Brief review of Quantum Mechanics (QM)

Note: This is a collection of several formulae and facts that we will use throughout the course.
It is by no means a complete discussion of QM, nor will I attempt to justify any of it.
Let me start by discussing systems with a single quantum particle (we will call this the
electron, although it could be any quantum particle of nite mass).
Important facts:
the state of such a system is described by a wavefunction (r, t), which is generally a complex
function. Its physical meaning is that:
|(r, t)|
2
dV
is the probability to nd, at time t, the electron within a small volume dV centred at point r. From
this we can infer that the wavefunction must be continuous (probabilities cannot change abruptly)
and also normalized: _
dV |(r, t)|
2
= 1
The integral is over the entire space, and it simply says that the particle is certainly somewhere, so
the total probability to nd it in the universe must be 1.
the evolution of the wavefunction is governed by the Schr odinger equation (we will deal exclu-
sively with non-relativistic particles; for relativistic ones, one should use the Dirac equation):
i h

t
(r, t) =

H(r, t) =
_

h
2
2m

2
+ V (r)
_
(r, t)
Here, h = h/(2) is related to Plancks constant h,

H is an operator known as the Hamiltonian.
Operator simply means that it acts on the wavefunction and does something to it what precisely
is shown by the nal part. We see that there are two terms, the rst is related to the kinetic energy
and consists of taking derivatives from the wavefunction. More specically:
y
x

e
e
e

z
r
Fig 1. Spherical coordinate system.

2
=

2
x
2
+

2
y
2
+

2
z
2
in cartesian coordinates where r = xe
x
+ ye
y
+ ze
z
.
If wed rather use, for instance, spherical coordinates
where r = re
r
is specied by r, , (see gure), then:

2
=
1
r
2

r
_
r
2

r
_
+
1
r
2
sin

_
sin

_
+
1
r
2
sin
2

2
The second term in the Hamiltonian describes the po-
tential energy V (r) in which the particle moves we will
always take this to be time-independent in this course.
Much of what I say next is only valid for such time-
independent potentials, so be careful if you wish to ex-
tend this to time-dependent potentials.
1
for reasons which will become apparent soon, in this course we will mainly wish to know the
eigenstates and eigenfunctions of this Schr odinger equation. They have the general form:

E
(r, t) = e

i
h
Et

E
(r),
where E is an energy (known as an eigenenergy) for which the time-independent Schr odinger equa-
tion:

H
E
(r) = E
E
(r)
has a physically acceptable solution for
E
(r) (continuous, normalizable, etc). Note that |
E
(r, t)|
2
=
|
E
(r)|
2
, so if the particle happens to be in such a state, the probability to nd it at various places
does not change in time (this is one of the reasons why these states are very important).
Since there will generally be many possible solutions, we need to index the wavefunctions somehow.
Above I indexed them by their energy, but typically we will use some quantum numbers to tell us
which specic solution we are talking about.
The collection of all allowed values of E is called the spectrum. If a given E has a single
corresponding eigenfunction
E
(r), we call that state non-degenerate. If there are several linearly
independent such eigenfunctions, we call that level degenerate. I will use state, level, eigenstate
interchangeably, and all mean essentially the same thing. The state with the lowest energy is called
the ground-state; all other states are excited states.
Before going on, let me give you two simple examples:
(a) 1D particle in an innite well
1D means that we consider only motion along one axis, usually x-axis. This is a convenient
simplication, since once we know this solution, we can easily generalize to 2D and 3D motion in
similar potentials.
The innite quantum well is dened by:
V (x) =
_
0, 0 < x < L
, otherwise
and is a simple way of describing the particle in a box. Of course, an innite potential outside the
box is physically impossible, but this turns out to be a good approximation if that potential is much
much larger than the typical energy of the particle in the box.
Lets solve this. First, we note that the particle cannot be outside the box (no way to give it innite
energy), so the wavefunction must vanish there (no probability to nd the particle in those regions).
Since the wavefunction is continuous, it follows that we must have (x = 0) = (x = L) = 0.
Inside the box, the Schr odinger equation reads:

h
2
2m
d
2
(x)
dx
2
= E(x)
so were trying to nd for what values of E we have acceptable solutions. You should convince
yourself that there is no acceptable solution if E < 0 (ask me if this gives you trouble!). So then lets
look at E =
h
2
k
2
2m
> 0, with k > 0. The ODE is now trivial:
d
2
(x)
dx
2
= k
2
(x) (x) = a cos(kx) + b sin(kx)
2
where we must also have (x = 0) = (x = L) = 0. The rst condition gives us a = 0. We then
must have b = 0, otherwise = 0, which is not normalizable. For the second boundary condition to
work when b is nite, we must have sin(kL) = 0 kL = n, where n = 1, 2, ... is an integer. Note
that I threw n = 0 out. Is that right?
So the allowed solutions are indexed by n (this is our quantum number for this problem), and
read:
k
n
=
n
L
E
n
=
h
2
k
2
n
2m
= n
2
h
2

2
2mL
2
,
n
(x) =

2
L
sin(k
n
x)
The constant in
n
(x) is such that the eigenfunction is normalized. If you need some practice with
integrals, check that. If you need some practice with QM, solve this problem for a 2D and then a 3D
innite quantum well. In the latter cases, gure out what are the degeneracies of the various levels.
(b) 1D simple harmonic oscillator
In this case, V (x) =
m
2
x
2
2
is the potential of a harmonic oscillator, which, in the classical case,
would oscillate with period T = 2/. Here the particle can be anywhere < x < , so we
dont have explicit boundary conditions like in the previous problem. The solution still must be
normalizable, which means that (x) cannot blow up as |x| quite the opposite, it should
decrease fast enough so that the integral of |(x)|
2
is nite. This condition will now select the allowed
eigenfunctions and eigenenergies.
The only diculty here is to gure out what are the solutions of the ODE:

h
2
2m
d
2
(x)
dx
2
+
m
2
x
2
2
(x) = E(x)
The details are pure math and you can follow them in any standard QM textbook (this is why you
have to study ODEs and PDEs, though, to be able to go through this step). The end result is that
we again have a single quantum number, which is also an integer but now including 0: n = 0, 1, 2, ....,
the allowed energies are:
E
n
= h
_
n +
1
2
_
and the eigenfunctions are:

n
(x) =
1

2
n
n!
_
m
h
_1
4
e

mx
2
2 h
H
n
__
m
h
x
_
where H
n
(x) = (1)
n
e
x
2
d
n
dx
n
_
e
x
2
_
are the Hermite polynomials. This might look a bit scary, but if
you plug some values in, youll nd that H
0
(x) = 1, H
1
(x) = x, H
2
(x) = x
2
1, ... so they are simple
polynomials. The exponential that appears in the wavefunctions insures that they decrease fast as
|x| , as we know they have to. The rest is the normalization coecient.
In case youre experiencing some ODE-solving-induced fears just now, let me assure you that in
this course, we wont need to solve any ODE more complicated than in problem (a), which is the
simplest possible. For more complicated examples, like (b), I will either just give you the results
and youll have to wait to study the details in QM, if you havent already; or well work through it
together if they are simple enough. I dont expect this to be any problem, in the end.
In more complicated case, we may have more quantum numbers and more complicated spectra,
but the basic idea is the same.
3
I dont think we will use this at all, but just for completeness let me mention that the set of
eigenstates is complete, meaning that any (reasonably well-behaved) function can be written as a
linear combination of these eigenfunctions, i.e. there exist a unique set of numbers c

such that
f(r) =

(r),
where here would list all possible quantum numbers. By reasonably well-behaved I mean that f
should be continuous or at most have a few jump discontinuities, i.e. not do anything too crazy.
The reason why this property is very useful is because we can apply it to the initial wavefunction,
which is well-behaved:
(r, t = 0) =

(r) (r, t) =

i
h
Et

(r),
The second equality is a consequence of the linearity of Schr odingers equation, and shows that once
weve gured out the eigenfunctions and the spectrum, and what the constants c

are for the original

, we use the fact that the eigenfunctions are orthonormal:

(r)

(r) =
,
where
,
= 1 if all the quantum numbers for the two eigenfunctions are equal to each other, and
zero otherwise. You should convince yourself that this gives:
c

