Вы находитесь на странице: 1из 51

Quantum Mechanics

Christian Maes1
1 Instituut

voor Theoretische Fysica, KU Leuven


(Dated: October 3, 2016)

What follows are lecture notes with the course on quantum mechanics in the first
semester of the Physics Master at the KU Leuven. More traditional subjects and
the introductory formalism are not fully treated here. We refer to the textbooks
Quantum Mechanics (Second Edition) by Bransden and Joachain, to Quantum Mechanics, a modern development by Ballentine and to Modern Quantum Mechanics
(Second edition) by Sakurai as reference material for these subjects (as also treated
briefly in the notes that follow). The present text is the version of September
2016; a corrected enlarged version will come later as the beginning of a
more complete treatise for future use.
Keywords: quantum mechanics in its modern view

Electronic address: christian.maes@kuleuven.be

ii
Contents

I. Introduction
II. Structure of non-relativistic quantum mechanics

1
2

A. Wave mechanics

B. Diracvon Neumann formalism

1. Hilbert space

2. Operators

3. Dirac formalism

10

C. Composite systems

11

D. Space-time structure of quantum mechanics

12

1. Space-time symmetries

13

2. Displacement and momentum

14

3. Angular momentum

16

E. Density matrix

17

1. Mixing

18

2. Restriction

19

3. Conditional

20

F. Entanglement

21

III. History and challenges


A. Problems with the classical theory

22
23

1. Black body radiation

23

2. Photo-electric effect

24

3. Spectroscopic data

26

B. Historical line

27

C. Challenges of today

28

IV. Exercises
V. Spin

29
32

A. Pauli equation

32

B. Stern-Gerlach in the pilot-wave picture

33

iii
C. Qubit

35

D. Addition of angular momentum

37

1. Addition of two spins 1/2

38

2. Addition of two spins 1

39

3. Spinorbit addition

40

E. Spin resonance phenomena


VI. Exercises
VII. Perturbation theory

40
41
43

A. Rayleigh-Schrodinger perturbation theory

43

B. Variational method

45

C. WKBJ-method

45

D. Time-dependent perturbation theory

45

VIII. Density functional theory

48

A. Thomas-Fermi theory

48

B. Hartree-Fock

48

C. Kohn-Sham

48

IX. Semi-classical approximation

48

Lecture One
I.

INTRODUCTION

Quantum mechanics is part of the revolution in the physics of the twentieth century. As
a summary of systematic departures from classical mechanics that are manifested through
quantum mechanics we can quote
its radical particle point of view, called quantization, elucidating the corpuscular nature of light and interaction, etc. That is for example responsible for the discrete
nature of spectral lines, for the photo-electric effect, etc.
the very nonclassical one-particle behaviour allowing tunnelling, overcoming potential
barriers, important for decay, scattering and radiation processes.
the presence of discrete inner degrees of freedom that interact with electromagnetic
fields, via the so called spin, whose origin is relativistic.
new properties of matter, like magnetism, superconductivity, superfluidity,...
the deviation from Maxwell-Boltzmann statistics, realizing Bose-Einstein and FermiDirac distributions. Think of ultracold gases with the phenomenon of Bose-Einstein
condensation or about the theory of electron conduction, or about neutron stars etc.
its weird nonlocal multiple body influence, where entanglement and many-body coherence reigns.
Since its beginning around 1900, much of the understanding has evolved and many things
and perspectives have changed and have been influenced by other developments. The present
course turns towards the future, as we feel much more is to be explored. The true quantum
revolution is yet to start, especially where it concerns many-body effects, critical phenomena,
nonequilibrium complexity and space-time structure. Also concerning quantum technology,
like for quantum computing and transport, hold Bachmans words you aint seen nothing
yet. Here we restrict to nonrelativistic quantum mechanics thus not treating the origin of
spin, and not formulating the theory on a relativistic space-time.

2
II.

STRUCTURE OF NON-RELATIVISTIC QUANTUM MECHANICS


A.

Wave mechanics

We are dealing here with N point particles with masses m1 , m2 , . . . , mN . We would like
to give a theoretical framework to specify their positions as time proceeds in the background
of classical (Galilean) space-time. That turns out to be more difficult and maybe even
rather weird compared to the classical equations of motion we are used to. We will find
for example that the dynamics of the positions is (or, can be) nonlocal, which is perhaps
pointing to some even more involved (not understood) quantum aspect of space-time.
One source of inspiration, experimentally realized for example in diffraction experiments
with electrons1 is that even material points appear to be guided in their motion by
specifying some wavefront. That was the idea of de Broglie who associated first in a rather
ad hoc way a wavelength = h/p to a particle with momentum p. Schrodinger took the
ideas of Hamilton in unifying geometric optics with the laws of classical mechanics to its
logical conclusion and, from the de Broglie formula and combining with physical optics,
actually found the equation that governs the wave. Thus enters the wave function ; that is
a complex function on the coordinates x1 , x2 , . . . , xN R3 of the particles2 . If we have two
wave functions 1 and 2 , then any linear combination a1 1 + a2 2 should count as a wave
function also, which is typical for waves in fact. Yet we do not need to think that the wave
function is really representing a physical wave, nor we do not need to think of as some
field like in electromagnetism, but rather something that is part of the description of how
nature works. After all is a simultaneous function of all the particle positions (defined
on configuration space, not on physical space), and there need not be any factorization of
into one-particle functions like in (x1 , x2 ) = (x1 )(x2 ) for two particles.
The positions of the particles x1 , x2 , . . . , xN will of course be governing most of the physics
as about all possible observations or measurements will proceed via uncovering their values
1

Most famous are the 1927 experiments by Davisson and Germer in which slowly moving electrons hit a
crystalline nickel target. The same diffraction pattern as for X-rays came out.
Sometimes, especially important for interacting with electromagnetic fields one must remember that the
particle also carries internal degrees of freedom, called spin. Let us forget for the moment about the
physical origin of spin, but its representation in the wave function is simply to make it spinor-valued. For
example, for a spin 1/2 particle we would have (q) = (1 (q), 2 (q))T C2 . In what follows shortly we
continue with spinless particles, and come back later to the case with spin.

3
at certain moments with possibly huge detectors. The Born-rule says that the positions are
distributed with probability density = ||2 when the wave function of the system is ;
its complex conjugate is written as .

Classically again, the state of a system evolves according to the Newton equation of
motion, once we specify the forces. It is a second order differential equation for the position,
and its study makes what is called analytical mechanics. In quantum mechanics, we get
first order differential equations. Here is the Schrodinger equation, first for one particle with
mass m

~2 d2

(x, t) =
(x, t) + V (x) (x, t)
(II.1)
t
2m dx2
where in three dimensions with Cartesian coordinates x = (x(1), x(2), x(3)) we understand
~

2
2
2
d2
+
+
x = 2 =
dx
x(1)2 x(2)2 x(3)2
The Schrodinger equation (II.1) is a differential equation and the basic object of study in
(nonrelativistic) quantum mechanics. The mathematics is very much related to classical
subjects in analysis and algebra. After all, the linear differential operator in the righthand side (called, somewhat confusingly, the Hamiltonian H = ~2 /2m + V ) determines
the physics once we specify initial and boundary conditions. Supposing that (x, t) =
eiEt/~ (x), we get the eigenvalue equation
H = E
called the stationary Schrodinger equation, which naturally leads to quantization conditions
as imposed by boundary values for . Standing waves is just one example. Here the potential
V is realvalued and supposedly reflecting the physical situation (for example, respecting
the possible space-time symmetries); it is in many cases just what you would write for the
(classical) potential energy3 .
For N particles, we get essentially the same
" N
#
X ~2

i~ (q, t) =

xk + V (q) (q, t)
t
2m
k
k=1
3

(II.2)

The question of what we mean by energy is perhaps not so trivial, especially not in quantum mechanics.
It refers to many things at the same time, like related to conservation laws, to the generation of timeevolution and to the relation with work in thermodynamics. In general its measurement is indirect, via
heat, color, etc. and the way of speaking about it is mostly classical.

4
with q = (x1 , x2 , . . . , xN ). It is important to remark that the Schrodinger equation is linear,
(1)

(2)

i.e., when we take as initial condition the combination 0 = 0 + 0 , then for all times
t 0,
(1)

(2)

t = t + t
(k)

is a solution of the Schrodinger equation when t

(k)

is started from 0 , k = 1, 2. Thus, the

Schrodinger equation preserves superpositions (later more on that). We already mentioned


however that we better do not insist on thinking of as just a wave, again because it is not
defined in physical space R3 but on configuration space R3N .
We will also have to check in what sense that Schrodinger equation (II.2) respects the
classical space-time symmetries. For example, for preserving time-reversal symmetry it
appears we have to change the order of time together with taking complex conjugation (and
we need that V is real-valued). In general we will see that in order to have the Galilean
(classical) space-time invariance, we need to require that wave functions which differ by a
positionindependent phase factor describe separately the same physics.

So far we have described what will govern the particle dynamics (the Hamiltonian and
the wave function). Not it is time to specify the dynamics of the positions themselves as
describing events in classical space-time. Particle trajectories are a priori not problematic
at all4 ; on the contrary, they are very much part of the story as they complete the physical
picture. To understand the particle dynamics, we first make a remark. Define the velocity
field with mass matrix m for wave function as
v (q, t) :=

~
Im
(q, t),
m

Im z :=

1
(z z )
2i

and note that it enters the continuity equation

||2 + (v ||2 ) = 0
t
for conservation of probability. Indeed, for conservation of probability (following the Born
4

There is interesting mathematics to study the well-behavedness of solutions on which we do not dwell here.
As for the concept of particle position itself many people appear to say we do not need these trajectories
at all. To write the equations of motion for the positions is then called Bohmian mechanics see for
example Making Sense of Quantum Mechanics, Springer, 2016, by Jean Bricmont.

