subscripts a and b are not summed over 1, 2, …, n as in previous

formulas. An example of this theorem is to let


LA be the observable of energy

states Ei of an atom. Then and a = Ei. If Bob’s states are also energy

states, then and (7.22) becomes . Thus the

subscript “a” is summed over E1, E2, …, En. As another example, were LA =
a
then would be summed over .

Theorem 7.4. Alice's and Bob’s systems are pure iff the composite wave function

factorizes (i.e., ). In this case Alice's wave function can be


expressed

(7.25)

Proof. Suppose . Then are the


states of Alice and Bob, respectively. Then

That is, and, by definition (7.12b), Alice’s system is pure. ✔


Similarly for Bob. ✔

Conversely, suppose A and B are pure. Then there are state vectors

and . So

By definition (7.17), for a composite system,

70
.

So

2 2
This represents n equations (i.e., ) in n unknowns (i.e.,

). Clearly is a solution. Thus,


factorizes. ■

Definition. If the composite wave function does not factorize, we say that AB is
(and Alice and Bob are) entangled. AB is maximally entangled if all of Alice's
(and Bob's) probabilities P(k) in equation (7.12a) are equal.

We can restate Theorem 7.4 using this definition: If Alice and Bob are entangled
then Alice does not have a state space. That is, AB is entangled implies A is
entangled.

Summary – What Alice knows about her system and the composite system
when AB is pure

 Alice has complete knowledge about the composite system because she

knows composite state vector . She knows everything that can be


known.
 As for her own system:

o At one extreme, if factorizes, then from Theorem 7.4 her system

is pure. She knows her state vector . Alice has complete


knowledge of her system.
o At the other extreme, if AB is maximally entangled, she knows
nothing of her state (because all outcomes are equally likely)
though she has maximum information about AB.
 Einstein had difficulty accepting this.

Example. If the composite state vector is a singlet or a triplet, then AB is


maximally entangled.

Proof. We must show that all of Alice's probabilities are equal; that is,

71
Suppose . Then . Thus,

Using we get

So . Thus, AB is
maximally entangled. ✔

We showed in Exercise 7.8(b) for that

T2 and T3
and we can similarly show this for . ✔ ■

We next prove the footnote on p. 108 of the book that states that any Hermitian
matrix can be diagonalized by a change of basis. There are insights to be gained,
including the corollary to the theorem.

Theorem 7.5. If M is an n x n Hermitian matrix, there is a unitary matrix U such



that U MU is a diagonal matrix having the eigenvalues of M on its diagonal.

Proof. The matrix M represents an operator in some basis . Because M is

Hermitian there is an orthonormal basis of eigenvectors of M and a set

72
of corresponding eigenvalues. Write , and let U be the matrix
whose columns are the eigenvectors:

. (A)
So

. (B)

M acting on column j of U gives

. (C)

† †
This implies that the i j-element of U MU , row i of U (expression B) times
column j of MU (expression C), is

. (D)

. ✔
U is unitary because

73
. ■
Corollary. Let be Alice's density matrix in a pure composite system AB, and

denote the eigenvalues of by . Let U be the matrix whose columns are the

eigenvectors of . Then is also a density matrix as well as a

diagonal matrix whose diagonal entries are . Thus the


eigenvalues are their own probabilities of occurrence when is the observable.

Proof. Let denote orthonormal bases for Alice, Bob,


and the composite system, respectively. Since AB is pure, there is a state vector

, (a)
and by definition (7.17) Alice’s density matrix is

where . (b)

We can express the eigenvectors of  in terms of the basis :

. (c)

The matrix U is

, (d)
and so

. (e)
By Theorem 7.5 U is unitary,
. (f)

74
Also by Theorem 7.5, is a diagonal matrix having the eigenvalues of on its
diagonal:

. ✔ (g)

In order to show that is a density matrix we must find a wave function

such that . We will find by rewriting in terms of the

eigenvector composite basis .

a
We start by writing Alice’s basis elements in terms of the eigenvector basis

elements .

