Y. D. Chong Ch.

2: Resonances | Graduate Quantum Mechanics

Chapter 2: Resonances

Sagredo—Even as a boy, I observed that one man alone by giving

these impulses at the right instant was able to ring a bell so large
that when four, or even six, men seized the rope and tried to stop it
they were lifted from the ground, all of them together being unable
to counterbalance the momentum which a single man, by
properly-timed pulls, had given it.
Salviati —Your illustration [can also] explain the wonderful
phenomenon of the strings of the cittern or the spinet, namely, that
a vibrating string will set another string in motion and cause it to
sound... these vibrations cause the immediately surrounding air to
vibrate and quiver; then these ripples in the air expand far into
space and strike not only all the strings of the same instrument but
even those of neighboring instruments.

Galileo Galilei, Two New Sciences

2.1. BOUND STATES AND FREE STATES

A curious feature of wavefunctions in infinite space is that they come in two distinct
varieties: (i) bound states that are localized to one region, like the ground state of a
harmonic oscillator, and (ii) free states that extend over all space, like a plane wave.
Both kinds of states can co-exist in a single system. This is demonstrated by a simple
model called the 1D finite square well, whose Hamiltonian is
p̂2
Ĥ = − U Θ(a − |x̂|), (2.1)
2m
where x̂ and p̂ are 1D position and momentum operators, m is the particle mass, U and a
are positive real parameters, and Θ denotes the step function (i.e., 1 if the input is positive,
and 0 otherwise). As shown below, the potential well has depth U and width 2a.

We can solve the time-independent Schrödinger wave equation for this Hamiltonian using
a simple technique called the transfer matrix method. Some key aspects of the calculation
are described below; the rest of the details are given in Appendix B.

19
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

Outside the potential well, the Schrödinger wave equation reduces to

ℏ2 d2 ψ
− = Eψ(x) (for |x| > a). (2.2)
2m dx2
For E < 0, we can find discrete solutions of the form
r
2mE
ψ(x) = c± e∓κx , where c± ∈ C, κ = − ∈ R+ . (2.3)
ℏ2
By choosing e−κx for x > a, and e+κx for x < −a, the wavefunction vanishes exponentially
away from the potential well, so it is normalizable to unity. These solutions are called
bound states. There is also a lower energy bound, E ≥ −U , stemming from the variational
principle. Hence, the bound states form a discrete set of eigenenergies within the range

−U ≤ E < 0. (2.4)

For E > 0, we cannot construct exponentially localized solutions. We instead consider


ikx −ikx
α− e + β− e
 , x < −a r
2mE
ψ(x) = (something), −a < x < a where k = 2
∈ R+ . (2.5)
α eikx + β e−ikx ,
 ℏ
+ + x > a,
R∞
Such a solution is called a free state. It cannot be normalized to unity since −∞ |ψ(x)|2 dx
diverges. However, it is “almost normalizable” in the sense that |ψ|2 does not blow up as
|x| → ±∞. We can thus treat it as a valid quantum wavefunction, on the same footing as
plane waves (the wavefunctions for momentum eigenstates).
For each E, we can find coefficients α± and β± for Eq. (2.5) so as to satisfy the Schrödinger
wave equation. It turns out that these four coefficients are not independent; for example, if
we fix α± , that uniquely determines β± (see Appendix B). As this can be done for all E ≥ 0,
Eq. (2.5) represents a continuous family of free state solutions.
The following figure shows numerically obtained results for a square well with U = 30
and a = 1 (in units where ℏ = m = 1). The energy spectrum is shown on the left side.
There are five discrete bound states; their plots of |ψ|2 versus x are shown on the right side.
These results were computed using the transfer matrix method described in Appendix B.

20
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

The number of bound states is not fixed. In the figure below, we vary the potential
minimum −U (with fixed a = 1), and plot the bound state energies:

For U = 30, there are five bound states. As the well gets shallower, these disappear one by
one, with only one surviving as U → 0. (There is a theorem stating that any 1D attractive
potential, no matter how weak, supports at least one bound state; see Exercise 1.)
Many features of this simple model are generalizable to more complicated cases, including
higher spatial dimensions. One notable difference is that in 3D, it is possible for an attractive
potential to be too weak to support a bound state (i.e., its binding energy cannot overcome
the zero-point energy of a prospective ground state). The figure below shows an example
based on a uniform 3D spherically symmetric well (the l labels are angular momentum
quantum numbers). To learn more about this phenomenon, refer to Exercise 2.

