Class Notes for Modern Physics, Part 2

J. Gunion
U.C. Davis

9D, Spring Quarter

J. Gunion
 The particle nature of matter

Most of you are already convinced that matter is composed of particles,

but it is useful to at least briefly recall how our current understanding arose
historically.
There were four major items in making this case:
1. Around 1833, Faraday performed a series of electrolysis experiments.
these established three basic things:
(a) that matter consists of molecules and that molecules consist of atoms;
(b) that charge is quantized, because only integral numbers of charges are
transferred between the electrolysis electrodes;
(c) and that the subatomic parts of atoms carry positive and negative
charges.
However, he was unable to directly determine the masses of these
subatomic particles, but it seemed clear that they were related to the
atomic weights that were known from chemistry.
Also, the absolute size of the charge of these subatomic particles could
not be determined from electrolysis — only that charge was quantized.

J. Gunion 9D, Spring Quarter 1

 Considerable time would pass before the next major input.

2. Around 1897, Thomson identified cathode rays as something with the

same sign as the negative charges seen by Faraday.
And, he found that all negative particles emitted from a cathode had
identical e/me values, where e was the charge.
He postulated that whatever this object was, it was probably a fundamental
constituent of matter. We know it as the electron.
A few years later, he was able to use measurements in a viscous cloud
chamber to roughly determine the magnitude of the charge separately.
He found that “e is the same in magnitude as the charge carried by the
hydrogen atom in the electrolysis of solutions.”

3. In 1909, Millikan was able to obtain a much more precise measurement

of the electronic charge.
This could be combined with the e/me value obtained by Thomson to
obtain a value for me that was about 1000 times smaller than the mass
of the Hydrogen atom (the latter being close to the proton mass, mp),
which had been known from atomic weights and chemistry.

J. Gunion 9D, Spring Quarter 2

4. Finally, in 1913, Rutherford and co-workers established the nuclear model
of the atom by scattering fast-moving α particles (charged Helium nuclei)
from metal foil targets.

He showed that atoms consist of a compact positively charged nucleus

(with diameter about 10−14 m) surrounded by a swarm of orbiting
electrons (with the electron cloud diameter being of order 10−10 m.)

Here, I will try to say a few additional words about the Thomson and
Rutherford experiments. Please read the material in the book on the
Millikan experiment.

Thomson

The apparatus and schematic are shown below. We consider a B ~ field

pointing into the page and a E ~ field in the plane of the page. When
~E = −eE
present, these produce forces F ~ (upwards) and F~B = −e(~ ~
v × B)
(downwards) on the e−.

J. Gunion 9D, Spring Quarter 3

 Fig. 4−5, p. 111

Click to add title

Fig. 4−6, p. 111

J. Gunion 9D, Spring Quarter 4

 ~ field. Initially, upon entering from the left, only
First, turn on just the E
vx is non-zero, but upon exiting vy = ay t, where

F eE Ve l
ay = = = and t= . (1)
me me me d vx

This gives

 
V le vy Vl e
vy = , tan θ = = . (2)
mevxd vx vx2 d me

So a measurement of θ gives us a value for mee provided we can determine

vx. Thomson determined vx, which remained the same if he kept his
accelerating anodes at the same voltages, etc., by adding to E ~ the B~
field. The forces exactly balance (and the e− is undeflected) when (for
any q, including q = −e)

E V
qE = qvxB , ⇒ vx = = . (3)
B dB
J. Gunion 9D, Spring Quarter 5
 ~ that gives exact balance, we get
Thus, using the B

e vx2 d tan θ V tan θ

= = . (4)
me Vl B 2ld

Thomson obtained a result of ∼ 1.0 × 1011 C/kg (vs. really accurate

data of 1.76 × 1011 C/kg). This was clearly much larger than the
electrolysis values which were probing charge/proton mass. Thomson
had clearly demonstrated the existence of a new elementary particle with
mass about 1000 times smaller than the mass of the proton (or hydrogen
atom from the atomic weight / electrolysis point of view).

Rutherford

Based on his own experiments and those of others, in which it was clear that
an atom was not a simple object, but had balancing negative and positive
charged particles in it, with the negative one having a much smaller mass
than the positive one, Thomson proposed the “plum-pudding” picture of
the atom.

J. Gunion 9D, Spring Quarter 6

 The atom was visualized as a homogeneous sphere of uniformly distributed
mass and positive charge in which were embedded, like rasins in a plum
pudding, negatively charged electrons, which just balanced the positive
charge to make the atom electrically neutral.

Of course, this picture failed to explain the rich line spectra that people
were finding for excited atoms, in particular even the simplest Hydrogen
atom.

Rutherford and collaborators had noticed that a beam of α particles

(i.e. Helium ions, 2p2n) broadened upon passing through a metal foil,
indicating that the foil was quite easily penetrated and yet at the same
time caused significant scattering. This was hard to explain in the pudding
model where the positive charge was spread all over the pudding.

After experimentation from 1909 to 1913, to be described, Rutherford

concluded that all the positive charge, and most of the mass, was
concentrated in a central nucleus of the atom. In particular, this
picture was the only one that produced events in which the α particle
was scattered at a very big angle, occassionally even backwards. The
experimental apparatus and schematic picture of what is going on is
depicted on the following page.

J. Gunion 9D, Spring Quarter 7

 Fig. 4−10, p. 120

Fig. 4−11, p. 121

J. Gunion 9D, Spring Quarter 8

 In order to account for the occassional large angle, including backwards,
deflections, Rutherford pictured the atom as having a central charged
nuclear core and employed nothing more than Coulomb’s force law

(2e)(Ze)
F =k (5)
r2
(where α has charge 2e in magnitude, the nucleus has charge Ze in
magnitude, and k is Coulomb’s constant). He predicted the following
result for scattering:

k2Z 2e4N nA
∆n = 4 , (6)
4R2( 12 mαvα
2 )2 sin (φ/2)

where R and φ appear in the figure, N is the number of nuclei per unit
area of the foil (and is thus proportional to the foil thickness), n is the
total number of α particles incident on the target per unit time, ∆n is
the number of α particles entering the detector per unit time at an angle
φ, and A is the area of the detector. The velocity vα is determined from
the accelerating potential difference between the α emitter and the gold
foil (or other) target: Kα = 12 mαvα 2
= (2e)V (non-relativistic ok here,
and use charge of α = 2e).

J. Gunion 9D, Spring Quarter 9

 The agreement with the φ dependence was spectacular.

Fig. 4−12, p. 123

In fact, the normalization of the line allowed a determination of the then

poorly determined Z values for different nuclei.

J. Gunion 9D, Spring Quarter 10

 Rutherford also used the exactly back-scattered α particles to estimate
the size of the nucleus. If dmin is the distance of closest approach of
the α particle to the nucleus, and we know the kinetic energy of the α
particle (which we do from the accelerating potential), then we may use

1 2 (Ze)(2e)
mαvα =k (7)
2 dmin

to solve for dmin. When Rutherford’s prediction of eq. (6) starts to fail,
the corresponding dmin presumably is the point at which the α is actually
impinging into the nucleus itself rather than just Coulomb scattering from
it.

Example
Estimate the radius of the Aluminum nucleus.
In 1919, Rutherford was able to show a breakdown in eq. (6) for 7.7 MeV
α particles scattered at large angles from aluminum (Z = 13). Estimate
the radius of the Al nucleus.
Answer: assume all the α kinetic energy, Kα, goes into potential energy

J. Gunion 9D, Spring Quarter 11

 at dmin. Then,

2Ze2
dmin = k
Kα
2(13)(1.6 × 10−10 C)2
= (8.99 × 109 N · m2/C 2)
(7.7 × 106 eV )(1.60 × 10−19 J/eV )
= 4.9 × 10−15 m . (8)

Spectral Lines and Balmer

As mentioned earlier, many spectral lines had been seen, coming from the
sun, coming from excited atoms, and so forth. There was no explanation
yet. ***Do spectral line demo.***

A particularly famous result was the one obtained by Balmer in 1885.

He managed to “fit” the results of Angstrom’s measurements of the
wavelengths of the spectral lines from excited Hydrogen. These are
displayed on the next page.

J. Gunion 9D, Spring Quarter 12

 Fig. 4−20, p. 129

Balmer noted that the line wavelengths took the form:

 2 
n
λ(cm) = C2 , n = 3, 4, 5, . . . (9)
n2 − 22

where C2 = 3645.6 × 10−8 cm, a constant called the convergence limit

J. Gunion 9D, Spring Quarter 13

 to which one tends as n → ∞. He further speculated that there would
be found other spectral line series of Hydrogen that would be fit by the
general form (equivalent for nf = 2 to the previous form)
!
1 1 1
=R − (10)
λ n2f n2i

where ni > nf and R = 1.0973732 × 107 m−1 is the Rydberg constant

and is the same for all the different Hydrogen series lines. These series
came to be named after the experimentalists that were first to see them:
Balmer: nf = 2 (visible and near UV); Lyman: nf = 1 (more UV and
harder to see); followed by Paschen, Brackett and Pfund (nf = 3, 4, 5)
in the IR.

The groundwork was now in place for

Bohr

To understand how revolutionary Bohr’s ideas were, consider the conundrum

that the atomic physics people found themselves in. The picture of a

J. Gunion 9D, Spring Quarter 14

 positively charged nucleus with e−’s circling around it was now firmly
established.
But, according to classical Maxwell, the centripetal acceleration the e−’s
were continually undergoing would cause them to radiate E&M waves.

Fig. 4−21, p. 131

Figure 1: Classical model of the nuclear atom.

J. Gunion 9D, Spring Quarter 15

 Classical theory would then imply that as the e−’s lose energy they would
move closer to the nucleus. Further, the spectrum of radiation from this
continuous process would change continuously, and one should not see
sharp spectral lines.

In fact, as the e− gets closer to the nucleus it is moving faster and

faster in a stronger and stronger Coulumb field and the frequency of the
radiation would get higher and higher.

Of course, we would not be around to see all this in any case.

Bohr followed the lead of Planck and Einstein by assuming that if light
was quantized then why shouldn’t atomic electronic orbits be quantized
in some way. Then, spectral lines could arise when an electron jumped
from one such electronic orbit to another orbit (of lower energy) by
emitting a photon of definite frequency given by ∆E = hf .

Armed with the picture of the atom just developed by Rutherford, in

1913 Bohr published a 3-part paper in which he postulated that electrons
in atoms are confined to stable, nonradiating energy levels and orbits
known as stationary states.

J. Gunion 9D, Spring Quarter 16

 Fig. 4−22, p. 132

Figure 2: Bohr’s model of the Hydrogen atom. The radius is assumed to

be constant because of “quantization”.

As just stated, Bohr realized that the spectral lines corresponded to

photons of a definite wave length and definite frequency and so he knew
that the separation between his stationary states should come in units of

J. Gunion 9D, Spring Quarter 17

 ∆E = hf = h λc .
The only remaining question was how to match this on to Balmer’s
formula. He discovered that this matching worked if he hypothesized
that the electron’s orbital angular momentum about the nucleus had
to be an integral multiple of h̄ ≡ 2hπ (obviously, h had to be involved
somehow)
mevr = nh̄ , n = 1, 2, 3, . . . (11)

Let us see why this works. (For atoms, we can use non-relativistic
procedures with adequate accuracy.)
1. The electric potential energy of the e− is U = −ke2/r, where
k = 1/(4π0).
2. The total energy of the atom is the sum of the potential and kinetic
energies,
1 2 e2
E = K + U = me v − k . (12)
2 r
3. Meanwhile, Newton’s force law says force=centripetal acceleration, or

ke2 v2 1 2 ke2
= me , which ⇒ K= me v = . (13)
r2 r 2 2r

J. Gunion 9D, Spring Quarter 18

 4. Putting this result into the equation for E, eq. (12), gives

ke2
E=− . (14)
2r
5. Next, we solve for v in terms of n and r using the equations above,
1 2 ke2
mevr = nh̄ and 2 mev = 2r , to obtain

n2h̄2
rn = , n = 1, 2, 3, . . . (15)
me ke2

r1 is often denoted by a0, and is called the Bohr radius,

h̄2
a0 = = 0.0529 nm . (16)
me ke2

6. Finally, substitute the form of rn into the equation for E just above,
i.e. eq. (14), to obtain
2  
ke 1 13.6
En = − =− eV . (17)
2a0 n2 n2

J. Gunion 9D, Spring Quarter 19

 The values n are called the quantum numbers characterizing different
states.
The lowest state E1 = −13.6 eV is called the ground state.
The n = 2 state is the 1st excited state and has energy E2 = −3.4 eV,
and so forth.
At this point, we can explain the Balmer formula.
1. A spectral photon is emitted when the atom drops from a state with
a high n = ni to a state with more negative energy i.e. with smaller
n = nf .
2. The different series are obtained using nf = 1, nf = 2, . . . for the
final lower-n state.
3. In other words, we have
!
2
1 f Ei − Ef ke 1 1
= = = − . (18)
λ c hc 2a0h n2f n2i

One finds that ke2/(2a0h) = R, the Rydberg constant and one gets a
theoretical post-diction of the Balmer formula.

J. Gunion 9D, Spring Quarter 20

 This is depicted in Fig. 3.

Fig. 4−24, p. 134

Figure 3: Bohr’s explanation of the various Hydrogen spectral series.

An Example
Suppose the stellar atmosphere has a temperature of order T =
79, 000 K. (a) Is it reasonable to expect that a lot of the Hydrogen atoms

J. Gunion 9D, Spring Quarter 21

 will be excited to the first excited state? (b) What is the wavelength
of the light emitted when these excited atoms decay back to the n = 1
ground state level?
First, we compute the average thermal energy per atom:

3
kB T = (1.5)(8.62 × 10−5 eV /K)(79000) = 10.2 eV . (19)
2

This, we must compare to the energy of excitation,

E2 − E1 = −3.4 eV − (−13.6 eV ) = 10.2 eV . (20)

Since these are comparable, we expect substantial excitation.

The wavelength could either be obtained from the Balmer formula or we
can return to Bohr’s basic model according to which

hf = E2 − E1 ⇒
hc (4.136 × 10−15 eV · s)(3 × 108 m/s)
λ = =
E2 − E1 10.2 eV
= 1.22 × 10−7 m = 122 nm , (21)

J. Gunion 9D, Spring Quarter 22

 well into the ultraviolet.

Bohr immediately realized that all of this could be extended to ions

obtained from an element with a given nuclear charge Z by removing all
but one of the e−’s.
Such an ion has a single e− orbiting about a nuclear charge of +Ze.
Proceeding as above, but using the higher nuclear charge, one obtains
2  2
a
2 0 ke Z
rn = n , implying En = − 2
, n = 1, 2, 3, . . . (22)
Z 2a0 n

When applied to He+, several previously unexplained spectral lines in

radiation from the sun were explained.
An Example
Pickering, in 1896, observed unexpected spectral lines in the light from
ξ-Puppis, a star.
He found that these lines fit the spectral formula
 
1 1 1
=R 2
− 2
, (23)
λ (nf /2) (ni/2)

J. Gunion 9D, Spring Quarter 23

 where R is the Rydberg constant. We can easily check that these lines
are simply those associated with He+ as follows.

Since He+ has nuclear charge of Z = 2, we have energy levels given by

2  
ke 4
En = . (24)
2a0 n2

Using hf = Ei − Ef , we then have

1 f Ei − Ef
= =
λ c hc !
2
ke 4 4
= −
2a0hc n2i n2f
ke2
 
1 1
= − (25)
2a0hc (ni /2)2 (nf /2)2

where ke2/(2a0hc) = R is precisely the Rydberg constant.

J. Gunion 9D, Spring Quarter 24

 An interesting question from the class
During the lecture on this material, an interesting question was asked.
This concerned why we don’t see atoms absorbing starlight. Well, in
fact we do. As we discussed, one should visualize a distant star sending
light towards the earth, with some dust or gas cloud (for example) in
between the star and the earth. If the cloud mainly contains Hydrogen,
for example, then starlight with the right frequency to excite a Hydrogen
atom from a low energy (e.g. ground) state to a higher state will often
get absorbed and not make it through the cloud. We get what are called
absorption spectra (that can be used to help determine the red-shift of
the star relative to the cloud and of the cloud relative to us).
The energy of the star radiation can even be sufficient to completely
ionize the Hydrogen atom if the cloud is close to the star or the star is
of a particularly energetic type.
A second question was why one doesn’t get radiation, coming from the
excited atom when it falls back to its ground state, that fills in the
absorption line.
In fact, there is such radiation, but it goes in all directions (not just
towards the earth) and so the amount headed towards earth is greatly

J. Gunion 9D, Spring Quarter 25

 diminished.
An important dimensionless ratio
It turns out to be interesting to consider the ratio

vn=1 1 h̄
= from mvr = nh̄
c c mr1
1 ke2 h̄2
= from r1 = me ke2
c h̄
ke2 1
= ≡α= . (26)
h̄c 137
Note how small v1/c is. The non-relativistic approximation employed by
Bohr was ok.
The quantitiy α is sometimes called the fine structure constant. It is a
very useful characterization of the strength of the E&M force. Other
forces, such as the strong and weak forces that we will learn more about
late in the quarter, have different strengths.
Such dimensionless ratios constructed using known physical constants
(here, h̄, c, k and e) are typically of deep theoretical significance. To
construct α, we needed the new fundamental constant h̄.

