Dynamical Decoupling

Shreya JVS, Sayan Biswas, Duha Alsurdi, Aritrika

December 8, 2025

1 Project II: Dynamical Decoupling:

Problem
Consider two two-level systems (TLSs) that interact via a potential:

V̂ = V0 |e1 e2 ⟩⟨e1 e2 |.

In addition, the excited states experience different detunings (“inhomogeneous broadening”), ∆1 and
∆2 , and there is one laser that drives both TLSs.

1. Explain what inhomogeneous broadening is, and how it differs from homogeneous broadening.
Give examples.

1.1 Definitions and Examples:

Homogeneous broadening:
Spectral line is broadened by effects experienced equally by all atoms. As an example, the
spontaneous emission case we took in class, we saw how the spectral line of emission showed a
Lorentzian shape at resonance, and broadening by γr , the radiative decay rate. The broaden-
ing here is natural, meaning it’s just a result of Heisenberg uncertainty principle which tells us
that the inherent width of spectral lines is caused by the finite lifetime of an atom’s excited state.

Inhomogeneous broadening:
An increase in the width of a spectral line of absorption or emission that occurs because different
atoms in a medium have slightly different transition frequencies, that may happen due to dif-
ferences in the local environment every atom lives in, that may cause energy shifts. The overall
observed spectrum is an average of these individual, slightly shifted, spectral lines, resulting in
a broader Gaussian peak.
Example on Inhomogeneous:
Doppler broadening is a prominent example on inhomogeneous broadening. It is considered
inhomogeneous because atoms in a Maxwell–Boltzmann thermal distribution move at different
velocities, so each atom experiences a different Doppler shifted frequency of the incoming light.

1
2. Formulate the Hamiltonian of the system and derive the equations of motion H.

1.2 Hamiltonian and Equations of Motion

We consider two driven TLSs, each described in the way we learned in class for a single TLS, j:
ℏδj (j) ℏΩ (j)
Ĥ (j) = −
σ̂ − σ̂ , j = 1, 2, (1)
2 z 2 x
where δj is the detuning of TLS j (δj = ωL −ωo,j ), Ω is the Rabi frequency (driving field identical
(j)
for both TLSs), and σ̂x,z are Pauli matrices acting on TLS j only.
The two TLSs interact via:
V̂ = V0 |e1 e2 ⟩⟨e1 e2 |. (2)
And as evident by the form of interaction, it’s only ”on” when both the electronic populations
of the TLSs are in the excited states.

Full Hamiltonian
The total Hamiltonian is
ℏδ1 (1) ℏδ2 (2) ℏΩ (1)
σx + σx(2) + V0 |e1 e2 ⟩⟨e1 e2 |


Ĥ = − σz − σz − (3)
2 2 2

Equations of Motion
We use the basis
|g1 g2 ⟩, |e1 g2 ⟩, |g1 e2 ⟩, |e1 e2 ⟩,
and expand the wavefunction as:
|ψ(t)⟩ = cgg (t)|g1 g2 ⟩ + ceg (t)|e1 g2 ⟩ + cge (t)|g1 e2 ⟩ + cee (t)|e1 e2 ⟩. (4)
Corresponding to the different possibilities os states that can be occupied.
Applying Schrodinger’s equation iℏψ̇ = Hψ, we obtain, for the RHS:

H|ψ⟩ = cgg H|gg⟩ + ceg H|eg⟩ + cge H|ge⟩ + cee H|ee⟩. (5)
Insert each expression:
 
