Isotope effect, Cooper pairs and BCS theory
Rishi Paresh Joshi
March 2025
1 Phonons and isotope mass
The variation in phonon frequency with the inverse square root of isotope mass,
ωphonon ∝ √1M , arises from the physics of lattice vibrations in a crystal, where
phonons are quantized vibrational modes. This relationship emerges naturally
when considering the dynamics of atoms in a crystal lattice, modeled as a system
of masses connected by springs (the harmonic approximation). Let’s derive this
step by step.
1.1 Understanding Phonons and Lattice Vibrations
Phonons are quantized collective vibrational modes of atoms in a crystal lattice.
In a solid, atoms are arranged in a periodic lattice and interact with their neigh-
bours via interatomic forces, often called harmonic (spring-like). The frequency
of these vibrations depends on the mass of the atoms and the strength of the
interatomic interactions. When we consider the effect of isotopic substitution,
we’re changing the mass ( M ) of the atoms in the lattice (e.g., replacing 12 C
with 13 C in a carbon-based material like diamond). Isotopes have the same
electronic structure, so the interatomic forces (and thus the ”spring constant”
of the bonds) remain unchanged, but the mass of the atoms changes.
1.2 Model the Lattice as a Harmonic Oscillator
To understand phonon frequencies, start with a simple model: a one-dimensional
chain of atoms. Imagine a chain of atoms with mass M , connected by springs
with spring constant k. This spring constant k represents the strength of the
interatomic forces, which depends on the electronic structure and is unaffected
by isotopic mass. For a single harmonic oscillator (one atom vibrating against
a fixed point), the frequency of oscillation is given by:
r
k
ωphonon =
M
TC M 1/2 = constant.
1
This is the classical frequency of a mass M on a spring with spring constant
k. Notice the √1M dependence: as the mass increases, the frequency decreases
because a heavier mass oscillates more slowly for the same restoring force.
1.3 Extending to a Crystal Lattice
In a crystal lattice, the atoms are coupled, and the vibrations are collective
modes (phonons). Consider a one-dimensional lattice with atoms of mass M ,
lattice spacing a, and nearest-neighbor interactions modeled by a spring con-
stant k. The equation of motion for the displacement un of the n-th atom
is:
d2 un
M 2 = k(un+1 − un ) + k(un−1 − un ) = k(un+1 + un−1 − 2un )
dt
Assume a traveling wave solution for the displacement, typical for phonons:
un = u0 ei(qna−ωt)
where q is the wavevector, ω is the phonon frequency, and a is the lattice spacing.
Substitute into the equation of motion:
−M ω 2 u0 ei(qna−ωt) = k eiqa + e−iqa − 2 u0 ei(qna−ωt)
Cancel the common exponential terms:
−M ω 2 = k eiqa + e−iqa − 2 = k(2 cos(qa) − 2) = k · 2(cos(qa) − 1)
qa
M ω 2 = 2k(1 − cos(qa)) = 4k sin2
2
r
4k qa
ω= sin
M 2
This is the dispersion relation for phonons inqa 1D lattice. The key point is the
ω∝ √1 dependence, which arises from the k
term.
M M
1.4 Application to Isotopes
When the mass M changes due to isotopic substitution (e.g., from M1 to M2 ),
the spring constant k remains the same because the electronic interactions (bond
strength) are unchanged—isotopes differ only in nuclear mass, not in chemical
properties. Thus, the phonon frequency scales as:
r
k 1
ω∝ ∝√
M M
For two isotopes with masses M1 and M2 , the ratio of their phonon frequencies
is: r
ω2 M1
=
ω1 M2
If M2 > M1 , then ω2 < ω1 , meaning the heavier isotope has a lower phonon
frequency.
2
1.5 Physical Interpretation of phonons
The √1M dependence makes sense physically: heavier atoms oscillate more slowly
because their inertia is greater, but the restoring force (from the interatomic
potential, i.e., k) remains the same. In a 3D crystal, the same principle applies.
The phonon dispersion relation depends on the direction and branch (acoustic
or optical), but the frequency always scales with √1M when considering the effect
of mass.