J. Gunion 9D, Spring Quarter 26

 The correspondence principle
One justification given by Bohr for his angular momentum quantization
condition is that it is required if we demand a correspondence principle
according to which

lim [quantum physics] = [classical physics] (27)

n→∞

where n is a typical quantum number of the system such that large n

corresponds to a limit in which one should approach a classical type of
situation, such as long wavelengths.
We will not go into the details of this in class.

Franck and Hertz

Of course, the critical assumption made by Bohr in his explanation of the

spectral lines was that an electron could be in a higher n state and that
when it “fell” down to a lower state it emitted a single photon.
A direct verification that the photon energies corresponded to the
separation between energy levels of the electron of the atom was needed.

J. Gunion 9D, Spring Quarter 27

 Franck and Hertz provided an explicit experimental demonstration that
this was indeed the case.
They sealed some Mercury Hg inside a tube and accelerated e−’s through
the tube using a voltage V . These e−’s then collide with the Hg atoms
inside the tube, possibly giving energy to them.

Fig. 4−27, p. 141

Figure 4: The Franck-Hertz apparatus.

For small V , these collisions were elastic and the e−’s retained most of

J. Gunion 9D, Spring Quarter 28

 their kinetic energy (very little is taken by the much more massive Hg
atoms in an average collision). Even after many collisions, the e− arrives
at the accleration grid with energy of about eV .
Following this acceleration, their apparatus had a retarding voltage gap
between the accelerating grid and the following collector plate of about
1.5 V . Thus, some e−’s will be collected if V > 1.5 V .

Fig. 4−28, p. 142

Figure 5: The Franck-Hertz current as a function of accelerating voltage V .

As V is increased, more and more e−’s make it to the collector until

J. Gunion 9D, Spring Quarter 29

 the energy eV that the electrons have acquired matches the energy
difference between two atomic energy levels. At this point, the collision
between some of the accelerated e−’s and the Hg can be inelastic —
the Hg atom absorbs the K = eV energy of the e− when one of its own
electrons is excited to a higher n level.. There is a sudden dip in the
current reaching the collector.

When V is increased further, more and more electrons reach the collector
until once again the current suddenly dips. What is happening is that V
is large enough for the accelerated e−’s to have two inelastic collisions
with two subsequent Hg atoms.

The separation between the dips was found to be ∆V ∼ 4.9 V . They

interpreted e∆V as being the energy difference between the ground state
of low n for one of the Mercury orbiting electrons and the next excited
state of this same electron.

How could they check this? Well, if they really had excited the Hg atomic
electron to a higher level, it should emit a photon of the corresponding
frequency when this electron fell back down to its original lower level.

J. Gunion 9D, Spring Quarter 30

 The expected wave length of the photon was therefore given by

hc hc 1240 eV · nm
e∆V = ∆E = hf = , ⇒ λ= = = 253 nm .
λ ∆E 4.9 eV
(28)
This is the precise wavelength they observed. In 1925, they were awarded
the Nobel prize for this confirmation of Bohr’s theory.

Connection of Bohr quantization to the wave nature of matter

The next big question was why should angular momentum be quantized
in the manner proposed by Bohr?
What turns out to be the fundamental idea was that developed by de
Broglie in 1925.
He speculated that if light, a wave phenomenon originally, also had a
particle-like nature, then why not the reverse?
He also was looking for a way to explain the integers and quantization
that emerged in Bohr’s atomic theory, which concerned electrons circling
a nucleus. The only way that integers had cropped up in the past was

J. Gunion 9D, Spring Quarter 31

 in wave interference phenomena and normal modes of vibration (such
as simple string standing waves). He decided that periodicity should be
assigned to electrons under appropriate circumstances such as in atomic
orbits.
For this, he needed a wavelength and a frequency to associate with
particles. In analogy with light, he postulated

h E
λ= and f = . (29)
p h

(We will return to the problem with this that arises if you compute wave
velocity as f λ = E/p using the relativistic formulae, p = γ(u)m0u and
E = γ(u)m0c2, which would give E/p = c2/u > c.)
De Broglie noticed that if we employ the photon-like formula p = mve =
h/λ and plug this into mevr = nh̄, we find

h h 2πr
r=n ⇒ = n. (30)
λ 2π λ
In other words, the circumference of the electron orbit must contain an
integer number of electron wavelengths, which, in turn, implies that the

J. Gunion 9D, Spring Quarter 32

 electron wave pattern will repeat by matching onto itself after going
around a full orbit. In other words, an electron orbit should correspond
to a standing, self-reinforcing wave pattern, much like a plucked guitar
string.

Fig. 5−2, p. 153

Figure 6: De Broglie’s explanation of Bohr quantization for the case of

n = 3, showing self-reinforcing standing wave pattern for an e− around the
nucleus with three wave lengths fitting into 2πr circumference.

We will soon turn to more discussion and eventual direct confirmation of

J. Gunion 9D, Spring Quarter 33

 the association of a wavelength with a massive particle using λ = h/p.
In the end, we will see that this standing wave pattern and the whole Bohr
picture is not an accurate point of view. However, it was very critical to
developing the correct view that we now call Quantum Mechanics.

J. Gunion 9D, Spring Quarter 34

 Wave Equations and Fourier Ideas

Since I want to make sure we are all on the same page, I will give a very
brief review of wave equations and the E&M wave equation in particular.
The latter material was in sections 32.2 and 32.3 of University Physics
Part 2 by Young and Freedman, which is I believe the text employed for
your earlier courses. Had you covered chapters 35 and 36 of this text,
which I understand you probably did not, you would have seen a very
detailed discussion for light of the interference phenomena and so forth
that we have already talked about. But, you did, I believe, cover these
phenomena for mechanical waves, which is where we begin.

First, recall the differential equation that you studied and understood for
mechanical waves on a string or in water.

∂ 2y(x, t) 1 ∂ 2y(x, t)
= , (31)
∂x2 v2 ∂t2
where y(x, t) is the displacement of the string, or ... at location x and
time t, and v is the velocity with which the wave moves along the string.

J. Gunion 9D, Spring Quarter 35

 Solutions of this equation take the form f (x − vt) or f (x + vt). In
particular, denoting the argument of f as θ, for either θ = x − vt or
θ = x + vt we have

∂ 2f ∂ 2f ∂ 2f 1 ∂ 2f
= f 00(θ) and = v 2f 00(θ) ⇒ = . (32)
∂x2 ∂t2 ∂x2 v2 ∂t2

A particularly simple choice for f is one such that f 00 ∝ f . A possible

example of this type is

 
2π
f = sin (x − vt) (33)
λ

where, of course, you recognize λ as the wavelength such that if

x → x + λ the shape of the wave repeats. Of course, there is also a
frequency of repetition intrinsic to the above form given by

2π 1 v
vT = 2π , or f = = . (34)
λ T λ
J. Gunion 9D, Spring Quarter 36
 Often, it is more convenient to write

2π
(x − vt) ≡ kx − ωt , (35)
λ

where k ≡ 2π/λ and you can check that ω = 2πf .

We should also recall that one can superimpose different solutions of the
wave equation and still get a solution. For example,

ei(kx−ωt) and e−i(kx−ωt) (36)

are both also solutions to the wave equation and the earlier sin form can
be written as (using eib = cos b + i sin b)

1 h i
sin(kx − ωt) = ei(kx−ωt) − e−i(kx−ωt) . (37)
2i

Of course, in the case of a real observable thing like the string

displacement, when we superimpose these complex exponentials, we
should always do so in such a way that the superposition has a real value.
But, it is nonetheless convenient to use the complex exponentials. In

J. Gunion 9D, Spring Quarter 37

 fact, as we shall see, matter waves are intrinsically complex and can be
so because the waves have no direct physical manifestation. It is only
their |amplitude|2 that can be interpreted as probability.

We can also superimpose solutions with different wave lengths and

frequencies (always holding f λ = v fixed for a given wave velocity). This
amounts to giving a Fourier representation of the wave solution: e.g.

+∞
1
Z
f (x − vt) = √ dkfe(k)eik(x−vt) , (38)
2π −∞

where k can run over negative as well as positive values and ω(k) = vk
is implicitly required by this form. The functional form fe(k) defines√the
Fourier decomposition of the wave solution f (x − vt). The inverse 2π
is just a conventional choice.

The above might be complex. So, for a real observable like string
displacement, we would take the real part of f (x − vt). It will still be
a solution of the wave equation since it will still be a function only of
x − vt.

J. Gunion 9D, Spring Quarter 38

 A useful example

Consider a “square-wave” shape for fe(k)

1
fe = 0, k < k0 − ∆k (39)
2
1 1
fe = 1, k0 − ∆k ≤ k ≤ k0 + ∆k (40)
2 2
1
fe = 0, k > k0 + ∆k . (41)
2

This fe(k) is plotted below. (The book, Example 5.7, uses the notation
fe(k) = a(k).)

J. Gunion 9D, Spring Quarter 39

 Fig. 5−23, p. 172

Figure 7: Input k-space function.

The resulting analytic form for f (x − vt) at t = 0 (derived a bit later) is

∆k sin(∆k · x/2) ik0x

f (x) = √ e , (42)
2π (∆k · x/2)

the real part of which is plotted in the figure.

J. Gunion 9D, Spring Quarter 40

 Fig. 5−24, p. 173

Figure 8: Output Re[f (x)] at t = 0.

What you see in the figure for Re[f (x)] is the rapid oscillation of
the cos(k0x) = Re[eik0x] factor within an envelope described by the
sin(∆k x/2)/(∆k x/2) factor. The latter has its first nodes at ∆k x/2 =
±π, i.e. x = ±2π/∆k. (There is no node at x = 0 because

J. Gunion 9D, Spring Quarter 41

 limx→0 sin x/x = 1.) The full width between the two nodes is thus
∆x = 4π/∆k.

What we wish to particularly point out is the relationship between the

width of the input bump in k space to the width of the output wave form
in x space. We have
∆x∆k = 4π . (43)
The smallest value for this product occurs if a Gaussian form (fe(k) ∝
1 2 2
e− 2 (k−k0) /(δk) ) is employed. The output then has a similar Gaussian
1 2 2
shape in x (f (x) ∝ e− 2 (x−x)) /(δx) ), with δx = 1/δk, or δxδk = 1.
With a certain
√ “formal” defintion√ of ∆x and ∆k that we will come to,
∆x = δx/ 2 and ∆k = δk/ 2, and we obtain

1
∆x∆k = . (44)
2

Thus, for any possible form of fe(k), we have

1
∆x∆k ≥ . (45)
2

J. Gunion 9D, Spring Quarter 42

 The Heisenberg Uncertainty Relation for Photon Waves

So what? What is the physical impact? First, as we shall remind ourselves

in more detail in a moment. Light obeys the same kind of wave equation
just considered, with v = c.
Next, let us input the light wave / photon relation that

2π 2πp p
k= = = , (46)
λ h h̄
eq. (45) can be rewritten as

h̄
∆x∆p ≥ . (47)
2
This is the famous Heisenberg uncertainty principle that was first proposed
for matter, and only later was it realized that it was already present in
the description of light waves as photon packets with p = h/λ.
We will return to a thorough discussion of the implications of this kind of
uncertainty principle. However, you should at this point take note of the

J. Gunion 9D, Spring Quarter 43

 fact that it is simply a mathematical result that follows from combining
wave propagation ideas with quantization of the wave into particles, in
the light case the particles being the photons.
We now derive f (x) (at t = 0) using the input fe(k). The only thing you
need to know is that e dk = eak/a, where a = ix in our case.
R ak

Z +∞
1 ikx
f (x ) = √ fe(k)e dk
2π Z−∞
k0 +∆k/2
1 ikx
= √ e dk
2π k0−∆k/2
1 1 h i(k0+∆k/2)x i(k0 −∆k/2)x
i
= √ e −e
2π ix
1 eik0x
 
1
= √ 2 sin ∆k x
2π x 2
∆k sin(∆k · x/2) ik0x
= √ e . (48)
2π (∆ k · x/ 2)

Another way of understanding the uncertainty relation in the case of light

waves / photons is to return to the single slit wave experiment. There,
we “recalled” that a slit of size D gave a first diffraction minimum at
θ ∼ λ/D.

J. Gunion 9D, Spring Quarter 44

 Derivation

D (D/2)sin θ
2
θ

Path difference between top of slit and middle of slit = (D/2)sinθ = λ/2
for full cancellation requires θ=λ/D

Figure 9: Derivation of single-slit diffraction minimum location.

We can give a simple derivation based on Huygen’s principle, something I

hope you are familiar with. It says that the propagation of a wave can be
constructed by dividing up the wave into many different little (circular in

J. Gunion 9D, Spring Quarter 45

 a planar configuration) wavelets emanating from any well defined surface
(curve in a planar configuration).
Using Huygen’s principle, consider a wavelet emanating from the top of
the slit and one from the midpoint of the slit. The diagram shows that
these two wavelets will be precisely 1/2 wavelength out of phase (i.e.
they will cancel) when sin θ ∼ θ = λ/D. The same will apply to a
wavelet emanating from  below top and  below the midpoint, and so
forth. Thus, θ ∼ λ/D is the condition for a minimum in the diffraction
pattern.
Demonstration
Take a red laser, λ ∼ 650 nm = 0.65 × 10−6 m. Take a slit of
order D = 0.2 mm = 2 × 10−4 m. The first minimum will be at
θ ∼ 0.0032. Place a screen about l = 10 m away and the distance
between the two first minima on either side of the maximum should be
about d = 2lθ ∼ 0.065 m = 6.5 cm.
Implications for photon momenta
For the photon to have travelled there, it must pick up a momentum
py perpendicular to the initial (upwards) momentum of px ∼ p = h/λ.

J. Gunion 9D, Spring Quarter 46

 Thus, we have

py py λ θh λh h
θ∼ ∼ ⇒ py ∼ ∼ ∼ (49)
px h λ Dλ D

from which we find (can’t expect to get 2π type factors right here)

h
∆py ∆y ∼ py D ∼ D ∼ h. (50)
D

Once again, the uncertainty relationship emerges. Here, we have tried

to confine the E&M wave to a location of size D in the y direction as
it propagates to the right in the x direction, and, in so doing, we have
generated a substantial uncertainty in py . The more we try to define the
wave location in a certain direction, the greater the uncertainty in the
momentum in that same direction.

J. Gunion 9D, Spring Quarter 47

 Electromagnetic Waves

Let me now give a brief review of the E&M wave equation. Had I known
that this was only given very brief attention in your previous course, I
would have surely begun this quarter’s lectures with the following review.
One starts with the two Maxwell equations:

d
I Z
~ · d~l = −
E ~ · n̂ dA
B (51)
C dt S
d
I Z
~ · d~l = µ00
B ~ · n̂ dA
E (52)
C dt S

where the latter assumes no source current I, as appropriate for

propagation in a vacuum. The vector n̂ represents a unit vector normal
to the surface S at any given point. The closed loop C runs along the
boundary of the surface S, and the orientation of n̂ relative to C is given
by the right-hand rule.
In principle, you have read the material in University Physics Section 32
~
to learn that these reduce (for a wave traveling in the x direction with E

J. Gunion 9D, Spring Quarter 48

 ~ pointing in the z direction) to
pointing in the y direction and B

∂Ey ∂Bz
= − (53)
∂x ∂t
∂Bz ∂Ey
= −µ00 (54)
∂x ∂t

I give a brief derivation of the first equation. We apply eq. (51) to the
case of E~ = ŷEy and B ~ = ẑBz . The equation says that a time varying
Bz (using n̂ = ẑ and very small size dx by dy loop in the x, y plane) can
generate an E ~ field that circulates around the small loop. Applying this
to the E~ = ŷEy case, we find that Ey must vary with x. In short, the
time varying Bz field is generating a spatial variation of Ey as a function
of x. Of course, Ey will end up with time dependence that matches that
of the time derivative of Bz . A figure showing how this application works
is below.

J. Gunion 9D, Spring Quarter 49

 y E loop integral = E_y(x+dx)dy − E_y(x) dy = (dE_y(x)/dx)dxdy
B surface integral = B_z(x) dx dy

E_y(x) dy
n E_y(x+dx)

B_z(x)

dx

z
x
Figure 10: Set up for deriving eq. (53).