5
rule) we get directly from the Schrodinger equation (II.2) the quantum flux equation

||2 + J = 0,
t

J := v ||2

with components
Jk =

~
( k k )
2imk

Another way of writing that is to put the polar decomposition = R exp[ ~ S] with R and
S real functions, so that v = S/m, determined by the phase factor S.
Therefore, we can safely and nicely put the differential equation for the positions
d
Q(t) = v (Q(t), t),
dt

Q(t) = (X1 (t), . . . , XN (t))

(II.3)

or, for each particle


d
~
k
Xk (t) =
(Q(t), t),
Im
dt
mk

k = 1, . . . , N

(II.4)

By the above structure it is assured that if at some initial moment the particle positions
are distributed with |(q, t = 0)|2 , then at any later moment t, as they have followed (II.3),
they are distributed with density |(q, t)|2 where (q, t) is the solution of the Schrodinger
equation (II.2) with (q, t = 0) as initial condition. The equation (II.4) is a first order
differential equation where the right-hand side depends however a priori on all the other
particle positions. That makes the dynamics very non-classical. Note also that the velocity
vector v is certainly not linear in 5 .
Most importantly, by the above structure it is assured that if at some initial moment the
particle positions are distributed with |(q, t = 0)|2 , then at any later moment t, as they
have followed (II.3), they are distributed with density |(q, t)|2 where (q, t) is the solution
of the Schrodinger equation (II.2) with (q, t = 0) as initial condition.

That is all for the structure of the dynamics,... the rest is computation and exploration.
As a simple example, we can see what happens in the double slit experiment. In Fig.1 you see
the particle trajectories coming from the double slit as in the standard Young experiment.
Think of electrons, coming from the left, say at a rate of one every year. Each electron follows
5

The velocity will be determined by that part of the wave function where the particles actually are. That
is an example of effective collapse of the wave function.

FIG. 1: 80 quantum trajectories going through the two-slit experiment. Taken from
Observing the Average Trajectories of Single Photons in a Two-Slit Interferometer in
Science 332, 11701173 (2011).
a very non-classical trajectory, passing through one of the holes and making a scintillation
point on the right screen. As the initial position of each electron is different and drawn
following Borns rule from an initial wave packet, there are also different trajectories. Each
trajectory is numerically computed using (II.3) where the wave function evolves from (II.2)
with boundary conditions taking into account the double slit architecture. The resulting
picture (of all the arrival points) makes the typical interference pattern on the screen.
Note that the wave function is symmetric relative to the axis of the arrangement while all
the same the particles pass one by one through only one slit. Another trivial example where
trajectories need not have the symmetry of wave function is the emission of an particle
from a nucleus. The wave function can very well be spherically symmetric like
1
2
(x, t) = eikrik t/(2m) ,
r

with k > 0, r = |x|

and yet each actual particle moves radially away from the origin in one specific direction.
I hope nobody is surprised; the wave function is not everything just like the Hamiltonian is
not everything.

7
B.

Diracvon Neumann formalism


1.

Hilbert space

The state space for the quantum system has been discussed above in the context of wave
R
mechanics. There the wave function is required to be square integrable, d3N q |(q)|2 <
, and we really care only about rays, i.e., constant phase factors can be ignored. More
generally one thinks of a Hilbert space to accommodate these wave vectors. A Hilbert
space H is a complex vector space, so you can add elements in it, and it is equipped with a
scalar (inner) product, h|i : H H C satisfying for all , H,
h|i 0

with h|i = 0 = = 0

h|1 + 2 i = h|1 i + h|2 i

(II.5)

h|i = h|i for all C


h|i = h|i
P
An easy example is Cn with hz|wi = nk=1 zk wk for zk , wk C. Or for the space of continuous
functions on the interval [a, b] we can take
Z b
dxf (x) g(x)
hf |gi =
a

We say that and are orthogonal if h|i = 0. An orthonormal sequence (kN ) has
hj |k i = jk .
A Hilbert space is a complete normed space with norm |||| coming from an inner product
p
|||| = h|i. We will always work with separable Hilbert spaces, allowing a countable
orthonormal basis. For our wave mechanics, the Hilbert space is
L2 (Rn , dn x) = { : Rn C with |||| < }
with n = 3N , but we can also work on bounded spaces of course. Note that the plane waves
eikx are not in the Hilbert space for unbounded domains, but mostly that is not a problem
as we we work in reality with wave packets.
The Fourier transform is defined on integrable functions via
Z
1 n/2

f (k) = ( )
dn x eikx f (x)
2

8
Remember for example that the Fourier transform of the gradient of f is obtained by multiplying the function f with the vector ik. Multiplying and taking derivative is then somehow
dual under Fourier transform. Similarly, there are various reciprocities; for example, the
variance of a Gaussian gets inverted via Fourier transform. That leads to the mathematical
representation of the so called uncertainty relations, which have however a more general
algebraic expression for operators6 .

2.

Operators

It is part of the usual quantum formalism to speak naively about observables as


self-adjoint operators. We need not do that, but at any rate it is interesting to see some
formal aspects of quantum mechanics in which operators enter.

An operator is called Hermitian (or symmetric) if


h|A|i = h|A|i

(II.6)

in analogy with a Hermitian matrix. Then, all its eigenvalues are real and eigenvectors
corresponding to distinct eigenvalues are orthogonal. A Hermitian operator in an infinitedimensional vector space may or may not possess a complete set of eigenvectors.
6

Speaking of that, for any two operators A and B, the Schrodinger uncertainty relation (1930) is that
2 1
2
1
(A)2 (B)2 h{A, B}i hAi hBi + [A, B]
2
2i
in terms of the commutator [A, B] := AB BA, the anticommutator {A, B} := AB +BA and the variance
(A)2 := hA2 i hAi2 . Follows the (Robertson) uncertainty relation (1929),
A B

1
|h [A, B] i
2

Heisenbergs relation (1927, with formal result obtained by Kennard and Weyl in 1928) uses A = X and
B = P to obtain a relation between variances of position and momenta operators which, for the context
of wave mechanics, is well known in Fourier theory. Since Heisenbergs attempt to refute the introduction
of unobservable variables in quantum mechanics, noncommutativity has been claimed to be the central
formal feature of quantum mechanics. Searches for quantum geometry or even quantum logic appear to
follow that line of thinking. It is important not to confuse the uncertainty relations, which are general and
useful inequalities, with other notions such as quantum uncertainty (related to the universal validity of the
Born rule, or to the notion of vacuum fluctuations) or such as quantum contextuality (which emphasizes
when needed the importance of the set-up of the measurement of an observable beyond the observables
mathematical representation).

9
If the adjoint of A, A = A, in the sense of (II.6), plus that they have the same domain as
operators, then we say that A is self-adjoint.

To each self-adjoint operator A corresponds a unique family of projection operators


E(), R with E() projecting onto the subspace corresponding to eigenvalues less or
equal to . We then have for all functions f ,
Z
f ()dE()
f (A) =

That becomes much simpler for matrices A, say when we have an orthonormal basis (|ei i)
of eigenvectors with eigenvalues i and corresponding projectors Pi = |ei i hei |:
f (A) =

f (i ) Pi

which is the simplest example of a spectral decomposition.

We often act with an operator on the wave function. The Hamiltonian is a differential
operator and we have already heard about the momentum operator (on which more later).
If we want to see what to expect for the position x of a particle we use the Born rule giving
probability density |(x)|2 . So expectations are
Z
d3 x f (x)(x) (x) = h|f |i
where we can think of the multiplication operator (x) 7 f (x)(x) symbolised by f |i and
hx|f |i = f (x)(x). Let us now consider more generally
Z
h|A|i = d3 x (x) A(x)
for an operator A = A which is self-adjoint. For example, we will like the operator P ,
later to be called the displacement operator, or momentum operator, the generator of space
translations,
Z
h|P |i =

d3 x (x) P (x),

P = i~

Clearly, as the wave function evolves according to the Schrodinger equation, t = U (t)
with U (t) = exp[itH/~] unitary, we have at time t,
ht |A|t i = hU (t)|A|U (t)i = h|U (t)AU (t)|i

10
and we get the Heisenberg picture for the evolution of operators
A At := eitH/~ A eitH/~

(II.7)

From here we can understand that operators and their algebra can go on to live their own
life, even far from any wave function. For example, when A commutes with H, [A, H] = 0
we obviously have At = A and the expectation ht |A|t i cannot depend on time, whatever
initial wave function .
The Heisenberg equation of motion underlies (II.7),
i

d
At = [H, At ] + At
dt
~
t

(II.8)

where the last term takes the internal time-dependence of the operator (like for a potential
that is time-dependent).

3.

Dirac formalism

For a given linear vector space H there is the dual space of linear functionals F on H,
for , H,
F (a + b) = aF () + bF ()
These linear functionals form again a linear vector space by defining (F1 + F2 )() := F1 () +
F2 (). The Riesz theorem states that there is a onetoone correspondence between linear
functionals F and vectors , such that all linear functionals have the form
F () = h|i
where is some fixed vector.
Dirac has been calling the vectors in H ket vectors, written then as |i, and the functionals
are called bra vectors, written hF |. The Riesz correspondence is antilinear in the sense that
c hF | c|F i
We can start to use that notation in the wildest senses. For example we write for the wave
functions (q) = hq|i, (q) = h|qi, and hq|q 0 i = (q q 0 ) with |qi hq| the orthogonal
projection on |qi. The one-particle wave function is then a superposition as
Z
Z
3
|i = d x |xi hx|i,
d3 x |xi hx| = 1

11
The position operator x (where the hat has nothing to do with Fourier transform now) is
introduced via its action on the basis |xi,
x|xi := x|xi
and differentiation operators are likewise
hx0 ||xi := (x0 x)x ,

hx0 ||xi := (x0 x)x2

If we write the momentum operator now like p = ~/i, then the commutator gives
[
x, p] = i~
proportional to the unit matrix in R3 . Furthermore, i[H, x] = p/m and therefore
pt
d
xt =
dt
m
Clearly that formalism can make abstraction of the specific representation of the Hilbert
space, and is very flexible and easy to use. It would however be a sad mistake to identify
quantum mechanics as physical theory with these manipulations.