1.6 Optical Phonons and Diatomic Lattices
For a diatomic lattice (e.g., with two different masses M1 and M2 ), the optical
phonon frequency at the zone center ( q = 0 ) also shows this dependence. The
optical mode involves the two atoms vibrating out of phase, and the frequency
is determined by the reduced mass µ = MM11+M M2
2
. If one mass is fixed and the
other changes due to an isotope, the frequency still scales as √1µ , which reduces
to √1 for the varying isotope.
M
Conclusion
The phonon frequency varies as √1M because the frequency of lattice vi-
brations depends on the inverse square root of the atomic mass in a harmonic
system, while the interatomic force constants remain unchanged with isotopic
substitution. This is a direct consequence of the dynamics of coupled harmonic
oscillators in the lattice, as seen in both simple 1D models and more complex
3D crystals.
2 Isotope effect in superconductors
The Isotope Effect in Superconductivity The isotope effect refers to the observed
dependence of a superconductor’s critical temperature (Tc ) on the isotopic mass
of its constituent atoms. Discovered experimentally in the 1950s, this phe-
nomenon provided critical evidence for the role of lattice vibrations (phonons)
in superconductivity.
Theoretical Basis In a conventional superconductor, the transition to the
superconducting state occurs when electrons form pairs that move without re-
sistance. The isotope effect arises because the pairing mechanism involves in-
teractions mediated by phonons, which are quantized lattice vibrations. The
frequency of these phonons (ω) depends on the ionic mass (M ) of the lattice
atoms. In a simple harmonic oscillator model, the phonon frequency scales as:
1
ω∝√ .
M
The critical temperature Tc is related to the strength of the electron-phonon
interaction. In BCS theory, Tc is approximately given by:
Tc ∝ h̄ωe−1/(N (0)V ) ,
3
where h̄ is the reduced Planck constant, N (0) is the electronic density of states
at the Fermi level, and V is the effective electron-phonon coupling strength. If
V is independent of isotopic mass, then:
Tc ∝ ω ∝ M −1/2 .
Thus, Tc varies inversely with the square root of the isotopic mass, leading to
the relation:
Tc M 1/2 = constant.
Experimental Evidence The isotope effect was first confirmed in mercury by
measuring Tc for different isotopes (e.g., 200 Hg and 204 Hg). Experiments showed
that Tc ∝ M −α , with α ≈ 0.5 for many elemental superconductors, consistent
with the phonon-mediated pairing hypothesis. Deviations from α = 0.5 occur in
some materials due to variations in N (0) or V , but the effect strongly supports
the idea that lattice dynamics are integral to superconductivity.
Significance The isotope effect provided a key clue that superconductivity
is not purely an electronic phenomenon but involves the lattice. This insight
guided the development of BCS theory, which explicitly incorporates phonon-
mediated electron pairing.
3 Reference system: System of degenerate states
Consider a system of four states |1⟩, |2⟩, |3⟩, |4⟩ forming a basis in which the
Hamiltonian 0 is diagonal, with four identical eigenvalues ϵ. We then add an
interaction potential V̂ that has matrix elements on this basis that are all iden-
tical and negative, with value V . ⟨j|V̂ |i⟩ = Vji = V . Finding the new energy
levels, i.e. the new eigenvalues of the Hamiltonian, involves solving the system
of equations
ϵ−V −E V V V c1 0
V ϵ − V − E V V c2 0
= .
V V ϵ−V −E V c3 0
V V V ϵ−V −E c4 0
The solutions are the values of Emaking the determinant vanish or,
(E − ϵ)3 (E − ϵ + 4V ) = 0
which leads to three levels of energy equal to the initial energy and a fourth
lower energy level, E1 = ϵ, E2 = ϵ, E3 = ϵ; Eα = ϵ − 4V . The eigenvector |ψ⟩
associated with this last level is the linear symmetric combination of the states
|1⟩, |2⟩, |3⟩ and |4⟩,
1
|ψ⟩ = (|1⟩ + |2⟩ + |3⟩ + |4⟩).