ℏ ℏΩ

H|ψ⟩ = cgg (δ1 + δ2 )|gg⟩ − |eg⟩ + |ge⟩
2 2
 
ℏ ℏΩ ℏΩ
+ ceg − (δ1 − δ2 )|eg⟩ − |gg⟩ − |ee⟩
2 2 2
 
ℏ ℏΩ ℏΩ
+ cge (δ1 − δ2 )|ge⟩ − |gg⟩ − |ee⟩
2 2 2
 
ℏ ℏΩ

+ cee − (δ1 + δ2 )|ee⟩ − |ge⟩ + |eg⟩ + V0 |ee⟩ . (6)
2 2

Grouping terms with the same states/kets, we obtain

h i
H|ψ⟩ = ℏ2 (δ1 + δ2 )cgg − ℏΩ 2 (ceg + cge ) |gg⟩
h i
+ − ℏΩ2 cgg − ℏ
2 (δ 1 − δ 2 )ceg − ℏΩ
2 c ee |eg⟩
h i
+ − ℏΩ2 cgg + 2 (δ1 − δ2 )cge − 2 cee |ge⟩
ℏ ℏΩ

h
i
+ − ℏΩ2 (ceg + c ge ) + − ℏ
2 (δ 1 + δ 2 ) + V 0 cee |ee⟩. (7)

2
On the RHS, we have:
d
iℏ |ψ⟩ = iℏċgg |gg⟩ + iℏċeg |eg⟩ + iℏċge |ge⟩ + iℏċee |ee⟩. (8)
dt

Equating both sides and reading off the equations of motion, we obtain

iℏċgg = ℏ2 (δ1 + δ2 )cgg − ℏΩ

2 (ceg + cge ) (9)

iℏċeg = − ℏΩ
2 cgg − 2 (δ1 − δ2 )ceg −
ℏ ℏΩ
2 cee (10)

2 cgg + 2 (δ1 − δ2 )cge −

iℏċge = − ℏΩ ℏ ℏΩ
2 cee (11)


iℏċee = − ℏΩ
2 (ceg + cge ) + − 2 (δ1 + δ2 ) + V0 cee .
ℏ
(12)

Canceling the ℏs:

1 Ω
iċgg = (δ1 + δ2 )cgg − (ceg + cge ), (13)
2 2
Ω 1 Ω
iċeg = − cgg − (δ1 − δ2 )ceg − cee , (14)
2 2 2
Ω 1 Ω
iċge = − cgg + (δ1 − δ2 )cge − cee , (15)
2 2 2
 
Ω 1 V0
iċee = − (ceg + cge ) + − (δ1 + δ2 ) + cee . (16)
2 2 ℏ

Or, more compactly,

d
iℏ |ψ(t)⟩ = Ĥ |ψ(t)⟩, (13)
dt
which in components reads    
ċgg cgg
ċ  c 
 eg   eg 
iℏ   = Ĥ   , (14)
ċge  cge 
ċee cee
with the Hamiltonian matrix in the ordered basis {|gg⟩, |eg⟩, |ge⟩, |ee⟩} given by

δ1 + δ 2 Ω Ω
 
 2 − − 0
 2 2 

 −Ω
 δ1 − δ2 Ω 
− 0 − 

Ĥ =  2 2 2 
. (15)
 −Ω
 δ1 − δ 2 Ω 
0 − 

 2 2 2 

 Ω Ω δ1 + δ2 V0 
0 − − − +
2 2 2 ℏ

3
3. If δ1 = δ2 , show how the interaction can realize a CZ operation on the atoms, leaving them in
an entangled state.

In this part, we assume that the two atoms have identical detunings δ1 = δ2 , (to be under-
stood why later), and we want to show two things:

(a) The interaction behaves like a CZ gate. A CZ gate is a two-qubit gate that acts as follows:

CZ |gg⟩ = |gg⟩,
CZ |ge⟩ = |ge⟩,
CZ |eg⟩ = |eg⟩,
CZ |ee⟩ = − |ee⟩.

i.e;
CZ = diag(1, 1, 1, −1).
That is, the CZ gate leaves |gg⟩, |ge⟩, |eg⟩ unchanged, but multiplies |ee⟩ by a minus sign.
Which fits the interaction Hamiltonian, since it’s only ”on” when it acts on |ee⟩, but we’re
left with preparing the correct state to be acted on.
(b) Acting with this CZ gate creates entanglement from a non-entangled state. If we start in
a suitable superposition (not entangled), after the CZ gate the system becomes entangled.
The given interaction Hamiltonian,

Ĥint = V0 |ee⟩⟨ee|,

acts only when both atoms are excited. We want to show that if we prepare a suitable
superposition containing |ee⟩, then letting the system evolve under Ĥint for the right amount
of time produces exactly the CZ phase on |ee⟩.