C.

Composite systems

It is often essential to have a convenient and physically good way to include new degrees
of freedom. For example we need to add another system of particles or we must include
internal degrees of freedom. For that we make use of tensor products. Say for two particles,
we take the space of all linear combinations of products of wave functions. Of course,
a typical element of the tensor space cannot be written as a product but is of the form
P
P
(x, y) = k,` k` k ` (x, y) = k,` k` k (x)` (y). That is basically all, but let us add
some mathematics.
Suppose we have two Hilbert spaces H1 and H2 . There is the cartesian product H1 H2
with elements (1 , 2 ) for 1 H1 and 2 H2 . We define then a bilinear form 1 2 on
that cartesian product via
1 2 (1 , 2 ) = h1 , 1 i h2 , 2 i
with 1 H1 , 2 H2 . An inner product hi is defined from
h1 2 , 10 20 i = h1 , 10 i h2 , 20 i

12
and makes a new Hilbert space.
We often try to forget about these tensor products because of natural isomorphisms like
nk=1 L2 (R, dxk ) = L2 (Rn , dn x)
and the same with spinorvalued wave functions representing for example L2 (R3 , d3 x) C2
for one spin 1/2 particle. It will be interesting to learn to add spins, see further in Section
??.

D.

Space-time structure of quantum mechanics

Physical theories try to exploit essential symmetries of space-time. Classical space-time


is Galilei-invariant which gives essential constraints to the type of interactions, Hamiltonians
etc. The question appears how to take that into the quantum formalism. How is the waveparticle-structure of quantum mechanics compatible with the classical Galilei symmetries?
The general idea must be the following. Suppose we have a particle trajectory solution of
(II.4) and we consider a transformation of space-time. That transformation also transforms
the trajectory. Is that transformed trajectory again an allowed trajectory solving (II.4) for
the same Hamiltonian and transformed wave function? Here is an example, concerning timereversal symmetry. Suppose that a particle follows trajectory x(t) over times 0 t T . If
we reverse the time we see the trajectory x (t) = x(T t). That time-reversed trajectory
solves the equation of motion (II.4) for guiding wave . The complex conjugate solves
the Schrodinger equation (II.1) with the same Hamiltonian as originally when the potential
V is real. So we have time-reversal invariance when the potential is real, and we have understood why time-reversal induces complex conjugation in the Schrodinger representation.
To put it more formal, we have the lawmakers (H, ) (Hamiltonian and wave function) determining trajectories Q(t) (see (II.4)). A space-time transformation changes that trajectory
into Q0 (t) and we are asked to find a transformed wave function 0 so that (1) it satisfies the
Schrodinger equation for (the original) Hamiltonian H and (2) governs the dynamics of Q0
via (II.4). Then, (H, 0 ) give rise to the trajectory Q0 , and we say that the transformation
is indeed a space-time symmetry.

13
1.

Space-time symmetries

The traditional continuous symmetries7 form a ten-parameter family of rotations, translations, time-shifts and boosts. These act on the space-time coordinates which induce a
transformation for the wave function. Let us see how it remains a solution of the same
Schrodinger equation. To do that we forget now about the potential V in (II.2); it must be
physical and hence Galilean invariant (and real to have time-reversal symmetry) and better
takes care of itself. Let us look at the one-particle free equation instead
i~

~2

=
(q, t)
t
2m

(II.9)

For the space and time shifts it is easy to see that


0 (q 0 , t0 ) := (q 0 a, t0 s)
also satisfies the equation (II.9) but now in the primed variables. Same thing for
0 (q 0 , t0 ) := (Rq, t0 )
where R is a rotation; simply use that = is a scalar. For Galilean boosts we move to
a frame with relative velocity u, q 0 = q + ut, t0 = t. Let us again be more serious and start
from what needs to be done. For a given trajectory (say in one dimension) (X(t), 0 < t, T )
we define a new (candidate) trajectory Y (t) = X(t) + ut. Then, if the wave functions (x, t)
govern x(t) via (II.4), then (y, t) = exp i[ky (k)t] (y ut, t) (for no matter what (k))
will govern the motion of y(t), because substituting that in (II.4) we find
d
d
~k
Y (t) = X(t) +
dt
dt
m
which works when ~ k/m = u. So far we did not need a form for (k); that will come next.
We need to conclude by checking that is also a solution of (II.9). In higher dimensions
that becomes
0

0 (q 0 , t0 ) = ei[kq (k)t ] (q 0 ut0 , t0 )


where we will be using the choice m u = ~k. We calculate now (with q = q 0 ut0 )
i~

0 0 0
0
0
(q , t ) = i~ 0 {ei[kq (k)t ] (q 0 ut0 , t0 )}
0
t
t


~2
i[kq 0 (k)t0 ]
0
0
0
= e
~(q, t )
(q, t ) i~u (q, t )
2m

(II.10)
(II.11)

Mirror (or parity) symmetry, charge conjugation and time-reversal symmetry are not included here.

14
The last line indeed equals ~2 0 0 (q 0 , t0 )/2m when ~(k) = mu2 /2 or (k) = ~ k 2 /(2m).
Let us now look closer at these transformations on wave functions, which represent the
Galilei transformations. For example, the time-shift is generated by the Hamiltonian; that
is basically what the Schrodinger equation is saying. We can equivalently write
i~

= H t = eitH/~ 0
t

and the time-evolution is given by the bounded linear operator U (t) = eitH/~ . In fact
it defines a unitary one-parameter group U (t) : H H on the Hilbert space H of wave
R
functions with norm ||||2 = dq |(q)|2 , for which then
t 7 U (t) is continuous for each
U (t + s) = U (t) U (s)
kU (t)k = kk for all t R
and

U (t) = H U (t)
t
For the other transformations, there are similar representations. In fact there is a theorem
i~

by Wigner (1931) which says that if there is a transformation T : H H which is a


symmetry in the sense that |hT |T i| = |h|i|, then, essentially, T is either unitary
(linear) or anti-unitary (antilinear). So also rotations, space-displacements and boosts have
generators (writing the (anti)unitary operator as the exponential of a generator), and we
could investigate their algebra. Let us start with concentrating on the displacement operator
for now.

2.

Displacement and momentum

The displacement operator P acts on wave functions as the differentiation


P = i~

(II.12)

In rectangular coordinates, we have components Pj = i~/x(j). The reason is seen via


Taylor expansion, for a single particle,
i
(x + a) = (x) + a P (x) + O(a2 )
~

(II.13)

15
naturally.

We can now ask whether that operator has anything to do with momen-

tum, as we know it from classical mechanics.

In fact, it has been already tempting

to do that from the moment we have called H the Hamiltonian. After all, in classical
mechanics for a standard conservative dynamics the Hamiltonian is p2 /2m + V , and
comparing with the Hamiltonian in the Schrodinger equation we are led to believe there
is a relation between p2 (momentum squared) and ~2 ; (II.12) is exactly identifying those.

Consider therefore the spreading of a wave packet. Again we take the free evolution
(II.9), with ~ = m = 1. The initial wave function is a wave packet (x, 0) = 0 (x), x R3 .
The solution of the Schrodinger equation is
Z
(x, t) = d3 y 0 (y)

1
2
ei(xy) /t
3/2
(2it)

(II.14)

pretty much like you would solve the diffusion equation (but do not forget the imaginary
unit). You can try to continue for Gaussian wave packets, but in general getting an exact
solution becomes tedious. Here we are only interested in the large time asymptotics and we
can show that as t ,
(x, t) '

1
ix2 /2t x
e
0 ( )
(it)3/2
t

(II.15)

where the Fourier transform 0 of 0 has appeared8 . But what is then the probability
distribution of the asymptotic velocity X(t)/t of the particle t ? Well, expectations of
a function of that asymptotic velocity can now be computed by inserting (II.15) into
Z

x
d x g( ) |(x, t)|2 '
t
3

d3 x x x 2
g( ) |0 ( )| =
t3
t
t

d3 v g(v) |0 (v)|2

so that it follows that |0 |2 is the probability density in velocity space. For example, the
average asymptotic velocity is
Z

d3 v v |0 (v)|2 = h0 |P 0 i

(II.16)

by a simple application of Fourier transforms. In that sense, the expectation of the


displacement operator gives information about the asymptotic velocity of the particle. That
8

We do not give the derivation here of (II.15). One method is referred to as the stationary phase argument.
The reader should check the Gaussian case.

16
momentum need not be related (or proportional) to the instantaneous velocity.

Such relations can be lifted to include many other observables and the student should
understand from here the attempted identification of certain classes of operators with observables. That interpretation must be taken with the necessary care, because the way in
which an experiment or observation is actually carried out can greatly influence the type of
operator we must consider9 .

3.

Angular momentum

See for example, https://en.wikipedia.org/wiki/Angular momentum operator (last


access 19/09/2016). Or, see your standard textbook on quantum mechanics (e.g. read
Chapter 6 in Bransden & Joachain).

For rotation about axis = 1, 2, 3 in three space dimensions over angle we use the (rotation) operators x R ( )x on physical space, lifted to the unitary operators exp[i J ]
on function space. Again we find the J from Taylor expansion for small angles just like
in (II.13). The operators J are called angular momentum operators because they turn out
to be
J =QP
where Q is the position and P is the momentum (previous, displacement) operator. We can
now compute the commutation relations of the various components and with the square J 2 .
They are themselves defining an algebra of angular momenta which can have much broader
significance. In particular here is an opportunity to introduce (quantum) spin (angular
momentum) when we no longer forget about the internal contribution. For spinors, meaning
a ncolumn vector of complex-values wave functions, there appears indeed the possibility to
rotate also in the spinor-space, thus basically any transformation on Cn . That is no longer a
space-time symmetry however, so we need to postpone the discussion to a later lecture. We
will also need to discuss there how to add angular momenta, in Section V D, as important
for the composite systems of Section V D.
9

That is especially true for the quantum world where it got called contextuality.