=
_
dV

(r)(r, t = 0)
as the correct coecients in the decomposition of the initial wavefunction.
So far I only discussed the spatial location of the particle. It turns out that quantum particles
also have spin this is a purely quantum mechanical property, without classical analog, but for
our purposes you can think of it as a tiny magnetic moment associated with the particle. Its value
is quantized, for example the electron has spin S = 1/2. In general, S can be any integer or half-
integer number, and for a given S, the projection of the spin along an axis can only take the values
S
z
= S, S + 1, ..., S 1, S, i.e. 2S + 1 values are possible. If there is no magnetic eld (which is
what we will assume for most of the course) all these orientations of the spin have the same energy.
(Well, one can add terms in the Hamiltonian to describe something known as spin-orbit coupling, in
which case states with dierent spin projections will have dierent energies, so my last statement is
not always true. However, this spin-orbit coupling is a relativistic correction, and we wont bother
with such fancy things in this course).
Going back to the electron, one usually calls the state with s
z
= +1/2 as spin up, and the one
with s
z
= 1/2 as spin down. The existence of both possibilities leads to extra degeneracies of
the spectrum. For instance, a complete specication of the state of an electron in the innite well
is characterized by both n and s
z
: I could have a spin-up electron in state n = 3, or a spin-down
electron in that state, and these are distinct because of the spin, even though they have precisely the
same spatial wavefunction. In terms of notation, I would write the complete wavefunction as:

(r, t) = (r, t)

where the last bit is dierent if = or =, i.e. it is a way to show which spin we have for this
electron. One could go into many more details, but I think this will suce for what we need.
4
Now, a few facts about systems with more identical quantum particles. If there are N
particles, the general state of the system will be described by something like:
( r
1
, ..., r
N
, t)

1
,...,
N
where the rst part has information about the probability of nding the various particles around their
corresponding locations, and the second part tells us what are their spins doing. In the absence of
magnetic elds, the spins cannot change and the evolution of the spatial part is given by a generalized
Schr odinger equation.
If there are no interactions between particles, then the Hamiltonian for the full system is

H =

N
i=1

H
i
, where

H
i
would be the Hamiltonian if only particle i existed, and depends only on r
i
.
If there are interactions, then we need to add interaction energies that depend on the positions of
two or more particles, so we can no longer split the Hamiltonian in a simple sum.
We will generally assume our electrons to be non-interacting (this should sound very crazy to
you, because we know that there are strong Coulomb repulsions between charged particles like the
electrons, so how come we can ignore them? Ill try to give you some physical answers as we go along,
but ask me more when we get there, if youre not satised). In any event, if we can take

H =

N
i=1

H
i
,
then you can convince yourself that the eigenfunctions for the whole system are simply products of
the eigenfunctions for each particle, and the eigenenergies are the sums of the individual particle
energies (as expected if there are no interactions).
For example, suppose we have two electrons in the innite well, one in level n and the other one in
level m. This is an eigenstate of the whole (two-particle) problem, with energy E
n,m
= E
n
+E
m
. You
might guess that the eigenfunction is simply
n,m
(x
1
, x
2
) =
n
(x
1
)
m
(x
2
), since I said that electron
1 is in level n and electron 2 is in level m. Mathematically, this guess satises the Schr odinger
equation and is normalizable etc. Physically, though, it has a major problem, namely that electrons
are indistinguishable so I cannot possibly say which one is in which level. In other words, I might
have, just as well,
m
(x
1
)
n
(x
2
) as a solution, but this is not distinguishable from the rst.
The way around this is that for electrons (and any other particles with half-integer spin, which
are called fermions), the correct wavefunction is the fully antisymmetrized combination of all per-
mutations we can make between which particle is in which state. For two electrons, this means that
the eigenfunction is, actually:

n,m
(x
1
, x
2
) =
1

2
[
n
(x
1
)
m
(x
2
)
m
(x
1
)
n
(x
2
)]
so that
n,m
(x
1
, x
2
) =
n,m
(x
2
, x
1
). This last equality shows that if I interchange the particles,
the wavefunction must change the sign. Note that the probability stays the same (because of the
square), as it should. For more particles (if they are fermions), I have to make sure that any time I
exchange any two of them, the wavefunction changes its sign.
For your information: if the particles have integer spin, they are called bosons and then we need
to take the symmetric combination, so that
n,m
(x
1
, x
2
) = +
n,m
(x
2
, x
1
) (but we wont deal with
many bosons in this course).
Back to our electrons: what I just said seems to imply that I cannot have both electrons in the
same level, since if n = m, obviously
n,n
(x
1
, x
2
) = 0. This is not quite right, and that is because we
forgot that we also have to specify the spin in order to fully specify the state.
So, lets rewind. Suppose both electrons have spin up, and one is in level n and one in m. Then,
the wavefunction could either be
n
(x
1
)
m
(x
2
)
1