Claim: (h)

Define and , vectors composed of row


vectors.

for a = 1, …, n

75
.

for a = 1, …, n.

for a = 1, …, n. ✔

We can use (h) to write the composite state vector in terms of the basis

. (i)

We have succeeded in writing in terms of the eigenvector composite basis.

Next define as the coefficient of in equation (i) :

. (j)
So

(k)

Replacing a’ by a in (j) yields

. (l)
Hence

. (m)

We can now express in terms of a wave function.

76
Claim : (n)

, which proves the claim. ✔


Equation (n) means that is a density function. ✔

More specifically, is generated from per equation (7.17).

Lastly, we show that . Our approach is to carefully use Theorem 7.3.

First, replace
LA in the theorem by . Since the eigenvectors of are with
a
associated eigenvalues , we must replace the eigenvectors by and the

eigenvalues a by . Then the composite state vector , in terms


ab
of the basis , must be replaced by , expressed in terms of

the basis . (Note that the subscript a need not be replaced by subscript ).

Consequently, RHS of equation (7.17) must be changed from to

77
. The latter sum equals by equation (n), so we replace on

LHS of (7.17) by . That is, we replace Alice’s density matrix in the

theorem by the density matrix . The conclusion to the theorem, equation

(7.22), then becomes ■

Lessons Learned about  and :

1. Diagonal elements of are probabilities

2. Diagonal elements of are { }, the eigenvalues of

3. is a probability but not generally equal to

Theorem 7.6. If A and B are matrices then Tr AB = Tr BA.

Proof. and . So

.
(*) Replace i by k and k by i. ■

The next theorem applies to as well as .

Theorem 7.7. Let be the wave function for a pure composite system AB,
and let  be the density matrix for system A.
(a)  is Hermitian
(b) (Exercise 7.6) Tr () = 1
(c) If  is an eigenvalue of , then 0 ≤  ≤ 1
(d) If any eigenvalue equals 1 then all the others are zero

(e) For a pure state A,

(f) For an entangled state A,

Proof.

(a) and

78
(b) ✔

(c) From the Corollary to Theorem 7.5,  is a probability. ✔

(d) By (b) and (c), if one diagonal element  = 1, then the rest are zero. ✔

(e) For a pure system A, by definition there is a state vector , and is the

projection operator: . So

. ✔
Also

. ✔

(f) For a mixed state A, where and

. So

. ✔

Also, denote . Then

✔ ■

Theorem 7.8. Suppose Alice's density matrix  represents a pure system with

state vector . Then

(a) is an eigenvector of  with eigenvalue +1

(b) If then is an eigenvector of  with eigenvalue 0


(c)  has exactly one non-zero eigenvalue , and  = +1

79
Proof.

(a) Since is pure, by Theorem 7.4 . So

(b) ✔

(c) By (a), has one eigenvalue equal to 1. By Theorem 7.7 (d), if any
eigenvalue is 1, all the others are 0. ■

Corollary. If the state vector of a composite system is a product state vector,


then (a–c) above hold.

An example of the concepts of the past few pages is helpful: Alice's and Bob's
state vectors, wave functions, observables and expected values; density matrix
and a change of basis to diagonalize it; product state and wave function; and
correlation of the observables.

80
Example.

Given

 A state space basis of for both Alice and Bob,


 Alice's system is prepared in the direction

 Bob's system is prepared in the generic direction

 Alice and Bob measure spin along the z-axis


 The composite system is a product state

Compute/Identify

 Find Bob's state and wave function, observable and its


eigenvectors and eigenvalues, and use equation (4.13) to compute

 Find Alice's state, wave function, observable, eigenvectors and

eigenvalues, and compute .

 Find Bob's density matrix , confirm that it is the projection operator, and

compute .