17
E.

Density matrix

There is another object for quantum description that we need to introduce, called the
density matrix. Mathematically there is little problem. Think about a n n matrix
= (aij ), i, j = 1, . . . , n, with elements aij C for which aij = aji (Hermitian), for which
all eigenvalues are non-negative, and for which the trace is one. Those are called density
matrices.

You can use these density matrices to make normalized positive linear functionals on the
space of all Hermitian matrices10 : A 7 (A) TrA. The unit matrix gets mapped on the
number 1 and if A 0, then (A) 0. If you have an orthonormal basis (ei ) in Cn for
which is diagonal is, then
(A) =

n
X

pi hei |Aei i

i=1

where pi is the eigenvalue of with eigenvector ei . The inner product has been introduced
at (II.5). If also A is diagonal in the basis (ei ), with Aei = i ei , i R, then
(A) =

n
X

i pi

(II.17)

i=1

is the expectation value of the function : {1, . . . , n} R for the probability distribution
(pi ). It would appear then from the mathematics that a density matrix extends to the
quantum world the notion of probability, but that is only a formal analogy. Note also
that we get a pure state if and only if 2 = , then (A) = h, Ai really for some vector |i.

There are extensions of the previous math for infinite-dimensional spaces. We then have
a system with Hilbert space H and we call every non-negative bounded self-adjoint operator
: H H with trace TrD
= 1 a density matrix. Take for example the Hilbert space
D
H = L2 (B) of the complex square integrable functions on some domain B. Consider also
the space End(C) of linear maps (endomorphisms) C C. You can then see the density
on H and
matrix as a function D : B B End(C). The relation between the operator D
the function D on B B is

D(q)
=

dq 0 D(q, q 0 ) (q 0 )

B
10

In that sense, density matrices are superoperators as they act on operators.

18
is truly
You can, as inverse problem, try to find conditions here on the function D so that D
a density matrix but let us continue with giving some idea about possible physical meanings.

1.

Mixing

A first instance where we meet density matrices is in the case of statistical mixtures. By
some ignorance or uncertainty we do not know the wave function of the system and we do
not want to guess too much. That means that the wave function is (effectively) random
with a probability distribution (d) on the unit sphere S(H) of the Hilbert space H. The
corresponding statistical density matrix11 is
Z
=
D
(d) |ih|
S(H)

It is good to realize that different probability distributions possibly lead to the same
That is not a problem for
density matrix. For different (d) you then get the same D.
the usual statistical analysis of experiments. Imagine you want to measure something that
you have been able to associate to an operator A. The probability to find an outcome for
measuring A in an interval Z R equals
Z
A (Z)]
(d) h|PA (Z)|i = Tr[DP
Prob[Z] =

(II.18)

S(H)

where PA (Z) is the projection on Z in the spectral decomposition of A. The reason for
(II.18) is that, with wave function , that probability first equals h|PA (Z)|i and we then
average with . Let us illustrate that in the context of (II.17).
We take again the set-up with Hilbert space H = Cn and the observables are n n Hermitian matrices. We choose a basis (|ei i) and we have a probability law (pi ). The Hermitian
matrix A has eigenvalues (i ), A|vi i = i |vi i. The eigenvectors (|vi i) form an orthonormal
basis. There is possibly degeneracy; suppose that the m n vectors |vj1 i, |vj2 i, . . . , |vjm i all
belong to eigenvalue (one of those i ): A |vjk i = |vjk i, k = 1, . . . , m. The projection on
the eigenspace for that is
PA () =

m
X

|vjk ihvjk |

k=1

11

That goes back to J. von Neumann, Wahrscheinlichkeitstheoretischer Aufbau der Quantenmechanik,


G
ottinger Nachrichten 1(10), 245272 (1927).

19
The probability that we measure for the observabele A the value , when the state of the
P
system is |ei i, is equal to k |hei |vjk i|2 . If the probability for |ei i equals pi , then we find the
probability for measuring to be given by
Prob[] =

pi

|hei |vjk i|2

A ()] as it should for the quantum formalism with


and that exactly equals Tr[DP
=
D

pi |ei ihei |

Physical realizations of mixtures are easily found in thermal physics. Quantum statistical
mechanics uses density matrices for calculating magnetizations etc. Take light; it can be in
various pure states of polarization. These are superpositions of the two possible helicities for
photons, right and left circular polarization. We can for example look at the state of vertical
polarization as a superposition of left and right polarization, which will pass unabsorbed
through a vertical linear polarizer while the right polarized state would be partly absorbed.
However, unpolarized light (such as the light you are now using) will be absorbed anyhow
and cannot be described as a superposition but rather as a mixture of left and right polarized
light, or, empirically equivalent, as a mixture of vertically and horizontally polarized photons.
Of course, we can turn a mixed state into a pure state by filtering.

2.

Restriction

A second situation where a density matrix is relevant is when a system S1 possibly gets
entangled with a second system S2 . That means you start from a composite system
S1 S2 with wave function (x1 , x2 ) in H1 H2 , and there is a priori no wave function that
corresponds only with H1 . If you want to get a state for the first system, you can integrate
out the degrees of freedom belonging to the second system S2 . You then look at the reduced
density matrix of the subsystem S1 which can be associated with tracing out the second
subsystem,
= Tr2 |ih|
D
For wave functions, with q2 Rn ,
D(q1 , q10 )

Z
=
Rn

dq2 (q1 , q2 ) (q10 , q2 )

20
If you measure something that only involves the first system S1 , then you can directly use
to compute the statistical outcome12 . Note however that the wave
the density matrix D
in
function evolves according to the Schrodinger equation while the density matrix D
general has no autonomous dynamics (if S1 and S2 are truly coupled). Today people often
use the von Neumann entropy
:= TrD
log D

s(D)
to get a measure of entanglement between the two subsystems. It can be a function of
time of course but it is also fascinating to discover its dependence on the architecture
depends on the volume of
and geometry (e.g. whether that entanglement measure s(D)
subsystem 1 or rather on the surface between subsystem 1 and 2, etc.).

Looking back at the example of polarized light we can have after some decay process two
photons travelling in opposite directions which are however entangled. Their joint state
is for example
1
(|R, Li + |L, Ri)
2
That is a pure state. However, if we only look at one photon, investigating the statistics of
one side, we see unpolarized light, that is the density matrix
1
(|RihR| + |Li hL|)
2
You can of course also start from a statistical density matrix for the composite system, as
in the previous section for mixtures, and then take the restriction to a subsystem. That is
a combination of mixture and restriction13 .

3.

Conditional

Imagine now again a coupled system S1 S2 for which you have the wave function
(x1 , x2 ); think of two particles simply. Imagine we measure the position corresponding
12

13

That construction goes back to L. Landau, Das D


ampfungsproblem in der Wellenmechanik. Z. Physik
45, 430441 (1927).
You can find it for example on p. 424 in J. von Neumann, Mathematical Foundations of Quantum
Mechanics, Princeton University Press, Princeton, New Jersey (1955). Translation of Mathematische
Grundlagen der Quantenmechanik, Springer-Verlag, Berlin (1932).

21
to the second particle to be X2 = Z. Here Z is some spatial point where we found the
second particle. You can then produce a density matrix by taking the restriction to the first
particle (with x1 R3 ),
(x1 , Z) (x01 , Z)
dx1 (x1 , Z) (x1 , Z)
R3

D(x1 , x01 ) = R

which of course depends on Z. You are not integrating out here as in the previous section;
rather, you condition. If Z would be time-dependent (e.g. via (II.4)), you will generate
a time-dependent density matrix even though the wave function is static. That is one
reason how it becomes a useful concept14 , to introduce time in a description which is timeindependent for the total wave function.

F.

Entanglement

An entangled system refers to a composite system, typically made of individual particles


that are thought of as being spatially separated. The state is entangled if it cannot be
factorized into a product of the particle states. As consequence, one particle (or spin etc.)
can then not be specified in its state without considering the other(s). That is again an
instance of the fact that the wave function is typically not a product of individual oneparticle wave functions.
It remains true that the state of the composite system can be a sum or superposition of
products of states of local constituents. However not all states are separable. That means
the following. Consider the state
|i =

ck,` |ki |`i

k,`

in a two-particle composite system. The state is separable if ck,` = ak b` ; inseparable states


are called entangled states.

Let us take two qubits (say, two spin 1/2 systems referred to as A and B-subsystems)
with basis vectors |0iA , |1iA and two basis vectors |0iB , |1iB . An entangled state is
1
(|0iA |1iB |1iA |0iB )
2
14

See for example in D. D


urr, S. Goldstein and N. Zangh`, Quantum equilibrium and the origin of absolute
uncertainty. J. Stat. Phys. 67, 843907 (1992).

22
Now we cannot attribute to either system A or system B a definite pure state; the restriction
to A is described by density matrix

1
2

(|0iA h0| + |1iA h1|). For example the von Neumann

entropy of the subsystems is greater than zero, a mark of being entangled.


Suppose Alice observes system A, and Bob is an observer for system B. Let Alice make a
measurement in the eigenbasis |0i, |1i of A. She will measure two possible outcomes with
equal probability. If it is 0, sure Bob measures 1 at the other side, and same for measuring
1. But if Alice measures spin in another basis, then Bob will measure 1/2 of the times 0 and
1/2 of the times 1. So what Bob measures is influenced by the type of measurement Alice is
making in flight and even if the systems A and B are spatially separated. That is related
to the so called EPR paper that we will discuss later. Remember though that the outcome
of Alices measurement remains random and no information can be transmitted to Bob by
acting on her system.