So the overall logic is:

π/2 pulses int Ĥ
|gg⟩ −−−−−−−→ product superposition −−− → entangled state.

The story start when both atoms are in the ground states in the two emitters, now, we want to
generate the right state, whatever that state is, it should contain the (|ee⟩), so that we can act
on it by the interaction Hamiltonian.

Hamiltonian without interaction

At the beginning, both atoms are in the ground state (|gg⟩). We can consider each atom as an
independent two-level system driven by the laser:

(j) ℏδj (j) ℏΩ (j)

Ĥsingle = − σ − σ , j = 1, 2, (16)
2 z 2 x
If we ignore the interaction Hamiltonian in the question at this stage, the total Hamiltonian is
just the sum of the two single-atom Hamiltonians:
(1) (2)
Ĥno int = Ĥsingle ⊗ I2 + I1 ⊗ Ĥsingle , (17)

where (I1 , I2 ) are identity operators on atom 1 and 2, respectively.

Later, when we turn off the driving (set Ω = 0) and consider only the interaction, the relevant
Hamiltonian will simply be
Ĥint = V0 |ee⟩⟨ee|. (18)
This term does nothing unless both atoms are in |e⟩.

4
Single-atom Bloch vector and π/2 pulse
The general pure state of a single driven two-level atom can be written on the Bloch sphere as
   
θ θ
|ψ(θ, ϕ)⟩ = cos |g⟩ + eiϕ sin |e⟩, (19)
2 2
where
• θ ∈ [0, π] is the polar angle,
• ϕ ∈ [0, 2π) is the azimuthal (phase) angle.
If we start in |g⟩, in the south pole of the Bloch sphere, a pulse of area θ = Ω ∗ t rotates the
Bloch vector by an angle θ and produces
   
pulse with angle θ θ θ
|g⟩ −−−−−−−−−−−→ |ψ(θ, 0)⟩ = cos |g⟩ + sin |e⟩, (20)
2 2

For a π/2 pulse the final state becomes:

1
|ψπ/2 ⟩ = √ (|g⟩ + |e⟩). (21)
2

Thus, a π/2 pulse acting on an atom initially in |g⟩ creates a superposition of |g⟩ and |e⟩ with equal
amplitudes (up to an overall phase). On the other hand, a π pulse acting on an atom initially in
|g⟩ creates an atom that’s in |e⟩, and tensor product this |e⟩ with the second |e⟩coming from the
second |g⟩ will result in non entangeled state that can’t be entangled by applying the interaction
Hamiltonian, and that’s why we need the π/2 pulse.

Two-atom state after π/2 pulses

We start with both atoms in the ground state:

|ψinitial ⟩ = |g1 g2 ⟩ = |gg⟩. (22)

We apply the same π/2 pulse (with appropriate phase) to each atom:
π/2 |gj ⟩ + |ej ⟩
|gj ⟩ −−−→ √ , j = 1, 2. (23)
2

Therefore the joint state becomes the tensor product of two single-atom superpositions:
   
|g1 ⟩ + |e1 ⟩ |g2 ⟩ + |e2 ⟩
|ψafter pulses ⟩ = √ ⊗ √ . (24)
2 2
Expanding this product,
1  1 
|ψafter pulses ⟩ = |g1 g2 ⟩ + |g1 e2 ⟩ + |e1 g2 ⟩ + |e1 e2 ⟩ = |gg⟩ + |ge⟩ + |eg⟩ + |ee⟩ . (25)
2 2

This state is still a product of two single-atom superpositions (no entanglement), but importantly
it has non-zero amplitudes for all four basis states. In particular, the |ee⟩ component is present,
so the interaction Hamiltonian Ĥint = V0 |ee⟩⟨ee| will now have something to act on.

Interaction, CZ, and entanglement

After the π/2 pulses, the system is prepared in a product superposition. In this section we
compute the full time evolution under the Hamiltonian (with the laser turned off), first for the
general case δ1 ̸= δ2 , and then show how the special case δ1 = δ2 produces a clean CZ operation
and generates entanglement.