 Find the unitary matrix U of Theorem 7.5 such that is a diagonal


matrix and use it to help show in 4 different ways that Bob's system is pure
 Find similar information for Alice's density matrix

 Find the product state and the wave function


 Show that Alice's and Bob's observables are uncorrelated

81
Solution.

Bob

Spin is prepared in direction n̂ and measured along z-axis. Thus, Bob’s spin

basis is and his observable is .✔


u and d
The eigenvectors are with corresponding eigenvalues 1 and -1. ✔

Denote Bob's prepared state by . From the spherical


coordinates of n̂ we deduce that the angle between n̂ and the z -is and we
call the phase angle. The conditions for equations 1.02 and 2.2 apply. We use

them to solve for in terms of and :

A general solution for is , .

The phase factor must be included or else .

The general expressions for are , :

The conditions for equation (4.13) hold. Hence

Check #1: ✔
(definition of expected value)

Check #2: (cosine of angle between prep and measurement) ✔

82
Bob's wave function is . ✔
Alice

From Bob's generic results we deduce that

. (Alice's phase factor is zero because .) Also

, , , and

. Thus, and her wave function is

Density Matrix

Bob's system is pure since AB is a product space. Thus, his density matrix is

where :

Therefore

Check #1: . ✔

Check #2: From (7.12b), should be the projection operator :

83
Check #3:

Even though the formula above for in Theorem 7.2 pertains to mixed
systems, it is true also for a pure system, like this one, because pure is a special
case of mixed where P(a) = 1 for one state a and P(a') = 0 for all other states a'.

We next diagonalize and confirm (per the Corollary to Theorem 7.5) that its

diagonal entries are the eigenvalues of . (We would also diagonalize but it
is already diagonalized and we observe that its diagonal entries are indeed its
eigenvalues.)

We begin by using the characteristic polynomial to find the eigenvalues of

. (a)

We find the eigenvectors of by solving , resulting in

and .

84
Let U be a unitary matrix whose columns are the eigenvectors of :

.
Then

So

Thus, ✔

We show below in four different ways that Bob's system is pure:


85
Plugging Alice's values into Bob's generic formulas yield

.
So

Composite System

The product state is the tensor product

. ✔
The product state wave function is

. ✔

86
Therefore ✔

As expected, are uncorrelated. ■

Note 1. Exercise 7.9 is practically a subset of this exercise except that Alice's as
well as Bob's state is generic.

Note 2. Had the system been prepared in the z-direction and measured in some
direction n̂ (instead of vice-versa), the problem would essentially be the same

but we would not have been able to use key formulas like 2.2 [ ], 4.13

[ ], or 7.13 [ ]. That is, care must be taken to measure


spin in the z-direction to use many of the formulas in the book.

We saw in Bob's generic system that his density matrix has one eigenvalue
+1 and the other 0. This behavior is required by Theorem 7.8.

87
Summary – What Alice knows about her system and Bob's system if the
composite system is pure
 Alice knows the wave function (or, equivalently, the composite state) of
the composite system AB since it is pure
 At one extreme, if the composite state is a product state (i.e., not
entangled) then
o Alice knows everything about her own state, namely the probability
of each possible outcome.
o But she has no knowledge of Bob's system since the two systems
are uncorrelated.
 At the other extreme, if the composite state is maximally entangled, then
o Alice knows nothing about her own state because the probability of
each outcome is the same as every other
o But after she performs a measurement of her state she knows
everything about Bob's state because the two states are fully
correlated. For example, if Alice measures up then she knows that
Bob will measure down.

Section 7.8 Quantifying the spin measurement apparatus A

We assign the spin apparatus A three states.


b
= Blank = Read-out in apparatus window prior to a measurement

= Read-out of +1

= Read-out of -1
u and d
Alice's states are .
The basis for the composite system (consisting of Alice plus the apparatus) is

Let L be the apparatus Hermitian measurement operator, defined as

(i)
After a measurement, if A reads +1 then the spin is up and if it reads -1 then the
spin is down.