III.

HISTORY AND CHALLENGES

Quantum mechanics is one of the main scientific revolutions of the 20th century15 . It
mostly addresses the (sub)atomic world.

It has started from problems with empirical

facts, mostly related to the interaction of light with matter, and inspired by the rise of the
atomistic picture of matter.

Various new discoveries were done towards the end of the 19th century. The majority
of physicists was experimental, occupied with electric measurements, heat phenomena,
spectroscopy and electric discharges in gases.
At the end of 1895 Wilhelm Rontgen discovered the famous X-rays (Nobel prize 1901).
Radioactivity was discovered a year later by Henri Becquerel (Nobel prize 1903). Even more
important is the year 1897, the official birth year of the electron. There had been of course
different theoretical indications for the existence of that first elementary particle. The most
important indication (especially worked out by Lorentz) was the so called Zeeman-effect
where the influence of a magnetic field causes a splitting of the spectrum of light emitted by
some gases (Nobel prize 1902). The electron got discovered finally by Joseph Thomson and
15

H. Kragh, Quantum Generations, a history of physics in the twentieth century, Princeton University Press,
Princeton, New Jersey (1999).

23
got elevated to be the universal elementary particle for all chemical elements. The electron
theory for metals was launched by Paul Drude (1863-1906). He combined there the study
of optical, electric and thermal properties of metals.

Specific examples include the problem of black body radiation, the specific heat of solids
and the discrete spectra of certain light sources. In general one distinguishes the old quantum
mechanics (till around 1925) from the new quantum mechanics, where phenomenological
rules of thumb were extended into systematic analysis of equations. A standard case is the
treatment of the hydrogen atom. Already in 1913 Niels Bohr gave a model of the atom from
which one could infer the frequency of emitted light, but it was only from the Schrodinger
equation (1926) that the energy spectrum could be derived.

A.

Problems with the classical theory

Quantum mechanics did not arise out of a theoretical idea which was unrelated to experimental evidence. Around 1850-1900 various problems appeared in the interaction of light
with matter. As a matter of fact, the three major revolutions of physics around the turn of
the 20th century all related to problems at intersections of classical pillars. Between electromagnetism and mechanics, relativity found its reason for existence as it was exceedingly
unclear how to unify classical mechanics with the motion of light and objects in electric and
magnetic fields. Statistical mechanics arose at the intersection of thermodynamics and mechanics for one wanted to understand the more microscopic and dynamical meaning of heat
and entropy. The laws of thermodynamics were no longer seen as absolute but emergent and
derivable from more mechanical considerations. Similarly, quantum mechanics originates in
the intersection between thermodynamics and electromagnetism. We give some examples.

1.

Black body radiation

Heating an object creates colours. One can explore that phenomenon more systematically
for objects that are very good emitters of light, so called black bodies. They are kept at
a temperature T and one looks at the (volume or) intensity of the outcoming light as a
function of the color. Since colour is really related to the frequency of the light, we get plots

24
of the intensity of the black body radiation as function of frequency. Such experiments were
carefully done and compared to theoretical predictions. The latter, based on a formula by
Rayleigh and Jeans, showed serious problems with the high frequency part of the radiation;
it is the famous ultraviolet catastrophe which predicts that almost all radiated power will
be in the high frequency domain. That was not the case observed experimentally.
It was Max Planck16 (1900) who discovered a way out, by postulating a quantization of
the energy, in that way breaking the continuum aspect of electromagnetic fields. The new
theoretical formula fitted the experimental curve very well with just one fitting parameter,
the so called Plancks constant h, with dimension of energy x time, or distance x momentum.
That new constant of nature will survive further developments and gives a measure of the
separation between the quantum world and the classical world.

2.

Photo-electric effect

This time we shine light on a metal plate to try to free electrons from the material
surface. That requires a certain energy, the so called work function or escape potential. The
electrons that are freed make an electric current that can be measured. Each electron has
a certain kinetic energy. A lot depends of course on the material but it can happen that if
we start with red light (low frequency), the electrons do not get out even when we increase
the intensity of the red light. That is strange if we think of light as a wave; strong waves
would break any wall. It turns out there is a minimal frequency and only from then on are
electrons able to escape from the material. So what did not work for red light, does work
for blue or violet light. As we further increase the intensity of the light, more electrons get
free but their individual kinetic energy does not increase. What is happening?
Einstein17 got the idea to think of monochromatic light as consisting of particles (light
particles, later called photons) that each have an energy proportional to the frequency,
E = h, for E energy, and the frequency. Each photon can free an electron, but it needs
sufficient energy, hence a sufficiently high frequency. That explains the phenomenon of the
16

17

The book Black-Body Theory and the Quantum Discontinuity, 1894-1912 from T.S. Kuhn in 1978 is an
extensive study of the research programme of Planck and of the history of the first quantum theory.
Ueber einen die Erzeugung und Verwandlung des Lichtes betreffenden heuristischen Gesichtspunkt in
Annalen der Physik 17, 132148 (1905)

25
photo-electric effect.
That light consists of particles was an old idea; Newton already entertained that idea
and thought of light as granular with different particles for different colours. Yet, many
experiments (diffraction and interference) had convinced scientists that light was a wave.
To say that light consists of particles appeared therefore as a serious challenge and started
the many reflections on the wave versus particle properties of light, and in fact of everything.

Let us visit therefore the initial arguments of Einstein.


The approach of Einstein differed greatly from that of Planck. Einstein concentrated himself
on the experimental part of the radiation law that was in strong conflict with the classical
calculations. The origin of the problem was the classical thinking together of two quite
different physical objects, particle on the one hand and the electromagnetic field on the
other hand. The latter has many more, a continuum of, degrees of freedom. For equilibrium
and equipartition, all excitement enters the field18 .
When h/kB T 1, big frequencies, we get Wiens law
e(T, ) '

8h 3 h/kB T
e
c3

(III.1)

for the energy density at temperature T and at frequency . Einstein writes the total entropy
Z +
s(, e)d
(III.2)
Stot = V
0

as an integral over entropy densities corresponding to frequencies and energy density


e = e(T, ). We can compute that entropy density from the thermodynamic relation
1
s(, e)
=
T
e(T, )

(III.3)

The result is
s(, e) =


kB e
c3 e
log
1
3
h
4h

(III.4)

Let us now work mono-chromatically. The entropy in volume V is then S = V s,


S = S(, E, V ) =
18


kB E
c3 E
log
1
3
h
4h V

The role of thermodynamics in the thinking of Einstein is summarized in M.J. Klein, Thermodynamics
in Einsteins Thought, Science 157, 509516 (1967).

26
and E = E(T, ) = V e is the energy in the cavity at frequency . Comparing that with the
entropy for a subvolume V0 , we get the difference
S S0 =

V
kB E
log
h
V0

(III.5)

Einstein remembers that for N particles in a volume V , the (ideal gas) entropy equals
S = kB N log V , and for a change V0 V ,
S S0 = N kB log

V
V0

(III.6)

Einstein concludes that the entropy of monochromatic radiation varies with the volume as
the entropy of an ideal gas, upon identifying
E = N h
The N is now the number of light quanta19 with frequency .

3.

Spectroscopic data

When atoms get excited, they emit light. For example, a hydrogen gas, when placed
in a tube with electric discharges will start to emit light. It turns out that the emitted
light consists of a discrete number of frequencies and moreover, that the lines (the various
frequencies) are forming a pattern their separation follows a specific mathematical rule.
Clearly that must be related to the structure of the atom. In fact, with the rebirth of the
atomic hypothesis came the question of how the atom looks like. A first idea was that
electrons are spread throughout an atomic matter. Yet, Rutherfords experiments showed
that the atom must have a nucleus. The atom could then be imagined as consisting of a
positively charged nucleus surrounded by negatively charged electrons.
Niels Bohr (1913) proposed a scheme that followed the old quantization methods. The
mathematics can even be followed easily. Imagine a nucleus with charge Ze (where Z is the
number of protons) and an electron with charge e in circular orbit with radius r around
it. For mechanical stability the centripetal force must equal the Coulomb force, or
Ze2
mv 2
=
r2
r
19

The word photon was introduced only in 1926 by the American chemist Gilbert Lewis.

27
As planets around the sun, also electrons are imagined to move in a plane with conserved
angular momentum, that is
mvr = n

h
2

is an integer (n) multiple of Plancks constant h divided by 2 and m is the mass of the
electron. This quantization rule appears ad hoc but let us see what it implies. First of all, as
mvr can only take a certain sequence of values, it implies that not all orbits will be possible;
they are quantized at certain distances from the nucleus. To move from one to another orbit
is not quite described but at the same time, the electron orbit is now stable. Solving for v
we get
Ze2 = mrv 2 =

mrn2 h2
n2 h2
=
4 2 m2 r2
4 2 mr

and so
r=

n2 h2
,
4 2 mZe2

v=

2Ze2
,
nh

n = 1, 2, . . .

The total energy of the electron is kinetic energy plus potential energy,
E=

mv 2 Ze2
mv 2

=
2
r
2

and hence
E=

2 2 m2 e4
,
n2 h2

n = 1, 2, . . .

We see that the quantization of the angular momentum has lead to the quantization of
energy. If an electron jumps between orbits it will emit the energy difference corresponding
to two quantum numbers n. That seemed to go into the direction of an explanation of
the particular form of lines in the hydrogen spectrum. Obviously however the derivation
starting from quantizing some sort of planetary motion around the nucleus is rather naive
and possibly misleading.

B.