The 2nd of the integral-form Maxwell equations (applied with n̂ = ŷ and

a very small loop in the x, z plane of size dx by dz) implies that a spatial
variation of Bz as a function of x will be generated by a time variation

J. Gunion 9D, Spring Quarter 50

 of Ey . A set up analogous to that depicted in Fig. 10 would give you
eq. (54).
Equivalently, you may have seen the two integral equations rewritten
using the famous general theorem, called Stoke’s theorem, which states:
I Z
~ · d~l =
F ~ ×F
(∇ ~ ) · n̂ dA (55)
C S

where F~ is any arbitrary vector “field” and ∇

~ ×F
~ denotes the “curl” of
~ . The two important components of the definition of the curl are
F

∂Fy ∂Fx
~ ×F
(∇ ~ )z = − (56)
∂x ∂y
∂Fx ∂Fz
~ ×F
(∇ ~ )y = − ; (57)
∂z ∂x

these will be needed in the differential forms of eqs. (51) and (52),
respectively.
Using this theorem in eqs. (58) and (59) and the fact that the surface
S, and its normal n̂, can be thought of as being arbitrary, the integrands

J. Gunion 9D, Spring Quarter 51

 must be equal so that eqs. (51) and (52) imply

~
∂B
~ ×E
∇ ~ = − (58)
∂t
~
∂E
~ ×B
∇ ~ = µ00 , (59)
∂t

respectively. We now assume, as above, that B ~ = ẑBz only and

~ = ŷEy only. In this case, we are interested in the z component of
E
eq. (58) and the y component of eq. (59). We then employ the curl
defintions of eqs. (56) and (57) to obtain

∂Ey ∂Bz
~ × E)
(∇ ~ z= , ~ × B)
(∇ ~ y=− . (60)
∂x ∂x
Substituting the above into eqs. (58) and (59), respectively, we get
eqs. (53) and (54), repeated below.
∂Ey ∂Bz
= − (61)
∂x ∂t
∂Bz ∂Ey
− = µ00 (62)
∂x ∂t
J. Gunion 9D, Spring Quarter 52
 Let us now consider the equation that can be derived from eqs. (53) and
∂ ∂ 2 Ey ∂ ∂Bz ∂Bz
(54). Take ∂x eq. (53) ⇒ ∂x2 = − ∂t ∂x
and substitute for ∂x
using
∂Ey
eq. (54), which states ∂B
∂x
z
= −µ0 0 ∂t to obtain:


∂ 2 Ey ∂ 2 Ey
= µ00 . (63)
∂x2 ∂t2

This matches the mechanical wave equation provided the velocity is

v 2 ≡ c2 = µ010 . Following a similar procedure we also find

∂ 2 Bz ∂ 2 Bz
= µ00 . (64)
∂x2 ∂t2

Thus, the E and B oscillations are continually feeding one another

through Maxwell’s laws and as a result the wave propagates in the x
direction.
If we employ a form

Ey = A sin(kx − ωt) , (65)

J. Gunion 9D, Spring Quarter 53

 then we can check that the associated form for Bz must be

1
Bz = A sin(kx − ωt) , (66)
c

by employing either eq. (53) or eq. (54). For example, eq. (54) states
1 ∂Ey
that ∂B
∂x
z
= − c2 ∂t
. Substituting in the above forms we get

∂Bz 1
= A k cos(kx − ωt)
∂x c
1 ∂Ey 1
− = − 2 A(−ω) cos(kx − ωt)
c2 ∂t c
1 ω
= A cos(kx − ωt)
c c
1
= A k cos(kx − ωt) using ω/k = c . (67)
c

The fact that Ey and Bz are exactly “in phase” all the time, is one of
the remarkable features of E&M radiation. But, it had to be true in
order for one to “feed” the other.

J. Gunion 9D, Spring Quarter 54

 Energy carried by an E&M wave

It is also useful to remind ourselves about the amount of energy carried

by an E&M wave. You need to remember that the energy density stored
in the E~ and B~ fields is given by

1 ~ ·B
1B ~
uE = ~ ·E
0 E ~, uB = , (68)
2 2 µ0

~ = |E|/c.
respectively. As we have seen above, for the travelling wave, |B| ~
So, u = uE + uB can be written in a variety of forms:

1 ~ 2
1 |B|
u = ~ 2+
0|E|
2 2 µ0
~ 2
= 0|E|
~ 2
|B|
=
µ0
0
r
= ~ B|
|E|| ~ . (69)
µ0

J. Gunion 9D, Spring Quarter 55

 And, we should also remember that for the E&M wave, travelling with
velocity c, the amount of energy transported through a surface area
perpendicular to the wave’s direction of travel (e.g. a surface in the
y, z plane for travel in the x direction) is simply S = cu, which has
the correct dimensions since c = m/s while u = energy/m3 so that
S = energy/m2/s.
An Example

Suppose the maximum |E| ~ value for a traveling sinusoidal wave, moving in
the x direction is Emax = |E|~ = 100 N/C and occurs at t = 0, x = 0.
Give a value for the amount of energy impacting a screen perpendicular
to the x axis per unit area per unit time at t = 0, x = (2/3) × λ.
Answer: Since the field is maximum at t =
0, x = 0, it is convenient
to use the form Ey = Emax cos 2λπ (x − ct) . Substituting t = 0, x =
(2/3)λ gives Ey = Emax cos(4π/3) = − 21 Emax. From our earlier
equations, we have

S = cu = c0Ey2
1
= (3 × 108 m/s)(8.85 × 10−12 C 2/N · m2)(− 100 N/C)2
2
J. Gunion 9D, Spring Quarter 56
 = 6.6375 J/(m2 · s) . (70)

Of course, as time passes, at this same location, the S value will oscillate
up and down and so the average energy per unit area per unit time will
be Saverage = 21 Smax = 12 c0Emax2
. This is what we usually call the
intensity of the E&M wave, but we see that a more accurate name
would be average intensity.
We do not know how many photons this corresponds to (on average)
until λ is specified. Also note that we could compute (at any instant)
Bz using Bz = Ey /c.

Momentum Carried by an E&M wave

We have stated that the relation between the energy and the momentum
carried by an E&M wave is p = E/c, where in the continuous wave
vision E is the same as S. p will then be so much momentum per unit
area per unit time.
To derive this relation between the energy and momentum carried by
an E&M wave, is a bit of an exercise. I give it below in case you are
interested.

J. Gunion 9D, Spring Quarter 57

 Consider a test charge Q (of unit area) on which the wave impinges. The
wave will start this test charge moving, to begin with in the y direction
as a result of Fy = QEy . Once Q has some vy , the magnetic field of
the wave will act on it to produce a force in the x direction

Fxx̂ = Qvy ŷ × Bz ẑ = x̂Qvy Bz . (71)

Since Fx = dpx/dt, we get momentum being fed to the charge at the

rate of (using Bz = Ey /c, as above)

dpx Ey
= Qvy Bz = Qvy . (72)
dt c

Meanwhile, starting from vx = 0 (so that Fx is not doing any x direction

work yet) potential energy is being added to this charge, because it is
moving against the electric field, according to

dU
∆U = QEy ∆y , ⇒ = QEy vy . (73)
dt

Substituting the result of solving this equation for vy into the previous

J. Gunion 9D, Spring Quarter 58

 equation gives  
dpx dU/dt Ey 1 dU
=Q = , (74)
dt QEy c c dt
implying that on a per second basis the amount of energy being supplied
by the E&M wave and the amount of x momentum being supplied by
the wave must be related by p = U/c. But, the amount of energy being
supplied by the E&M wave (all this is per unit area per unit time, recall)
is simply U = S. We have been denoting S by E, which, to repeat,
for a wave is the amount of energy per unit are per unit time passing
a certain perpendicular plane. And, of course, if both the momentum
and the energy are being carried by photons, then the energy per photon
must be related to the momentum per photon by p = E/c.
Return to previous example
At t = 0, x = (2/3)λ, how much momentum is being transferred to
the screen per unit area per unit time, assuming that all the radiation is
being absorbed?
Answer: Using E = S = cp, we compute

S 6.6375J/(m2 · s)
p= = = (2.212 × 10−8 kg · m/s)/(m2 · s) . (75)
c 3 × 108 m/s

J. Gunion 9D, Spring Quarter 59

 Hopefully, the demonstration of a little set of vanes inside a vacuum
container is something you have seen?

General Lesson

Thus, just as in the case of a wave on a string, the E ~ and B~ fields

contained in a light wave have real physical implications. E ~ could
~ could deflect a moving charged
accelerate a charged test particle and B
test particle.

A Lesson in Wave Amplitudes and Probabilities

We have seen that a single slit will have a minimum at sin θ = λ/D
coming from the complete cancellation of various Huygen’s wavelet
amplitudes from different parts of the slit. At θ = 0, all the wavelets
arrive in phase at the central point of the screen and simply add up to
give you a maximum E field, call this maximum E0. E0 will have some
wave-like form, of course, and so will oscillate up and down as time passes
according to some form like
 
2π
E0 = Emax sin (xscreen − ct) . (76)
λ

J. Gunion 9D, Spring Quarter 60

 There will be an instantaneous intensity I0 = c0|E0|2 and the average
intensity will be hI0i = 12 I0max = 12 c0|Emax|2.

One can also add up the wavelets for any other θ. I will not go through
the derivation, but the result is

 2
sin[πD(sin θ)/λ] sin[πD(sin θ)/λ]
E = E0 , ⇒ I = I0 (77)
πD(sin θ)/λ πD(sin θ)/λ

Now let us consider the case of D  λ. Then, since limx→0 sinx x = 1

we have E = E0, independent of θ. That is, uniform intensity on the
detecting screen.

Next consider the case of two such slits. If we cover up either one
of the slits, we will get a uniform E0 on the detecting screen and the
corresponding I0 ∝ |E0|2, just as discussed above. But if we now open
up both slits, we will get our two-slit interference pattern. Let us call the
distance from the upper slit, slit #1, to the screen L. Then the distance
from the lower slit #2 to the screen is L + S sin θ.

J. Gunion 9D, Spring Quarter 61

 L

L + S sin θ
S θ
S sin θ

Path difference between top slit and bottom slit = S sin θ

Figure 11: Two (narrow) slit interference.

From this picture, we obtain
1+2 1 2
E = E +E    
2π 2π
= Emax sin (L − ct) + sin (L + S sin θ − ct) .
λ λ

First, let us note that the two waves cancel when S sin θ = (n + 12 )λ,

J. Gunion 9D, Spring Quarter 62

 since the arguments of the sin’s would differ by π, i.e. one-half cycle.
Thus, the two-slit pattern will have minima with zero intensity at such
angles. Correspondingly, the two-slit pattern maxima occur at S sin θ =
nλ. At such angles, constructive addition of the two waves is perfect.

Of course, for general θ, L = xscreen/ cos θ, but this is not needed for
the above discussion.

Now, let us return to the question posed at the end of the last class, for
which we focus on the case of θ = 0. Then,

 
2π
E 1+2 = 2Emax sin (xscreen − ct) , ⇒ I 1+2 = 4c0|E0|2
λ
(78)
implying that I 1+2 = 4I0, where I0 is that from just one slit! Here, I0
denotes the instantaneous intensity. For the average intensity, the same
applies:
 
1
hI 1+2i = 4hI0i = 4 I0max . (79)
2

J. Gunion 9D, Spring Quarter 63

 What about matter waves?

In contrast, as I have said before, the matter waves that we shall come
to do not have any such direct physical interpretation.
~ and B
Matter waves are sort of like putting the E ~ together in the form

~ =E
F ~ + icB
~. (80)

~ , we get
If we take the absolute square of this F

~ | 2 = (E
|F ~ + icB)
~ · (E
~ − icB)
~ = |E|
~ 2 + c2|B|
~ 2 (81)

which is indeed proportional to the intensity of the electromagnetic wave.

We have learned that it is the intensity that tells us the probability of
finding a photon at a certain point in space at a certain time.
For matter waves, the wave will usually be denoted by Ψ. Ψ can always
be decomposed into its real and imaginary parts:

Ψ = Ψ1 + iΨ2 . (82)

J. Gunion 9D, Spring Quarter 64

 For matter waves, Ψ1 and Ψ2 do not have any direct physical manifestation
analogous to the way in which E ~ and B ~ can impact a test charge. There
is no test probe that one can employ. The only interpretation of Ψ is
that
Probability of finding particle ∝ |Ψ|2 = |Ψ1|2 + |Ψ2|2 . (83)

Ad hoc derivation of E&M wave equation

Before ending this “review”, let me note an amusing “derivation” of the

E&M wave equation.

First, we note again that the wave equation for X being either Ey or Bz
takes the form
∂ 2X 1 ∂ 2X
2
= 2 2
. (84)
∂x c ∂t
Suppose we write the energy momentum relationship for light in the form
2 E2
p = c2 , multiply this times X and then make the replacements

∂ h̄ ∂
E → ih̄ and p→ . (85)
∂t i ∂x
J. Gunion 9D, Spring Quarter 65
 Then,
E2 ∂ 2X 1 ∂ 2X
 
p2X = 2
X ⇒ −h̄2 = , (86)
c ∂x2 c2 ∂t2
which contains our wave equation. A hand-waving motivation for these
identifications is to note that for a wave solution of the type eik(x−ct)
(our general form for the case of v = c) that we were discussing earlier,
it is certainly the case that

h̄ ∂
eik(x−ct) = h̄keik(x−ct) = peik(x−ct) , (87)
i ∂x
where we used
h 2π h
h̄k = = =p (88)
2π λ λ
for a photon within an E&M wave. Similarly,

∂
ih̄ eik(x−ct) = h̄kceik(x−ct) = Eeik(x−ct) , (89)
∂t

where we used (see above)

h̄kc = pc = E (90)

J. Gunion 9D, Spring Quarter 66

 for a photon within an E&M wave. (Note how we had to use a
combination of wave and photon ideas for this little game.)
The replacements of eq. (85) turn out to also be applicable for particles
with mass. In a very real sense, the replacements of eq. (85) are all that
are required to formulate the theory of Quantum Mechanics that we now
turn to. But, we will approach QM from the beginning and only come
back to these considerations after a while.

J. Gunion 9D, Spring Quarter 67

 Matter Waves

More de Broglie

We have already discussed that λ = h/p explains Bohr’s quantization

via 2πr = nλ. This was a non-relativistic case. A natural question is
whether we should use the relativistic momentum of Einstein in the more
general situation. Answer=Yes!
For example, if we accelerate an electron through a large voltage V , it
will acquire kinetic energy K = eV .
How do we p get the momentum? Remember that E = K + mec2 and
that cp = E 2 − m2ec4. Plugging in the form just given for E, we
obtain (writing in a form that displays the small eV limit)

1q 2 2 2 4
1p 2 2
p = (eV + mec ) − mec = e V + 2eV mec2
c
√ s c
2eV mec2 eV
= 2
+ 1. (91)
c 2mec

J. Gunion 9D, Spring Quarter 68

 From this, we obtain

h hc 1
λ = =√ q
p 2eV me c2 eV
+1
2m e c 2
 
h 1 1
= √ p q . (92)
2me × 1eV V (volts) eV
+1
2me c2

We evaluate the factor out in front as

h 6.63 × 10−34 J · s
√ = p
2me · 1eV 2(9.11 × 10−31 kg)(1.6 × 10−19 J )
= 1.227 nm . (93)

To use the previous formula, V should be given in terms of volts, since

1V was taken inside the square root.

Remembering that mec2 = 0.511 M eV , we see that if the kinetic energy

eV from acceleration is more than a small fraction of an M eV , we will
need to use the full expression.

J. Gunion 9D, Spring Quarter 69

 Davisson-Germer

The experimental confirmation of the de Broglie hypothesis was due to

Davisson and Germer in 1927. Their apparatus is depicted below.

Fig. 5−4, p. 156

Figure 12: The Davisson-Germer apparatus.

J. Gunion 9D, Spring Quarter 70

 Except for an accident in which they created a single large crystal at
the surface of their Nickel target, they would never have seen the effect.
Checking de Broglie was not actually the original goal of their experiment,
but they were smart enough to realize what was going on when they saw
sharp variations in the intensity of the “reflected” electrons.

Click to add title

Fig. 5−6, p. 157

Figure 13: Electron scattering from an atomic lattice.

The correct picture is that the electron wave scatters off the top layer
of Nickel atoms on the surface (the electrons had low energy and did

J. Gunion 9D, Spring Quarter 71

 not penetrate beyond the surface). Because these were part of a single
crystal, they had a very regular spacing, as depicted in the figure.
The electron waves arrive in phase (assuming 90◦ incident angle) and
are then scattered at angle φ. As they leave the surface, the waves
from different atoms are out of phase by an amount given by AB =
d sin φ. Only if d sin φ = nλ will the different scattered waves all
constructively interfere. Using the formula we just derived in the non-
relativistic approximation, and an accelerating voltage of 54 V , the
electron wavelength will be

1.227 nm
λ= √ = 1.67 × 10−10 m . (94)
54

What did they see? From X-ray measurements DG knew that their
atomic spacing was d = 2.15 × 10−10 m. As illustrated in the next
figure, they found constructive interference for φ = 50◦ corresponding to

λ = d sin φ = 2.15 × 10−10 m sin 50.0◦ = 1.65 × 10−10 , (95)

in excellent agreement (given experimental errors) with the prediction

above of de Broglie’s formula for λ for the given momentum.

J. Gunion 9D, Spring Quarter 72

 Fig. 5−5, p. 156

Figure 14: Scattered intensity vs. scattering angle for 54 eV electrons

incident at 90◦.

If one employs higher acceleration voltages, then the e− will penetrate

further into the surface, and the e− waves will see many layers of the
crystal structure. The picture is below.