If the initial spin is in a superposition state then

, (ii)

88
an entangled state where P(+1) = . It is maximally entangled if in the

singlet state of :

Moreover, over time the system evolves via a unitary operator U operating on

the initial state .

Even though we will only be concerned with measurements taken when we turn
the apparatus on, that is when the screen is initially blank, to verify L is a
bonafide operator we proceed to see if we can generate a consistent Hermitian
matrix for L.

In the 1st matrix below we have labeled the rows and columns of L. For example,

the (u +1 , u b) element of L is called . As another example, the (d b) element


of the initial and final state vectors are labeled and m4 respectively. The matrix
equation provides the general formula for L acting on an arbitrary state vector.

To solve for the elements of L we begin with the two equations in formula (i).

89
90
To determine additional elements of L we must define additional actions by L.
When the on-off button is pushed a 2nd time we define the result to be “off” (i.e, a
blank screen) while preserving the system state. That is, we define

(iii)

91
The four elements yet to be defined actually don’t matter since

the states and can never be obtained. The simplest


solution is to define elements (3,3) and (5,5) to be 1 and (3,5) and (5,3) to be 0.
That is, starting in either of these states means staying in the same state.

That completes the matrix L:

92
Returning to entanglement and collapse, at first the apparatus knows the spin
state and is entangled, but Alice has not yet looked at the display so she is not
yet entangled. After she looks at the apparatus, she becomes entangled, and
from her perspective the spin wave function has collapsed. From Bob's
perspective the system has not collapsed; it is just a 3-way entangled system
(spin, apparatus, Alice).

Thus, an entangled entity observes a particle while an unentangled entity


experiences a wave. As in Special Relativity where there is no contradiction that
two observers see a different speed for an object, there is no contradiction that
entangled Alice sees a particle yet unentangled Bob sees a wave.

We next further clarify this in terms of locality.

Section 7.9 Entanglement and Locality

Definition. A system is local if no information can be sent faster than light.

Extend Alice’s system A to contain her measuring apparatus and Alice herself,

and the same for Bob’s system B. Let be a basis for A and be a basis

for B. Then is a basis for AB. Let be the composite system wave
function, possibly entangled:

Alice’s complete description of her system is contained in her density matrix :

(7.31)

We will show there is nothing Bob can do to instantly change , thus proving that
even if AB is maximally entangled, no faster-than-light signal can be sent by Bob
to change Alice’s system. That is, quantum mechanics does not violate “locality”.

Bob’s system, including any changes he makes, must be described by a unitary


matrix:

93
.

The corresponding unitary matrix for the composite system is

When U acts on the initial wave function it creates a new, or final, wave function

. (7.32)

Changing , the corresponding bra vector is

. (7.33)

Also

94
.

So

. (7.34)

Then Alice’s final density function is

So nothing Bob does affects Alice’s density function.

Section 7.10 Bell's Theorem

Susskind does not state Bell's Theorem. He argues that it violates a different
definition of locality (that he also does not state).

Example 1. (QM: waves do not have definite properties, not even hidden
properties, until they collapse). Entangle 2 particles to have opposite spins. Let
Alice and Bob make repeated spin measurements in the directions 0°, 120°, and
-120°, each choosing directions randomly. Compare their first measurements,
their 2nd measurements, their 3rd measurements, etc. Find the expectation that
compared spins are in opposite directions.

Solution. First, observe that for any direction, .

Suppose Alice measures +1 in one of the directions. Then,

95
.

When Bob randomly chooses his direction, there is a 1/3 chance he picks the
same direction as Alice and 2/3 chance he picks a different direction.