Historical line

As clear from the above, quantum mechanics started from specific problems in observations where light interacts with matter. The old quantum mechanics extends till
the mid-1920s, and consisted mostly of direct quantization attempts of an otherwise
classical mechanics. That changed around 1924-1926 with the birth of a new quantum
mechanics, based on dynamical equations. On the one hand, that was the matrix mechanics

28
of Werner Heisenberg, and on the other hand there was the wave equation of Erwin
Schrodinger. Around the middle of the 20th century appears the first theories of quantum
fields and the standard model of elementary particles is ready by the 1970s. A return to
interpretational questions, aided by new quantum technology, quantum information theory
and quantum optics appears from the 1980s, with for example the Nobel Prize in Physics
2012 awarded jointly to Serge Haroche and David J. Wineland for ground-breaking experimental methods that enable measuring and manipulation of individual quantum systems.

Each of these periods had a certain style and interpretation of what it all means. The
old school is referred to as the Kopenhagen interpretation (after the Danish physicist) Niels
Bohr, where Paul Dirac and John von Neumann were instrumental for its mathematical
formulations. Today, even though not fully aware of it, many physicists appear to adhere to
the so called many worlds interpretation, which goes back to Hugh Everett in 1957. Perhaps
even more widespread remains the attitude that quantum mechanics provides a workable
theory with extraordinary predictive power - no further questions asked.

C.

Challenges of today

The quantum revolution still has to start in many ways. As a caricature, the twentieth
century dealt with single or independent quantum particles; the twenty-first century is
challenged by many body quantum effects. There are multiple ways of illustrating that and
there are various manifestations.

In condensed matter physics there are many challenges related to strongly correlated
electrons.

The idea of quantum matter and quantum criticality requires fundamental

departures from the approximations that have been used so far. A great tool will be
quantum simulations but much of that is still plagued with many types of restrictions.
An example for condensed matter problems is high temperature superconductivity and the
combination of semi-conductor tools with quantum optics. For example, manufacturing
of ion traps to simulate quantum qubits, to make quantum memories and quantum error
correction codes is all part of the ongoing ambition.
A related field is quantum information; in quantum computational physics one attempts

29
to push the limits of decoherence. One likes to preserve the true quantum nature despite
contacts with large thermal baths. That is essential for making quantum machines.
Quantum nonequilibria are other examples where we are mostly left in the dark. One
needs good models for quantum transport and dissipation which show typical and useful
quantum properties. Phenomena like many body localization, quantum relaxation and
thermalization are fully under attention these days.
For large scale physics, there is another strong coupling problem which is referred to as
the problem of quantum gravity. After all we want physical cosmology, based on general
relativity, to be compatible with quantum mechanics. It is unclear how to construct a
quantum space-time that unifies these theories.

IV.

EXERCISES

1. Give an example of two different probabilities on the quantum state space that give
the same statistical density matrix = I/n on the Hilbert space Cn .
2. Show that the Born rule is dynamically compatible with the evolution (II.3).
3. Give the detailed calculations of the Galilei invariance of the Schrodinger equation.
Detail the (II.10)-calculation to the end.
4. In what sense is Plancks constant small? Compared to?
5. Go over the formul (II.14), (II.15) and (II.16) to check carefully. Do it for a Gaussian
wave packet.
6. Prove that the von Neumann entanglement entropy is equal for both subsystem A and
subsystem B.
7. Give the derivation of the Heisenberg equation of motion (II.8).
8. Calculate the entanglement entropy when you trace out one of the spins in a general
state in the subspace of a two spin one-half system where the two particles have
opposite spin. For which of the states in this subspace are the two spins maximally
entangled (i.e. the Von Neumann entanglement entropy is maximal)?

30
9. We consider a spin 1 particle (like which?). We pick a basis corresponding to the
three eigenvectors of the z-component sz , with eigenvalues +1, 0, 1 respectively. An
ensemble is described by density matrix

2 1 1
1

= 1 1 0
4

1 0 1
Is that really a density matrix? [Check!] Is it describing a pure or a mixed state?
What is the average value of sz and what is the standard deviation?
10. Argue that an arbitrary linear operator A on a tensor product Hilbert space H =
H1 H2 is a linear combination of operators like Z = X Y .
11. Xavier and Yves have a light source which emits two photons at a time, backtoback
in the laboratory frame. The ensemble is given by
1
= (|LLihLL| + |RRihRR|)
2
where Lis lefthanded polarization, and R means righthanded polarization. Yves
stays with the light-source at home while Xavier goes a lightyear away. The light
source is on. Yves measures the polarization of his photon and finds it is left-handed.
He quickly sends a note to Xavier that he is going to see a left-handed photon, sometime
later. Indeed finally Xavier sees his photon, and measures its polarization. He sends
a note back to Yves that his photon was lefthanded (which Yves already knew).
Now Yves changes the apparatus, without telling anyone, so that the ensemble is
1
= (|LLi + |RRi)(hLL| + hRR|)
2
Again the experiment is done and the result is identical to the first experiment.
Explain.
12. Consider a Hermitian matrix A, somehow associated to a physical quantity like for
spin measurement, with ei being the basis of eigenvectors,
Aei = i ei

31
Suppose the quantum vector of the system is
w=

cj ej

When we couple with an apparatus, we consider the quantum state as the tensor
product, say of the form
0 = 0 (z) w
with 0 (z) the wave function of the measuring device with variable z. We think of
z as the (macroscopic) position of the pointer; e.g. z = 0 means that the pointer is
centered. We take 0 as the initial quantum state where 0 is centered around z = 0.
Now the Schrodinger evolution runs with coupling
H = i A

= iA z
to see that the wave function t at
Solve the Schrodinger equation i t

time t is a superposition with macroscopically disjoint pointer positions (when t is not


too small).
13. Check explicitly the Heisenberg uncertainty relation. Relate it also to an inequality
between Fourier transforms.
14. Check the commutation relations between the components Lx , Ly , Lz of the angular
momentum. Introduce L2 and show also that it commutes with the three components. Introduce the raising and lowering operators L = Lx iLy and compute the
commutator between L+ and L .
15. Explain: when two Hermitian operators commute, a common set of eigenfunctions
exist.

32

Lecture Two
V.

SPIN

Spin is a typical quantum mechanical feature, perhaps most intrinsic to quantum


mechanics and, following the words of Pauli (1925) for the spin of the electron, giving a
mysterious Zweideutigkeit. It arises really in the interplay between relativity and quantum
mechanics, and it reaches specific Hamiltonians and theoretical discussions via the Dirac
equation (for four-component spinors and which is fully relativistic). The Pauli equation
is the non-relativistic limit of the Dirac equation20 , and here we see already something
spectacular: the spin interacts with an external electromagnetic field. That is why we talk
of spin as an intrinsic angular momentum. On the experimental side the Stern-Gerlach
experiment (1922) takes a beam of particles in an inhomogeneous magnetic field and finds
that the beam gets split into two or more discrete sub-beams. There we see experimentally
the interaction of some sort between the magnetic field and the spin, which indeed
appears to be well reproduced in the way of thinking of an intrinsic angular moment and
corresponding magnetic moment.

A.

Pauli equation

In the non-relativistic theoretical description one thinks now of the wave function as a
multi-component spinor, say a vector. The number of components is related to the type of
particle (its spin). We start with a particle of mass m and charge q and spin 1/2. So let us
look at the Pauli equation (the two-component Schrodinger equation in an electromagnetic
field) for vector potential A and scalar electric potential :



1
2
( (i~ qA)) + q = i~
2m
t
20

Dirac found that making the Schr


odinger equation compatible with special relativity requires introducing
multiple components for the wave function, i.e., spin.

33
where = (x , y , z ) are the Pauli matrices and

Since ( a)( b) = a b + i (a b) we can rewrite that as


H = i~

,
t


1
(i~ qA)2 q~ B + q
2m

for H =

in magnetic field B = A. The spin only affects the motion of the electron in the presence
of a magnetic field, and we speak about the induced magnetic moment = q~/(2m). Note
that the charge refers to the electron; for example, a silver atom is electrically neutral but
does inherit a magnetic moment from the spin of its one valence electron. The fact that the
silver beam splits in two when passing a strong heterogeneous magnetic field was important
proof of intrinsic angular momentum of the electron.

Remark: We can wonder whether there is another more mathematical reason why spin is
associated to angular momentum. The latter is directly connected (for its orbital version,
see Section II D 3) to rotation in physical space. But that orbital rotation induces a rotation
in the space of Hermitian matrices:
h(Rq) = U () h(q) U ()
where for q = (x, y, x) R3 ,

h(q) =

x iy

z
x + iy

is the Hermitian matrix associated to q; Rq is its rotation through an angle 2 about the
z-axis and

U () =

B.

ei

0
i

0 e

SU(2)

Stern-Gerlach in the pilot-wave picture

The trajectories are given from a generalization of (II.3), summing over the spin index.
We now have as velocity field
v =

~
(, )
Im
,
m
(, )

for (, ) = +
+ +

34

FIG. 2: from the paper of Travis Norsen.


It is again compatible/expressing the Born rule that (, ) is the conserved probability
density for the particle. Note that particles have position (in time) but no spin values; spin
is a property of the wave function (which becomes a spinor of some type).
Consider for example21 a wave function in the (yz)plane,
(y, z) = Aeiky (c+ +z + c z ) ,
0

= Aeik y

y < 0, w/2 < z < w/2



eiz c+ +z + eiz c z , y > 0, (y, z) T

(V.1)
(V.2)

where T stands for a triangular region around y = 0 = z with apex towards the positive
ydirection, where the field is interacting with the incident beam (plane-wave packet)
21

From T. Norsen, The Pilot-Wave Perspective on Spin. American Journal of Physics 82, 337348 (2014).