J. Gunion 9D, Spring Quarter 73

 D cos θ
D θθ

Figure 15: Multi-layer diffraction of deeply penetrating beam.

Because of equal entry and exit angles for the e− (or any other particle,
e.g. neutron), the waves from any two atoms on any one horizontal crystal

J. Gunion 9D, Spring Quarter 74

 layer will always be in phase. However, waves from atoms on different
crystal layers are not necessarily in phase. The picture shows just one
such pair of waves. There are many. One gets strong cancellation among
the many unless the path differences are all an integer number of wave
lengths. This leads to Bragg’s law:

2D cos θ = mλ (96)

The advantage of the multilayer diffraction type of probe is that the

cancellation among the many different wavelets is so complete at any
angles other than the Bragg angles that very precise information about
the crystal structure can be obtained.

Indeed, crystal diffraction is an indispensable tool in the study of solids;

the details of the diffraction patterns provide much information about
the crystal’s microscopic geometry.

An example of the very narrow constructive interference zones that

emerge from e−’s penetrating a thick crystal appears in the following
figure.

J. Gunion 9D, Spring Quarter 75

 Fig. 5−7, p. 157

Figure 16: Bragg diffraction of 50 keV electrons from a 4000 nm thick

single crystal of CU3Au.

More Examples

J. Gunion 9D, Spring Quarter 76

 (I) If moving with v = 300 m/s, what would be the wavelength (a) of an
18, 000 kg airplane, and (b) of an electron?
Answer:

6.63 × 10−34 J · s
λairplane = = 1.23 × 10−40 m
(18000 kg)(300 m/s)
6.63 × 10−34 J · s
λelectron = = 2.43 × 10−6 m .(97)
(9.11 × 10−31 kg)(300 m/s)

The latter is something you can hope to measure using the kind of
techniques just described. The former is not something you could ever
measure — we do not need to worry about wavelengths and wave patterns
in our everyday world!

(II) Consider a two-slit experiment using electrons. The slits are assumed
to be very narrow compared to the wave-length of the electrons. Beyond
the slits is a bank of e− detectors. At the center detector, directly in the
path the beam would follow if unobstructed, 100 electrons per second
are detected. Suppose that as the detector angle varies, the number per

J. Gunion 9D, Spring Quarter 77

 unit time of e−’s arriving varies from a maximum of 100/s to a minimum
of 0. Suppose the electrons have K = 1.0 eV of kinetic energy and the
narrow slits are separated by S = 0.020 µm. (a) At what angle, θX , is
the detector X located where the minimum is reached? (b) How many
electrons would be detected per second at the center detector if one of
the slits were blocked? (b) How many electrons would be detected per
second at the center detector and at detector X if one of the slits were
narrowed to 36% of its original width?

Answers:
h h
(a) At the minimum, we require S sin θX = 12 λ. We need, λ = p
= mv
and we get v from

1 −19 1
K= 2
me v , ⇒ (1 eV )×(1.6×10 J/eV ) = (9.11×10−31 kg)v 2
2 2
(98)
which gives v = 5.93 × 105 m/s. From this we get

p = mv = (9.11×10−31 kg)(5.93×105 m/s) = 5.40×10−25 kg·m/s ,

(99)

J. Gunion 9D, Spring Quarter 78

 and

h 6.63 × 10−34 J · s
λ= = = 1.23 × 10−9 m. (100)
p 5.40 × 10−25 kg · m/s

Inserting this into our requirements gives

!
−9
1 1.23 × 10 m
sin θX ∼ θX = = 0.031 or 1.76◦. (101)
2 0.020 × 10−6 m

(b) In the discussion that follows, we do not write the wave form explicitly.
But, there is always a wave form present. In the present case of e−
waves the wave form would be something like

Aei(kx−ωt) = A [cos(kx − ωt) + i sin(kx − ωt)] , (102)

initially, i.e. before passing through a slit, and would afterwards be similar
in form with kx − ωt replaced by kr − ωt, where r is the distance from
a slit. The important point will be that whatever the wave form, the
two slits will have equivalent wave forms, that are in phase at a central

J. Gunion 9D, Spring Quarter 79

 detector location, or exactly out of phase at the first minimum. The
fluxes referred to below can be thought of as the average intensity of the
oscillating waves.
With both slits open, the electron flux (electrons per second) is 100/s
at the central detector. But, the electron flux is proportional to the
probability of detection and therefore to the square of the amplitude of
the total matter wave (from both slits):

|ΨT |2 ∝ 100/s ⇒ |ΨT | ∝ 10 . (103)

Since the two slits are very narrow and the waves from the two slits add
equally at this point of constructive interference, the amplitude of either
individual wave must be half the total:

|Ψ1| ∝ 5 ⇒ |Ψ1|2 ∝ 25/s . (104)

With one slit closed, the electron flux at the central detector would be
1/4 the two slit flux, or 25/s. Note the importance of assuming that the
slits are really narrow. In this case, this same flux would apply for all the
electron detectors, regardless of angle, when only one slit is open.

J. Gunion 9D, Spring Quarter 80

 You might say, what happened to conservation of probability, or of the number
of photons?. We have not violated anything here. The 100/s when both
slits are open applies only to the central detector. As one moves
away from the central detector, the intensity varies (see E&M two-slit
discussion) from a maximum of 100 to a minimum of 0 so that averaging
over the screen we get 50/s. This is precisely 2× the single-slit uniform
intensity, as required by photon number conservation!

(c) If one slit were open and its width (already very narrow) were reduced
to 0.36 of its original size, all detectors would register an electron flux
that is 0.36 × 25/s, or 9/s. In equation form, this means that

|Ψ01|2 = 0.36 × |Ψ1|2 ∝ 0.36 × 25/s = 9/s , ⇒ |Ψ01| ∝ 3. (105)

This is 60% of the original amplitude.

With both slits open (but with slit #1 at only 36% of its original size),
we have two waves of different amplitudes, one proportional to 5 (slit
#2 of original width) and one proportional to 3 (slit #1). At points of
constructive interference, such as the central detector, where the waves

J. Gunion 9D, Spring Quarter 81

 add, the total amplitude will be ∝ 5 + 3:

|Ψ0T |constructive ∝ 8 , ⇒ |Ψ0T |2constructive ∝ 64/s. (106)

At points of previously complete destructive interference, where the two

waves are 180◦ out of phase, such as at detector X, the cancellation
would no longer be complete. The waves still come in with opposite
signs for the amplitudes so that the amplitude is proportional to 5 − 3,
leading to

|Ψ0T |destructive ∝ 2 , ⇒ |Ψ0T |2destructive ∝ 4/s . (107)

The average electron flux is

1
(64/s + 4/s) = 34/s , (108)
2

i.e. the sum of the 9/s and the 25/s expected from each slit alone.
To repeat, to find the probability (or flux) at a given location, we do
not add the probabilities (or fluxes) from each slit at that location; these

J. Gunion 9D, Spring Quarter 82

 are always positive, so they cannot cancel. Rather, we add the wave
amplitudes, which may add constructively or destructively, to find the
total wave, and then square the total wave to find the probability.
*** There is a special problem assigned to cover this material: see web
page. You will be tested on some quiz or exam on this kind of thing. ***

The electron microscope and related devices

Recall the formula for the e− wavelength in terms of the acclerating

voltage
1.227 nm 1
λe = p q . (109)
V (volts) eV
2 + 1
2m e c

An electron microscope makes use of an accelerating voltage of V ∼

100000 volts, leading to λe ∼ 0.003 nm, as compared to typical
light wavelengths in the visible spectrum of ∼ several hundred nm.
Thus, electrons have the potential of far greater resolution capable of
revealing much finer structures. Magnification, however, is not directly
related to λ, being limited by other things such as appertures and
“optics” of the device. In practice, the best that can be achieved is

J. Gunion 9D, Spring Quarter 83

 a magnification of 10,000 to 100,000 with resolution of 0.2 nm, as
compared to magnification and resolution of ∼ 2000 and ∼ 100 nm for
optical microscopes. The electron microscope allows pictures of individual
DNA strands, bacteria and the like. These developments were crucial to
modern biology, ....
Other devices include scanning electron microscope (SEM) and scanning
tunneling microscope (STM) and atomic force microscope (AFM). These
involve further applications of QM to which we shall turn in later chapters.
The latest device for studying structures, especially of germanium crystals
and other semi-conductors, is a light source of very high energy γ-rays.
These are beams of photons with energies beyond even the X-ray range.
Typical energy is ∼ 10 to 50 × 109 eV . The wavelength that one is
talking about is

ch 1.24 × 103 eV · nm
λ= ∼ = 1.24 × 10−7 nm . (110)
E 10 × 109 eV

J. Gunion 9D, Spring Quarter 84

 More on the Heisenberg Uncertainty Principle

Let us review once more the HUP. We have found by example that for any
wave pattern it is always true that

1
∆k∆x ≥ , (111)
2
where I have stated that the minimum arises for Gaussian wave packets.
We then input either Planck (photons) or de Broglie (matter waves) via
the relation
h 2π 1
p = = h̄ = h̄k , ⇒ ∆p∆x ≥ h̄ . (112)
λ λ 2
Another uncertainty relation involves the uncertainty in energy of a wave
packet, ∆E, and the time, ∆t, taken to measure that energy. Using
Gaussian or other wave forms that are functions of kx − ωt and that are
of finite extent in ∆t, we can derive the wave result that

1
∆ω∆t ≥ , (113)
2
J. Gunion 9D, Spring Quarter 85
 where once again the minimum is for Gaussian forms.
We now input the relation

1
E = hf = h̄(2πf ) = h̄ω , ⇒ ∆E∆t ≥ h̄ . (114)
2

This result states that the precision with which we can know the energy
of some system is limited by the time available for measuring the energy.

The Mechanistic Point of View of the HUP

∆px∆x
Here we consider an idealized (thought) experiment in which we try
to measure the position of a particle using photons. A more careful
treatment is given in the book. Here, I just give the idea of the
argument.
h
• The photon carries momentum given by p = λ .
• The matter particle tends to pick up some portion of this momentum
(depending upon angle of incidence which is determined by size of lens

J. Gunion 9D, Spring Quarter 86

 of microscope employed — see book) so

h
(∆p)particle being probed ∼ . (115)
λ
• Also, the position of the particle can not be determined to any greater
precision than the wavelength λ of the light:

(∆x)particle being probed >

∼ λ. (116)

• Multiplying, we get

(∆p∆x)particle being probed >

∼ h. (117)

This shows in a mechanistic way that any attempt to improve your

measurement of ∆x by employing smaller λ necessarily increases the
amount of momentum that the photon will typically transfer to the
particle being probed (the direction being unpredictable) as a result of
the higher momentum being carried by each photon.
The key physics ideas that lead to the uncertainty principle from the
mechanistic point of view are:

J. Gunion 9D, Spring Quarter 87

 1. There is an indivisible nature of the light particles (photons) and
nothing less than one photon can be used to perform a measurement
of the momentum or energy of another particle.
2. There is a wave nature of light that even a single photon cannot evade.
3. These lead to the impossibility of predicting or measuring the precise
(classical) path that a single scattered photon will follow, which in turn
implies inability to determine precisely the momentum transferred to
the electron.

∆E∆t
A similar argument is possible for the ∆E∆t relation.
• Consider a wave of frequency f incident on a particle at rest.
• Suppose that the minimum uncertainty in the number of waves we can
count is 1 wave.
# we count
Since f = , we get
time interval

1
∆f = , (118)
∆t

where ∆t is the time interval available for counting the waves.

J. Gunion 9D, Spring Quarter 88

 • We now wish to employ the photon scattering off the particle to
determine the particle’s energy. The photon is bringing in an amount
of energy that is uncertain by the amount (from Planck formula for
photon)
1
∆E = h∆f = h , ⇒ ∆E∆t ∼ h . (119)
∆t
Another approach to this same energy-time uncertainty is the following:
• a photon with E = cp = hc λ
hits a particle in a powerful microscope.
• The best that you can do to determine time is specified by the arrival
of the photon wave front. When this wave front arrives is known no
better than λc (i.e. the time separation between two bumps in the light
wave intensity). Thus,
λ
∆t ∼ (120)
c
is the smallest amount of time that you are using to perform your
energy measurement.
• Meanwhile, the photon impact changes the energy of the particle it is
probing by an amount of order

hc
∆E ∼ ; (121)
λ
J. Gunion 9D, Spring Quarter 89
 i.e. if you want to not change the energy of the particle the photon is
probing, you must keep λ large. But, then this means it takes longer
for the wave front arrival to be clearly defined. The result is

hc λ
∆E∆t = ∼ h. (122)
λ c

Heisenberg Uncertainty Principle (HUP) Examples

e− in a Hydrogen atom
Is there any relation between the energy levels of the Hydrogen atom
and the uncertainty principle? Let’s see.
We suppose that the electron is confined in a one-dimensional sense
to a region of order ∆x. Then, let us employ the HUP in the form
∆px ∼ h̄/∆x. (I have chosen the numerical factor to give me the
prettiest results.) The associated kinetic energy is

(∆px)2 h̄2
K ≥ (K)∆px ≡ > . (123)
2me 2me (∆x)2

J. Gunion 9D, Spring Quarter 90

 Let us demand that this kinetic energy not exceed significantly the
negative potential energy associated with this same distance scale.

h̄2  ke2 
 
(K)∆px ∼ ∼ − . (124)
2me(∆x)2  ∆x 

This gives us,

h̄2 a0
∆x ∼ = . (125)
2kme e2 2
What this is telling us is that it is very difficult to confine the electron to
a distance much smaller than a0 using the electromagnetic force. If we
scale up the potential energy using Z for a charged ion, the ∆x cannot
decrease faster than 1/Z without violating the HUP. The same argument
a0 1 ke2 Z
would give us ∆x > 2 Z . Plugging this into − ∆x gives us energy levels
that should scale as Z 2, as they do.
We can actually go further. It is apparent that the typical potential
energy for an e− confined to a region of size ∆x is

ke2
U =− , (126)
∆x
J. Gunion 9D, Spring Quarter 91
 which can be combined with our minimum K for the particle to compute
the total energy:

h̄2 ke2
E =K+U = − . (127)
2me (∆x)2 ∆x

Note that E → 0 for ∆x → ∞, has a minimum somewhere and then

E → +∞ for ∆x → 0. The most likely value of ∆x is the value that
minimizes E. Taking derivatives, this gives

∂E h̄2 ke2
= 0 = −2 + (128)
∂∆x 2me (∆x)3 (∆x)2

which is solved by
h̄2
∆x = = a0 , (129)
me ke2
which, after substitution into the above form for E gives

k2e4me
E=− 2 (130)
2h̄
J. Gunion 9D, Spring Quarter 92
 which is precisely the E0 energy level of the first Bohr orbit.
The Unstable Z boson
The Z boson is an unstable (i.e. a particle that decays) with mass
mZ ∼ 91 × 109 eV . The average lifetime of the Z is

τZ = 2.9 × 10−25 s . (131)

This lifetime is determined by how many different types of particles it can

decay into and what the strengths of those decays are. One important
type of particle is something called a “neutrino”, ν. There are potentially
many different types of neutrinos. The more Z → νν channels there
are, the shorter the Z lifetime. Since we cannot see ν’s directly (they
are very weakly interacting and have zero charge), it is important to
determine the Z lifetime to indirectly determine how many ν’s there are.
However, the above τZ is far too short to actually measure directly. So,
how do we determine it. Answer: use Heisenberg uncertainty principle
for theoretically predicted shape of “mass spectrum”.
If we attempt to measure the mass of the Z by using e+e− → Z
collisions with different values of the e+e− total energy, what do we

J. Gunion 9D, Spring Quarter 93

 expect to see? The HUP says we should expect to see a distribution of
mass values of a certain shape.

35
N=2 DELPHI
30 N=3
N=4
25

20

15

10

5
88 89 90 91 92 93 94 95
Energy, GeV
Figure 17: e+e− → hadrons as a function of Me+e− . The Z peak is
centered about mZ = 91.13 × 109 eV = 91.13 GeV and has a width of
roughly 2 to 3 × 109 eV .

J. Gunion 9D, Spring Quarter 94

 The resonance picture shows that we cannot make a precise determination
of the mass. As stated, this is required by the uncertainty principle which
says that we would need an infinite amount of time to get a precise mass
determination, whereas the resonance disappears quickly. The HUP says

1 6.582 × 10−16 eV · s
∆E ≡ ∆mZ ∼ h̄ = ∼ 2.3×109 eV = 2.3 GeV ,
τZ 2.9 × 10−25 s
(132)
and this is what is explicitly seen in the plot. The plot also shows how
the peak would get narrower (broader), relative to its height, if certain
decay modes are eliminated (added).

Relation of HUP to Two-Slit Interference Pattern

Things we know
Assume equal slit widths.
1. It is only when we have both slits open that the interference pattern
develops.
2. Even if we send only one e− at a time, if both slits are open the e−
hits at the detector bank will accumulate where the interference wave

J. Gunion 9D, Spring Quarter 95

 prediction has a maximum and no e−’s will hit at the destructive wave
pattern cancellation minima.
3. We cannot be sure where any given e− will end up; only the final
average pattern can be predicted with certainty.
4. If we close one slit, the accumulation pattern changes to approximately
uniform (for very narrow slits).

