Matthew Schwartz

Lecture 20:
Quantum mechanics

1 Motivation: particle-wave duality

This has been a course about waves. We have seen how light acts like a wave in many contexts –
it refracts, diffracts, interferes, disperses, etc. Light also has a paricle-like nature. Newton was
the first to propose this, based mostly on observations about ray-tracing in geometrical optics.
For example, that shadows had sharp edges were evidence to Newton that light was made of
particles. This was in the 17th century. However, all of Newton’s observations are perfectly con-
sistent with light as a wave. Geometric optics is simply the regime when the wavelength of light
is much smaller than the relevant scales of the apparatus. Over the next 200 years, as the wave
theory was developed by physicists such Fresnel, Huygens and Maxwell, the particle-like nature
of light was basically accepted as wrong.
However, light is made of particles. An easy way to see this is with a digital camera. Digital
cameras are wonderfully more sensitive than film cameras. As we discussed in Lecture 17 on
color, as you pump up the “ISO” on a digital camera, you are pumping up the gain on the CCD.
That is, you change the mapping from intensity to the number 0-255 recorded at each pixel. At
very high ISO such as 6400, you can be sensitive to single photon depositions on your sensor.
Suppose we take a light source and add more and more filters in front of it to reduce the inten-
sity and take pictures of it. The result looks like this

Figure 1. Going from bottom right to top left, nine photos of the same scene are taken with more and
more light filtered out. The graininess of the images is mostly due to shot noise: there are a finite number
of photons hitting the camera.

The type of graininess that results from the finite number of photos in light is called shot noise
(There are other sources of graininess, such as sensor imperfections, but shot noise dominates at
low light).
If you thought light was just a wave, when you lower the intensity, you would expect the
photo to just get dimmer and dimmer at lower light. That there are dots in discrete places is
evidence that the intensity is produced by particles hitting the sensor. If you take the same pic-
ture again at low light, the dots show up in different places. So there is an element of random-
ness to the locations of the photons.

1
2 Section 1

We have seen that classical electromagnetic waves carry power. This is obvious from the fact
that they can induce currents. By setting electrons into motion, the energy of the wave goes into
kinetic energy in the electron motion. Electromagnetic waves also carry momentum. Again, this
is easy to see classically: if a wave hits an electron at rest, the oscillating electric field alone
would only get it moving up and down, like a cork on water when a wave passes. However, once
the electron is moving, the magnetic field in the wave pushes it forward, through the F = eB × v
force law. The electron is pushed in the direction of the wavevector ~k , thus the momentum in a
plane wave is proportional to ~k .
So if light is made of particles, those particles must be carrying momentum ~p, and this ~p
must be proportional to the wavevector ~k . We write ~p = ~ ~k . By dimensional analysis, this con-
h
stant ~ must have units of J · s. We write ~ = 2π and the value observed is

h = 6.26 × 1019 J · s (1)