35
see Fig. 2.
The z are eigenfunctions of the Pauli matrix in the z-direction, hence proportional to
the eigenfunctions of the interaction Hamiltonian (spin part)
~ = bzz (y)
H = ~ B
~ = bz z (y) the inhomogeneous magnetic field in the z-direction, concentrated around
for B
y = 0 (here formally represented by the delta-function (y)). The coefficients c make the
linear combination. For the incident beam (y < 0) the particle will drift in the positive
y-direction,
~
~k
dX
=
y
dt
m
Then it encounters the field around y = 0 and we have
~

~k 0
~
dX
=
y +
|c+ |2 |c |2 z
dt
m
m
That has to be complemented with the initial position of the particle, in particular with
its initial zlocation in the beam, as distributed according to the incoming wave packet.
There are then two types of trajectories. If the particle enters at z > w(1/2 |c+ |2 ) it gets
deviated up, and otherwise it gets in the downward beam; see Fig. 2.
Note that when c+ = c in the initial wave packet, particles with initial lateral position
z > 0 will always be deflected upward whether b > 0 or b < 0. In other words, the same
initial wave function and the same particle with the same initial location will be said to
have up-spin in the z-direction when doing the experiment with b > 0 and will be told to
have down-spin when doing the experiment with b < 0. Spin values therefore cannot be
considered intrinsic states of the particle; the exact same initial state can yield either of the
two possible measurement outcomes, depending on which of the two possible experimental
devices is chosen for performing the experiment.

C.

Qubit

When forgetting about spatial freedom (and spreading of wave functions) we can treat
a system of spins (only). That is a typical set-up for the study of magnetism in condensed
matter physics. For example the qubit has state vector in C2 . Particles with spin 1/2 are
the proton, the neutron, the electron, the neutrino, etc. The basis is |0i and |1i representing

36
(1, 0) and (0, 1) respectively. An arbitrary state vector is |i = |0i + |1i with , C
and ||2 + ||2 = 1, or
|i = cos

|0i + sin ei |1i,


2
2

0 , 0 < 2

The state vectors thus span the unit sphere in R3 , which is also called the Bloch sphere.
N

For N spin 1/2 particles the (spin-)Hilbert space is C2 ; note how the dimension grows
exponentially with the number N of particles.

The most general density matrix for a single spin 1/2 system is
1
= (1 + a )
2
where a is a vector whose length is not greater than 1, and is the vector of the three Pauli
matrices.
Suppose that the system has a magnetic moment = ~ /2 and is in a constant magnetic
field B. Let us take the constant magnetic field in the z-direction so that H = B =
~B
z . Then, defining := B, we see that (t) = 12 (1 + at ) is given by
2
(t) = eiHt/~ eiHt/~ = eitz /2 eitz /2 .
Substituting and making use of the identity

eiz = cos + i sin z =

0
i

0 e

we find that


it/2
it/2
e
0
1 + az ax i ay
e
0
1


at = Tr
2
0 eit/2
ax + i ay 1 az
0 eit/2
= (ax cos t + ay sin t, ay cos t ax sin t, az ) .
Note that we have used Tr[ ] = 0 when 6= , so that Tr[(t)] = at .
x and y (numbers now, not to be confused with the Pauli matrices).
x = 1. In other words, that value of y does not imply anything about outcomes for
x . One is inclined to say here that it is the measurement of y that must be
x . We now empirically find that in one hundred percent of the cases x = 1 always.
y = 1 and 50 percent of the electrons had y = 1. By bringing them together again in

37
the x -apparatus, then 50 percent of those with y = 1 will show the value x = 1 y = 1
or y = 1, only that its wave function is in a symmetric combination of y = +1 and
y = 1.

D.

Addition of angular momentum

In many interesting cases, for example in studies of magnetism, the spin degree of
freedom decouples from the translation degree of freedom and we are able to speak about
spin (and their finite-dimensional Hilbert space) without referring anymore to the physical
space R3 .
As a major application of working with composite systems (cf. Section V D) let us discuss
here the addition of angular momenta (see Section II D 3). The main point of departure
is that the sum of angular momenta is an angular momentum in the tensor product.
By angular momentum we mean each time, abstractly, a triplet of Hermitian operators
(J (1) , J (2) , J (3) ) satisfying the commutation relations
[J (i) , J (j) ] = i~ ijk J (k)
See for example https://en.wikipedia.org/wiki/Clebsch-Gordan coefficients
(last access 27/09/2016).
We have a priori a system with angular momentum J1 with eigenvectors |j1 m1 i and a second
system with angular momentum J2 with eigenvectors |j2 m2 i. We add J1 +J2 = J for the composite system. There we have of course the tensor product basis |j1 j2 m1 m2 i = |j1 m1 i |j2 m2 i.
But we also have a new (other basis) corresponding to the total angular momentum J; we
denote it by |j1 j2 jmi. To go from one basis to the other we need to compute the ClebschGordan coefficients
hj1 j2 m1 m2 |j1 j2 jmi
nothing but the coefficients of the unitary basis transformation.
Note that |j1 j2 jmi is an eigenstate of Jz with eigenvalue m1 + m2 , thence m = m1 + m2
necessarily for that coefficient not be zero. As a consequence, the maximal allowed value of
m is j1 + j2 . As always m can only take the values j, j + 1, . . . , j so that the maximal

38
allowed value for j is j1 + j2 . In fact, for j = j1 + j2 , m = j1 + j2 we have
|hj1 j2 j1 j2 |j1 j2 j1 + j2 j1 + j2 i| = 1
Take now the state with m = j1 + j2 1 so that there are two possibilities for m1 , m2 , these
are m1 = j1 , m2 = j2 1 and m1 = j1 1, m2 = j2 . That is compatible with j = j1 + j2 and
j = j1 + j2 1 (only). We continue in that way. Note also that the minimal value of j must
be |j1 j2 | because
j1 +j2

(2j + 1) = (2j1 + 1))(2j2 + 1)

j=|j1 j2 |

is the required dimension. To determine the remaining Clebsch-Gordan coefficients we can


use the raising and lowering operators J = Jx iJy to obtain recursion relations. We
illustrate that now.

1.

Addition of two spins 1/2

Let S = S1 1 + 1 S2 = S1 + S2 correspond to adding the spin of two spin 1/2 objects.



The composed system C2 C2 has four dimensions. We can make a basis by taking products of the one-particle basis, say (1), (1) for the first spin and (2), (2) for the second
spin, respectively for spin m1 = 1/2 and m1 = 1/2, and spin m2 = 1/2 and m2 = 1/2.
We have then basis vectors (1)(2), (1)(2), (1)(2) and (1)(2), with m = 1, 0, 0, 1
respectively. Remember that m = m1 + m2 and m1 = 1/2, m2 = 1/2.
The allowed values of the total spin are s = 0 and 1 (because |j1 j2 | j j1 + j2 in
general, and for each j we must have m = j, j + 1, . . . , j.)
Now go to the normalized eigenstates of S12 , S22 , S 2 and Sz , called s,m . We have the antisymmetric singlet state (s = 0)
1
00 = [(1)(2) (1)(2)]
2
and the triplet (s = 1) of symmetric eigenstates 11 = (1)(2),
1
10 = [(1)(2) + (1)(2)]
2
and 1,1 = (1)(2).

Note that S 11 = 2~ 10 , S 10 = 2~ 1,1 for the lowering operator S = S1, + S2, .


That is in fact how we find 10 and 1,1 from 11 .

39
2.

Addition of two spins 1

We are to add angular momenta j1 = 1 and j2 = 1 to form j = 2, 1, and 0 states, that is


expressing all (nine) |j mi eigenkets in terms of |j1 j2 m1 m2 i.
Beginning with the highest weight state, |j1 = 1, j2 = 1; j = 2, m = +2i := |2, +2i and
writing the tensor-product states as |j1 = 1, j2 = 1; m1 , m2 i := |m1 , m2 i with mi {+, , }
(in terms of which |2, +2i = |+, +i), we find that acting with the raising/lowering operator
on these states gives
p
j(j + 1) m(m 1)~ |j, m 1i ,
p
p
J |m1 , m2 i = 1 2 m1 (m1 1)~ |m1 1, m2 i + 1 2 m2 (m2 1)~ |m1 , m2 1i .
J |j, mi =

Therefore,
J |2, +2i = 2~ |2, +1i = J |+, +i =

2~ |+, i +

2~ |, +i

and so
1
1
|2, +1i = |+, i + |, +i .
2
2

Similarly, we find J |2, +1i = 6~ |2, 0i and


J |+, 0i =

2~ |, i +

2~ |+, i ,

J |0, +i =

2~ |, +i +

2~ |, i

which yields

1 
|2, 0i = |+, i + |, +i + 2 |, i .
6
Continuing on, we find that
1
1
|2, 1i = |, i + |, i ,
2
2

|2, 2i = |, i

in a similar fashion.
The highest-weight state in the j = 1 sector is a bit subtle because it is formed from a
linear combination of |+, 0i and |0, +i. Following the CSW conventions (detailed on p. 319
of Bransden & Joachain), we take the coefficient of |+, 0i to be real and positive:
1
1
|1, +1i = |+, i |, +i .
2
2

40
(That this is the appropriate relative combination of each can be verified by acting on each
side with J+ ; the overall normalization is set by requiring |h1, +1|1, +1i|2 = 1.) Acting on
the above with J as before, we find
1
1
|1, 0i = |+, i |, +i ,
2
2

1
1
|1, 1i = |, i |, i
2
2

as well.
Finally, the lone j = 0 state can be found by considering a combination
|0, 0i = a1 |+, i + a2 |, i + a3 |, +i
such that J+ |0, 0i = 0, |h0, 0|0, 0i|2 = 1, and a1 is real and positive. Doing so gives

1 
|0, 0i = |+, i + |, +i |, i .
3
3.