Now try to do better.

1. Suppose you have both slits open, but you try to measure unambiguously
which slit a given e− passes through.
⇒ you disturb the e−.
2. For example, place some detecting particles on the right side of the
slit. Use the recoil of one of these particles to determine which slit the
e− goes through.
3. To decide which slit, need to measure the detecting particle’s position
with ∆y  D (D in the figure is the separation between slits, not the
size of an individual slit).
4. During the collision, the detecting particle suffers a change in momentum
∆py , equal and opposite to the change in momentum experienced by
the e− passing through the slit.

J. Gunion 9D, Spring Quarter 96

 Fig. 5−33, p. 184
Figure 18: Determining the slit the e− goes through.
5. An undeviated e− landing at the first minimum and producing an
interference pattern has

py h
tan θ ∼ θ = = . (133)
px 2pxD

J. Gunion 9D, Spring Quarter 97

 (This is our old 12 λ path difference requirement.)
6. Thus, we require that an e− scattered by a detecting particle has

∆py h h
θ= or ∆py  (134)
px 2pxD 2D

if the interference is to not be distorted.

7. Because
(∆py )e− = −(∆py )detecting particle , (135)
h
(∆py )detecting particle  (136)
2D
is also required.
8. Altogether, for the detecting particle, we require

h h
(∆py ∆y)detecting particle  ·D = . (137)
2D 2
This is in clear violation of the uncertainty principle.
If ∆y is small enough to determine which slit the electron goes through,
∆py will be so large that the e−’s will be deflected all over the place
and the interference pattern will be destroyed.

J. Gunion 9D, Spring Quarter 98

 Matter Probability Waves and the Schroedinger
Equation

I will be, and have been for that matter, kind of combining the material
appearing in Chapters 5 and 6. You should read this material as a unit. For
example, I have set up the ingredients for writing down the Schroedinger
equation that first appears in Sec. 6.3, where it is introduced more or less
by fiat. I will now go a little bit beyond just simply writing it down and tell
you one way that you can kind of understand it.

The Schroedinger Equation

The Schroedinger equation is the wave equation for matter probability

waves when the non-relativistic limit is appropriate. A different equation
must be used if the matter particles are in a situation where they typically
have relativistic velocities.

Recall the game we played for “deriving” the light-wave equations.

We wrote E 2 = c2p2, multiplied this equation by some Ey or Bz

J. Gunion 9D, Spring Quarter 99

 electromagnetic field and then replaced

h̄ ∂ ∂
px = , E = ih̄ . (138)
i ∂x ∂t

In this case, we write (non-relativistically and neglecting any potentials

or such for the moment — i.e. we consider a “free” particle)

2  2 
2 p 2 p
E = m0 c + , ⇒ EΨ = m0 c + Ψ
2m 2m
∂Ψ h̄2 ∂ 2Ψ
⇒ ih̄ = m0 c2 Ψ − . (139)
∂t 2m0 ∂x2

In the text, the constant m0c2 is absorbed into an overall redefinition of

the energy scale in this NR limit, but this is really misleading when it
comes to considering how the particle is moving, as we shall see.

The very simplest solution to this equation is the exponential form:

Ψ = Aei(kx−ωt) , (140)

J. Gunion 9D, Spring Quarter 100

 where we compute ω(k) from the SE (Schroedinger equation) requirement

i(kx−ω (k)t 2
2 2
Aei(kx−ω(k)t) ,


ih̄(−iω(k))Ae = m0c − h̄ (ik) (141)

implying
m0 c2 h̄k2
ω(k) = + . (142)
h̄ 2m
Note how this is consistent with the de Broglie / Planck relations

2π 2πp p E E
k= = = , and ω = 2πf = 2π = (143)
λ h h̄ h h̄

or equivalently

h 2π
p= = h̄ = h̄k , E = hf = h̄2πf = h̄ω (144)
λ λ

being substituted into E = m0c2 + p2/2m:

(h̄k)2
E = h̄ω = m0c2 + . (145)
2m0

J. Gunion 9D, Spring Quarter 101

 The m0c2/h̄ part of ω(k) is normally dropped in the non-relativistic limit,
as it amounts to an irrelevant redefinition of the absolute energy scale.
In the relativistic limit, it cannot be dropped.
Note that a sin or cos function form does not solve the SE. For example,
if we tried the sin function form, the single time derivative would give
us a cos function, whereas the double space derivative would give us
back − sin. In the free-particle case at any rate, we must employ the
intrinsically complex “plane wave” form

Aei(kx−ωt) = A cos(kx − ωt) + iA sin(kx − ωt) . (146)

This is still a traveling complex wave moving in the +x direction because

of the kx − ωt argument which says that if I move in the direction x
by an amount ∆x then I can compensate by advancing t by an amount
∆t = k∆x/ω. We shall soon consider whether the velocity that you
might compute from

∆x ω(k) m0 c2 h̄k m0 c2 p
= = + = + (147)
∆t k h̄k 2m0 p 2m0

has any meaning. The answer is no. We will have to deal with wave

J. Gunion 9D, Spring Quarter 102

 packets and the concepts of phase and group velocity to which we shortly
turn.

The HUP (again).

As stated, the above ei(kx−ωt) solution to the SE is called a plane wave

solution.

The particle described by this solution has a precisely defined momentum

(in the x direction) of p = h̄k, as computed above.

If the HUP is correct, ∆x should be infinite! and it is!

This is because
|Ψ|2 = A2 (148)
is completely independent of x and so the particle has a uniform
probability of being anywhere along the x axis! Obviously, it is nonsense
to discuss the velocity of a uniform probability distribution.

J. Gunion 9D, Spring Quarter 103

 How fast is the matter particle moving?

We must now face the subtle issue of how to construct a wave form
that can describe an actual physical particle and how it is we determine
the velocity of the particle. This will bring us to consider the difference
between group and phase velocity.
We considered in Fig. 8 and surrounding material how to create a photon-
like object by adding together E&M type wave patterns. There, the
group and phase velocities were both equal to c and we did not distinguish
or even discuss. For massive particles one must be careful.
The book has a discussion using two sin waves. However, I prefer to
use the plane wave form we have just been discussing, which is an actual
solution of the SE (unlike the sin or cos forms alone).
To define a “particle” we clearly cannot use a single plane-wave solution
for which the particle has no preferred location.
We must superimpose a least two plane-wave solutions.
Let us begin with exactly two and see what happens.

J. Gunion 9D, Spring Quarter 104

 • Write
Ψ(x, t) = Aei(k1−ω1t) + Aei(k2−ω2t) (149)
where we have taken in to account the fact that when k changes a
little bit then so must ω.
• Let us assume we only change k by a small amount and write

k1 = k0 + dk , k2 = k0 − dk , ω1 = ω0 + dω , ω2 = ω0 − dω .
(150)
• Then,
h i
Ψ = A ei[(k0+dx)x−(ω0+dω)t] + ei[(k0−dx)x−(ω0−dω)t]
h i
i(k0 −ω0 t) i[(dk)x−(dω )t] −i[(dk)x−(dω )t]
= Ae e +e
= Aei(k0x−ω0t)2 cos [(dk)x − (dω)t] . (151)

• We have a complex exponential that moves at speed

ω0 h̄ω0 E0
vp ≡ = = , (152)
k0 h̄k p0

J. Gunion 9D, Spring Quarter 105

 where vp is called the phase velocity appropriate to these central
values, modulated by a cosine function that moves at speed

dω
vg ≡ , (153)
dk
which is called the group velocity.
• It is this latter term that defines the envelope of the wave and tells us
where the probability is. Indeed,

|Ψ|2 = 4A2 cos2 [(dk)x − (dω)t] . (154)

The phase velocity has disappeared. Since a physical particle moves

with its probability, the actual physical speed of the particle is vg .
• Of course, this two-plane-wave superposition is still a bit too simple.
It is really a sequence of peaks that we observe passing by. To get a
localized wave form we have to form the kind of superposition discussed
earlier: Z ∞
i(kx−ω (k)t)
Ψ(x, t) = ψ(k)e
e dk . (155)
−∞
We will do another explicit construction of this type shortly. The
important point is that since the sum must include plane waves of

J. Gunion 9D, Spring Quarter 106

 various k (= p/h̄) and various ω(k) (= E(p)/h̄) values, neither the
momentum nor the energy of the group is well defined.
• All of this applies to photon waves as well as particle waves. It’s
just that ω(k) = ck for a photon wave is a much simpler formula
2 2
than ω(k) = mh̄0c + 2h̄km0
which is the formula that applies for a NR
particle with mass. For E = pc, vp = E/p = c and for ω = ck,
vg = dω/dk = c also.
• It is useful to next generalize to the proper relativistic result:
q
E(p) = p2c2 + m20c4 . (156)

The phase velocity is, as above,

s  2 s  2
E(p) m0 c m0 c
vp = =c 1+ =c 1+ . (157)
p p h̄k

From this we see that for a massless photon, vp = c, but that for a
massive particle vp > c. However, this is not a problem since vp does
not describe where the probability is!

J. Gunion 9D, Spring Quarter 107

 • Instead, we must look at vg = dω/dk. We have (using our time-tested
relation f = E/h)

E(p = h̄k) 1
q
ω(k) = 2πf = = h̄2k2c2 + m20c4 ,
⇒ h̄ h̄
dω(k) 1 h̄2kc2
vg = = q
dk h̄ h̄2k2c2 + m2c4
0
c c2
= q = , (158)
1+ m 0c

2 vp
h̄k

implying that vg < c, given that vp > c as just derived above.

• We can now check that vg is the actual speed of the particle by using

E γ(u)m0c2 c2 c2
vp = = = ⇒ vg = = u! (159)
p γ(u)m0u u vp

More on wave functions and propagating waves

J. Gunion 9D, Spring Quarter 108

 It was Max Born who in 1925 zeroed in on the interpretation of Ψ as a
probability amplitude.
He had in mind in particular the case where we want to describe a single
particle that is sufficiently localized (unlike our nice plane wave solution,
more like our wave packet type solutions) that you can (and should)
define a normalizable probability. Then, he said that

P (x)dx = |Ψ(x, t)|2dx (160)

should be the probability that the particle will be found in the infinitesimal
interval dx about the point x. Once again, I stress that Ψ itself is
not something you can observe and even |Ψ|2 is only observable in a
probabilistic sense.
Because of its relation to probabilities, we will typically insist on the
following.
1. Ψ should be single valued and a continuous function of x and t. In this
way, no ambiguities will arise concerning the predictions of the theory.
2. In fact, we will typically require that not only should Ψ be continuous,
but it should also be smooth in the sense that its first derivatives are
finite.

J. Gunion 9D, Spring Quarter 109

 However, there are certain specialized situations where this is not
appropriate. But, for most of the physical situations we shall discuss
in this course, this requirement should be met. The exception is when
there is some discontinuity in some potential.
3. Next, the total probability of finding a particle somewhere should be
unity. Since |Ψ|2 is the probability distribution for the particle in
question, this means
Z ∞
|Ψ(x, t)|2dx = 1 for any t . (161)
−∞

Similarly, Z b
Pab = |Ψ(x, t)|2dx (162)
a
should be the probability of finding the particle in the interval a ≤ x ≤
b.
In this formulation of QM, the fundamental problem of QM is the
following: Given the wavefunction at some initial instant, say t = 0, what
is the wavefunction at any subsequent time t.
To answer this question requires a dynamical equation for Ψ(x, t). We

J. Gunion 9D, Spring Quarter 110

 have already written this dynamical equation down. It is Schroedinger’s
(wave) equation.
We have discussed it in the free-particle case, mostly focusing on the
free-particle plane wave solution. However, the function Ψ = Aei(kx−ωt)
is not normalizable in the way we would like. There is uniform probability
everywhere.
However, we have also learned how to proceed. We form an appropriate
superposition of plane waves, sometimes called a wave packet. As stated
earlier, the Gaussian type of superposition is the most ideal. Let us go
through the details now.

1. We start with Z ∞
ikx
Ψ(x, 0) = Ψ(k)e
e dk (163)
−∞
√
with Ψ(k) = (Cα/ π) exp[−α2k2].
e
Since Ψ(k)
e is centered about k = 0, we will be constructing a wave
packet that has a central momentum and velocity of 0 (motionless
particle), although there will be a spread of velocities and momenta
about this central value.
2. We can perform the k integral if we carefully examine it and use the

J. Gunion 9D, Spring Quarter 111

 process called “square completion”. We need

Cα ∞ x2
Z 2
−(αk− 2ix ) −
Ψ(x, 0) = √ dke α 4α2
π −∞
C −x2/4α2 ∞ −z2
Z
= √ e e dz defining z = αk − ix/(2α)
π −∞
2 2 2
= Ce−x /4α = Ce−(x/2α) . (164)
R∞ 2 √
Above, we used −∞
e−z dz = π — see any integral table.
A Gaussian form has begot another Gaussian form.
3. We now state that the appropriate way to define the width of a
Gaussian form is to take the probability and write it in the form

−x2 /2(∆x)2 −k2 /2∆k)2

P (x) ∝ e , or P (k) ∝ e , (165)

for distributions centered about x = 0 and k = 0.

2 2 2 2
So, using |Ψ(k)|
e 2
∝ e−2α k and |Ψ(x, 0)|2 ∝ e−2x /(4α ) we identify

k2 1 x2
2α2k2 = 2
, and 2 2
x 2
= (166)
2(∆k) 4α 2(∆x)2

J. Gunion 9D, Spring Quarter 112

 from which we find
1
∆k = , and ∆x = α , (167)
2α
implying ∆k∆x = 12 . That these are the appropriate definitions of
∆k and ∆x will be left until a later time. But, assuming this, we are
back to our minimum HUP after multiplying by h̄ and converting to
p = h̄k.
4. Finally, we can normalize this Ψ(x, 0) by requiring
Z ∞
1 = |Ψ(x, 0)|2dx
Z−∞
∞
2 − 2x 2 / 4α 2
= C e dx
−∞ Z ∞
2
√ −y 2
√
= C ( 2α) e dy using y = x/( 2α)
−∞
2
√ √ 1
= C ( 2α) π , ⇒ C= √ , (168)
(2π)1/4 α

where we have claimed that α = ∆x for an appropriate definition of

∆x.

J. Gunion 9D, Spring Quarter 113

 We should now discuss how it is that this wavepacket evolves with time.

We find that it disperses, getting broader in space as time passes. This

is because it was initially made up of plane waves with many different k
values (both positive and negative).

The steps and results are as follows:

1. For t 6= 0, use
Z ∞
i(kx−ω (k)t)
Ψ(x, t) = Ψ(k)e
e
−∞  
Z ∞ h̄k 2
i kx− 2me t
= Ψ(k)e
e dk
−∞Z
∞
C −α2 k2 +ikx−ih̄tk2 /(2me )
= √ e dk (169)
π −∞

where we inserted the expression for ω(k) that is required by Schroedinger’s

equation.
2. We can again complete the square. In the exponent we have

J. Gunion 9D, Spring Quarter 114

 ikx − k2[α2 + ih̄t/(2me)]
 s 2
ih̄t ix x2
= − k α2 + − q  −  .(170)
  
2me 2 α2 + ih̄t
4 α2 + ih̄t
2m e 2m e

I will not go through the details of the variable shift and integration.
The fact that the k2 coefficient is complex is not any particular
problem. You just R∞have to trust your√square completion process and
use the standard −∞ exp[−y 2]dy = π at the appropriate point.
3. The important component of interest is the residual coming from the
last term above which says that
 
x2
Ψ(x, t) ∝ exp −  . (171)
 
ih̄t
4 α2 + 2m e

4. As always, in defining a width we are interested in the probability and

J. Gunion 9D, Spring Quarter 115

 not the amplitude. We have

" # " #
2 2
x x
2
|Ψ(x, t)| ∝ exp − 14
ih̄t
× exp − 14
2
α + 2m e α2 − 2im
h̄t
   e
1 2 1 1
= exp − 4 x 2
+ 2
 α + ih̄t/(2m
e) α − ih̄t/(2me)
2α2
= exp − 14 x2 
  
h i2 
α4 + 2h̄t
me
 
2
 1 x
= exp −

i2 
2 α2 + h̄t
h
2m e α
 2 
1 x
≡ exp −
2 [∆x(t)]2
 2
h̄t
⇒ [∆x(t)]2 = [∆x(0)]2 + , (172)
2me∆x(0)

where we used α = ∆x(0). In getting from the 2nd to the 3rd line

J. Gunion 9D, Spring Quarter 116

 above, we used
1 1 2a
+ = . (173)
a + ib a − ib a2 + b2

Non-stationary case
If we want a particle whose central location is moving, we simply modify
our input form of Ψ(k)
e so that it is a Gaussian (or other choice) centered
about k = k0. The packet would then move with a group velocity given,
in the NR case, by
    2 
dω(k) d h̄k h̄k0
vg = = = . (174)
dk k = k0 dk 2m0 k = k0 m0

The wave packet would, just as in the k0 = 0 case discussed above, spread
out as time passed due to the presence of a distribution of momenta and
velocities for individual subcomponents of the wave packet.