2.2. QUASI-BOUND STATES AND RESONANCES

For the 1D finite square well, there is a clear distinction between bound and free states.
However, certain states, called “quasi-bound states”, can straddle the two cases. As we
shall see, they play a particularly important important role in scattering experiments.
The figure below shows an example of a potential function that hosts quasi-bound states.
In the exterior region |x| > b, the potential vanishes. Between x = −b and x = b, there is a

21
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

barrier of height Vb , in the middle of which is a well of depth U < Vb and width 2a. Since
V ≥ 0 everywhere, there are no bound states: all eigenstates must be free states.

There is something intriguing about the central well. Consider an alternative potential
(
Vb − U, |x| < a,
Valt (x) = (2.6)
Vb , otherwise,

which is a finite square well. This should have at least one bound state, whose energy lies in
the range Vb − U < E < Vb , and whose wavefunction ψ(x) decays exponentially away from
the well. Since Valt differs from the original potential V only in the region |x| > b, where
ψ(x) is small, this should also serve as an approximate solution for V !
In the context of the original scattering potential V , such a solution is called a quasi-
bound state. It acts like a bound state, but is not actually a bound state for V (since it
is not an exact eigenstate). In fact, as noted above, V lacks true bound states.
We can also analyze the situation using the scattering framework from Chapter 1. Con-
sider an incident particle of energy E > 0 whose wavefunction is

ψi (x) = Ψi eiki x . (2.7)

This produces a scattered wavefunction, which takes the following form in the exterior region:
(
f− e−iki x , x ≤ −b
ψs (x) = Ψi × (2.8)
f+ eiki x , x ≥ b.

We can obtain f± via the transfer matrix method (see Appendix B). The figure below shows
numerical results obtained for U = 20, Vb = 30, a = 1, and b ∈ {1.2, 1.4}, with ℏ = m = 1.

22
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

The vertical axis shows the transmittance |1 + f+ |2 , which is the probability for the
particle to pass through the scatterer. The horizontal axis is the particle energy E. Unsur-
prisingly, for E < Vb − U the transmittance approaches zero, and for E ≳ Vb it approaches
unity. For Vb − U < E ≲ Vb , the transmittance forms a series of narrow peaks; for larger b
(i.e., when the central well is more isolated from the exterior space), the peaks are narrower.
At the top of the figure, we have also plotted the bound state energies for the square well
potential Valt (x). These energies closely match the locations of the transmittance peaks!
Looking at the wavefunction reveals other interesting features. Below, we plot |ψ(x)|2
versus x at the energies of the first three transmittance peaks, along with the corresponding
quasi-bound state wavefunctions. At each peak, observe that: (i) |ψ(x)|2 is very large in the
potential region, and (ii) its shape is very similar to the corresponding quasi-bound state.

We interpret the situation as follows: at certain energies, an incident particle ends up

spending a long time trapped inside the scatterer, taking on many of the characteristics of
a bound state. But unlike a true bound state, it is not trapped forever. Eventually, it leaks
out of the scatterer and escapes to infinity.
The enhancement of |ψ|2 by the presence of a quasi-bound state is called resonance,
and it is closely analogous to the phenomenon of the same name in classical mechanics.
When a damped harmonic oscillator is subjected to a periodic driving force, it undergoes
steady-state oscillation at the driving frequency. If this matches the frequency of one of the
oscillator’s “normal modes”, the oscillation amplitude grows large, as Galileo noted in the
epigraph of this chapter.
In the quantum scattering experiment, the incident wavefunction plays the role of the
driving force, E acts as the driving frequency, and quasi-bound states act like the normal
modes. Oftentimes, scattering experiments are conducted for the express purpose of locating
and studying resonances. When a resonance is found, it can be used to deduce various
important information about the scatterer, as we shall see in the next few sections.

23
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

2.3. GREEN’S FUNCTION ANALYSIS OF SCATTERING RESONANCES

The Green’s function formalism from Chapter 1, Sec. 1.5, provides a powerful and general
way to understand quasi-bound states and resonances. Notably, this framework applies not
only to 1D models, but works for higher dimensions as well.
Let Ĥ = T̂ + V̂ be the Hamiltonian of a system supporting resonances, where T̂ is the
kinetic energy operator and V̂ is the potential operator. We decompose the potential into
V̂ = V̂0 + V̂1 , (2.9)

where V̂0 is a “confining potential” that supports a bound state, and V̂1 is a “deconfining
potential” that turns the bound state into a quasi-bound state. For example, the potential
functions for the 1D model of the previous section are shown below:

(Note that this 1D case is just an illustration; Eq. (2.9) applies to higher dimensions too.)
When the potential is just V̂0 , let there be a bound state |φ⟩ with energy E0 . Furthermore,
let us assume that the potential supports a continuum of free states {|ψk ⟩} with energies
{Ek }, where k is some continuous index for labeling the free states (we will discuss what
this index might represent later). The states satisfy the Schrödinger equation


T̂ + V̂0 |φ⟩ = E0 |φ⟩ (2.10)


T̂ + V̂0 |ψk ⟩ = Ek |ψk ⟩, (2.11)
along with the orthogonality and completeness relations
X
⟨φ|ψk ⟩ = 0, |φ⟩⟨φ| + ˆ
|ψk ⟩⟨ψk | = I. (2.12)
k
P
Here, k represents a sum over all free states. Since the free states form a continuum, this
ought to be expressed as an integral, similar to how we treat momentum eigenstates (see
Chapter 1). But, for the moment, we write it as a sum for ease of presentation.
Let us treat V̂ as a scattering potential, and compute the Green’s function. According
to Dyson’s equations, the Green’s function for the full system is
Ĝ = Ĝ0 + ĜV̂1 Ĝ0 , (2.13)
where  −1
Ĝ0 (E) = lim+ E − T̂ − V̂0 + iε (2.14)
ε→0

is the causal Green’s function. Note that Eq. (2.13) is exact, not an approximation. If we
can obtain Ĝ(E), we can determine the scattering amplitudes.

24
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

Let us find the matrix elements of Ĝ(E) by using |φ⟩ and {|ψk ⟩} as a convenient basis.
Note, however, that this is not the energy eigenbasis of Ĥ.
As usual when dealing with Dyson’s equations, we must beware of the fact that Ĝ appears
on both the left- and right-hand sides. We judiciously insert a resolution of the identity as
follows:
!
X
Ĝ = Ĝ0 + ĜIˆV̂1 Ĝ0 = Ĝ0 + Ĝ |φ⟩⟨φ| + |ψk ⟩⟨ψk | V̂1 Ĝ0 . (2.15)
k

We then compute the matrix element ⟨φ| · · · |φ⟩ for both sides:
X
⟨φ|Ĝ|φ⟩ = ⟨φ|Ĝ0 |φ⟩ + ⟨φ|Ĝ|φ⟩ ⟨φ|V̂1 Ĝ0 |φ⟩ + ⟨φ|Ĝ|ψk ⟩ ⟨ψk |V̂1 Ĝ0 |φ⟩
k
  X
⟨φ|Ĝ|φ⟩ 1 − ⟨φ|V̂1 Ĝ0 |φ⟩ = ⟨φ|Ĝ0 |φ⟩ + ⟨φ|Ĝ|ψk ⟩ ⟨ψk |V̂1 Ĝ0 |φ⟩
k
! !
⟨φ|V̂1 |φ⟩ 1 X
lim ⟨φ|Ĝ|φ⟩ 1 − = lim+ 1+ ⟨φ|Ĝ|ψk ⟩ ⟨ψk |V̂1 |φ⟩
ε→0+ E − E0 + iε ε→0 E − E0 + iε
k
  X
lim+ ⟨φ|Ĝ|φ⟩ E − E0 − ⟨φ|V̂1 |φ⟩ + iε − ⟨φ|Ĝ|ψk ⟩ ⟨ψk |V̂1 |φ⟩ = 1. (2.16)
ε→0
k

Similarly, computing the matrix element ⟨φ| · · · |ψk ⟩ gives

X
⟨φ|Ĝ|ψk ⟩ = ⟨φ|Ĝ0 |ψk ⟩ + ⟨φ|Ĝ|φ⟩ ⟨φ|V̂1 Ĝ0 |ψk ⟩ + ⟨φ|Ĝ|ψk′ ⟩ ⟨ψk′ |V̂1 Ĝ0 |ψk ⟩
k′
!
X
= lim+ (E − Ek + iε)−1 ⟨φ|Ĝ|φ⟩ ⟨φ|V̂1 |ψk ⟩ + ⟨φ|Ĝ|ψk′ ⟩ ⟨ψk′ |V̂1 |ψk ⟩ .
ε→0
k′