If he chooses the same direction, then his result is -1 since the particles are
oppositely entangled. That is, his conditional probability of picking the opposite
spin is

If he chooses a different direction than Alice, then we use (1.02) to compute


Alice’s spin probability in this direction:

.
So, Bob’s conditional probability is

Thus,

The calculation is the same if Alice measures -1. Thus,


1
E(Opposite spins) = 2 . ■

In the next example, the doors represent directions, the color blue represents +1
spin, and red represents -1 spin.

Example 2. (Classical Mechanics, or QM where waves have definite properties).


Suppose there are 3 doors and behind each door are 2 colored balls. Two of the
doors have a blue ball in front and a red behind, and the other door has a red ball
in front and a blue ball behind. Alice and Bob each randomly select a door. Alice
always selects the ball in front and Bob always selects the ball in back. Find the
expectation that the balls selected are different colors.

Solution.

Event 1 P(Event 1) Event 2 P(Event 2) P(Opp Contribution to


colors E(Outcome)
)

96
Alice picks 2/3 Bob picks 1/3 1
Blue same door
Alice picks 2/3 Bob picks 2/3 1/2
Blue diff door
Alice picks 1/3 Bob picks 1/3 1
Red same door
Alice picks 1/3 Bob picks 2/3 0
Red diff door

E(opposite colors) ■

The issue being investigated in examples 1 and 2 above is whether or not


entangled particles have a definite spin prior to being measured. Einstein and
Bohr famously argued about this but since the calculated result of a single
measurement is the same under either assumption, they thought this to be
simply a philosophical disagreement. However, years later Bell generated an
inequality based upon examples similar to the two above. His inequality shows
there is indeed a measurable consequence for each assumption.

Suppose the particles do have a definite spin prior to measurement. If the front
balls are all blue (or all red) then E(opposite color) = 1. Suppose there are balls
of both colors. There must be 2 blue balls and 1 red, or vice-versa. Example 2
5
proves that E(opposite spin) = 9 . Thus, no matter what, E(opposite spin)  .

However, if the particles have no definite spin until the moment of measurement,
1
then Example 1, which uses QM calculations, predicts that E(opposite spin) = 2 .

Theorem (Bell's Inequality). Two particles are entangled with opposite spin. They
are repeatedly measured along axes chosen randomly and independently from 3
different directions. If they have definite spins (even if they are hidden values that

we cannot access), then E(opposite spin) .

Years after Bell developed his famous theorem, the technology became available
to make a large number of spin measurements. The measured result was ½.
Thus, it was found that particles do not have definite spin until the moment a
measurement is made.

97
Example 3. (Single spin measurement). A classical computer can simulate
quantum spin.

Solution. We assume as usual there is an apparatus with a Measure Button and


a Display Window that can be placed along any axis in 3-space. We imagine a

classical computer that initially stores 2 complex numbers in memory,

that are normalized as usual by . The computer solves the


Schrödinger equation to update the 's. The computer also stores a unit vector
representing the apparatus' 3D orientation. Finally, the computer stores +1 or -1
and displays the number in the Display Window.

Alice selects an orientation and pushes the Measure Button on the apparatus.
The computer uses a random number generator to output +1 or -1 with

probabilities and , respectively. Then the Schrödinger equation takes


over until the apparatus is (possibly) re-oriented and the button is pushed again.

There is no known experiment that Alice can perform to distinguish that the
computer is not a quantum computer. ■

Example 4. (Multiple measurements). A classical computer can also simulate a


pair of separated, entangled particles assuming information from the computer
travels instantaneously to each apparatus.

Solution. Denote the composite system by

Let
LA and LB be Alice's and Bob's respective measurement operators. Then

and similarly for Bob . For

example, .

The computer initially stores 4 complex numbers: and it


updates them using the Schrödinger equation for the combined system. Alice's
Display Window shows only her spin (+1 or -1) as does Bob's. Each apparatus
can be independently oriented and each has its own Measure Button. When
Alice pushes her button, the composite computer uses a random number
generator to instantly display +1 or -1 on Alice's Display Window according to the

probabilities , and similarly for Bob.