2
(the last bits just show that both electrons
have spin up), or we could have
m
(x
1
)
n
(x
2
)
1

2
. The only way to make a fully antisymmetric
5
wave-function, now, is:

n,m,,
(x
1
, x
2
) =
1

2
[
n
(x
1
)
m
(x
2
)
m
(x
1
)
n
(x
2
)]
1

2
In other words, since both electrons have spin up, changing one with the other wont aect the spin
part but it will exchange x
2
x
1
, so the wavefunction can only be antisymmetric due to its spatial
part. Now we see that indeed, we cannot have n = m because then the wavefunction vanishes (so
it is not normalizable). In other words, we cannot put two spin-up electrons in the same level, but
theres no problem putting them on dierent levels n = m (in case youre wondering, the 1/

2 factor
is to normalize the total wavefunction. Check it!). Very similar arguments apply to two spin-down
electrons, and we arrive at the same conclusion: they cant both occupy the same level.
What if one electron has spin-up, and one has spin-down? Well, now I could have 4 possible
choices:
n
(x
1
)
m
(x
2
)
1

2
,
n
(x
1
)
m
(x
2
)
1

2
,
m
(x
1
)
n
(x
2
)
1

2
,
m
(x
1
)
n
(x
2
)
1

2
, i.e. ei-
ther electron 1 is in state n with spin up and then 2 is in m with spin-down, or all other permutations.
It turns out that we can make two fully anti-symmetrized functions out of these pieces. The
rst is known as the singlet, and is given by:

n,m,singlet
(x
1
, x
2
) [
n
(x
1
)
m
(x
2
) +
m
(x
1
)
n
(x
2
)] [
1

2
]
Here, the spatial part is unchanged if I interchange 1 2, but the spin part gets a minus sign, so
indeed the whole wavefunction is antisymmetric if I exchange the particles (exchanging means both
locations and spins).
Note that in this singlet state, I am allowed to put both particles in the same level n = m! So,
we are allowed to put two electrons in the same level, if one has spin up and one has spin down, and
the spins are in a so-called singlet (that antisymmetric combination).
I can also get an antisymmetric total wavefunction by making the spin part symmetric and the
spatial part antisymmetric. This belongs to the triplet:

n,m,triplet
(x
1
, x
2
) [
n
(x
1
)
m
(x
2
)
m
(x
1
)
n
(x
2
)] [
1

2
+
1

2
] .
In this case n = m is not allowed!
If we have time to discuss magnetic properties towards the end of the course, well discuss a bit
more about singlets and triplets, and spins in general. For most of the course, what I need you to
remember is that we can place at most 2 electrons in a non-degenerate level, one with spin-up and
one with spin-down. This is an illustration of the Pauli principle, which says that a quantum level
(where spin projection must also be specied) can be occupied by at most one fermion.
This short discussion should give you a hint why we prefer to work with non-interacting particles.
In this case, if we know the spectrum for a particle, we know what happens for any number of
particles: we just have to occupy levels making sure to not violate Paulis principle, and properly
antisymmetrize the products of individual states to get the wavefunction. The energy is just the sum
of the individual electron energies.
For interacting particles, we need to start from scratch and recalculate the spectrum by solving the
full Schr odinger equation any time we add a new particle, because the new particle will change the
states of all previously existing particles through interactions. So, the wavefunctions dont factorize
into simple products (plus the antisymmetrization) and the total energy is not simply a sum of
all the individual energies. Things are much much more dicult in that case, and in fact how to
properly deal with systems where interactions are big (i.e., roughly speaking, we cant treat them
by perturbation theory) is still the biggest challenge in condensed matter physics actually, probably
in all of physics.
6