1.

Fundamental Concepts

1.1

Introduction

Theoretical framework of classical physics:

state (at xed t) is dened by a point {xi , pi } in phase space
F G F G
(at space, symplectic
Poisson bracket {xi , pj } = ij {F G} = i x
i pi pi xi
md)
Observables are functions O(xi , pj ) on phase space [ex: x2 + y 2 + z 2 , p2 /2m, . . .]
Hamiltonian H(x, p) denes dynamics

q = {q, H} for q = xi , pj , . . .












I





I


(x, p)

qM

z
phase space

qM

Framework describes all of mechanics. E & M including uids, materials, etc. . . .

Simple, intuitive conceptual framework
Deterministic
Time-reversible dynamics
Example: Classical simple harmonic oscillator (SHO)
H=

1 2
k 2
p
x +
2
2m
p






!
'


X
"

1
p
m
p = {p, H} = kx
(= m
x)

x = {x, H} =

Theoretical framework of quantum physics

state dened by (complex) vector |v in vector space H. (Hilbert space)
Observables are Hermitian operators Odagger = O
Dynamics: |v(t) = eiHt/|v(0) (H hermitian, constant)
Collapse postulate
[(subtleties:
(simple version)

If | = i |i,
j
with A|i = i |i , i =
2
Then with prob. |i | , measure A = i ,
| = |
i after measurement

degeneracy,
cts. spectrum)]

This framework [(with suitable generalizations, i.e. eld theory)] describes all quantum sys
tems, and all experiments not involving gravitational forces.
Atomic spectra (quantization of energy)
Semiconductors transistors, (quantum tunneling)
Counterintuitive conceptual framework
Nondeterministic
Irreversible dynamics
Both classical & quantum conceptual frameworks useful in certain regimes.

Classical picture not fundamental replaced by QM.

Is QM fundamental?

Perhaps not, but describes all physics of current relevance to technology & society.

Simple example of QM: 2-state system (spin

1
2

particle)

Stern & Gerlach 1922

 S


---

'

-Ag
atoms
Oven 


























IB 

N
screen

Silver: 47 electrons, angular momentum = (mag. dipole moment) from spin of 47th
electron.

U = B
U
Bz
Fz =
= z
Z
Z
Gives force along z depending on z .
Energy

Classically, expect

&

Actually see

#
%

#
%

e
e

Sz
2mc
mc

[ = 1.0546 1027 ergs]

=
2

z
Sz

So measuring Sz = discrete values (2 states)

Can build sequential S G experiments, using components

Sz = +
splitter

Z
Sz =

lters

Sz = +
















Sz =

(Can also form splitter, lters on x-axis, etc.)

Single particle experiments
i)


Sz =+
Z

ii)








Sz= +
Z

Sz =+
Z











Sx= +
X

Sx =

100%
0%

50%
50%

iii)


Sz =+
Z

Sx= +
X







Sz= +
Z

Sx =

Sz =

100%

0%

BUT
iv)


Sz =+
Z

v)


Sx= +
X














Sz =+
Z














Sx =

Sz= +
Z

Sz =

Sz= +
Z

Sz =

50%
50%

50%

50%

Cannot simultaneously measure Sz , Sx

incompatible observables

S z Sx =
Sx S z

Analogous to 2-slit experiment for photons.

In iii), not measuring Sx .

in iv), v) measuring Sx .

iii) needs interference of probability wave no classical interpretation.

iii), iv) = Irreversible, nondeterministic dynamics (assuming locality)

1.2
1.2.1

Mathematical Preliminaries
Hilbert spaces

First postulate of QM:

The state of a QM system at time t is given by a vector (ray) | in a complex Hilbert
space H. [[will state more precisely soon.]]
Vector spaces
A vector space V is a collection of objects (vectors) | having the following proper
ties:
6

A1: | + | gives a unique vector | in V .

A2: (commutativity) | + | = | + |

A3: (associativity) (| + |) + | = | + (| + |)

A4: vector | such that | + | = | |

A5: For all | in V , | is also in V so that | + (|) = |.

[(A1A5): V is a commutative group under +]

For some Field F (i.e., R, C, with +, dened) scalar multiplication of any c F with any
| V gives a vector c| V . Scalar multiplication has the following properties
M1: c(d|) = (cd)|
M2: 1| = |
M3: c(| + |) = c| + c|
M4: (c + d)| = c| + d|.