As in Example 3, there is no known experiment that Alice or Bob can perform to
distinguish that this is not a quantum computer. ■

98
The computer in the above example is non-local because it sends signals faster
than the speed of light. Yet this system is local with respect to our definition of
locality because even instantaneous information from Bob about his wave
function cannot affect Alice's density function (which encapsulates everything she
can know about her system). This is the argument that Susskind makes to
support his claim that Bell's Theorem (and also the Einstein-Bohr argument) was
about a computer simulation and not about the real world. He does not explicitly
give his alternate definition of “simulation locality” nor does he describe how it
relates to Bell Theorem (which he doesn’t state).

Example 5. (Multiple measurements). A classical computer cannot simulate a


pair separated, entangled particles if information from the computer travels to
each apparatus at less than or equal to the speed of light. Thus, the combined
system is local with respect to definition 2.

Solution. First of all, if the computer is situated, say, with Bob, then the slower-
than-light updating of her Display Window is a clear giveaway to Alice that the
computer is classical.

But perhaps part of the composite computer resides with Alice and part with Bob.
They can now see instantaneous results in their Display Windows.

As usual let Alice's and Bob's system states be and

, respectively. Alice's part of the computer can only generate

and, similarly, Bob's part can only generate . The composite

computer cannot instantly determine as would be


necessary if, say, the spins were entangled in the singlet state (opposite spins).

The computer can only use for Alice and for Bob.
So, how do Alice and Bob figure out the computer is not a quantum computer?
Suppose that Alice and Bob make a large number of measurements,
independently and randomly choosing from directions 0°, 120°, and -120°. Since
1
the spins have distinct values, according to Bell's Theorem, E(opp spin) > 2 . But,
also according to Bell's theorem, a true quantum computer would yield
1
E(opp spin) = 2 . So, Alice and Bob are able to determine that the computer is
not a quantum computer. (They could also just compare their spins and observe
that more than half of them are not opposites.) ■

99
Chapter 8. Particles and Waves

TABLE 8.1

Concept Discrete Continuous


Discrete set of outcomes Continuous set of outcomes
Observable L
(eigenvectors) (eigenvectors)
Vector Space Finite dimensional Vector space of complex-valued
of States complex vector space functions f

State Vector
(8.1) (A)

Bra Vector

Wave
Function

Probability (3.11)
(B)
Inner
Product
(8.2)
Normaliza-
tion (8.3)
 is the function that satisfies
Kronecker
Delta & Dirac
Delta (8.4)
Functions for every continuous function F having
compact support

Integration
By Parts (F)
for wave functions F and G

Basis Vector
(G) (H)
Inner
Product with (1.5) (8.13)

100
a Basis
Vector

Quantum Mechanics is not so much about particles and waves as it is about the
set of non-classical principles given in Chapter 3 that govern their behavior. We
now extend the principles and concepts from the discrete systems we have so far
studied to continuous systems (where we will at last develop wave examples.)

Table 8.1 provides a side-by-side overview of the discrete and continuous cases.
It also includes key equation numbers, both mine and the book’s. For easier

comparison, for the discrete case I have changed the notation from to .
The sections below provide clarification and detail.

In Table 8.1, the values x could be points in or even other types of


more abstract spaces. For this book, however, it is sufficient to consider just the
case where .

Vector Space

It is easy to show that all the conditions of a vector space hold if we define

addition of two functions to be and scalar multiplication by

a complex number to be .

Probability

For the discrete case, if is a basis for the state , then

(8.1)
and probability is determined by

. (3.11)

For the continuous case there are both a probability density function and a

cumulative distribution function. If is a basis for the state , then

. (A)
The density function is

. (B1)
The cumulative distribution function can act on a finite or infinite interval:

101
. (B2)

Inner Product

Let .

The discrete form of the inner product is

. (1.14)

The continuous form is

. (8.2)

Dirac Delta

Definition. A function F has compact support if there is some closed interval


[a,b] outside of which F(x) = 0.