Definition of ∆x?
So, let us now use the probability interpretation of Ψ(x, 0) to actually

J. Gunion 9D, Spring Quarter 117

 define an appropriate definition of ∆x. We employ (from eq. (164))

2 2
 2
 2 2
P (x) = |Ψ(x, 0)|2 = Ce−x /4α  = C 2e−x /2α . (175)


We could then define the average value of x as

Z ∞
hxi ≡ x P (x) dx , ⇒ hxi = 0 (176)
−∞

for the particular form of P (x) above due to the fact that P (x) is even
in x → −x, whereas as x changes sign under x → −x.
We now come to the “standard” definition of ∆x:

2 2
(∆x)2 = h(x − hxi)2i = hx2i − 2hxhxii + hxi = hx2i − hxi . (177)

Let us compute hx2i.

Z ∞
hx2i ≡ x2 P (x) dx
−∞

J. Gunion 9D, Spring Quarter 118

 Z ∞
2 2 −x2 /2α2
= x C e dx
−∞ √
2 3
R ∞ −ax2 √
= C 2πα using −∞ e dx = π/(2a3/2)
2 2
√
= α using C = 1/( 2πα) . (178)

Thus, using this definition of ∆x, we do indeed find that ∆x = α as

claimed earlier.
You should remember how we proceeded here. We could compute other
average values using P (x), and these will have real physical meaning in
various situations, just as ∆x really describes the size of a wave packet
in a quantitative way.

Another example: a structure localized in time

This time, we consider a probability that is somewhat localized in time.
In particular, we consider the case of an unstable particle produced at
t = 0 which then decays according to a probability distribution

dN
P (t) ≡ = N0e−t/τ , (179)
dt
J. Gunion 9D, Spring Quarter 119
 where by convention τ is referred to as the particle lifetime.
Of course, we must remember that P (t) is notpthe amplitude, but rather
the probability. The amplitude will be f (t) ∝ P (t) ∝ e−t/2τ .
If we are talking about solutions to the SE, it must be that this time
structure is a superposition of plane-wave type solutions (these are all we
have — they form a complete set). In the case of a time structure, this
means that there must exist a fe(ω) such that
Z ∞
f (t) = dω fe(ω)e−iωt . (180)
−∞

The appropriate fe(ω) is found by the inverse relation

∞
1
Z
fe(ω) = dteiωtf (t) (181)
2π −∞

as we easily show.
Proof

J. Gunion 9D, Spring Quarter 120

 We use a “dummy” variable t0 to define the fe(ω) integral form and then
write:

∞ Z ∞ 
1
Z
0
f (t) =? dω dt0eiωt f (t0) e−iωt
2π −∞ −∞
Z ∞ Z W
1 0 0 iω (t0 −t)
= dt f (t ) lim dωe
2π −∞ W →∞ −W
Z ∞ iW (t0 −t) −iW (t0 −t)
1 e − e
= dt0f (t0) lim
2π Z−∞ W →∞ i(t0 − t)
∞
1 0 0 2 sin[W (t0 − t)]
= dt f (t ) lim
2π −∞ W →∞ t0 − t
= f (t) . (182)

The last step takes a bit of work. Note that if t0 6= t then the
sin[W (t0 − t)] is very rapidly oscillating as a function of t0 (for
R very large
W ) and in the large W limit nothing will survive in the dt0 integral
for any tiny dt0 interval where t0 6= t. Assuming that f (t) is a smoothly

J. Gunion 9D, Spring Quarter 121

 varying function on the scale of 1/W , this allows us to write

∞
1 2 sin[W (t0 − t)]
Z
dt0f (t0) lim 0−t
2π −∞ W →∞
Z ∞ t
0
1 0 2 sin[W (t − t)]
= f (t) lim dt
2π W →∞ −∞ t0 − t
= f (t) , (183)

where for the last step we have simply looked up the integral in a table
and found that its value is 2π independent of t0 − t and W .
QED
So, now let us return to the computation of interest:

∞
1
Z
fe(ω) = dteiωtf (t)
2π Z−∞
∞
1
∝ dteiωte−t/2τ
2π Z0
∞
1
= dtet(iω−1/2τ )
2π 0

J. Gunion 9D, Spring Quarter 122

 " #∞
t(iω−1/2τ )
1 e
=
2π iω − 1/2τ
 0
1 1
= . (184)
2π 1/2τ − iω

From this we compute the P (ω) distribution as

1
P (ω) = |fe(ω)|2 ∝ 1 . (185)
ω2 + 4τ 2

If we want to make connection with the Z resonance example given

earlier, see Fig. 17, we would write h̄ω = E − mZ c2 and the functional
form given above is precisely that used to draw the plotted curves, one of
which fits perfectly (for the correct choice of τ = τZ ) the experimental
data. The quantity h̄/τ that determines the width of the E distribution
is sometimes written as Γ = h̄/τ . For the precise τZ given earlier we
have
h̄
ΓZ = = 2.3 GeV , (186)
τz
as computed earlier, eq. (132).

J. Gunion 9D, Spring Quarter 123

 The Schroedinger Equation in the Presence of
Forces/Potentials

We begin by following our previous route of writing an equation for the

energy, but now including a potential energy term, U (x), which is
associated with a force, F (x) = −dU/dx:

p2
E= +U. (187)
2m0

We will consider only cases where U = U (x). Now, multiply by Ψ(x, t)

∂
and make the replacements E → ih̄ ∂t ∂
, p → h̄i ∂x to obtain the SE:

∂Ψ(x, t) h̄2 ∂ 2Ψ(x, t)

ih̄ =− + U (x)Ψ(x, t) . (188)
∂t 2m0 ∂x2

Although we have motivated the above replacements on the basis of the

HUP and wave ideas and analogies to E&M wave regarding momentum

J. Gunion 9D, Spring Quarter 124

 and energy, the SE and the probability interpretation of Ψ must really
be regarded as a basic new law that we must repeatedly test against
experiment.

Is there a solution analogous to ei(kx−ω(k)t) to the SE in the presence of

U?
When U does not depend explicitly on t, we can always write

Ψ(x, t) = ψ(x)φ(t) (189)

which is to say the time and space dependence can be separated. (In
the U = 0 case, ψ(x) = eikx and φ(t) = e−iωt.) Substituting this form
into eq. (188) and dividing by ψφ, we obtain (I will drop the subscript 0
on m0, and simply write m in what follows.)

h̄2 ψ 00(x) φ0(t)

− + U (x) = ih̄ . (190)
2m ψ(x) φ(t)

The lhs depends only on x and the rhs depends only on t. This is only
possible if both sides are equal to the same constant, which we call E

J. Gunion 9D, Spring Quarter 125

 (of course). We can immediately integrate the rhs:

dφ(t) −i E
ih̄ = Eφ(t) : ⇒ φ(t) = e h̄ t = e−iωt . (191)
dt

This leaves us with solving the time-independent lhs

h̄2 d2ψ(x)
− + U (x)ψ(x) = Eψ(x) , (192)
2m dx2

which is often rewritten in the form

d2ψ 2m
= 2 [U (x) − E]ψ(x) . (193)
dx2 h̄

If U = 0, this reduces to our free-particle situation with ψ(x) = eikx

and E = h̄2k2/(2m). Even in the case of U 6= 0, E will be the total
energy, and will be a constant describing the state in which the particle
resides.
2
The solution ψ(x) will depend upon U , but for P (x) = |ψ(x)| to be
physically interpretable, ψ(x) should be finite everywhere, single-valued

J. Gunion 9D, Spring Quarter 126

 dψ
and continuous. Further, ψ(x) should be smooth in that dx
should be
d2 ψ
normally be continuous. (Just from the SE itself, dx2
can only be infinite
if U is infinite.)

Finally, note that since |φ(t)|2 = 1, we have

2 2
P (x, t) = |Ψ(x, t)| = |ψ(x)| (194)

which is to say that P (x, t) is independent of time. This is why such

solutions are referred to as stationary states — all probabilities are static.

Particle in a box

We envision an electron or other particle confined absolutely to a region

between x = 0 and x = L by a potential


0, 0≤x≤L
U (x) = . (195)
∞ , x < 0 or x > L

J. Gunion 9D, Spring Quarter 127

 Fig. 6−6c, p. 201

Figure 19: The infinite square-well potential.

1. Because the particle is absolutely confined by the infinite potential, we

must have ψ = 0 for x < 0 and x > L.
2. By continuity, we must then have ψ = 0 at x = 0 and x = L.

J. Gunion 9D, Spring Quarter 128

 3. For 0 ≤ x ≤ L, U = 0 and we must solve (defining E = h̄2k2/(2m)
as usual in the NR case)

d2ψ 2mE
=− 2 ψ(x) = −k2ψ(x) . (196)
dx2 h̄

4. The solution can be a combination of sin kx and cos kx:

ψ(x) = A sin kx + B cos kx , for 0 < x < L . (197)

5. At x = 0, ψ(0) = 0 ⇒ B = 0.
At x = L, with B = 0, and assuming A 6= 0, ψ(L) = 0 ⇒ kL = nπ.
This latter condition can be written in what should now be a not-
unexpected manner. Namely:

nπ 2π nπ n
kL = nπ ⇒ k= ⇒ = → λ = L,
L λ L 2
(198)
which is to say that we must fit a half-integer number of wave-lengths
in between the box edges.

J. Gunion 9D, Spring Quarter 129

 Fig. 6−9a, p. 204 Fig. 6−9b, p. 204

Figure 20: The infinite square-well potential wave-functions and

probabilities.
The wave functions are:
 
nπx
ψn(x) = A sin , for 0 < x < L and n = 1, 2, . . . (199)
L

RL 2
p
To normalize, we require 0
|ψn(x)| dx = 1 which gives A = 2/L.

J. Gunion 9D, Spring Quarter 130

Figure 21: Scanning tunneling microscope (STM) pictures of electron
patterns in approximately infinite well set ups.
Above, are two pictures of the electron density for two approximately
infinite well set ups. For example, the 2nd picture has a one-dimensional
oval well set up. The regular peaks along the ring of atoms that create
a “well” like our square well are the STM measurements of the electron
density. Note the wave pattern. So, you should believe this stuff!

J. Gunion 9D, Spring Quarter 131

Figure 22: (a) The discovery of the STM’s ability to image variations in the density
distribution of surface state electrons created a compulsion to have complete control of
not only the atomic landscape, but the electronic landscape also. Here we have positioned
48 iron atoms into a circular ring in order to ”corral” some surface state electrons and
force them into ”quantum” states of the circular structure. The ripples in the ring of
atoms are the density distribution of a particular set of quantum states of the corral. The
STM measured distributions agree with those found for the corral by solving the classic
eigenvalue problem in quantum mechanics – a particle in a hard-wall box. (b) Here we have
another STM measurement where iron atoms are shaped into a stadium shaped structure.
J. Gunion 9D, Spring Quarter 132
 6. The corresponding energy values are:

h̄2k2 n2π 2h̄2

E= = , n = 1, 2, . . . (200)
2m 2mL2

Fig. 6−7, p. 202

Figure 23: The infinite square-well potential energy levels.

The lowest energy level is the n = 1 state, which we call the ground

J. Gunion 9D, Spring Quarter 133

 state.
π 2h̄2
E1 = . (201)
2mL2
7. Note that n = 0, corresponding to E = 0 is not a possible solution.
This means the particle can never be at rest.
This is what we expect from the HUP. A particle confined to ∆x ∼ L,
should have ∆p ∼ h̄/L and K ∼ (h̄/L)2/(2m), which is precisely the
kind of result that we get!
E1 is sometimes referred to as the zero-point energy.

Example 1
Could we ever expect to see any of this stuff for a macroscopic object?
Answer: no!
For example, consider a 1.00 mg object confined to move between two
rigid walls separated by 1.00 cm. (a) Calculate the minimum speed of
the object. (b) If the speed of the object is (an observable amount of)
3.00 cm/s, find the corresponding, value of n.
Solution:
(a) Treating this as a particle in a box, the energy of the particle can

J. Gunion 9D, Spring Quarter 134

 only be one of the values given by Eq. (6.17), or

n2π 2h̄2 n2h2

En = = . (202)
2mL2 8mL2

The minimum energy results from taking n = 1. For the above m and
L, we calculate

(6.626 × 10−34 J · s)2

E1 = = 5.49 × 10−58 J . (203)
8.00 × 10−10 kg · m2

Because the energy is all kinetic (U = 0 inside well), E1 = 12 mv12, or

q
v1 = 2(5.49 × 10−58 J )/(1.00 × 10−6 kg) = 3.31 × 10−26 m/s ,
(204)
an immeasurably small speed. The object would take 3 × 1023 s, or about
1 million times the present age of the universe, to move the 1.00 cm
between the walls.

J. Gunion 9D, Spring Quarter 135

 (b) If the speed is v = 3.00 cm/s, then the particle’s energy is

1 1
E=K= 2
mv = (1.00×10−6 kg)(3.00×10−2 m/s)2 = 4.50×10−10 J .
2 2
(205)
This too, to be allowed, would have to be one of the En values. To
determine which one, we solve for n, and obtain
√
8mL2E q
n= = (8.00 × 10−10 kg · m2)(4.50 × 10−10 J ) = 9.05×1023 .
h
(206)
This is an enormous number. Indeed, the value of n is so large that we
would never be able to distinguish the quantized nature of the energy
levels. The difference between the energies for n = 9.05 × 1023 and
n0 = 9.05 × 1023 + 1 is only about 10−33 J , much too small to be
detected experimentally.
This is an example that illustrates Bohr’s correspondence principle, which
asserts that quantum predictions must agree with classical results for
large masses and lengths.

Example 2

J. Gunion 9D, Spring Quarter 136

 An electron is confined in an infinite well of 30 cm width.
(a) What is the ground-state energy?

(1)2π 2(1.055 × 10−34 J · s)2

E1 = = 6.77 × 10−37 J . (207)
2(9.11 × 10−31 kg)(0.3 m)2

(b) In this state, what is the probability that the e− would be found
within 10 cm of the left-hand wall?
2
We have P (x) = |ψ(x)| . For the general case,
r !2
2 2 nπx
|ψn(x)| = sin . (208)
L L

So,
  L/3
L 2 nπx
Z
2
P 0≤x≤ = sin dx
3 L L
Z0 L/3   
2 1 2nπx
= 1 − cos dx
0 L 2 L

J. Gunion 9D, Spring Quarter 137

 " #L/3 " 2nπ (L/3)
#
2nπx
2 x sin L
2 L L
= −L = − Lsin
L 2 4nπ L 6 4nπ
0
1 1 2nπ
= − sin . (209)
3 2nπ 3

The width L has canceled as we know it must on dimensional grounds.

For n = 1, we find a result of 1/3 − 0.137 = 0.196. This is to
be expected since P (x) for n = 1 peaks in the center of the well.
In fact, the probability for the electron to be in the center 1/3 is
1 − 2 × 0.196 = 0.609.
(c) If the e− instead has an energy of 1.0 eV , what is the probability
that it would be found within 10 cm of the left-hand wall?
We must first determine the n value for this energy by solving

n2π 2(1.055 × 10−34 J · s)2

1.0 eV × 1.6 × 10−19 J/eV = (210)
2(9.11 × 10−31 kg)(0.3 m)2

which yields n = 4.89 × 108. Then, using the general result above, we

J. Gunion 9D, Spring Quarter 138

 find
1 1 2(4.89 × 108)π 1
P (0 ≤ x ≤ L/3) = − sin ' .
3 2(4.89 × 108)π 3 3
(211)
Clearly, we are approaching a classical limit, even for an electron of
modest energy.
(d) For the 1 eV electron, what is the distance between nodes and the
minimum possible fractional decrease in energy?
The distance between nodes of the sin nπx
L
function is L/n. For
n = 4.89 × 108, this is
0.3 m
' 0.6 nm . (212)
4.89 × 108
This a very fine spacing indeed and would be very hard to detect.
The fractional spacing between energy levels is

(n−1)2 π 2h̄2 n2 π 2h̄2

En−1 − En 2mL2
− 2mL2
(n − 1)2 − n2 2 1
= = =− +
En n2 π 2h̄2 n2 n n2
2mL2

J. Gunion 9D, Spring Quarter 139

 2
' − ' −4 × 10−9 . (213)
4.89 × 108
Thus, the energy levels are very densely spaced and it would be very
difficult to detect a transition from one to the next at high n.

Interpreting The Wavefunctions

Notice that for n ≥ 2 there are places where |ψn(x)|2 = 0. This means
we can never find the particle at such a location.
You could ask, how can a particle get from one side of a box to the other
without passing such a zero-probability point?
The answer is that you cannot picture a particle moving in the classical
way that this question envisages. The wave picture is in contradiction to
such a particle picture.
Of course, for very large n, there are many zeroes but also many maxima.
These are so closely spaced that you cannot resolve the minima and
maxima, and you can only in practice see an average value of the
probability. Thus, what you will see will look like the particle can be at
any location it likes and can travel back and forth between the walls, but
that is not really what is correct at the underlying level.