V is called a vector space over F .

F = R:
real v.s.

F = C:
complex v.s.

Examples of vector spaces

a) Euclidean D-dimensional space is a real v.s.

a1
..
.
a
D

b) State space of spin 12 particle is 2D cpx v.s.

states: c+ |+ + c |, c C [[note: certain state are physically equivalent]]
c) space of functions f : [0, L] C
Henceforth we always take F = C
Subspaces
V W is a subspace of a v.s. W if V saties all props of a v.s. and is a subset
of W .
[suces for V to be closed under +, scalar mult.]
Ray
A ray in V is a 1D subspace {c|}.

Linear independence & bases

|1 , |2 , . . . , |n are linearly independent i
c1 |1 + c2 |2 + + cn |n = 0
has only c1 = c2 = = cn = 0 as a solution.
If |1 , . . . , |n in V are linearly independent, but all sets of n + 1 vectors are linearly
dependent, then
V is n-dimensional (n may be nite, countably , or uncountably .)
|1 , . . . , |n form a basis for the space V .
If |1 , . . . , |n form a basis for V , then any vector | can be expanded in the basis as

ci |i .
(Thm.)
| =
i

Unitary spaces
A complex vector space V is a unitary space (a.k.a. inner product space)
if given |, | V there is an inner product | C with the following
properties:
I1:

| = |

I2:

|(| + | ) = | + |

I3:

|(c|) = c|

I4:

| 0

I5:

| = 0 i | = |0

| is a sesquilinear form (linear in ), conj. lin. in |

Examples:
z1
..
. , zi C.
zN

wN
z|w = zI w1 + z2 w2 + + zN

a) V = CN , N-tuples |z =
b) V = {f : [0, L] C}
L
f |g = 0 f (x)g(x) dx
Terminology:

| = norm of | (sometimes, |

if | = 0, |, | are orthogonal

Dual Spaces
For a (complex) vector space V , the dual space V is the set of linear functions | : V C.
Given the inner product |, can construct isomorphism.
| |
V V
note: c| c | .
Physics notation (Dirac bra-ket notation)
| V
| V

ket
bra

Hilbert space
A space V is complete if every Cauchy sequence {|n } converges in V
N : |n |m < m, n > N (i.e., | : limn | |n = 0.)
A complete unitary space is a (complex) Hilbert space.
Note: an example of an incomplete unitary space is the space of vectors with a nite number
of nonzero entries. The sequence { (1, 0, 0, . . . ),
(1, 12 , 0, 0, . . . ),
(1, 12 , 13 , 0, . . . )
...
}
is Cauchy, but doesnt converge in V . This is not a Hilbert space.
A Hilbert space can be:
a) Finite dimensional (basis |1 , . . . , |n )

Ex. spin 12 particle in Stern-Gerlach expt.

b) Countably innite dimensional (basis |1 , |2 , . . .)

Ex. Quantum SHO

c) Uncountably innite dimensional (basis |x , x R)

[Technical aside:
=V.
A space V is separable if countable set D V so that D
(D dense in V : , |x V |y D : |x |y < )
(a), (b) are separable, (c) is not.
non-separable Hilbert spaces are very dicey mathematically.
Generally, separability implicit in discussion e.g., label basis |i , i takes discrete
values ]
9

Orthonormal basis
An orthonormal basis is a basis |i with i|j = ij
Any basis |1 , |2 , . . . can be made orthonormal by Schmidt orthonormalization
|1
|1 =
1 |1

|2 = |2 |1 1 |2
|
|2 = 2
2 |2

|3 = |3 |1 1 |3 |22 |3
..
.
Gives |i with i|j = ij .

If {|i} are an orthonormal basis then for all |,

ci |i ,
ci = i | .
| =
i

Can write as | =

i |i i |.

(Completeness relation)

Schwartz inequality
|| |||2

|, | .

Proof: write | = | + |

| = | + | + x | + ||2|
|
set =
|
| ||
| = |
0
|

[Second postulate of QM:

Observables are (Hermitian) operators on H [self-adjoint]

10

1.2.2

Operators

Linear operators
A linear operator from a VS V to a VS W is a transformation such that
A| + | = A| + A|

|, | .