~
known as Planck’sp constant. Since ω = c k for electromagnetic radiation, and since light is
2 2 2 2
massless, mc = E − c ~p = 0 we find that E = c|p~ | = ~c ~k = ~ω = hν, so energy is propor-
tional to frequency for light.
Of course, there were no digital cameras when quantum mechanics was invented. The key
experimental evidence for photons came from careful studies of shining light on metals. By 1900,
it was known that shining light could induce a current in a conductor. Interestingly though, the
induced current depended not just on the intensity of the light, but also on its frequency. For
example, in potassium, it was found that low intensity blue light would induce a current but
high-intensity red light would not. The current is induced in a metal because light hits elec-
trons, which are then freed to move around. In some situations, an even more dramatic effect
occurs: the electrons get ejected from the metal. This is called the photoelectric effect. In
1902, Philipp Lenard found that the energy of the ejected electrons did not depend on the inten-
sity of the light, but only on its frequency. A high frequency light source would free lots of elec-
trons, but they would all have the same energy, independent of the intensity of the light. A low-
frequency light source would never free electrons, even at very high intensity.
These observations are explained by quantum mechanics. The basic observation is that elec-
trons have a binding energy Ebind in a metal. One needs a minimal amount of energy to free
them. This binding energy is itself not a quantum effect. For example, we are bound the the
1 2 km
earth; it takes a minimum amount of energy Em in = 2 m vesc with vesc = 11.2 s the escape
velocity to set us free. Quantum mechanics plays a role through the identification of frequency
E
with energy: because photons have energy E = hν it takes a minimum frequency νm in = bhi n d to
free the electron. After that, the electron will have kinetic energy Ekin = Ebind − hν. So one can
make a precise prediction for how the velocity of the emitted electron depends on energy
1
2
m v 2 = Ebind − hν. This relation between v and ν was postulated by Albert Einstein in 1905
and observed by Robert Millikan in 1914. Both were later awarded Nobel prizes (as was
Lenard).
Two other important historical hints that classical theory was wrong were the classical insta-
bility of the atom and the blackbody paradox. The classical instability is the problem that if
atoms are electrons orbiting nuclei, like planets orbiting the sun, then the electrons must be
radiating power since they are accelerating. If you work out the numbers (as you did on the
problem set), you will see that the power emitted is enormous. So this classical atomic model is
untenable and must be wrong.
The blackbody paradox is that classically, one expects a hot gas to spread its energy evenly
over all available normal modes. This evenness can actually be proved from statistical
mechanics; the general result is known as the equipartition theorem. Now say you have light
in a finite volume, such as the sun. It will have normal mode excitations determined by
boundary conditions. Treating the volume like a box whose sides have length L, these will have
2π
wavevectors ~k = L ~n, where ~n is a vector of integers, e.g. ~n = (4, 5, 6). Since ω = c ~k in three
dimensions there are many more modes with higher frequency than lower frequency. If these all
Interference 3

have equal energy, the total intensity at high frequency would grow without bound. More pre-
cisely, the classical result is that I(ω) ∝ ω 4. This is known as the ultraviolet catastrophe.
One can see this ω 4 behavior of the spectrum of a blackbody for low frequencies, but experimen-
tally it was know by 1900 at least that the spectrum turned over, as shown in Figure 2.

Figure 2. The classical prediction for the blackbody spectrum diverges at high frequency.

In 1900, Max Planck noted that the observed spectrum could be fit by the curve now known as
the blackbody radiation formula (see Lecture 17). He later justified this fit by postulating that a
normal mode of frequency ω has energy E = ~ω and redoing the statistical mechanical calcula-
tion of the intensity spectrum to get his curve.

2 Interference
So light is made of particles, but has wavelike properties. How can this be? Let’s reconsider the
double slit interference pattern, as in Figure 3.

Figure 3. Double slit setup. Light passes through two slits and we see an interference pattern. The top
image on the right is the single-slit pattern, which follows if one of the slits is covered up. The bottom
image on the right is the result if both slits are open.

Recall from Lecture 19 that when a plane wave passes through two slits, there will be an
interference pattern produced. More precisely, if we have two slits separated by a distance d and
place ascreen a distance L from the slits, the intensity pattern on the screen will be I(y) =
dy
4I0cos2 π λL where y is the distance up the screen. The oscillations with y are the small black
ay

stripes you see on the bottom right of Figure 3 (the longer scale oscillation is the sinc π λL
pattern, due to the slit having finite width a).
4 Section 2

Since intensity is just photon counts, we expect this pattern will slowly build up if we shine
the light at low intensity. That is just what happens, as can be seen in Figure 4:

Figure 4. Double slit interference pattern at low intensity. The pattern accumulates slowly, going from
panel 3, to panel 2, to panel 1. Eventually, the granularity in the intensity pattern would wash out and it
would match what we get with high intensity light over a shorter time.