4
In the language of quantum mechanics |ψ⟩ is a stationary state, which means
that:
4
• if the particle is in this state at time t = 0 it will stay in this state
indefinitely, with quantum phase factor e−iEt/h̄ ;
• if at time t = 0, the particle is, for example, in the state |1⟩, which is
not an eigenvector of the complete Hamiltonian, at later times, it will be
found with non-zero probabilities in the other states |2⟩, |3⟩ and |4⟩. The
time-dependent equation gives the evolution of the system
d X
i ci (t) = Vij cj (t)
dt j
where ci is the probability amplitude of finding the particle in the state
|i⟩.
4 System of non-degenerate states
System with non-degenerate states
We now consider the case where eigenvectors of the non-interacting system
|1⟩, |2⟩, |3⟩ and |4⟩ correspond to different energies ϵ1 , ϵ2 , ϵ3 and ϵ4 . Introduction
of the potential V̂ leads to the eigenvalue equation
ϵ1 − V − E −V −V −V c1 0
−V ϵ2 − V − E −V −V c2 0
=
−V −V ϵ3 − V − E −V c3 0
−V −V −V ϵ4 − V − E c4 0
i.e., to four equations with four unknowns
V (c1 + c2 + c3 + c4 ) = ci (ϵi − E); i ∈ [1, 4]
or, equivalently, to the four relations
ci c1 + c2 + c3 + c4
= ; i ∈ [1, 4].
V ϵi − E
Summing each of thePtwo sides over the index “i” we then obtain a general
4
relation, and cancelling i=1 ci :
4
1 X 1
= .
V ϵ −E
i=1 i
If ϵ1 = ϵ2 = ϵ3 = ϵ4 = ϵ, we recover the solution E = ϵ − N V , where N is
the order of the matrix.
When the energies ϵi differ (ϵ1 < ϵ2 < ϵ3 < ϵ4 ), the equation always has one,
and only one, energy lower than the lowest of the original energies as appears
clearly by examination of the energy levels. It is the only solution for which all
the ci are positive.
5
Figure 1: Bound state in an initially non-degenerate state An interaction that
adds a term −V to each element of an initially non-degenerate Hamiltonian
matrix leads to the emergence of one, and only one, energy level Eα below the
lowest non-perturbed level.
Figure 2: The system of N levels forms a quasi-continuum. With the interaction
-V, one, and only one, the state emerges with energy below the lowest value of
the continuum.
The eigenvector |ψ⟩ associated with the eigenvalue Eα can be written in the
basis of non-perturbed eigenvector as
|ψ⟩ = c1 |1⟩ + c2 |2⟩ + c3 |3⟩ + c4 |4⟩
The ci , components of the “bound” state |ψ⟩ of the eigenvectors of the non-
interacting Hamiltonian, satisfy the relations c1 > c2 > c3 > c4 . Therefore,
the states of lowest energy appear with a greater weight in the projection of
|ψ⟩. This means that in the dynamic picture described previously, according to
which the electron jumps between the stationary states of the non-perturbed
system with occupation probabilities c2j , the particle has a higher probability of
being found in the states of lowest energy.
6
5 Two body interaction hamiltonian
When the particles interact with each other, we must add an interaction term
V̂ (r1 , r2 ) to the two-particle Hamiltonian which becomes
h̄2 2 h̄2 2
Ĥ0 (r1 , r2 ) = − ∇1 − ∇ + V̂ (r1 , r2 ).
2m1 2m2 2
In this case, proceeding by separating variables is no longer possible. The search
for new energy levels requires diagonalization of the matrix ⟨km kn |V̂ |kp ks ⟩ that,
in the basis |km kn ⟩ of eigenvectors of the non-interacting system of two particles,
has in addition to the diagonal terms we have just seen, non-diagonal terms that
are the matrix elements R
⟨kp ks |V̂ |km kn ⟩ = V12 V (r1 , r2 )e−ikp ·r1 e−iks ·r2 eikm ·r1 eikn ·r2 d3 r1 d3 r2 .