We write A = B i A| = B| |.
A acts on V through (| A)| = |(A|) (acts on right on bras.)
Outer product
A simple class of operators are outer products ||
(||)| = ||
Adjoint
Recall correspondence

V V
| |

Given an operator A, dene A (adjoint of A) (Hermitian conjugate) by A | |A

Example (||) = ||. Follows that |A| = (|A|).
Hermitian operators
A is Hermitian if A = A ( self-adjoint)

Technical aside: mathematically, Hermitian called symmetric. Self-adjoint i symmetric

+A & A have same domain of denition, relevant to i.e., Dirac op. in monopole background
(symmetric op. with several self-adjoint extensions) more: Reed & Simon, Jackiw
0
2
Example of domain of def: Consider H = L2 (R) = {funs f : R C : f f < 0 ex H,
2
O = mult by ex

2
2
Oex
/ H, so ex not in domain D(O).
Linear operators A form a vector space under addition (+ commutative, associative)
(A + B)| = A| + B|
Mult. dened by
(AB)| = A(B|)
Generally AB
=

BA
But (AB)C = A(BC)
Note: (XY )+ = Y + X +
11

Identity operator: 11
11| = |.

Functions of one operator f (A) = Cn An can be expanded as power series (must be careful
outside ROC can do in general if diagonalizable)
Diagonalizable operators: can always compute f (A) if f dened for diagonal elements (eval
ues)
Functions f (A, B) must have denable ordering prescription. (e.g. eA BeA = B + AB
BA + )
Inverse A1 satises AA1 = A1 A = 11
Does not always exist. (Ex. if A has an ev. = 0.) Note: BA = 11 does not imply AB = 11
(Ex. later)
Isometries
U is an isometry if U + U = 11, since preserves inner product (|U + )(U|) = |
Unitary operators
U is unitary if U = U 1 .
Example: non-unitary isometries. (Hilbert Hotel)

Consider the shift operator S|n = |n + 1 acting on H with countable on basis {|n,

= 0, 1, . . . }

S = |n + 1n| satises S + S = 11 but not SS + = 11. (SS + = 11 |00|).

S has no (2-sided) inverse.

Projection operators
A is a projection if A2 = A.
Ex. A = || for | = 1.
Eigenstates & Eigenvalues
If A| = a| then | is an eigenstate (eigenket) of A and a is the associated eigenvalue.
Spectrum
The spectrum of an operator A is its set of eigenvalues {a}

[technical aside: this is the point spectrum, mathematically, spectrum of A = set of

: A 11 is not invertible]

12

Important theorem

If A = A , then all eigenvalues ai of A are real, and all eigenstates associated with distinct
ai are orthogonal.
Proof:
A|a = a|a a|A = a|a
b|(A A+ )|a = (a b )b|a = 0
if
if

a = a is real.
b|a = 0

a = b,
a
=

b ,

Consequence of theorem:
For any Hermitian A, can nd an O.N. set of eigenvectors |ai
(ai not necc. distinct
can be degenerate)

A|ai = ai |ai ,

[Proof: use Schmidt orthog. for each subspace of xed evalue a OK as long as countable #
of (indep.) states for any a (e.g. in separable H) [caution: this set spans space of eigenvectors,
but may not be complete basis]
Completeness relation
If i are a complete on basis for H.
| =

|i i|

| ,

(completeness)
so
i |i i | = 11
(sum of projections onto 1D subspaces)
Matrix and vector representations
If H is separable, a countable on
basis,|i
1 |

2 |
can write | = i |ii |

..
.
| =

|ii| (|1|2 )

13

|A|

|A|

1
1
1
2

A=
|i i|A|j j | 2 |A|1 2 |A|2
..
..
..
i,j
.
.
.
If ai | are a basis of O.N. eigenvectors w.r.t. A, ai |A|aj = ai ij

a1

a2

A=
|ai ai ai |

a3

..
.
Usual matrix interpretation of adjoint, dual correspondence
(adjoint = conjugate transpose)
i|A|j = j |A |i

c1

c2 dual

dual: | | (c1 c2 )
..
.

|
ci di
inner product.

d

1
c1 c2 d2
..
.
When do eigenvectors of A = A form a complete basis for H?
True when H is nite dimensional (explicit construction from diagonalization), not neces

sarily when H innite dimensional.