J. Gunion 9D, Spring Quarter 140

 Expectation Values and Wave Functions

We have also seen that the wave functions and associated probabilities
can be used to compute expectation values, such as that we considered
when computing (∆x)2 ≡ h(x − hxi)2i. In general, any such average is
an average of some dynamical property that can be measured, such as
position, momentum, energy, . . . . We always define
Z
hQi = Ψ∗(x, t)QΨ(x,
b t)dx. (214)
all x

The symbol Q b stands for the operator associated with the observable Q;
for each observable there is a unique operator. For position it is simply
x. But, in many cases Q b is a differential operator. Thus, its location
above is absolutely critical.

We have already seen two examples of derivative operators when we

constructed the Schroedinger Equations:

h̄ ∂ b = ih̄ ∂ .
pb = , E (215)
i ∂x ∂t
J. Gunion 9D, Spring Quarter 141
 We will not go through it here, but it can be shown that (dropping the
time dependent part of the wave function)

h̄ ∂
Z Z
dxψ ∗(x) ψ(x) = dkψ(k)h̄k
e ψ(k)
e . (216)
i ∂x

This means that pb = h̄i ∂x

∂
is the appropriate x-space representation for
the momentum operator h̄k, where the latter is the obvious choice in the
k-space integral form on the rhs above. Another way of deciding that
the lhs above is a correct way of computing hpi is to write

dhxi d
Z
hpi = m =m Ψ∗(x, t)xΨ(x, t)dx (217)
dt dt
dΨ dΨ∗ ∂ 2Ψ ∂ 2 Ψ∗
then use the SE to evaluate and
dt dt
in terms of ∂x2
and ∂x2
, and
then do some partial integrations.
p2
Let us use pb to compute hpi and h 2m i
for the n = 1 ground state of the
infinite well.
h̄ ∂
Z
∗
hpi = ψ (x) ψ(x)dx
i ∂x
J. Gunion 9D, Spring Quarter 142
 r ! r !
L
2 πx h̄ ∂ 2 πx
Z
= sin sin dx
0 L L i ∂x L L
r ! r !
L
2 πx h̄ π 2 πx
Z
= sin cos dx
0 L i LL L L
Z L
h̄ 2π πx
πx
= 2
sin cos dx = 0 (218)
iL 0 L L

which is the answer we expected since the particle is not going anywhere.
Classically, it is just bouncing back and forth. Quantum mechanically,
the particle is a stationary wave form in a fixed box.
Now let us compute

∂2
Z
2 ∗ 2
hp i = ψ (x)(−h̄ ) 2 ψ(x)
∂x
! !
Z L r 2
r
2 πx −π 2 πx
= sin (−h̄2) 2 sin dx
0 L L L L L
!2
2 2Z L
r
π h̄ 2 πx
= 2
sin
L 0 L L

J. Gunion 9D, Spring Quarter 143

 π 2h̄2
= × 1, (219)
L2

where the 1 just comes from the fact that the remaining integral is just
the total probability integral.
First, let us note that this result implies that

p2 h̄2π 2
hEi = h i= = E1 (220)
2m 2mL2
as we expect.
Next, we note that since hpi = 0, we have

πh̄
q
2 2
∆p = hp i − hpi = . (221)
L

With some work, we could compute

s    2
1 1 1
q
2
∆x = hx2i − hxi = L2 − − L = 0.181L (222)
3 2π 2 2

J. Gunion 9D, Spring Quarter 144

 Combining, we find

πh̄
∆x∆p = 0.181L = 0.568h̄ , (223)
L
which is pretty close to the smallest (Gaussian wave function) result. Of
course, as n increases, ∆x∆p increases, with

L
lim ∆x = √ . (224)
n→∞ 12

This latter limit is the same as the classical limit, in which we would
compute hx2i using the uniform probability P = L1

L
1 1
Z
2
hx i = x2 dx = L2 ,
Z0 L L 3
1 L
hxi = x dx =
0 L r 2
1 1 L
⇒ ∆x = L − =√ . (225)
3 4 12

J. Gunion 9D, Spring Quarter 145

 Eigenvalues and eigenstates
∂
This discussion of the operator pb = h̄i ∂x allows us to introduce the
concept of an eigenstate and an eigenvalue.

In general, we say that some wavefunction Ψ is an eigenstate of an

operator Q
b if
QΨ
b = qΨ , (226)
where q is just some constant number, called the eigenvalue. Using our
infinite square well wave functions, we can illustrate using the operator
p.
b r r
h̄ ∂ 2 nπx h̄ nπ 2 nπx
pψ
b n(x) = sin = cos . (227)
i ∂x L L i L L L
The change of the function form means these wave functions are not
eigenstates of p.
b What about pc2 = pbp.
b Using the result above, we have

 " r #
h̄ ∂ h̄ nπ 2 nπx
pbpψ
b n(x) = cos
i ∂xi L L L
 2 r
nπ 2 nπx
= (−h̄2) sin
L l L

J. Gunion 9D, Spring Quarter 146

 h̄2n2π 2
= − ψn(x) (228)
L2

which is a constant times the original wavefunction. Thus, ψn is an

2 h̄2 n2 π 2
eigenstate of p with eigenvalue − L2 .
c

b 2i =
b then hQ
Note the general result that if Ψ is an eigenstate of Q,
2
2
q = hQi implying that
b

2 2 2
(∆Q) = hQ i − hQi = 0 .
b b (229)

Similar manipulations apply for the energy operator. For stationary states,
energy is always an eigenvalue. Our time dependence is Ψ(x, t) =
ψ(x)φ(t) = ψ(x)e−iωt for which
∂
EΨ(x, t) = ih̄ Ψ(x, t) = h̄ωΨ(x, t) ,
b (230)
∂t
in agreement with the standard energy-frequency relationship. In
contrast, pb 2 does not always give a definite eigenvalue when operating
on a stationary state. It just happens to for this infinite well case, where
U = 0 is a constant inside the well.

J. Gunion 9D, Spring Quarter 147

 The Heisenberg Uncertainty and Operators

If a wave function, Ψ is such that both the momentum and the position
are well defined, in operator language this means that Ψ should be an
eigenstate of both the momentum and the position operators:

h̄ ∂Ψ
pΨ
b = = pΨ , x
bΨ = xΨ . (231)
i ∂x

In this case, it should not matter in which order we operate the momentum
and position operators on the wave function. That is, we would for
consistency have
 
h̄ ∂ h̄ ∂
pb
bxΨ = x
bpΨ
b , ⇒ [xΨ] = x Ψ ,. (232)
i ∂x i ∂x

But, by explicit calculation this is not true. Instead, we find

 
h̄ ∂ h̄ h̄ ∂
[xΨ] = Ψ+x Ψ (233)
i ∂x i i ∂x

J. Gunion 9D, Spring Quarter 148

 which we write in terms of what is called a commutator ([b
a, b
b] = ab− b
bb bb
a)
that applies whenever the particular combination of operators on a wave
function.
 
h̄ ∂ h̄
[p,
b xb] = ,x = . (234)
i ∂x i

This non-zero commutator implies that momentum and position cannot

be simultaneously perfectly well-known. In other words, it encodes the
Heisenberg Uncertainty Principle for position and momentum.

A similar computation gives

 
∂
[E, t] = ih̄ , t = ih̄
b b (235)
∂t

implying that energy and time cannot be both precisely known, since
a wave function Ψ(x, t) cannot simultaneously have precise E
b and tb
eigenvalues.

J. Gunion 9D, Spring Quarter 149

 The Finite Square Well

How does all this change if we make the well finite?

Fig. 6−15, p. 209

Figure 24: The finite square-well potential.

J. Gunion 9D, Spring Quarter 150

 We write

 Ae+αx , x<0
ψ(x) = C sin kx + D cos kx , 0 < x < L . (236)
Be−αx , x>L


1. In the region x < 0, the time-independent SE takes the form

d2ψ 2m
= 2 (U − E)ψ (237)
dx2 h̄

which, for E < U , is solved by

ψ(x) = eαx , with α = [2m(U − E)/h̄2]1/2 . (238)

The form e−αx is also a solution of the equation, but becomes infinite
if x → −∞. This choice would not allow us to normalize the wave
function — the net probability integral would diverge. Thus, we must
choose only the e+αx form for x < 0.
2. In the region x > L, we have exactly the same SE, but now we must
only have the e−αx solution which decays exponentially for x → +∞.

J. Gunion 9D, Spring Quarter 151

 3. In the 0 < x < L region, we have the same SE as for the infinite well
with the same possible sin kx and cos kx solutions with h̄2k2/(2m) =
E. Note that E must be the same for all these different regions.
So, we now impose our continuity and smoothness requirements.
1. Continuity of ψ at x = 0 ⇒ A = D.
2. Continuity of dψ
dx
at x = 0 ⇒ αA = kC.
3. If we divide the 2nd result above by the first, A is eliminated and we
find
C α
= . (239)
D k
4. Continuity of ψ at x = L ⇒ C sin kL + D cos kL = Be−αL.
5. Continuity of the derivative at x = L ⇒ kC cos kL − kD sin kL =
−αBe−αL.
6. Dividing the equation just above by the preceding equation eliminates
B and we may also replace C/D by α/k to obtain

(α/k) cos kL − sin kL α

=− , (240)
(α/k) sin kL + cos kL k

which we rewrite as
α2
 
2α
tan kL 1 − = . (241)
k2 k
J. Gunion 9D, Spring Quarter 152
 For specified U and L, this equation can only be solved for very special
values of E (which is contained in both α and k).
Given a value of E that solves eq. (241), let us examine the behavior of
ψ. The crucial point is that ψ “leaks” into the x < 0 and x > L regions
where classically the particle cannot go. The distance, δ, from one of the
side boundaries of the well at which ψ declines to 1/e of its value at the
boundary is called the penetration depth. That is, δ is defined by

e−αδ = e−1 , (242)

implying
1 h̄
δ= =p . (243)
α 2m(U − E)
Note that δ gets increases as E increases in magnitude towards U .

Example
An electron is in a potential well with L = 0.200 nm and U = 100 eV .
Find the possible values of E for which the electron is bound to the well.
We are looking for values of E < U that solve eq. (241). The left and
right hand sides of eq. (241) are plotted below.

J. Gunion 9D, Spring Quarter 153

 10
7.5
5
2.5

20 40 60 80 100
-2.5
-5
-7.5
-10

Figure 25: Plot of left (red) and right (blue) sides of eq. (241). The
horizontal axis is E in eV .

Solutions correspond to points of intersection between the blue curve

and the (non-vertical) branches of the red curve. The corresponding

J. Gunion 9D, Spring Quarter 154

 values of the energy are E = 6.555, 25.900, 56.735, 93.833 eV . (We
note that the iterative technique outlined in the book does not converge
to the precise lowest energy value given above.)
Thus, there are only 4 possible bound state energy levels. For the largest
E value, we have

h̄ (197.3 eV · nm/c)
δ = p =p
2m(U − E) 2(511 × 103 eV /c2)(100 eV − 93.833 eV )
= 0.07859 nm , (244)

a rather substantial fraction of the basic well width of 0.2 nm.

Once we have an allowed E value, we would get the shape of the wave
function by setting A = 1 (temporarily —R eventually we would determine
∞
the magnitude of A that would give us −∞ |ψ(x)|2 = 1). Condition 1
then gives D = A = 1. Condition 2 then gives C = (α/k)D = α/k
(both of which are known once E is determined). Condition 4 then
gives B = eαL [(α/k) sin kL + cos kL]. (Recall that k, L, and α are all
known.) The wave function shape is now determined. We would then
integrate the square of this shape to determine the appropriate value of
A2 .

J. Gunion 9D, Spring Quarter 155

Figure 26: The finite square-well potential wavefunctions. Note the
continuity of the wave function and its leakage into classically forbidden
regions.

We should now ask what happens if E > U . Classically, this is a situation

in which the particle can escape the potential well. In the QM approach,

J. Gunion 9D, Spring Quarter 156

 this should be, and is, apparent. For example, for x < 0, if we solve

d2ψ 2m
= 2 (U − E)ψ , (245)
dx2 h̄

with E > U , the solutions are now of the form eilx or e−ilx (it will
prove more convenient to use these exponential forms than the equivalent
A sin lx and B cos lx forms), with l2 = 2h̄m
2 (E − U ). Similar results apply
to x > L. In both regions we have oscillatory behavior and there is no
damping for |x| → ∞. We will have more to say about related situations
in the next chapter.

The Harmonic Oscillator

This is a very important example. The restoring force for a harmonic

oscillator is F = −kx, corresponding to U = 12 kx2.
Not only is this one of the few cases for which an exact solution can be
obtained, but this particular case has repeated applications in all kinds of
contexts. In particular, anytime a potential U (x) has a local minimum,
for small excursions relative to this minimum at point x = a, one can

J. Gunion 9D, Spring Quarter 157

 approximate the shape of U (x) using the harmonic form:

d2U 

1
U (x) = U (a) + k(x − a)2 , where k= (246)
2 dx2  x=a

We can then agree to redefine our energy scale so that we reference to

U (a) and redefine our coordinate axis so that the minimum is located at
x = 0.
Classically, for U (x) = 12 kx2, the particle would oscillate about x = 0
p
with an angular frequency ω = k/m and with maximum amplitude,
or excursion in x from x = 0, set by the initial conditions. Let the
maximum excursion correspond to x = ±A. The total energy would be
conserved and would be E = 12 kA2. We could take A to be as small as
we like, implying that there should be no minimum for E.
So, let us see what happens in QM. The time-independent SE takes the
form
d2ψ 2m 1 2 2
2
= 2 ( mω x − E)ψ(x) . (247)
dx h̄ 2
The general solutions of this equation can be obtained in closed form,
but the techniques go beyond what we wish to cover in this course. We

J. Gunion 9D, Spring Quarter 158

 focus on obtaining the solution that corresponds to the lowest possible
energy, i.e. the ground state.

1. We have seen that a typical ground state has no “nodes” where ψ

vanishes.
2. Also, the gs wavefunction typically is symmetric for a symmetric
potential, and so ψ = f (x2).
3. Of course, for a confined solution, ψ should vanish as |x| → ∞.

Let us try
2
ψ(x) = C0e−αx . (248)
For this form

d2ψ d h −αx2
i 2
= C0(−2αx)e = C0[−2α + 4α2x2]e−αx (249)
dx2 dx

which has the same structure as the other side of the SE equation,
 
2m 1 2
2 mω 2x2 − E C0e−αx , (250)
h̄ 2

J. Gunion 9D, Spring Quarter 159

 provided

2m 1 mω
4α = 2 ( mω 2) ,
2
or α=
h̄ 2 2h̄
2mE mω 1
2 = 2α = , or E = h̄ω . (251)
h̄ h̄ 2

In short, we have

1 − mωx
2
E0 = h̄ω , ψ0(x) = C0e 2h̄ , (252)
2

where C0 is to be determined by normalizating the integrated probability

to 1. ψ0 and its square are depicted below.

J. Gunion 9D, Spring Quarter 160

 Fig. 6−18, p. 214

Figure 27: Ground state wavefunction and probability density.

The most important point about the ground state is that E0 6= 0. That
is, there is a minimum energy for the QM oscillator state. This energy
is sometimes called the zero-point energy. It has important implications.
It means for example that a crystal lattice of ions, each of which can be

J. Gunion 9D, Spring Quarter 161

 thought of as being trapped relative to its presumed fixed location by a
harmonic oscillator type potential, is actually “humming” with the energy
of this huge number of harmonic oscillators in their ground states.
Should we have expected such a minimum energy? The answer is that
the HUP requires it. Suppose we say that the electron is confined to
|x| ≤ A. Then, the HUP requires minimum px ∼ h̄/A. This will imply
an excursion in x such

p2x h̄2 1 h̄2

K∼ ∼ ∼ mω 2x2max , ⇒ |xmax|2 ∼ . (253)
2m 2mA2 2 m2 ω 2 A2

Requiring, |xmax|2 < 2

∼ A (for consistency with our input assumption)
leads to

2 h̄2 2 2 h̄
|xmax| ∼ <
∼A , ⇒ A >
∼ mω , (254)
m2 ω 2 A2

which corresponds to α ∼ mω/h̄ (as found) and minimum energy

 
1 2 h̄ 1
E0 >
∼2 mω = h̄ω . (255)
mω 2

J. Gunion 9D, Spring Quarter 162

 We can continue this guessing game to the first excited level fairly easily.
ψ1(x) should have just one node, and be antisymmetric, suggesting
ψ1(x) ∝ x exp(−αx2). Substituting into the SE we find this form works
if α is the same as before (α = mω/(2h̄)) and E = 32 h̄ω.

The general result is

 
1
En = n+ h̄ω . (256)
2

This is depicted in the following figure.

The important aspect of this prediction is the equal spacing of the energy
levels. For this system, if the state is in a n 6= 0 state, when it decays to
its next lowest state, the frequency of the radiation will always be given
by hf = h̄ω, no matter what the starting n value.