Integration over r1 and r2 extends over the box’s volume. This matrix
element takes the value in a unit volume, assuming the size of the box is much
larger than the range of the potential (to avoid boundary effects) with r =
r2 − r1 : (
V (q) if q = ks + kn − km − kp ,
⟨kp ks |V̂ |km kn ⟩ =
0 otherwise,
where Z
V (q) = V (r)eiq·r d3 r.
The process of interaction can be interpreted as follows :
• two electrons are initially in states |m⟩ and |n⟩ (momenta km and kn );
• during the interaction they exchange a momentum q with probability
amplitude V (q);
• After this exchange, the electrons are to be found in states |p⟩ = |m − q⟩
and |s⟩ = |n + q⟩ with momenta (km − q) and (kn + q). Because this
exchange must be without any change in total momentum, only terms
satisfying the condition
km + kn = kp + ks
are nonzero.
7
Figure 3: Diagrammatic presentation of the interaction between two particles
In the initial state, the particles are in states |m⟩ and |n⟩. By exchanging
momentum q, they are scattered into the states |m − q⟩ and |n + q⟩ .
6 The BCS attractive interaction
The process is as follows:
1. an electron of initial momentum km “excites” a phonon of wave vector q
transferring a momentum q; its final momentum is (km − q);
2. the created phonon interacts with an electron of momentum kn that “ab-
sorbs” it, acquiring a momentum (kn + q).
The electrostatic interaction between electrons and crystal ions causes the ex-
citation and absorption of the phonon. The first electron (negative charge) dis-
places crystal ions (positive charge) and generates the phonon during its motion.
The moving ions will interact with the second electron by giving up their move-
ment. A detailed analysis reveals that to be non-vanishing, the non-diagonal
elements of the interaction matrix should satisfy two conditions:
• The initial and final states must be states of opposite wave vectors and
opposite spins. They are, therefore, of the form:
Vk′ ,k = ⟨k↑′ , −k↓′ |V̂k,k |k↑ , −k↓ ⟩
. Such states |k, k⟩ are called ’pair states’.
• As the particles exchanged during the interaction process are phonons, the
only transitions possible are between pair states involving energy transfers
less than the maximum phonon energy, which is order ED = h̄ωD (the
DEBYE energy 10 to 20 meV). As a consequence the only “pair states”
that can be involved in the transitions are those constructed out of wave
vectors of wavelength k whose energies k lie in a band of energies of order
ED around the FERMI energy EF . The simple BCS approximation then
considers that the non-zero matrix elements that describe the interaction
are all equal with
(
V if EF − ED < (k and k ′ ) < EF + ED
Vk ′ k =
0 otherwise
8
7 Form for the actual attractive potential
The coupling gq for longitudinal acoustic phonons in a jellium model (a simpli-
fied model of a metal) is related to the screened Coulomb interaction. It can be
approximated as:
s
h̄
gq = i qVq ,
2M N ωq
2
where Vq = q4πe2 +k 2 is the screened Coulomb potential, ks is the Thomas-
s
Fermi screening wavevector, M is the ionic mass, N is the number of ions, and
ωq = vs q (for acoustic phonons, with vs as the sound speed).
—
7.1 Second-Order Perturbation Theory
The effective interaction between two electrons arises from the virtual exchange
of a phonon, which we compute using second-order perturbation theory. The
second-order energy correction to a state |ψ0 ⟩ (two electrons in states (k1 , ↑
), (k2 , ↓)) is:
X |⟨n|Hint |ψ0 ⟩|2
E (2) = ,
E0 − En
n̸=0
Where:
Initial state: |ψ0 ⟩ = c†k1 ↑ c†k2 ↓ |0⟩, with energy E0 = ϵk1 + ϵk2 ,
Intermediate state |n⟩: One electron scatters by emitting a phonon, e.g.,
electron 1 from k1 to k1 + q, producing a phonon q, then electron 2 absorbs it,
scattering from k2 to k2 − q.