Defs. A is bounded i sup

|H
|=|0

|A|
|

<

A is compact if every bounded sequence {|n } (n |n < ) has a subsequence {|nk } so

that {A|nk } is norm convergent in H.
Facts:
A compact A bounded.
Every compact A = A has a complete set of eigenvectors. (compactness sucient)
not necessarily true for bounded A = A . (neither necessary nor sucient)
Ex. H = L2 ([0, 1])

A = x.

A is bounded, not compact. (|n = xn 2n + 1)

A has no eigenvectors in H.

For physics: Only interested in operators with a complete set of eigenvectors. These are
called observables. Observables need not be bounded or compact. (note: will reverse this
stance a bit for cts systems!)
14

Trace
The trace of an operator A is
Tr A =

i|A|i ,

ai =

|i ON basis

Aii .

(ai = eigenvalues of A)
Unitary transformations
If |ai , |bi are two complete ON bases, [(Ex. eigenkets of 2 Hermitian operators)]
can dene U so that U|ai = |bi (since |ai a basis denes U on all of H).
so bi | = ai |U .

U = U11 = U i |ai ai |

=
i |bi ai |

U =
i |ai bi |

so UU = i,j |bi ai |aj bj | = ij ij |bi bj | = 11 and UU = 11, so U 1 = U + , U unitary.

We have

Analogous to rotations in Euclidean 3-space M : M + M = MM + = 11. U are symme

tries of H.
Unitary transforms of vectors & operators
A vector | has representations in two bases as

di |bi .
| =
ci |ai =
How are these related?

dj |bj =

dj U|aj

dj |ai ai |U|aj

i,j

so ci = Uij dj ,

Uij = ai |U|aj are mtx elements of U in a rep.

Similarly, X = |ai Xij aj | = |bk Yk b | gives Xij = Uik Yk Ulj .

15

Diagonalization of Hermitian operators

Theorem. A Hermitian matrix (nite dim) Hij = i |H|j can always be diagonalized by
a unitary transformation.
Proof. if |i a general ON basis, ON eigenvectors |hi related to |i through
unitary .
|hi = U |i ,
hi |H|hj = ij hi = i|U HU |j
+
Hk Uj is diagonal.
so Uik
(generalizes to any observable)

Algorithm for explicit diagonalization of a matrix H (nite dimensional):

1) Solve det(H 11) = 0 for N N matrices, N solutions are eigenvalues of H.
2) Solve Hij cj = ci for ci s for each . N linear eqns. in N unknowns.
Gives eigenvalues & eigenvectors.
Invariants
Some functions of an operator A are invariant under U :

Tr A =
i |A|i ,
|i ON basis

Tr U AU =
Uij Ajk Ukj = jk Ajk = Tr A
i,j,k

[Technical note: careful for matrices need all sums converging.]

Another invariant: det A: det(U AU ) = det U det A det U = det UU det A = det A

[Note: full spectrum of evs is invariant!]

Simultaneous diagonalization
Theorem. Two diagonalizable operators A, B are simultaneously diagonalizable i
[A, B] = 0
say

A|i = ai |i ,
B|i = bi |i

AB|i = BA|i = ai bi |i .

Say AB = BA ,
A|i = ai |i .

AB|i = ai B|i ,

so B keeps state in subspace of e.v. ai . Thus, B is block-diagonal, can be diagonalized in

each ai subspace

16

1.3

The rules of quantum mechanics

[[Developed over many years in early part of C20. Cannot be derived justied by logical
consistency & agreement with experiment.]]
4 basic postulates:
1) A quantum system can be put into correspondence with a Hilbert space H so that
a denite quantum state (at a xed time t) corresponds to a denite ray in H.
so | c| represent same physical state
convenient to choose | = 1, leaving phase freedom ei |
Note: still a classical picture of state space (realist approach). Path integral approach
avoids this picture.
state really should apply to an ensemble of identically prepared experiments (pure
ensemble = pure state.)
Ex. states coming out of SG lter


Sz = +


Z

| = |+ .