5
Hamiltonian with the laser off
With the driving turned off (Ω = 0), the effective Hamiltonian is
ℏδ1 (1) ℏδ2 (2)
Ĥ = − σ − σ + V0 |ee⟩⟨ee| ≡ ĤZ + Ĥint . (26)
2 z 2 z
Since
[ĤZ , Ĥint ] = 0,
the evolution operator factorizes:
Û (t) = e−iĤt/ℏ = e−iĤZ t/ℏ e−iĤint t/ℏ ≡ ÛZ (t) Ûint (t). (27)

Action of ÛZ (t)

The energies of the computational basis states under ĤZ are:
ℏ
Egg = (δ1 + δ2 ), (28)
2
ℏ
Ege = (δ1 − δ2 ), (29)
2
ℏ
Eeg = (δ2 − δ1 ), (30)
2
ℏ
Eee = − (δ1 + δ2 ). (31)
2
Thus the phase evolution is:
ÛZ (t)|gg⟩ = e−i(δ1 +δ2 )t/2 |gg⟩, (32)
−i(δ1 −δ2 )t/2
ÛZ (t)|ge⟩ = e |ge⟩, (33)
ÛZ (t)|eg⟩ = e−i(δ2 −δ1 )t/2 |eg⟩, (34)
+i(δ1 +δ2 )t/2
ÛZ (t)|ee⟩ = e |ee⟩. (35)

Action of the interaction evolution Ûint (t)

The interaction Hamiltonian is a projector:
Ĥint = V0 |ee⟩⟨ee|. (36)

Expanding the exponential,

A2
eA = I + A + + ··· , (37)
2!
with
V0 t
A = −i |ee⟩⟨ee|, (38)
ℏ
gives  n
V0 t
An = −i |ee⟩⟨ee|. (39)
ℏ
Factoring out the projector,
 
Ûint (t) = I + e−iV0 t/ℏ − 1 |ee⟩⟨ee|. (40)

Thus,
Ûint (t)|gg⟩ = |gg⟩, (41)
Ûint (t)|ge⟩ = |ge⟩, (42)
Ûint (t)|eg⟩ = |eg⟩, (43)
−iV0 t/ℏ
Ûint (t)|ee⟩ = e |ee⟩. (44)

6
Full evolution for δ1 ̸= δ2
Acting with Û (t) on the initial state in Eq. 28,

1 h −i(δ1 +δ2 )t/2 i

|ψ(t)⟩ = e |gg⟩ + e−i(δ1 −δ2 )t/2 |ge⟩ + e−i(δ2 −δ1 )t/2 |eg⟩ + e+i(δ1 +δ2 )t/2 e−iV0 t/ℏ |ee⟩ .
2
(45)
If we let the system evolve for the right time, i.e, until |ee⟩ gets a minus sign; and others stay
the same, we get:
πℏ
e−iV0 t/ℏ = −1 =⇒ t = , (46)
V0
we obtain
1 h −i(δ1 +δ2 )t/2 i
|ψ(t)⟩ = e |gg⟩ + e−i(δ1 −δ2 )t/2 |ge⟩ + e−i(δ2 −δ1 )t/2 |eg⟩ − e+i(δ1 +δ2 )t/2 |ee⟩ . (47)
2

This shows explicitly that unequal detunings δ1 ̸= δ2 introduce different phases on each compo-
nent, spoiling the clean CZ structure.

Special case δ1 = δ2
Setting δ1 = δ2 = δ = 0 (so that what’s left is the pure interaction evolution, which is what the
question asks us to prove), Eq. 48 reduces to
1
|gg⟩ + |ge⟩ + |eg⟩ + e−iV0 t/ℏ |ee⟩ ,


|ψ(t)⟩ = (48)
2

At t = πℏ/V0 ,
1

|ψfinal ⟩ = |gg⟩ + |ge⟩ + |eg⟩ − |ee⟩ . (49)
2
The corresponding unitary is
 
πℏ
Ûint V0 = diag(1, 1, 1, −1), (50)

which is exactly the controlled-Z gate.