So far, the equations have been exact, but now we apply an approximation: in the last
line of the above equation, let the factor of ⟨φ|G|φ⟩ be large, so that the first term in the
sum is dominant. It will be shown below that ⟨φ|G|φ⟩ being large is precisely the resonance
condition, so this approximation will be self-consistent. Thus, we obtain
⟨φ|Ĝ|φ⟩ ⟨φ|V̂1 |ψk ⟩
⟨φ|Ĝ|ψk ⟩ ≈ lim+ . (2.17)
ε→0 E − Ek + iε
Combining this with Eq. (2.16) gives
" #
  X ⟨φ|Ĝ|φ⟩⟨φ|V̂ |ψ ⟩
1 k
lim ⟨φ|Ĝ|φ⟩ E − E0 − ⟨φ|V̂1 |φ⟩ + iε − ⟨ψk |V̂1 |φ⟩ ≈ 1.
ε→0+
k
E − Ek + iε
We finally arrive at the result

1
⟨φ| Ĝ(E) |φ⟩ ≈ (2.18)
E − E0 − ⟨φ|V1 |φ⟩ − Σ(E)
X |⟨ψk |V̂1 |φ⟩|2
where Σ(E) ≡ lim+ . (2.19)
ε→0
k
E − E k + iε

25
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

2.4. THE SELF ENERGY

The enigmatic Σ(E) appearing in the denominator in Eq. (2.18) is called the self energy.
If we put aside its dependence on E and treat it as a constant, Eq. (2.18) looks like the
Green’s function of a system with an energy eigenstate |φ⟩ of energy

E0 + ⟨φ|V̂1 |φ⟩ + Σ.

We can interpret this as the energy of the quasi-bound state. The first term is the energy
of the original bound state, the second term is a straightforward energy shift coming from
the deconfining potential V̂1 , and the third term is the self energy. According to Eq. (2.19),
Σ arises from the interplay between |φ⟩ and the free states {|ψk ⟩}. The quasi-bound state
disturbs the free states, which collectively exert a back-action that shifts its energy.
There is an important complication: Σ turns out to be complex, not real! From Eq. (2.19),
 
  X 2 1
Im Σ(E) = lim+ ⟨ψk |V̂1 |φ⟩ Im
ε→0
k
E − Ek + iε
  (2.20)
X 2 ε
=− ⟨ψk |V̂1 |φ⟩ lim 2 2 .
ε→0+ (E − Ek ) + ε
k

The quantity in the square brackets is a Lorentzian function of width ε. The area under the
curve is π, independent of ε, so we can use the limiting expression
ε
lim+ = πδ(x). (2.21)
ε→0 x2 + ε2
Therefore,
2
  X
Im Σ(E) = −π ⟨ψk |V̂1 |φ⟩ δ(E − Ek ) < 0. (2.22)
k

What does it mean for an energy to be complex? If a quantum state has energy E, its time
dependence goes as exp(−iEt/ℏ). For E = E + iΓ, the time evolution factor is

e−iEt/ℏ eΓt/ℏ .

Since Im[Σ] < 0, the quasi-bound state is decaying exponentially with time. This matches
our description of it as a localized state that is almost an energy eigenstate, but not exactly
one—so, after some time, it escapes the scatterer and disappears to infinity.

2.5. SCATTERING RESONANCES

In Chapter 1, Sec. 1.7, we derived the following relationship between the Green’s function
and the scattering amplitude f :

f (k → k′ ) ∝ ⟨k′ |V̂ |k⟩ + ⟨k′ |V̂ ĜV̂ |k⟩. (2.23)

Here, |k⟩ and |k′ ⟩ are incident and scattered plane wave states satisfying |k| = |k′ |. The
first term describes the lowest-order scattering process (the first Born approximation). The
second term contains second- and higher-order scattering processes.

26
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

By inserting resolutions of the identity between each V̂ and Ĝ operator in the second
term, we find that f contains a contribution of the form
⟨k′ |V̂ |φ⟩ ⟨φ|V̂ |k⟩
∆f (k → k′ ) ∝ ⟨k′ |V̂ |φ⟩⟨φ|Ĝ|φ⟩⟨φ|V̂ |k⟩ = , (2.24)
E − Eres − iIm[ Σ ]
where  
Eres ≡ E0 + ⟨φ|V̂1 |φ⟩ + Re Σ ∈ R. (2.25)
At resonance, the denominator is small so ∆f is the dominant part of f . It is worth noting
that ∆f contains contributions from all terms in the Born series, not just low-order terms.
The figure below shows the energy dependence of ∆f , according to Eq. (2.24):