2) Observable quantities correspond to Hermitian operators whose eigenstates form a

complete set.
Observable quantity = something you can measure in an experiment.
[[Note: book constructs H from eigenstates of A: logic less clear as HA = HB for some
A, B.]]
3) An observable H = H denes the time evolution of the state in H through

d
|(t + t) |(t)

|(t) = i lim
= H|(t) .
t0
dt
t

(Schrodinger equation)

17

4) (Measurement & collapse postulate)

If an observable A is measured when the system is in a normalized state |, where
A has an ON basis of eigenvectors |ai with eigenvalues ai .
a) The probability of observing A = a is

|aj ||2 = |Pa |

j:aj =a

where Pa =

j:aj =a |aj aj |

is the projector onto the A = a eigenspace.

b) If
becomes |a = Pa | =
A = a is observed, after the measurement the state
|a
a
|
(normalized
state
is
|

=
|
/
a |a ).
j
a
a
j:aj =a j
Discussion of rule (4):

Simplest case: nondegenerate eigenvalues

| =
ci |ai ,

ai =
aj .

Then probability of getting A = ai is |ci |2 .

2
Norm of | = 1

|ci | = 1.
After measuring A = ai , state becomes |
i = |ai .
This postulate involves an irreversible, nondeterministic, and discontinuous change in the
state of the system.
source of considerable confusion
less troublesome picture: path integrals.
alternatives: non-local hidden variables (t Hooft?), string theory new principles (?)
For purposes of this course, take (4) as fundamental, though counterintuitive, postulate.

To discuss probabilities, need ensembles.

Consequence of (4):

Expectation value of an observable A in state | is

2
|ai ai ai |.
|ci | ai = |A| since A =
A =
i

So for:
4 basic postulates of Quantum Mechanics:
18

1) State = ray in H

[incl. def of H space]

2) Observable = Hermitian operator with complete set of eigenvectors

3) i dtd |(t) = H|(t)
4) Measurement & collapse

Probability A = a:
|P a|

After measurement, system |

a = Pa |/ |Pa |

Pa =
|aj aj |
j:aj =a

Expectation value of A: A = |A| =

|ci |2 ai if | =

ci |ai

4
These are the rules of the game.
Rest of the course:
Examples of physical systems, tools to solve problems.
An example revisited in detail
Back to spin- 12 system.

State space

H = {| = C+ |+ + C | , C C}

P 1 Unit norm condition

| = |C+ |2 + |C |2 = 1

ei | , | are physically equivalent

Operators:
Sz =

2 z

Sx =

2 x

Sy =

2 y

1 0
0 1
0 1

measures spin along z-axis

measures spin along x-axis

1 0
0 i
i 0

measures spin along y-axis

For general axis n

:

Sn = S n

(HW #2)

2 .

has eigenvalues
eigenstates |Sn ; : Sn |Sn ; = 2 |Sn ; form complete basis.
19

|Sx ; = 12 |+
|Sy ; = 12 |+
|Sz ; = |+

1 |
2
i |
2

in Sz basis.

Some further properties of Si :

[Si , Sj ] = iijk Sk
{Si , Sj } = Si Sj + Sj Si = 12 2 ij

3
3 2 1 0
S 2 = S S = Sx2 + Sy2 + Sz2 = 2 11 =
0 1
4
4
And
[S 2 , Si ] = 0 .

Measurement
If | = c+ |+ + c |,

| = 1,
prob. that S
z =

prob. that Sz =

+
2

is
|
c+ |2
is
|
c |2

Consider single particle experiments from lecture 1.

Sz = + 2

i)

Sz = + 2

Sz










Sz

100%

0%

Sz = 2

| = |+

Repeated measurement of Sz gives + 2 100% of the time.

Sz = +

ii)


Sz










Sx

| = |+

|Sx ; =
so
so

1 (|+ |)

2
| = |+ = 1
2 [|Sx ; + + |Sx ; ]
prob. Sx = + 2 is 12 (50%)
prob. Sx = 2 is
12 (50%)

20

Sx = +

Sx =

50%
50%

iii)

S
z = +

Sx = +

Sz

Sz = +




S
x

| = |+

Sz

B S
x =

100%

0%

S
z =

| = 2 (|+ |)

|+ = 12 (|+ + |)

| = |+ + | = |+

Combined state | = |+ enters last measurement apparatus, since Sx not measured.

Gives Sz = + 2 100% of time.
iv)

Sz = +

Sx = +

Sz




1
2




S
x
|+ =

| = |+
state |+ =

Sz = +

1
2

Sz

(|+ + |)

50%

50%

Sz =

(|+ + |) enters last apparatus.

Prob.

Sx = + 2

: (50%)

Prob.