Spinorbit addition

To add the orbital and spin part of the angular momentum we take
J =L+S
We start with eigenfunctions
`sm` ms = Y`m` sms
for the operators L2 , S 2 , Lz , Sz , where Y`m` are the spherical harmonics and the sms are
column vectors with 2s + 1 entries having zeroes in all but one.
Moving to a basis for L2 , S 2 , J 2 , Jz we must write
jmj

Y`s

h`sm` ms |`sjmj i `sm` ms

m` ms

where the coefficients are the Clebsch-Gordan coefficients.

E.

Spin resonance phenomena

If we have a particle (like an electron again) with spin 1/2 there is a Zeeman effect in
which the energy levels split in the presence of a magnetic field. There is a separation of
energy
E = ge B B

41
for a free electron where ge = 2.0023 is the g-factor and B is the Bohr magneton. Such an
electron can now move between the two energy levels by abosrbing or emitting a photon of
energy h, with resonance condition
h = ge B B

(V.3)

If we expose a collection of paramagnetic centers, like free radicals, to microwaves at a fixed


frequency , we will resonance as we increase the magnetic field. Usually the lower energy
is more populated and a great absorbance will be measures when reaching the B for which
(V.3) holds.
Under thermodynamic equilibrium the distribution of radicals in the upper versus lower
level is governed by the Boltzmann distribution at a certain temperature.
Similarly, by using the spin of the nucleus one will see nuclear magnetic resonance.

VI.

EXERCISES

1. Estimate the strength of the magnetic field for obtaining spin resonance for an electron
under microwave radiation of 9400 MHz.
2. Consider the Hadamard gate (used in quantum information processing) which corresponds to the operator (on a qubit)
H=

|0i |1i
|0i + |1i

h0| +
h1|
2
2

Write it in terms of a 2 2 matrix in the |0i, |1i basis.


Show that
|0i + |1i

,
2
Now look at a bit-flip or NOT-gate,

H 2 |0i = |0i

H |0i =

x =

0 1
1 0

with x2 = I (identity), and find its square root S so that S 2 = x .


3. What is the unit of magnetic moment ?

42
4. Show in classical physics that any rotating distribution (on a ring) with uniform mass
and charge density will have a ratio

Q
=
L
2M
for its ratio of the magnitude of the dipole moment to the magnitude L of the
angular momentum, with Q the total charge and M the total mass.
5. Show that the (last) valence electron in the silver atom is in a 5s-state.
6. Show that two different Pauli matrices anticommute.
7. Show that the unique way to add two angular momenta to form a new angular momentum is to have = = 1 in J = J1 + J2 .

43

Lecture Three
VII.
A.

PERTURBATION THEORY

Rayleigh-Schr
odinger perturbation theory

We look at the eigenvalue equation


(Ho + H1 ) |n i = En |n i

(VII.1)

for the unperturbed Hamiltonian Ho for which we are supposed to know the eigenvectors
|ni and eigenvalues Eno :
Ho |ni = Eno |ni,

hn|n0 i = nn0

We assume smoothness around = 0 and we write


En = Eno + En(1) + 2 En(2) + . . .
|n i = |no i + n1 i + 2 |n2 i + . . .
Substituting that in the stationary equation (VII.1) we get
(H0 Eno ) |nr i = (En(1) H1 ) |nr1 i + En(2) |nr2 i + . . . En(r) |no i

(VII.2)

Write
|n i =

hn0 |n i |n0 i

n0

We deal first with the nondegenerate case where Eno 6= Eno0 for n 6= n0 .
Without loss of generality we choose
hno |n i = hn|n i = 1
(so that hn |n i 1). Since hn|ni = 1 we then have necessarily that hn|nr i = 0 for all
r 1.
As a consequence, acting with hn| on (VII.2),
En(1) = hn|H1 |ni,

|n1 i =

X |mi hm|H1 |ni


o
Eno Em
m6=n

44
En(2) =

X hH1 ||mi hm|H1 |ni


o
Eno Em
m6=n

and in general
En(r) = hn|H1 |nr1 i,

r>0

A simple but not very useful example is that of the charged harmonic oscillator
H=

P2
1
+ m 2 Q2 + Q
2m 2

where is proportional to the static electric field. The unperturbed oscillator has
1
Ho |ni = ~ (n + ),
2
and

r
hn|Q|n + 1i =

n = 0, 1, . . .

~
n + 1,
2m

hn|Q|n 1i =
(1)

~
n
2m

(2)

and all other matrix elements are zero. Therefore, En = 0 and En = 1/(2m 2 ) which is
natural since
H=

1
2
2
P2
+ m 2 (Q +
)

2m 2
m 2
2m 2

We go next to the degenerate case.


Here we write
Ho |nri = Eno |nri
where r = 1, 2, . . . , is the label of degeneracy. If we turn off the perturbation it is not clear
a priori whether and in what eigenstate of Ho we fall. Let us call these limiting states
o
|nr
i=

crr0 |nr0 i

r0

for unknown coefficients crr0 . From the first order we get


1
(1)
o
i
i = (Enr
H1 ) |nr
(Ho Eno )|nr

and hence
1
hns|Ho Eno |nr
i=

(1)
hns|Enr
H1 |nr0 i crr0

r0

or
X
r0

(1)
hns|H1 |nr0 i crr0 = Enr
crs

(VII.3)

45
That system has a nontrivial solution for the crr0 only if the determinant of the matrix
A is zero, with
(1)
us
Asu = hns|H1 |nui Ens
(1)

(1)

(1)

That characteristic equation has real roots, En1 , En2 , . . . , En , that we now need to
substitute in (VII.3) to find the crr0 .

B.

Variational method

If we have a self-adjoint Hamiltonian H and E0 is the lowest eigenvalue of H, then for


any we have the inequality
h|H|i
h|i
We can now take trial functions and parameterize them, and by varying them, see how to
E0

reach the lowest value.

C.
D.

WKBJ-method

Time-dependent perturbation theory

We go to the interaction representation by considering the states


|(t)iI = eiH0 t/~ |(t)i
for |(t)i solving the Schrodinger equation
i~

|(t)i = H|(t)i
t

Then, putting VI (t) := eiH0 t/~ V (t) eitH0 /~ , we check


i~

|(t)iI = VI (t)|(t)i
t

(VII.4)

Write now in the orthonormal basis of H0 ,


|(t)iI =

cn (t)|ni

with coefficients for which we find by inserting (VII.5) into (VII.4) that
i~ cm (t) =

X
n

Vmn (t) eimn t cn (t)

(VII.5)

46
with Vmn := hm|V (t)|ni, and mn := (Em En )/~ = nm .
(0)

(1)

(2)

Write cn (t) = cn +cn (t)+cn (t)+. . .. We want to obtain that expansion in the strength
of V .
Note that in the interaction picture
|(t)iI = UI (t, t0 )|(t0 )iI
and hence
i~

UI (t, t0 ) = VI (t) UI (t, t0 )


t

with UI (t0 , t0 ) = 1. Integrating we get the equivalent

Z t
i
UI (t, t0 ) = 1
dt0 VI (t0 ) UI (t0 , t0 )
~ t0
which can however be iterated to get
Z t
Z t
Z t0
i
1
0
0
0
0
UI (t, t0 ) = 1
dt VI (t ) 2
dt VI (t )
VI (t00 ) UI (t00 , t0 )
~ t0
~ t0
t0
and finally
UI (t, t0 ) =

n Z

X
i
n=0

t1

dt1

tn1

dt2 . . .

t0

t0

dtn VI (t1 ) VI (t2 ) . . . VI (tn )


t0

i
h
Rt
which defines the time-ordered exponential UI (t, t0 ) = T exp{ ~i t0 dt0 VI (t0 )} .
If now at time t = t0 we have the state |ii, then at time t
|i, t0 , ti = UI (t, t0 )|ii =

cn (t) |ni

n
(0)

with cn = ni ,
i
=
~

1
= 2
~

c(1)
n (t)

c(2)
n (t)

1
~2

i
dt hn|VI (t )|ii =
~
0

t0

t0

dt

t0

t0

XZ
m

dt00

t0

dt0

dt0 eini t Vni (t0 )

t0

hn|VI (t0 )|mi hm|VI (t00 )|ii

t0

t0

00

dt00 eimn t +imi t Vnm (t0 )Vmi (t00 )

47
etcetera. We have substituted Vnm (t) = hn|V (t)|mi and mn = (En Em )/~.
Note that |cn (t)|2 cannot be called the probability to jump from state |ii to state |ni;
rather it is the probability to find or to measure properties as being in the state |ni at the
later time t.

We often find perturbations with some goniometric periodicity. Let us take for some
time-interval 0 t T the harmonic perturbation
V (t) = heit + h eit
where h is some matrix. At times t > T we then have with respect to some final state |f i,
Z T
Z T
i
i
i(f i )t

c (t) = hf |h|ii
e
dt hf |h |ii
ei(f i +)t dt
~
~
0
0

i(f i )T
i(f i +)T
hf |h |ii 1 e
hf |h|ii 1 e
+
=
~
f i
~
f i +
(1)

Suppose now that |T |  1 (long duration) and take Ef > Ei so that f i > 0. Moreover
we make f i so that we keep only the resonant absorption. Then,
1
|c(1) (t)|2 = 2 |hf |h|ii|2
~

"

sin( 2 f i T )
1
( f i )
2

#2

is significant only if |Ef Ei ~| < 2~/T . Therefore, as T +,


1 (1) 2
2
|c (t)|
|hf |h|ii|2 (~ Ef + Ei )
T
~
It is however better to consider a transition in a continuum of states where n(E)dE is the
number of states in the energy range [E, E + dE]. We then conclude that we get a transition
rate of the form
2
|hf |h|ii|2 n(~ + Ei )
~

48
VIII.

DENSITY FUNCTIONAL THEORY


A.

Thomas-Fermi theory
B.
C.

IX.

Hartree-Fock
Kohn-Sham

SEMI-CLASSICAL APPROXIMATION

Вам также может понравиться