The graph of |∆f |2 versus E forms a Lorentzian curve centered at Eres . The width of the
peak can be characterized by the full-width at half-maximum (FWHM), the spacing between
the two energies where |∆f |2 has half its maximum value, which can be calculated to be
δE (FWHM) = 2 Im[Σ] . (2.26)
(Note: we showed earlier that Im[Σ] < 0.) The smaller the decay rate, the sharper the peak.
The phase arg[∆f ] also contains useful information. As E crosses Eres from below, the
phase increases by π. The shift also occurs over an energy width of ∼ |Im[Σ]|.
These resonant features—peaks and/or phase shifts—are sought after in many kinds of
real-world scattering experiments. Often, they are overlaid on a “background” caused by
non-resonant processes. This can be seen, for example, in the plot below from the CMS
experiment at the Large Hadron Collider (LHC), declaring the experimental observation of
the Higgs boson in 2012.
Weighted Events / (1.67 GeV)

© 2012 CERN. From CMS Collaboration website:

http://cms.web.cern.ch/news/observation-new-particle-mass-125-gev

27
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

2.6. FERMI’S GOLDEN RULE

In the previous sections, we have seen that the width of a resonance is determined by
the decay rate of a quasi-bound state, which is given by the imaginary part of its self
energy. Now, we will derive a simple and powerful formula for approximating it in many
circumstances, called Fermi’s Golden Rule.
Suppose we initialize the particle, at time t = 0, to a quasi-bound state |φ⟩. As |φ⟩ is
not an eigenstate of the Hamiltonian, the particle does not remain in that state forever. To
quantify the “decay” of the quasi-bound state, we imagine performing a measurement that
has |φ⟩ as one of its eigenstates, at time t > 0. The probability that the particle is observed
to still be in state |φ⟩ is
  2
P (t) = ⟨φ| exp −iĤt/ℏ |φ⟩ . (2.27)
To help us calculate P (t), let us define
(  
⟨φ| exp −iĤt/ℏ |φ⟩ e−εt , t ≥ 0
f (t) = (2.28)
0, t < 0,

where ε ∈ R+ . For t ≥ 0 and ε → 0+ , we see that |f (t)|2 → P (t). The reason we deal
with f (t) is that it is more well-behaved than the actual amplitude ⟨φ| exp(−iĤt/ℏ)|φ⟩.
The function is designed so that (i) it vanishes at times prior to start of the experiment,
and (ii) it vanishes as t → ∞ due to the regulator ε. The latter condition is based on
our expectation that the bound state should decay permanently into the continuum of free
states, and should never be repopulated by waves “bouncing back” from infinity.
To find f (t), we first take its Fourier transform,
Z ∞ Z ∞
F (ω) = iωt
dt e f (t) = dt ei(ω+iε)t ⟨φ|e−iĤt/ℏ |φ⟩. (2.29)
−∞ 0

Now insert a resolution of the identity, Iˆ =

P
n |n⟩⟨n|, where {|n⟩} denotes the exact
eigenstates of Ĥ (for free states, the sum goes to an integral in the usual way):
Z ∞ X
F (ω) = dt ei(ω+iε)t ⟨φ|e−iĤt/ℏ |n⟩⟨n|φ⟩
0 n
Z ∞    
X En
= ⟨φ|n⟩ dt exp i ω − + iε t ⟨n|φ⟩
0 ℏ
n (2.30)
X i
= ⟨φ|n⟩ En
⟨n|φ⟩
n
ω− ℏ
+ iε
 −1
= iℏ ⟨φ| ℏω − Ĥ + iℏε |φ⟩.

In the third line, the regulator ε removes the contribution from the t → ∞ limit. Hence,
lim F (ω) = iℏ ⟨φ|Ĝ(ℏω)|φ⟩, (2.31)
ε→0+

where Ĝ is our old friend the causal Green’s function (see Chapter 1, Sec. 1.6). Its appearance
can be traced to Eq. (2.28), which defines f (t) to be nonzero only for t ≥ 0.

28
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

In Section 2.5, we saw that when the system is at or close to resonance,

1
⟨φ| Ĝ(E) |φ⟩ ≈ , (2.32)
E − Eres − iIm[Σ]

where Eres is the resonance energy, and Σ is the self energy of the quasi-bound state. Using
this, we can perform an inverse Fourier transform to retrieve f (t):
Z ∞
dω −iωt
lim+ f (t) = lim+ e F (ω)
ε→0 ε→0 −∞ 2π
Z ∞
i e−iωt
≈ dω (2.33)
2π −∞ ω − (Eres + iIm[Σ])/ℏ
   
iEres t |Im[Σ]|
≈ exp − exp − t ,
ℏ ℏ

where Im[Σ] is to be evaluated at E = Eres . Thus,

2 Im[Σ(Eres )]
P (t) = e−κt , where κ = . (2.34)
ℏ
The quasi-bound state decays exponentially, with a decay rate proportional to |Im[Σ]|.
We previously derived a formula for Im[Σ], Eq. (2.22), reproduced here for convenience:
2
  X
Im Σ(E) = −π ⟨ψk |V̂1 |φ⟩ δ(E − Ek ). (2.35)
k