Sx = 2

: (50%)

Compatible vs. incompatible observables

Observables A, B are:

Compatible if [A, B] = AB BA = 0

incompatible if [A, B]
=

0.
Examples: S 2 , Sz
Sx , Sy

are compatible
are not compatible.

Theorem. Compatible observables A, B can be simultaneously diagonalized, and have eigen

vectors |ai , bi with
A|ai , bi = ai |ai , bi
B|ai , bi = bi |ai , bi .

21

(Proof in last lecture: AB| = aB| if A| = a| so B = Ha Ha , diagonalize in each

block.)
A complete set of commuting observables (CSCO) is a set of observables {A, B, C, . . .} such
that all observables in the set commute:
[A, B] = [A, C] = [B, C] = = 0
and such that for any a, b, . . . at most one solution exists to the eigenvalue equations
A| = a|
B| = b|
..
.

Tensor product spaces

useful for many-particle systems [+ quantum computing, . . . ]bigr)

Given two Hilbert spaces H(1) , H(2) , with complete ON bases |(1) i , |(2) j , the tensor
product
H = H(1) H(2)
is the Hilbert space with ON basis
(1)

(2)

|ij = |i |j
and inner product
(1)

(1)

(2)

(2)

i,j |k, = i |k 1 , j | 2 .
If H(1) , H(2) have dimensions N, M, then H = H(1) H(2) has dimension NM.
If H(1) , H(2) separable, H is separable.
[in particular, if either or both of H(1) , H(2) have countable basis & both countable or nite
H has countable basis.]

Tensor product of kets and operators

Kets:
If
| =
| =

(1)

ci |i H(1)
(2)

di |j H(2)
22

are kets in H(1) , H(2) .

then
| | =

ci dj |i,j H

i,j

is in H = H(1) H(2) . [Note: not all vectors in H are of tensor product form. Ex. |1,1 +
|1,2 + |2,1]
Operators:
If A, B are operators on H(1) , H(2) , then we can construct
A B as an operator on H = H(1) H(2) through
(1)

(2)

(A B)|i,j = (A|i ) (B|j ) .

[denes A B on all of H by linearity]

If A, B are observables, then A B is an observable.

Summary of Tensor product spaces

H = H(1) H(2)
dim H = (dim H(1) )(dim H(2) )
(2)

Basis: |i,j = |(1) |j

(2)

i,j = (1) | j |

(1)
(2)
dj |j
Kets: |
Ci |i | =

| | =
Ci dj |i,j
Bras same

(1)
(2)
Ci,j (A|i ) (B|j )
Operators (A B) Ci,j |i,j =
Simple class of operators on H:
A 11 ,

11 B .

If A, B act on H(1) , H(2) , will often refer to these as just A, B when context is clear.
Useful relation:
(A B) (C D) = (AC) (BD) .
23

Note: [(A 11), (11 B)] = 0.

Notation: in many books, tensor product symbol is omitted

| | ||
A B AB .
CSCOs in tensor product spaces
If {A1 , A2 , . . . , Ak } are a CSCO for H(1) , & {B1 , . . . , B } are a CSCO for H(2) ,

then {A1 , . . . , Ak , B1 , . . . B } are a CSCO for H(1) H(2)

[[Ex. of notation A1 = A1 11.]]

Example of tensor products:

Two spin- 12 particles
(1)

(2)

Consider two spin- 12 particles with Hilbert spaces H2 , H2 .

The two-particle Hilbert space is
(1)

(2)

H = H2 H2 .
A basis for H is:
|++
|+
|+
|

=
=
=
=

|+1 |+2
|+1 |2
|1 |+2
|1 |2

Operators:
A complete set of commuting observables is
Sz(1) = Sz(1) 11
Sz(1) = 11 Sz(2) .

24

Consider operators

Sz = Sz(1) + Sz(2) .

0 0 0
0 0 0 0

=
0 0 0 0
0 0 0
Sz(1) Sz(2) = (S (1) )z 11)(11 S (2) )
= Sz(1) Sz(2)

+1

2
1
=

1
4

+1

Incompatible observables
If [A, B] =

0, then cannot simultaneously diagonize A, B .

Experiments

A= a
A

















B= b1
B= b2
...