This figure is equivalent to zooming in on the central region in the bottom right panel of Figure
3 and lowering the intensity. It is also true that if you cover up one of the slits, you find the
single-slit interference pattern accumulating over time, as in the top right panel of Figure 3.
Now if you start to think about it is a little, this result is very weird. The intensity pattern
accumulates even if the intensity is very low, say one photon per minute. But if only one photon
is going through at a time, what is it interfering with? Certainly not other photons. The answer
is, it is interfering with itself! The photon must be going through both slits at once. Otherwise
the interference pattern couldn’t possibly be affected by closing off one slit.
To make this even clearer, you can do the same experiment with electrons. Exactly the same
thing happens: the electrons produce an interference pattern if both slits are open, but no inter- 
dy
ference pattern if one slit is closed. For the photon case, the pattern was I(y) = 4I0cos2 π λL
with λ the photon’s wavelength.
What is the “wavelength” of an electron? Well, for light momentum was ~p = ~k ~ and ~k = 2π .
λ
2π h
Thus, the wavelength of light can be extracted from its momentum as λ = ~ = |p~ | . It is natural
k
to expect that this equivalence will also work for electrons. This definition is called the de
Broglie wavelength
h
λ≡ (2)
|p
~|

This de Broglie wavelength correctly determines the scale for the interference pattern in the
electron double slit experiment.
1
Continuing this logic, the electron’s wavevector is ~k = ~ ~p and the electron’s angular fre-
1
quency is ω = ~ E, with E the electron’s energy. From special relatively, we know that the elec-
p
tron’s energy is E = m2c4 + p2c2 . Thus electrons satisfy the dispersion relation
p
2
~ω = m2c4 + ~2c2~k (3)

This has a low-frequency cutoff (like in a plasma): the frequency of the electron satisfies ω >
m c2
~
.
Another useful quantity with units of length is the Compton wavelength of the electron

h
λc = = 2.42 × 10−12 m (4)
mc
The Schrödinger equation 5

Unlike the de Broglie wavelength, the Compton wavelength does not depend on momentum.
The Compton wavelength is just a translation of the electron mass into length units and repre-
sents, roughly, the length scale where quantum mechanics becomes relevant. Calling it a wave-
length is somewhat inappropriate. It should more appropriatelly be called the Compton length
scale. The de Broglie wavelength really is the wavelength of the wave that a particle with
momentum p has. So all the interference effects and so on depend on the de Broglie wavelength.
The Compton wavelength does not directly characterize the scale for interference; it is merely a
length scale that comes up in a lot of quantum mechanics calculations, as in Eq. (17) below, so
it is handy to give it a name.

3 The Schrödinger equation

So light is made of particles that act like waves. Electrons are also particles which act like
waves. We will now focus mostly on electrons. The reason is that to we want to focus on just
one particle at a time, and with photons it’s hard to do that. Due to E = mc2 and that m = 0
for a photon, it is very easy to produce a lot of light. Indeed, there is no good regime where we
can study single photons without having to consider multiple photons. Quantum field theory
is quantum mechanics in regimes where energies are greater than the threshold for producing
pairs of new particles with E = mc2. For photons, this threshold energy is zero. For electrons, it
is twice the rest mass, 2me ≈ 1.5 × 10−13 J ≈ 1 MeV. This is an enormous amount of energy, not
generated in the laboratory until the 1930’s.
What is the wave equation for the electron? To deduce it, we can work backwards from the
dispersion relation. For electrons, plane waves ψ(x, t) = exp(iωt − ikx) with the dispersion rela-
tion in Eq. (3) would satisfy the wave equation

∂2
 
2 2~ 2 2 4
~ + ~ ∇ + m c ψ(x, t) = 0 (5)
∂t 2

with ψ(x, t) the amplitude of the electron, more commonly called its wavefunction. Eq. (5) is
called the Klein-Gordon equation.
The solutions to the Klein-Gordon equation are plane waves, or linear combinations of plane
waves. However, we are not usually interested in what happens to an electron pummeling
through free space. Instead, we like to study systems with lots of positively and negatively
charged particles, like atoms, molecules and solids. For example, around a proton, the electron is
q2
influenced by the background V (r) = 4πε r Coulomb potential.
0