In the sum over k, each term is a non-negative real number consisting of two factors: (i) a
k-dependent “weight” |⟨ψk |V̂1 |φ⟩|2 , and (ii) a delta function, which vanishes unless Ek = E.
Because only a subset of the free states are relevant to the sum, it is convenient to define
2
⟨ψk(E) |V̂1 |φ⟩ ,

which is the average weight over the participating free states (i.e., those with E = Ek ).
Then we can approximate Eq. (2.35) by pulling the weights out of the sum:
2
  X
Im Σ(E) ≈ −π ⟨ψk(E) |V̂1 |φ⟩ δ(E − Ek ). (2.36)
k

This allows us to write the decay rate in Eq. (2.34) as


2π  W = ⟨ψk |V̂1 |φ⟩
2
κ ≈ W ρ(Eres ), where ρ(E) =
X
δ(E − Ek ). (2.37)
ℏ 
k

This result is called Fermi’s golden rule. It states that the decay rate of a quasi-bound
state is determined by the product of two real factors:

29
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

• |W |2 , which is obtained by taking the absolute square of the transition amplitude

⟨ψk |V̂1 |φ⟩, and averaging over available free states (i.e., the free states meeting the
resonance condition Ek = Eres ). This characterizes how strongly the quasi-bound
state couples, on average, to the available free states.
• ρ(Eres ), which is the number of free states per unit energy at energy Eres . This says
that the decay rate is proportional to the number of available free states.
Typically, these two factors are estimated by making further approximations.
One very common approximation is to treat the free states as plane waves. In this case,
we take the k labels to be d-dimensional wave-vectors. As the wave-vectors are continuous,
we convert the sum over k into an integral in the usual way, by defining the discretization
step dk = 2π/L where L → ∞ (see Chapter 1, Sec. 1.2):
X
ρ(E) = δ(E − Ek )
k
dd k
Z
d
=L δ(E − Ek ). (2.38)
(2π)d
| {z }
≡ D(E)

The density of states D(E) counts the number of free states of energy E, per unit energy
and per unit volume. Note that it has different dimensions from ρ(E). We also need to
re-normalize the transition amplitude, via the usual delta-normalization convention:
 d/2
(new) L
|ψk ⟩ = |ψk ⟩(old) (2.39)
2π
 d/2
(old) 2π
⇒ ⟨ψk |V̂1 |φ⟩ = ⟨ψk |V̂1 |φ⟩(new) . (2.40)
L
Plugging Eq. (2.38) and (2.40) into (2.37), we can re-state Fermi’s golden rule as


 W = (2π)d/2 ⟨ψk |V̂1 |φ⟩
2π 2

κ ≈ W D(Eres ), where dd k (2.41)
Z
ℏ  D(E) = δ(E − Ek ).
(2π)d


2
Here, W is averaged over all plane waves satisfying the resonance condition Ek = E. Note
that the L factors in Eq. (2.38) and (2.40) do not appear in the result (2.41), as they have
cancelled each other out.

2.7. FERMI’S GOLDEN RULE IN A 1D MODEL

Let us apply Fermi’s golden rule to the simple 1D model of Section 2.2. The potential is

 V0 (x) = −U Θ(a − |x|)

V (x) = V0 (x) + V1 (x), where V1 (x) = Vb Θ(b − |x|) (2.42)

 0 < U < Vb .

30
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

The finite square well V0 (x) supports one or more bound states; for simplicity, we focus on
the ground state, whose energy is denoted by E0 (where −U < E0 < 0). Its wavefunction is
(
A cos(qx), |x| < a
φ(x) = (2.43)
B exp (−η|x|) , |x| ≥ a,
where A and B are constants to be determined, and
r r
2m 2m
q= 2
(E0 + U ), η = |E0 |. (2.44)
ℏ ℏ2
By matching φ(x) and dφ/dx across the x = a interface, we can derive E0 , A, and B. The
details are left as an exercise.
Having obtained φ(x), we want the transition amplitude ⟨ψk |V̂1 |φ⟩, where ψk (x) is a
representative free eigenstate of V0 (x) that the quasi-bound state decays to. We can estimate
this with a series of tricks. First,
Z ∞ Z b
∗
⟨ψk |V̂1 |φ⟩ = dx ψk (x) V1 (x) φ(x) = Vb dx ψk∗ (x) φ(x). (2.45)
−∞ −b