J. Gunion 9D, Spring Quarter 163

 Fig. 6−19, p. 215

Figure 28: Harmonic oscillator quantum energy levels.

Plots of the probability densities associated with these energy levels

are given in the next figure, where the dashed lines show the classical
probabilities for the same energy values. Note how the average classical
and quantum distributions get closer as n gets large.

J. Gunion 9D, Spring Quarter 164

 Fig. 6−20, p. 216

Figure 29: Harmonic oscillator probability densities, compared to classical

probabilities (dashed lines).
J. Gunion 9D, Spring Quarter 165
 Tunneling Phenomena

The square barrier

The classic example of a tunneling situation is the square barrier, depicted

in Fig. 30.

Figure 30: The square barrier.

J. Gunion Fig.
9D,7−1b, p.232
Spring Quarter 166
 Imagine a particle incident on the barrier from the left. Classically, if
E > U , then the particle will temporarily have reduced velocity as it
passes the barrier (there is a force to slow it down at the left-hand edge
of the barrier), but it will regain its initial velocity once it has passed the
barrier (there is a force to speed it up at the far right-hand edge of the
barrier).
If E < U , the particle cannot penetrate the barrier classically and the
particle will simply bounce off the barrier and reverse direction.
In QM, all regions are accessible to the particle even if E < U . The QM
penetration of the barrier is called tunneling.
To explore this quantitatively, we must set up the situation and impose
appropriate boundary condition requirements at the edges of the barrier.
We will imagine:
1. A wave is coming in from the left.
2. It hits the barrier.
3. Some of the wave is reflected, but some penetrates into the barrier.
4. The penetrating wave can make it over to the right-hand side of the
barrier with some reduced amplitude.
5. There is thus some wave moving to the right in the region past the

J. Gunion 9D, Spring Quarter 167

 barrier.
This is all depected in the figure below.

Fig. 7−2a, p.233

Figure 31: Transmission and reflection for the square barrier.

The above set-up is realized using the plane wave solutions of the SE as

J. Gunion 9D, Spring Quarter 168

 follows. See the figure below.

Fig. 7−2b, p.233

Figure 32: Transmission and reflection: plane wave set up

First, we must recall that a plane wave with perfectly defined momentum
moving to the right is proportional to ei(kx−ωt), where h̄k is the
momentum and E = h̄ω = h̄2k2/(2m) is the energy.

J. Gunion 9D, Spring Quarter 169

 Similarly, a plane wave moving to the left is proportional to ei(−kx−ωt).
Thus, on the left side of the barrier, where we have both incident and
reflected waves, we write

Ψ(x, t) = Aei(kx−ωt) + Bei(−kx−ωt) , x < 0. (257)

If reflection were complete, then |B| = |A|. In general, reflection is not

complete and we define the reflection coefficient by

(Ψ∗Ψ)ref lected B ∗B |B|2

R= = = . (258)
(Ψ∗Ψ) incident A∗ A |A|2

To the right side of the barrier, the physical setup envisioned means that
we should allow only the wave moving to the right:

Ψ(x, t) = F ei(kx−ωt) , x > L. (259)

The corresponding transmission coefficient is then given by

(Ψ∗Ψ)transmitted F ∗F |F |2
T = = = . (260)
(Ψ∗Ψ) incident A∗ A |A|2

J. Gunion 9D, Spring Quarter 170

 Since a given particle is either transmitted or reflected, these probabilities
should sum to unity:
R+T =1 (261)
is as sum rule that should emerge from the mathematics.

When the wave is in the barrier region itself, the SE takes a different
form.
d2ψ 2m(U − E) 2
2
= 2 ψ(x) ≡ α ψ(x) , (262)
dx h̄
where we are considering the situation with U > E so that the α2 on
the rhs above is positive (as opposed to the negative −2h̄mE2 that applies
outside the barrier region). The solutions of this equation are exponential
decay or increase. We should allow for both by writing

Ψ(x, t) = ψ(x)e−iωt = Ce−αx−iωt + De+αx−iωt 0 ≤ x ≤ L . (263)

Setting up the mathematics: boundary conditions

∂Ψ
We employ continuity of Ψ and of ∂x
at the right and left sides of the

J. Gunion 9D, Spring Quarter 171

 barrier to obtain:

A+B = C +D, (continuity of Ψ at x = 0)

ikA − ikB = αD − αC , (continuity of ∂∂x
Ψ
at x = 0)
Ce−αL + De+αL = F eikL , (continuity of Ψ at x = L)
(αL)e+αL − (αC)e−αL = ikF eikL , (continuity of ∂∂x
Ψ
at x = L)
(264)
These equations can be solved. It takes a number of pages. Perhaps I
will include the algebra in these notes at a later time. What one finds is
that
1
T (E) = h i , E<U. (265)
1 U2 2
1 + 4 E (U −E ) sinh αL
And, we can check that the sum rule R + T = 1 is true. In the above,
sinh x = (ex − e−x)/2. Note that for this case of E < U we find
T (E) < 1 always. If E > U , the wave behavior in the barrier region
is oscillatory, and there will generally be a series of E values for which
T (E) = 1. These are called transmission resonances.

Example
Two wires are separated by an insulating layer. Modeling the latter as a

J. Gunion 9D, Spring Quarter 172

 square barrier of height 10 eV , estimate the transmission coefficient for
penetration by 7 eV electrons: (a) for L = 5 nm and (b) for L = 1 nm.
We first must compute α.
p p
2m(U − E) 2(511 × 103 eV /c2)(3.00 eV ) ◦ −1
α= = ◦ = 0.8875 A .
h̄ 1.973 × 103 eV · A/c
(266)
The transimission coefficient is then
1
T = . (267)
h 2
i ◦ −1
1+ 1
4
10
7× 3
sinh2(0.8875 A L)

◦
For L = 50 A (i.e. 5 nm), we get T = 0.963 × 10−38, a very small
number.
◦
For L = 10 A, we get T = 0.657 × 10−7. Changing the barrier thickness
by just a factor of 5 has a huge effect.
Now suppose that we have the above situation and that a 1 mA current
of electrons is incident on the insulating barrier layer. How much of

J. Gunion 9D, Spring Quarter 173

 this current passes through the layer to the adjacent wire if the electron
energies are 7 eV and the layer thickness is 1 nm.

Answer: Each of the e−’s in the current has the probability given by
T = 0.657 × 10−7 to pass through the insulating layer. The cumulative
effect will be a transmitted current of

T ×1 mA = 0.657×10−7×1 mA = 0.657×10−7 mA = 6.57×10−11 A .

(268)

Heisenberg Uncertainty Intuition

Another way to understand the tunneling phenomenon is to focus on the

most important component of T (E), namely the exponential suppression
factor for large L from the sinh function:
√
T (E) ∼ e−2αL × less important factors ∼ e−2L 2m(U −E )/h̄
∼ e−2L/δ .
(269)
Here, δ has the same meaning as in the finite well bound state situation;
it is the length scale describing penetration of the wave into a forbiddne

J. Gunion 9D, Spring Quarter 174

 region. (The factor of 2 in the exponent is because T is the probability
rather than the amplitude.)

If we regard δ as a kind of ∆x over which we try to restrict the wave,

the HUP says ∆p ∼ h̄/δ, leading to

∆p2 h̄2 h̄2

∆K = = =   = U −E. (270)
2m 2mδ 2 h̄2
2m 2m(U −E )

This is just the right amount of kinetic energy to get the particle over
the U − E barrier; if δ = ∆x were any bigger then ⇒ not enough K to
get over barrier.

Example: large αL limit

It is possible to solve the boundary condition equations fairly easily in the

J. Gunion 9D, Spring Quarter 175

 case where e+αL is a big number. I repeat those equations here.

A+B = C +D, (continuity

of Ψ at x = 0)
ikA − ikB = αD − αC , of ∂∂x
Ψ
(continuity at x = 0)
Ce−αL + De+αL = F eikL , (continuity
of Ψ at x = L)
(αL)e+αL − (αC)e−αL = ikF eikL , of ∂∂x
Ψ
(continuity at x = L)
(271)
If e+αL is large, then D (but not DeαL) must be small (from 3rd b.c.
equation). The first two equations are then easily solved. Setting A = 1
(A normalization just sets overall flux), we find very easily (multiply 1st
equation by ik and add to 2nd equation to eliminate B terms) that

2ik 2
C= = α . (272)
ik − α 1− ik

If we define C 0 ≡ Ce−αL and D 0 ≡ De+αL, then the 3rd of the b.c.

1
equations (271) + ik × 4th b.c. equation yields

α
1 +1
C 0 + D0 = (−αC 0 + αD 0) , ⇒ D0 = C 0 ik
α
ik ik
−1

J. Gunion 9D, Spring Quarter 176

 ⇒ ! !
2α 2α
0 0 0 ik
2 −αL ik
C +D = C α = α e α (273)
ik
−1 1− ik ik
−1

Armed with this result, we can easily compute (see 3rd b.c. equation)

4α 2 2


0 0 2 −2αL 4kδ
2
|F | = |C + D | =  k
 e = e−2L/δ . (274)
α 2 2 1 + (kδ)2
1+ k2

Of course, the thing that is a bit tricky for you is all the use of absolute
squares of complex numbers. I go through this example so that you can
see that this complex stuff is essential for actual computations in QM.
In any case, you can now see that my earlier approximation of keeping
only the exponential factor is only a rough approximation. However, for
large L/δ, and (kδ)2 = E/(U − E) ∼ O(1), dropping the multiplicative
factor is a small perturbation on the very small exponential, thereby
justifying the use of
√
T (E) ∼ e−2L/δ = e−2L 2m(U −E )/h̄
. (275)

J. Gunion 9D, Spring Quarter 177

 Scanning Tunneling Microscope

The idea is simple. We want to examine the surface structure of a metal.

Inside the metal is an e− with some wave behavior inside metal.

There is a barrier that prevents e− from escaping the metal. This is

characterized by the U − K = eφ work function that we have discussed.

However, the real view is that the e− wave function actually penetrates
for a distance of order δ into the region outside the metal.

h̄
The e− has K = E inside the metal, so δ = √ .
2m(U −E )

The STM uses a probe to look at the exponential decay in this region
outside the metal.

This is depicted in the following figure.

J. Gunion 9D, Spring Quarter 178

 U=e φ + K

exp[−x/δ ]

probe

Figure 33: Scanning Tunneling Miscroscope setup. φ=work function. Ψ=

wave function.

Typical numbers would be U = 9 eV and E = 5 eV , ⇒ U − E = 4 eV .

J. Gunion 9D, Spring Quarter 179

 Then
h̄ h̄c
δ = p = p
2m(U − E) 2mc2(U − E)
◦
1.973 keV · A ◦
= p ' 1A . (276)
− 3
2(511 keV )(4 × 10 keV )

A “bias” voltage V would be introduced between the metal surface and

the probe to ensure that the probe is probing the metal and not the
reverse. The tunneling current would be given by

e2V
i= e−2L/δ , (277)
4π 2Lδh̄
and if L changes by even a small amount (i.e. if there is variation in the
surface defined by the e− wave functions), we will see a change in i.
◦
In the above example, if L → L + 0.01A then i changes by a fractional
◦
−0.01×2/1 A
amount of e ' 0.98. Such a 2% change in i is measurable
using appropriate amplification techniques. Thus, one can have sensitivity
◦
to surface details at the 0.01A level.

J. Gunion 9D, Spring Quarter 180

 Varying U (x)

In most physical cases, U is not actually a constant, but rather has some
non-trivial x dependence. We will do the example of α decay shortly.
The generalization of our approximate formula for T (E) that applies in
this case is
" √ #
2 2m
Z p
T (E) ∼ exp − U (x) − E dx . (278)
h̄ barrier

This reduces to our previous expression if the barrier region has length L
and U (x) = U is a constant in that region.

α decay simplified

Many nuclei heavier than lead naturally emit an α particle, but emission
rates vary by factors of 1013, whereas the energies of the α’s range only
from 4 to 8 M eV . Why?

J. Gunion 9D, Spring Quarter 181

 The basic picture is below.

Fig. 7−9, p.245

Figure 34: Picture of α decay.

J. Gunion 9D, Spring Quarter 182

 Consider the α-particle emission of 238U (nuclear charge of 92) which
turns into an unstable isotope of T h by emitting a 4.2 M eV α particle,
i.e. Eescape = 4.2 M eV .
We know that Rnucleus ∼ 7 × 10−15 m. This means that there is a
barrier height at R coming from the Coulomb attraction that is of order
(Z is the charge of the daughter nucleus in what follows))

2kZe2
Vcoulomb(R) =
R
2(90)(1.6 × 10−19 C)2(9 × 109 N · m2/C 2)
=
7 × 10−15 m
10−6 M eV
×
1.6 × 10−19 J
= 37 M eV . (279)

The α must tunnel through to radius

2kZe2 2(9 × 109 N · m2/C 2)(90)(1.6 × 10−9 C)2

R1 = =
E 4.2 M eV
37 M eV × R
= = 6.2 × 10−14 m = 62 f m . (280)
4.2 M eV
J. Gunion 9D, Spring Quarter 183
 Very crudely (see book for more precise formulae), assume

37 M eV
average potential height = ∼ 18 M eV ,
2
R1 − R (62 − 7) f m
average barrier width = = ∼ 28 f m .
2 2
(281)

Then,

1
T (E) = h i
U2
1 + 14 E (U −E ) sinh2 αL
 
4.2 M eV 4.2 M eV  −14

∼ 16 1− exp −2(2.8 × 10 m)/δ ,
18 M eV 18 M eV
(282)

where

h̄ 6.58 × 10−22 M eV · s
δ = p =p
2mα(V − E) 2(3727 M eV /c2)(18 − 4.2) M eV

J. Gunion 9D, Spring Quarter 184

 = 6 × 10−16 m = 0.6 f m . (283)

This gives

e−2·28/0.6 = 2.9 × 10−41 , ⇒ T = 1.6 × 10−41 . (284)

So, you might say how can it ever escape?

The point is that the above is the probability for tunneling in a single
collision of the α particle with the barrier. We must ask how many
collisions there are per second.
Inside the nucleus, the e− kinetic energy is roughly
1
K = mv 2 , ⇒
2
r s
2K 2(4.2 M eV ) 7
v = = 2
= 0.047 c = 1.4 × 10 m/s(285)
,
m 3727 M eV /c

where we should note that the output v implies that the NR approximation
is ok.

J. Gunion 9D, Spring Quarter 185

 Now, the nuclear diameter is 2R ∼ 1.4 × 10−14 m, so the α crosses the
nucleus in a time of about
1.4 × 10−14 m
tcrossing ∼ = 10−21 s . (286)
1.4 × 107 m/s
Now the lifetime, τ of the nucleus in the presence of α decay is given by
the point at which (# of times = number of collisions of the α particle
with the barrier for which the probability of one penetration becomes
equal to unity)

T (E) × # of times = 1 , (287)

where
τ
# of times = . (288)
tcrossing
Plugging this into the equation just above gives

T (E)τ 10−21 s
= 1, ⇒ τ = ∼ 1020 s . (289)
tcrossing 10−41

The actual 238U lifetime is τ = 4.5 × 109 yr ∼ 1017 s. Thus, our first
estimate, while very crude, is in the right ballpark.

J. Gunion 9D, Spring Quarter 186

 To get a more precise result (in better agreement with experiment), one
must actually carry out the integral

" √ #
2 2m
Z p
T (E) ∼ exp − U (x) − E dx . (290)
h̄ barrier

using the potential drawn. This gives (to a good approximation)

 r s 
E0 ZR
T (E) = exp −4πZ +8 , (291)
E r0

h̄2
where r0 = is a kind of a Z = 1 Bohr radius for the α particle:
mα ke2
r0 ∼ a0/7295 ∼ 7.25 f m since mα ∼ 7295me, and the associated
Z = 1 kinetic energy is

ke2 ke2 a0
E0 = = = 13.6 eV × 7295 = 0.0993 M eV . (292)
2r0 2a0 r0

J. Gunion 9D, Spring Quarter 187

 The more precise formulation of the lifetime is to use

ln 2
τhalf life ≡ τ1/2 = (293)
fcrossing T (E)

where fcrossing = 1/tcrossing = 1021 Hz.

Example
As an example of using the precise formula, consider T h which ejects
an α particle with energy E = 4.05 M eV . The nuclear radius is
R = 9.00 f m.
The daughter atomic number is Z = 88. We compute
" r r #
0.0993 9.00
T (E) = exp −4π(88) + 8 88
4.05 7.25
= exp[−89.542] = 1.3 × 10−39 . (294)

Taking the collision frequency to be fcrossing = 1021 (as is more or less

J. Gunion 9D, Spring Quarter 188

 applicable for all these heavy nuclei) then we find

fcrossing T (E) = 1.3 × 10−18/s (295)

and

0.693 0.693
τ1/2 = = = 5.4 × 1017 s = 1.7 × 1010 yr .
fcrossing T (E) 1.3 × 10−18
(296)
which compares favorably with the actual value of 1.4 × 1010 yr.

J. Gunion 9D, Spring Quarter 189