In this book it is assumed that all wave functions are continuous and have
compact support.

Intuitively the Dirac delta function is a density function at some value x.


In physics it is often defined informally as the function  that satisfies

(8.4)
for every continuous function F having compact support.

The reason this definition is “informal“ is that in fact there is no such function. We
give a more rigorous definition shortly.

The Dirac delta function can be visualized as the limit of a set of


shrinking step functions of increasing height as follows.

For define step function

Then

102
(C)

It is understood that the limit and integral sign cannot be freely interchanged.
This, too, doesn't make sound mathematical sense since it violates Fubini’s
theorem. Even worse, we do make the interchange when necessary or
convenient.

It follows immediately from (C) that

: (D)


and

: (E)

. ✔

When referring to equation (E) in light of (D), we informally say that 

approaches infinity at a rate that keeps the area under the curve
at unity.

Since, by (E), can equal infinity,  is not a (real-valued) function. This


is part of why the definition (C) does not make precise mathematical sense.

Thie informal definition (C)could also have been made in terms of the normal

distribution density function instead of . To see this, let

and compare . The graphs below


show the comparison for x = 1 and n = 20.

103
Unfortunately, the fact that the above definitions are informal and mathematically
incorrect makes proving some of its properties difficult. See, for example, the

argument below to normalize in equation (8.17).

A mathematically rigorous definition of  involves the concept of measure,


specifically Lebesgue measure. The Dirac delta “function” is not a function at all
but a measure. The argument of a measure, unlike a function, is a set. The Dirac

delta function is defined for any measurable set containing zero to be 1, and
to be 0 for any other measurable set. The Lebesgue integral with respect to 

satisfies or, by translation, ,


for any continuous function F with compact support. Note that the argument of 
is a set.

There is a cumulative distribution function associated with the density function 

where .

So,  is the derivative of  but only in the sense that


where these integrals are understood to be Riemann-Stieltjes integrals.

104
Integration By Parts

FG | ba
We are able to drop the term because wave functions are zero at :

(F)

Inner Product with a Basis Vector

x
Consider . When is one of the basis vectors used in the definition of

, the inner product takes a simpler form:

(8.13)

This equation is needed in the derivation of equation (8.18), which Susskind


leaves to the reader.

While equation (8.13) is perhaps an obvious generalization of the discrete


equation (1.5) shown in Table 8.1, Susskind mentions that (8.13) holds because
x
is represented by the Dirac delta function. This is worth showing and adds a
little more insight into (8.13).

In order to have a discrete process to imitate, we write the discrete basis vector
in terms of the Kronecker delta:

. (G)
Then,

(1.5)

In the continuous version of (G), the Kronecker delta is replaced by the Dirac
delta and takes the form

. (H)
Equation (8.2) can be rewritten

(I)
and so

105
. ✔

In other words, the wave function, , of a particle at position x is the

projection of a state vector onto the eigenvector of position.

Linear Operators

Recall that observables are represented by Hermitian linear operators. By


definition, L is a linear operator if L(x + y) = Lx + Ly.

Recall that L is Hermitian if .

In the two examples below,  is a complex-valued function of a real variable x.

Example 1. Multiplication Operator: (8.5)

(because since ).

Thus, by Theorem (3.06), X is Hermitian.

Example 2. Differentiation Operator: (8.6)

(8.7)

Thus, D is anti-Hermitian. But, we next show that this means that both i D and
- i  D are Hermitian. In particular, is Hermitian.

Theorem. If M is anti-Hermitian, then both i M and – i M are Hermitian.

Proof. Let . In particular,

106
(i)

(ii)

Let -complex matrix. Then

(iii)

(iv)

Suppose M is anti-Hermitian:

(v)
Then

. That is, i M is Hermitian. ✔

. That is, -i M is Hermitian. ✔■

Eigenvectors and Eigenvalues of the Multiplication Operator: Position

The multiplication operator X is the Hermitian operator that represents the


position observable. That is, a position is simply a number on the x-axis, and
multiplication by any real number is possible. So, an outcome of X is a real
number, a position.