Next, because ψk (x) and φ(x) are orthogonal,

Z ∞ Z b Z
∗ ∗
dx ψk (x) φ(x) = 0 ⇒ dx ψk (x) φ(x) = − dx ψk∗ (x) φ(x). (2.46)
−∞ −b |x|>b

The integration range consists of two pieces, x > b and x < −b, which can be interpreted as
the decay of the quasi-bound state to the left or right. We assume these contribute equally
to the escape probability, so
Z ∞ 2
2 2 ∗
⟨ψk |V̂1 |φ⟩ ≈ 2Vb dx ψk (x) φ(x) . (2.47)
b
Outside the scatterer, the escaping free state is approximately an outgoing plane wave,
eikx
ψk (x) ≈ √ , (2.48)
2π
where r
2m
Ek = Eres ≈ E0 + Vb ⇒ k≈
(E0 + Vb ). (2.49)
ℏ2
Plugging Eqs. (2.43) and (2.48) into Eq. (2.47), and solving the integral, gives
2 1 2 2 e−2ηb
⟨ψk |V̂1 |φ⟩ ≈ V B 2 . (2.50)
π b k + η2
The other quantity we need for Fermi’s golden rule is the density of free states. This can
be found by taking Ek ≈ ℏ2 k 2 /2m, and performing a change of variables:
Z ∞
ℏ2 k 2
 
dk
D(E) = δ E− (2.51)
−∞ 2π 2m
Z ∞
1 dk
=2· dE ′ δ(E − E ′ ) (2.52)
2π 0 dE ′
r
m
= . (2.53)
2π ℏ2 E
2

31
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

In Eq. (2.52), the factor of 2 accounts for the fact that for each E, both positive and negative
k contribute to the density of states.
We can now plug Eqs. (2.50) and (2.53) into Fermi’s golden rule, Eq. (2.41). The figure
below shows results for U = 6, Vb = 30, and a = 1, with computational units ℏ = m = 1.
The orange curve shows how κ, the decay rate from Fermi’s golden rule, varies with the
barrier thickness b − a. For comparison, the blue dots show the values of κ derived from
the resonant scattering peak (see Section 2.5), by computing the transmittance using the
transfer matrix method (see Appendix B) and numerically extracting the peak width.

The agreement is pretty good, especially considering the numerous approximations leading
to the derivation of Fermi’s golden rule! The behavior also matches our intuition: when the
barrier thickness is large, the quasi-bound state should indeed decay more slowly.
In this simple 1D example, Fermi’s golden rule is not especially useful since the scattering
problem can be solved easily (see Appendix B). However, in higher dimensions, nonpertur-
bative scattering problems are typically much less tractable. In such cases, Fermi’s golden
rule provides a very convenient way to estimate the widths of resonance peaks.

EXERCISES

1. Use the variational theorem to prove that a 1D potential well has at least one bound
state. Assume that the potential V (x) satisfies (i) V (x) < 0 for all x, and (ii) V (x) → 0
for x → ±∞. The Hamiltonian is
ℏ2 d2
Ĥ = − + V (x). (2.54)
2m dx2
Consider a (real) trial wavefunction
 1/4
2γ 2
ψ(x; γ) = e−γx . (2.55)
π
Note that this can be shown to be normalized to unity, using Gauss’ integral
Z ∞ r
−2γx2 π
dx e = . (2.56)
−∞ 2γ

32
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

Now prove that

Z ∞
⟨E⟩ = dx ψ(x) Ĥ ψ(x)
−∞
 2
2 Z ∞ Z ∞
ℏ dψ
= dx + dx V (x) ψ 2 (x) (2.57)
2m −∞ dx −∞
 Z ∞ 
√ √ −γx2
=A γ γ +B dx V (x) e ,
−∞

where A and B are positive real constants to be determined. By looking at the quantity
in square brackets in the limit γ → 0, argue that ⟨E⟩ < 0 in this limit. Hence, explain
why this implies the existence of a bound state.
Finally, try generalizing this approach
p to the case of a 2D radially-symmetric potential
well V (x, y) = V (r), where r = x2 + y 2 . Identify which part of the argument fails
in 2D. [For a discussion of certain 2D potential wells that do always support bounds
states, similar to 1D potential wells, see Ref. [4].]
2. In this problem, you will investigate the existence of bound states in a 3D potential
well that is finite, uniform, and spherically-symmetric. The potential function is

V (r, θ, ϕ) = −U Θ(a − r), (2.58)

where a is the radius of the spherical well, U is the depth, and (r, θ, ϕ) are spherical
coordinates defined in the usual way.
The solution involves a variant of the partial wave analysis discussed in Appendix A.
For E < 0, the Schrödinger equation reduces to
  p
 ∇2 + q 2 ψ(r, θ, ϕ) = 0 where q = 2m(E + U )/ℏ2 , for r ≤ a
  p (2.59)
 ∇2 − γ 2 ψ(r, θ, ϕ) = 0 where γ = −2mE/ℏ2 , for r ≥ a.