B = bk

(Assume A, B, C nondegenerate)
i) Allow all bi to combine without measuring B
Probability (C = c) = |c|a|2
[B not measured]

25

C= c
C





ii) Measure Bi & allow all parts to combine

Probability (B = bi ) = |bi |a|2
Prob. (C = c given B = bi ) = |c|bi |2

Prob. (C = c) =
|c|bi |2 |bi |a|2
i

[when B measured]
=

c|bi bi |aa|bi bi |c

zi zi ,

zi = a|bi bi |c

know (

zi )( zi ) = |a|c|2 .

So prob. (C = c) does not depend on measurement of B when

zi zi = (
zi )(
zj ) .
i

Sucient condition: only one zi

= 0,
so either a|bi = 0 or c|bi = 0 for all but one value of i
Sucient condition: either [A, B] = 0 or [B, C] = 0.

Dispersion
For A an observable, | a state,
dene A = A A
A2 is dispersion of A.
A2 = A2 2AA + A2
= A2 A2
is variance (a.k.a. mean square deviation) of A.

If A| = a|
A2 = a2 a2 = 0
Variance measures fuzziness of state.
26

Example:

In state |+

1
=
2
0
4
2

1
= 1 0 2
0
2

Sz2

1 0

2
4

2 2

= 0.
4
4

2
1
2
4
Sz = 1 0
2
0
4

0
1
2
= 1 0
0
0
2
=

2
.
=
4
Uncertainty relation
If A, B are observables,
A2 B2 14 |[A, B]|2
Proof.
Schwartz:

(on A|,B|)

(|A)(A|)(|B)(B|)
(|A)(B|)(|B)(A|)

2
= | 12 [A, B] + 12 {A, B} |

[A, B]
+
{A, B}
= 14

imaginary
([A, B] skew-Hermitian)
[prob. 11.]

real
({A,B} Hermitian)

1
4

|[A, B]|2 + 41 |{A, B}|2

1
4

|[A, B]|2 .

Example:

In state | = |+.
Sz2 = 0
2
Sx2 =
4
27

1
4

1.4

|[Sz , Sx ]|2 =

1
4

|Sy |2 = 0 .

Position, momentum and translation

Until now, all explicit examples involved nite-dimensional matrices.

Generalize to continuous degrees of freedom.

Want to describe particle in 3D by wavefunction (x, y, z)

Simply to 1D: (x)

Want |(x)|2 dx = probability particle is in region dx.

dx

|(x)|2


Natural Hilbert space: L2 (R):

Square integrable functions |(x)|2 < .

[[To precisely dene, need Lebesgue measure, . . . ]]
Can do QM in this framework.
L2 (R) is a separable Hilbert space.
Typical observables on L(2) (R):
P[a,b] projection on interval [a, b]

f (x) ,
axb
(P[a,b] f )(x) =
0,
otherwise
Quote from Von Neumann:
Dirac, in several papers, as well as in his recently published book, has given a
representation of quantum mechanics which is scarcely to be surpassed in brevity and
elegance, and which is at the same time of invariant character. It is therefore perhaps
tting to advance a few arguments on behalf of our method, which deviates considerably
from that of Dirac.
The method of Dirac, mentioned above, (and this is overlooked today in a great part
of quantum mechanical literature, because of the clarity and elegance of the theory)

28

in no way satises the requirements of mathematical rigor not even if these are
reduced in a natural and proper fashion to the extent common elsewhere in theoretical
physics. For example, the method adheres to the ction that each self-adjoint operator
can be put in diagonal form. In the case of those operators for which this is not
actually the case, this requires the introduction of improper functions with selfcontradictory properties. The insertion of such a mathematical ction is frequently
necessary in Diracs approach, even though the problem at hand is merely one of
calculating numerically the result of a clearly dened experiment. There would be no
objection here if these concepts, which cannot be incorporated into the present day
framework of analysis, were intrinsically necessary for the physical theory. Thus, as
Newtonian mechanics rst brought about the development of the innitesimal calculus,
which, in its original form, was undoubtedly not self-consistent, so quantum mechanics
might suggest a new structure for our analysis of innitely many variables i.e.,
the mathematical technique would have to be changed, and not the physical theory.
But this is by no means the case. It should rather be pointed out that the quantum
mechanical Transformation theory can be established in a manner which is just as
clear and unied, but which is also without mathematical objections. It should be
emphasized that the correct structure need not consist in a mathematical renement
and explanation of the Dirac method, but rather that it requires a procedure diering
from the very beginning, namely, the reliance on the Hilbert theory of operators.

29