The intermediate state is:
|n⟩ = c†k1 +q,↑ c†k2 −q,↓ a†q |0⟩,
where c†k represents electron creation operators and a†q represents phonon cre-
ation operators, with energy:
En = ϵk1 +q + ϵk2 −q + h̄ωq .
The energy difference is:
E0 − En = (ϵk1 + ϵk2 ) − (ϵk1 +q + ϵk2 −q + h̄ωq ).
7.2 Matrix Elements
The matrix element for the emission process is:
⟨n|Hint |ψ0 ⟩ = gq ,
9
Since Hint connects the initial state to the intermediate state by creating
a phonon and scattering an electron. The second step (absorption) leads to a
final state, but in the effective potential, we sum over intermediate states.
—
7.3 Effective Interaction Potential
The effective interaction potential Veff between two electrons is obtained by
interpreting the second-order energy shift as an interaction term. For electrons
scattering from (k1 , k2 ) to (k′1 , k′2 ), with momentum transfer q = k′1 − k1 =
k2 − k′2 , the effective potential is:
X |⟨final|Hint |n⟩⟨n|Hint |initial⟩|
Veff (q, ω) = ,
Einitial − En
intermediate states
where ω is the energy difference between initial and final states in frequency
units (h̄ω = ϵk1 + ϵk2 − ϵk′1 − ϵk′2 ).
The matrix elements give |gq |2 , and the energy denominator, for Cooper
pairs near the Fermi surface (ϵk1 ≈ ϵk2 ≈ EF ), simplifies. The energy difference
in the denominator becomes dominated by the phonon energy if the electron
energy difference is small:
ϵk1 + ϵk2 − ϵk1 +q − ϵk2 −q ≈ 0,
so:
E0 − En ≈ −h̄ωq .
However, including the frequency dependence (h̄ω) of the scattering process,
the denominator is:
h̄ω − h̄ωq .
Thus, the effective potential from phonon exchange is:
|gq |
Veff (q, ω) = .
h̄ω − h̄ωq
For the reverse process (phonon absorption then emission), the denominator
becomes h̄ω + h̄ωq , so the total contribution is:
1 1
Veff (q, ω) = |gq | + .
h̄ω − h̄ωq h̄ω + h̄ωq
Simplify:
1 1 2h̄ω 2ω
+ = 2 2
= 2 .
h̄ω − h̄ωq h̄ω + h̄ωq (h̄ω) − (h̄ωq ) ω − ωq2
Thus:
10
2ω
Veff (q, ω) = |gq | .
ω 2 − ωq2
— Computing |gq | Using the expression for gq :
s
h̄q 2 4πe2
gq = i ,
2M N ωq q 2 + ks2
4πe2
h̄
|gq | = q2 .
2M N ωq q 2 + ks2
For acoustic phonons, ωq = vs q, and M N = ρV (where ρ is the mass density,
V is the volume). However, in the jellium model, we approximate the prefactor.
The key term is:
q2 4πe2
|gq | ∝ .
ωq q 2 + ks2
The canonical result, after averaging over all electron modes, and approxi-
mating ω ≈ ωD gives:
h̄ωq2 4πe2 ωq2 4πe2
1
|gq | ≈ = .
2 q 2 + ks2 h̄ω 2ω q 2 + ks2
—
7.4 Final Form of the Second Term
Substitute |gq |:
4πe2
1 2ωq
Veff,perturbation (q, ω) = .
2 q 2 + ks2 ω2− ωq2
The total potential is adding the screening term:
4πe2 4πe2 ωq2
Veff,total (q, ω) = 2 2
+ 2 .
q + ks q + ks ω − ωq2
2 2
This is attractive for ωq > ω.
11
Figure 4: Graph of the interaction at the origin of BCS superconductivity Ini-
tially, two electrons occupy the pair state |k, −k⟩. At time t, the electron of
wave vector k creates a phonon of wave vector q and acquires the wave vector
k ′ = k − q. At a later time, the electron of wave vector −k absorbs the phonon
and its wave vector becomes k ′ = −k + q. The net effect is that the electron
pair has been scattered from the state |k, −k⟩ into the state |k ′ , −k ′ ⟩.
12