Since the final state (Eq. (49)) cannot be written as a tensor product of single atom states, the
interaction has generated entanglement!

7
(c) Realizing a CNOT gate for ∆1 = ∆2 and generating entangled states
When the drive is switched off (Ω = 0) and both atoms have the same detuning ∆1 = ∆2 ≡ ∆,
the rotating–frame Hamiltonian in the computational basis {|gg⟩ , |ge⟩ , |eg⟩ , |ee⟩} is
 
0 0 0 0
0 ℏ∆ 0 0 
H= 0 0 ℏ∆
. (51)
0 
0 0 0 2ℏ∆ + V0
It is convenient to split this into a local (single–qubit) contribution and a purely two–qubit
interaction,
H = Hloc + Hint , (52)
with    
0 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0
Hloc = ℏ∆ 
0
, Hint = V0  . (53)
0 1 0 0 0 0 0
0 0 0 2 0 0 0 1
In operator form,
Hloc = ℏ∆ (|e1 ⟩ ⟨e1 | + |e2 ⟩ ⟨e2 |) , (54)
Hint = V0 |e1 e2 ⟩ ⟨e1 e2 | . (55)
Both Hloc and Hint are diagonal in the same basis, so they commute,
[Hloc , Hint ] = 0. (56)

CZ gate in the interaction picture

We now move to the interaction picture with respect to Hloc . The interaction–picture evolution
operator is
 
  1 0 0 0
V 0 t 0 1 0 0
UI (t) = e−iHint t/ℏ = exp −i
 
|ee⟩ ⟨ee| = 
0 0
. (57)
ℏ 1 0 
−iV0 t/ℏ
0 0 0 e
Thus, the states |gg⟩, |ge⟩ and |eg⟩ are unchanged, while |ee⟩ acquires an additional phase.
Choosing the interaction time
πℏ
tCZ = (58)
V0
gives  
1 0 0 0
0 1 0 0 
UI (tCZ ) = 
0 0 1 0  = CZ.
 (59)
0 0 0 −1
In other words, for equal detuning the interaction produces a clean controlled–phase gate that
flips the sign of |ee⟩ and leaves the other basis states unchanged. Any additional phases from
Hloc are local Z rotations on the individual qubits and can be corrected by single–qubit pulses.

CNOT from CZ

A CNOT gate with qubit 1 as control and qubit 2 as target is related to CZ by

CNOT1→2 = (I ⊗ H) CZ (I ⊗ H), (60)
where  
1 1 1
H=√ (61)
2 1 −1
is the single–qubit Hadamard on the target qubit. Experimentally, this sequence corresponds to

8
(a) Apply a π/2 pulse (Hadamard) to qubit 2;
(b) Turn off the laser and let the system evolve freely for tCZ = πℏ/V0 (CZ gate);
(c) Apply a second π/2 pulse to qubit 2.

Entanglement from CNOT: one Bell state

To show that this interaction can generate entanglement, we start from the separable ground
state
|ψ0 ⟩ = |gg⟩ . (62)
First, we create a superposition on the control qubit by applying a Hadamard to qubit 1 (a π/2
pulse with appropriate phase):
1

|ψ1 ⟩ = (H ⊗ I) |gg⟩ = √ |gg⟩ + |eg⟩ . (63)
2
Next, we apply CNOT1→2 :

CNOT1→2 |gg⟩ = |gg⟩ , (64)

CNOT1→2 |eg⟩ = |ee⟩ , (65)

so that
1
|ψ2 ⟩ = CNOT1→2 |ψ1 ⟩ = √ |gg⟩ + |ee⟩ ≡ |Φ+ ⟩ ,


(66)
2
which is one of the four maximally entangled Bell states. The reduced density matrix of either
qubit is maximally mixed,

1
ρ1 = Tr2 |Φ+ ⟩ ⟨Φ+ | = (|g⟩ ⟨g| + |e⟩ ⟨e|) , (67)
2
confirming that |Φ+ ⟩ is entangled.