So, let’s adapt the Klein-Gordon equation a little so it’s more practical. First of all, in most
situations, the kinetic energy of the electron is much less than its rest mass (if this weren’t true,
we’d need quantum field theory). So it makes sense to Taylor expand the mass-energy relation:
p2
p
E = m2c4 + p2c2 = mc2 + 2 m + ···. To include the stuff around the electron, we can include the
fact that the energy of the electron can also be modified in the presence of some potential V (x),
p2
like around a proton or a nucleus. So E = mc2 + 2 m + V (x) + ···. In fact, since only energy dif-
ferences are ever measurable, we might as well absorb the rest mass into this potential. So we
~p2 ~p2 1
now have E = 2m + V (x). Note that since ~p = m ~v the 2m = 2 m ~v 2 is just the ordinary non-rela-
E ~p
tivisitic kinetic energy. In terms of the frequency and wavevector, ω = ~ and ~k = ~ , this
~ 2
k
becomes ~ω = ~2 2m + V (x) and the wave equation becomes

~2 ~ 2
 
∂
i~ + ∇ − V (x) ψ(x, t) = 0 (6)
∂t 2m
6 Section 4

This is called the Schrödinger equation.

4 The hydrogen atom

As an example of the kind of things that can be calculated with quantum mechanics, consider
the hydrogen atom. The hydrogen atom is just a proton and an electron. Since the proton is
q2
2000 times heavier than the electron, we treat it as static, generating a potential V (r) = 4πε r .
0
For the electron, we need to solve the Schrödinger equation in the presence of this potential.
I’m not going to solve it here, since it takes months. I’ll just state the result. The general
solution can be written as
∞
X
~ , t) =
ψ(x eiωntFn(x
~) (7)
n=1

for some frequencies ωn and some functions Fn(x~ ) of ~x only. These functions Fn(x
~ ) are products
of Laguerre polynomials and spherical harmonics, but right now we don’t care. What we do care
about is that all the time dependence in these solutions has the form of a phase factor. Of
course, we can always do a decomposition like this. If we forget about the ~’s, all we have done
is find the normal modes of this wave equation. ωn are the normal-mode frequencies. The fea-
ture particular to the hydrogen atom is what the frequencies are. For the hydrogen atom,

mec2α2 1 1
ωn = − = −13.6 eV , n = 1, 2, 3··· (8)
2~n2 ~ n2
e2 1
where α = 4πε ≈ 137 is the fine-structure constant. There are n2 different solutions for each
0
value of ωn.
Now, we already said that energy is ~ times angular frequency, so the hydrogen atom appar-
1
ently has discrete energy levels, with energies En = −(13.6 eV) n2 . The solution with n = 1 is
called the ground state, since it has the lowest frequency. n > 1 are called excited states.
Note that the energies are higher (negative numbers closer to zero) for higher n. E = 0 corre-
sponds to a free proton and electron. When the two are bound together, the energy is lower (or
else they wouldn’t bind!). This means that if a photon is to come in and knock one of the elec-
trons out, it needs at least as much energy as the ground state. Thus photons must have a min-
imal frequency to knock out electrons, which agrees with observations about the photoelectric
effect.
If you heat up hydrogen, you can get the electrons into the higher n energy levels. If a
hydrogen atom is in an excited state, it can drop down to a lower energy state and emit a
photon. Alternatively, hydrogen can absorb photons to transition between levels. By energy
conservation, the photon energy must be the difference in energies of levels of the hydrogen
atom. Moreover, since energy is proportional to frequency, we predict that of photons coming
out of (or going into) hydrogen should have frequencies given by
 
(13.6eV ) 1 1
ν= − , n = 1, 2, 3, 4, ... m = 1, 2, 3, 4, ... (9)
h n 2 m2

The associated wavelength for the m → n transition is

 
1 1 1 1
= − (10)
λ 91.17 nanometers n2 m2

Many of these frequencies are not visible. But some are. The n → 2 transitions make up the
Balmer series. For example, 3 → 2 has wavelength 656.4 nm, 4 → 2 has wavelength 486.24 nm,
and so on. These agree perfectly with the observed spectrum of hydrogen, as shown in Figure 5.
Wavepackets 7

Figure 5. Visible hydrogen emission spectrum.

It is perhaps useful to note that conservation of energy translates to conservation of fre-

quency. This isn’t that different from what we have seen classically: a driven oscillator moves at
the frequency of the driver. As waves pass across different media, the frequency is fixed, but the
wavelength changes.