Let be a state vector for the observable X, and


x0 an outcome. By Principle
2, an outcome is an eigenvalue of X. That is,

By the definition (8.5) of X, the LHS is

The RHS is

Thus,

107
a.e. (almost everywhere; i.e., except on a set of measure
zero)

a.e. (8.11)

This means that except on some set of measure zero, if .

We seek a function that has the property (8.11) and that also satisfies the
normalization requirement:

. (8.3)

We claim that the Dirac delta function meets these conditions. Let

(J)

If . Thus, satisfies (8.11). Also

So, intuitively satisfies (8.3):

This argument lacks rigor because we haven’t shown that is just


infinite enough that the area under the curve is unity.

Definition. We will refer to as the wave function in the position


representation.

Eigenvectors and Eigenvalues of the Multiplication Operator: Momentum

We move on to the concept of momentum. The momentum operator is defined


in terms of the differentiation operator:
. (8.14)

Plugging in the definition of D gives

108
(8.15)

Note 1. The factor is needed to provide units of momentum (mass times


velocity).

Note 2. The symbols “P” and “p” are used to denote probability (CDF and density
function). We use italics to denote momentum, “P ” and “p” (momentum operator
and momentum eigenvalue).

Experiment has shown that we cannot simultaneously measure position and


momentum. A state

can be expressed in term of either position or momentum. Of course, the


momentum wave function will be different than the position wave function. If we

label the momentum wave function , then the state written in terms of
momentum is:

. (K)
p
Since is one of the basis vectors in equation (K), formula 8.13 applies:

. (8.20)

We are now set to develop the equations that transform back and forth between

and .

p
Step 1. Since a momentum eigenvector can be considered to be a state of X,
x
it can be written in terms of the basis. It will have its own wave function that

we will call . That is,

. (L)

We combine this with equation (H) to get

. (M)

p
Step 2: Since is an eigenvector of P,

109
.

LHS: .

RHS: .

Solving the differential equation yields

a.e.

To solve for A we must apply normalization, but (8.3) doesn’t work:

I believe the problem arises from the impreciseness of the informal Dirac delta
definition. An approach that apparently works is to first assume that x is periodic

with period . This leads to discrete values


p n for p and use of the discrete

normalization equation in Table 8.1, above. Then taking , discrete   


continuous p, and Kronecker    Dirac , this finally leads to

.
Thus,

a.e. (8.17)

Therefore

(8.18)

Step 3. For a discrete basis,

110
. (8.21)

The continuous versions of this are

(8.21)
and

. (8.22)

Thus,

: (8.24)

Example. Alice measures the position wave form of some particle and she
wants to know the probability that she would have measured momentum p.

Solution. First, she calculates . Then . ■

Solving for in terms of is very similar.

: (8.25)

Equations (8.24) and (8.25) are reciprocal Fourier transforms, and are the
central equations of Fourier analysis.

Heisenberg Uncertainty Principle

Let be either a position or momentum wave function. By definition,

111
and

So,

,
or

. (8.29)

In classical physics, X P – P X = 0, but not in quantum mechanics where we see


from (8.29) that the commutator is not zero. By the Simultaneity Principle
developed in Chapter 5, this means that X and P cannot be simultaneously
measured. In fact, the Heisenberg Uncertainty Principle provides a floor for the
uncertainty.

Recall that we denote the uncertainty of an operator by its standard deviation.


is the standard deviation of X,
is the standard deviation of P.

When we apply the Heisenberg Uncertainty Principle developed in Chapter 5 to


X and P we get

.
No matter how cleverly we measure position and inertia, we can never drive the
uncertainty below this small quantity.

112

You might also like