For the first equation (called the Helmholtz equation), we seek solutions of the form

ψ(r, θ, ϕ) = f (r) Yℓm (θ, ϕ), (2.60)

where Yℓm (θ, ϕ) are spherical harmonics, and the integers l and m are angular mo-
mentum quantum numbers satisfying l ≥ 0 and −l ≤ m ≤ l. Substituting into the
Helmholtz equation yields

d2 f df 
r2 + 2r + q 2 r2 − l(l + 1) f (r) = 0,

2
(2.61)
dr dr
which is the spherical Bessel equation. The solutions to this equation that are non-
divergent at r = 0 are f (r) = jℓ (qr), where jℓ is called a spherical Bessel function
of the first kind. Most numerical packages provide functions to calculate these (e.g.,
scipy.special.spherical jn in Scientific Python).
Similarly, solutions for the second equation can be written as ψ(r, θ, ϕ) = g(r) Yℓm (θ, ϕ),
yielding an equation for g(r) called the modified spherical Bessel equation. The

33
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

solutions which do not diverge as r → ∞ are g(r) = kℓ (γr), where kℓ is called a mod-
ified spherical Bessel function of the second kind. Again, this can be computed
numerically (e.g., using scipy.special.spherical kn in Scientific Python).
Using the above facts, show that the condition for a bound state to exist is
qjℓ′ (qa) γk ′ (γa)
= ℓ , (2.62)
jℓ (qa) kℓ (γa)
where jℓ′ and kℓ′ denote the derivatives of the relevant special functions, and q and γ
depend on E and U as described above. Write a program to search for the bound
state energies at any given a and U , and hence determine the conditions under which
the potential does not support bound states.
3. In this problem, we will find the quasi-bound and free states for the model discussed
in Section 2.7, and use it to calculate the decay rate according to Fermi’s golden rule.
Let φ(x) be the bound state of a square well of width 2a and depth U , and let E0 be
the energy. This state satisfies the 1D time-independent Schrödinger wave equation,
ℏ2 d2
 
− + V (x) φ(x) = E0 φ(x), (2.63)
2m dx2
where 
 V0 (x) = −U Θ(a − |x|)

V (x) = V0 (x) + V1 (x), where V1 (x) = Vb Θ(b − |x|) (2.64)

 0 < U < Vb .
Take the ansatz (
A cos(qx), |x| < a
φ(x) = (2.65)
B exp (−η|x|) , |x| ≥ a.

(a) Using Eqs. (2.63)–(2.65), and the fact that both φ(x) and dφ/dx are continuous
at x = ±a, prove that E0 can be obtained by solving the transcendental equation
 r
2m
q= (E0 + U )


ℏ2
q tan(qa) = η, where r (2.66)
 η = 2m |E0 |.


ℏ2
(b) By using the fact that φ(x) is normalized to unity, prove that

 −1
2 exp 2ηa 1 + sin(2qa)/2qa 1
B = + . (2.67)
a cos2 (qa) ηa
(c) Write a program to compute the decay rate based on Fermi’s golden rule, by
combining Eqs. (2.41), (2.50), (2.53), (2.66), and (2.67). Hence, reproduce the
plot shown in Section 2.7.
(d) Write a program to extract the decay rate from the width of the transmission
peak, using the transfer matrix method (see Appendix B). Hence, investigate the
accuracy of the Fermi’s golden rule result for different values of U , Vb , and the
other model parameters.

34
Y. D. Chong Ch. 2: Resonances | Graduate Quantum Mechanics

FURTHER READING

[1] Bransden & Joachain, §4.4, 9.2–9.3, 13.4

[2] Sakurai, §5.6, 7.7–7.8

[3] R. Courant and D. Hilbert, Methods of Mathematical Physics vol. 1, Interscience (1953).
[link]

[4] B. Simon, The bound state of weakly coupled Schrödinger operators in one and two
dimensions, Annals of Physics 97, 279 (1976). [link]

35