5 Wavepackets
The wavefunction ψ(x, t) is the amplitude. So |ψ(x, t)|2 gives the intensity. As we saw in Sec-
tion 2, however, the intensity for single photons or single electrons does not show up as a very
faint smooth pattern, but rather as discrete dots. The reason behind this is one of the hardest
parts of quantum mechanics to get your head around. The way it works is that |ψ(x, t)|2 gives
the probability of finding an electron at a position x and t. This probability interpretation is
often taken as a postulate of quantum mechanics, although really it follows from careful consid-
eration of entanglement with the surroundings.
Let’s just take the probability interpretation as given. For V = 0, the Schrödinger equation is
simply
~2 ~ 2
 
∂
i~ + ∇ ψ(x, t) = 0 (11)
∂t 2m

This is called the free Schrödinger equation. It has plane wave solutions
~
~ , t) = eiωt−ik ·x
ψ(x ~
(12)
~k ~2
where ω and ~k satisfy the dispersion relation ω = 2m . Where are the electrons in the plane
wave? The probability of finding one at a point ~x at time t is |ψ(x ~ , t)|2 = 1. Thus they are
everywhere! Clearly plane waves are not useful for describing electrons in any regime where they
act particle-like.
There are lots of other solutions to the free Schrödinger equation besides the plane wave
ones. For example, we can construct wave-packets. As in Lecture 11, we can write at t = 0
 2
1 ~
x −x
~0
−2 ~
~)=e
ψ(x σx
ei k0 ·x~ (13)

this is a packet centered at ~x = ~x0 with width σx and a carrier wavelength ~k0. The Fourier trans-
form is
1 k −k0 2
 
σx −2
ψ̃(k) = e σk
e−ix0(k−k0) (14)
(2π)3/2
1
with σk = σ . This is a wavepacket centered on ~k0. But ~p = ~k
~ . So the momentum distribution
x
is Gaussian too.
8 Section 5

In fact, just as |ψ(x~ )|2 gives the probability that we find the electron at the position ~x at
time t, it is also true that ψ̃ ~k 2 gives the probability that we find the electron with

momentum ~p = ~k ~ . Thus the most likely value for momentum is ~p0 = ~ ~k0 and the width of the
~
momentum distribution is σ p = ~σk = σ . In particular, we find
x

σxσ p = ~ (15)
This is a special case of the Heisenberg uncertainty principle
(∆x)(∆p) > ~ (16)
~
(sometimes you see this relation with 2 on the left; the factor of 2 has to do with precisely how
∆x and ∆p are defined). This says that for any wavefunction, the uncertainty on position ∆x
times the uncertainty of momentum ∆p must be larger than or equal to ~. Gaussian
wavepackets saturate this inequality. Note that ~ = 1.04 × 10−34 J ·s is very small. So we can
actually know the position and momentum of the electron quite well at the same time – to less
than 10−17 each in SI units. Quantum mechanics only shows this weird uncertainty if you are at
very short distances.
Next, note that the carrier wavenumber k0 translates to the most likely momentum value. In
position space, we see from Eq. (13) that the phase of the wavefunction gives the momentum.
Thus a wavefunction at a given time has both position information (in the amplitude) and
momentum information (in the phase).
~2
~k
What happens to these electrons with time? Since the dispersion relation ω = 2m is not
linear, there is dispersion. That is, wavepackets broaden over time. As we saw in Lecture 11, the
second derivative of the frequency with respect to wavenumber determines the broadening rate.
~
In this case, Γ = ω ′′(k0) = m is independent of k0. In terms of the Compton wavelength in Eq.
λc
(4), Γ = 2πc . Then the width grows (see Eq. (36) of Lecture 11) as
s
λcct 2
 
λ
σ(t) = σx 1 + ≈ c ct (17)
2πσx2 2πσx

where the last form holds at late times. So no matter how localized the electron is at t = 0,
eventually, we will lose track of it. For example, suppose we start out knowing the electron
1
down to σx = λc = 10−12 m. Then at late time σ(t) = 2π c t. Thus the electron wavepacket is
broadening at the speed of light. Moreover, the smaller σx we start out with, the faster the
packet broadens. We can take σx large to slow the broadening, but for large σx, we don’t know
where the electron is to begin with. Electrons are very hard to pin down indeed!