Full sequence and generation of all four Bell states

The total unitary corresponding to the complete pulse sequence discussed above is

Useq = (I ⊗ H) CZ (I ⊗ H) (H ⊗ I) = CNOT1→2 (H ⊗ I), (68)

i.e. a Hadamard on the control followed by CNOT. Acting on the four computational basis states,
we obtain
1
Useq |gg⟩ = √ |gg⟩ + |ee⟩ = |Φ+ ⟩ ,


(69)
2
1
Useq |ge⟩ = √ |ge⟩ + |eg⟩ = |Ψ+ ⟩ ,


(70)
2
1
Useq |eg⟩ = √ |gg⟩ − |ee⟩ = |Φ− ⟩ ,


(71)
2
1
Useq |ee⟩ = √ |ge⟩ − |eg⟩ = |Ψ− ⟩ .


(72)
2
Thus, by starting from any of the four computational basis states and applying the same sequence

(i) H on qubit 1 → (ii) H on qubit 2 → (iii) CZ interaction → (iv) H on qubit 2,

we obtain all four Bell states {|Φ± ⟩ , |Ψ± ⟩}. This makes explicit how, for equal detuning ∆1 =
∆2 , the interaction V0 |ee⟩ ⟨ee| together with single–qubit rotations realizes a CNOT gate and
allows the preparation of maximally entangled two–qubit states.

9
Part (d): Effect of Inhomogeneous Broadening and Dynam-
ical Decoupling
Inhomogeneous Broadening and Phase Errors
During the entangling period, when the laser drive is turned off, the two-qubit system evolves
under the rotating-frame Hamiltonian
H = ∆1 Z1 + ∆2 Z2 + V0 |ee⟩ ⟨ee| . (73)
Using the identity |e⟩ ⟨e| = (I + Z)/2, the interaction term becomes
1
|ee⟩ ⟨ee| =
(I + Z1 + Z2 + Z1 Z2 ) , (74)
4
showing that the Hamiltonian contains detuning terms and the entangling term Z1 Z2 .
If ∆1 = ∆2 = 0, evolution for a time
πℏ
TCZ = (75)
V0
implements the ideal CZ gate,
CZ = diag(1, 1, 1, −1).
However, if ∆1 ̸= ∆2 , the state |ee⟩ acquires an unwanted phase e−i(∆1 +∆2 )t . Thus the prepared
Bell state
|gg⟩ + |ee⟩
|Φ+ ⟩ = √
2
evolves into
1  
|ψ(t)⟩ = √ |gg⟩ + e−i(∆1 +∆2 )t |ee⟩ . (76)
2
Averaging over detuning noise yields the fidelity
1h D Ei
F = 1 + e−i(∆1 +∆2 )t . (77)
2

Connection to Ramsey Interferometry

A single qubit in
|g⟩ + |e⟩
√
2
−i∆t
accumulates a phase e , giving the Ramsey fringe
1

Pe (t) = 1 + cos(∆t) . (78)
2
If ∆ fluctuates across the ensemble, the contrast decays. The same dephasing mechanism occurs
for the |gg⟩ ↔ |ee⟩ coherence.

Spin Echo for Two Qubits

A global π pulse about x satisfies
XZX = −Z,
so X ⊗ X flips both Z1 and Z2 , while
(X ⊗ X)Z1 Z2 (X ⊗ X) = Z1 Z2 .

Thus the echo-protected entangling gate is

(echo) πℏ
UCZ = e−iHT /(2ℏ) (X ⊗ X)e−iHT /(2ℏ) , T = . (79)
V0
Detuning effects cancel; the entangling term accumulates.

10
DD-Protected Entanglement Protocol Starting from |gg⟩
We summarize the full CNOT-based entanglement protocol, showing where the DD-protected
CZ block is inserted.

1. Prepare a superposition on qubit 1:

H1 1
|gg⟩ −−→ |ψ1 ⟩ = √ (|gg⟩ + |eg⟩) .
2

2. Begin constructing CNOT: apply Hadamard on qubit 2

H 2 1
|ψ1 ⟩ −−→ |ψin ⟩ = (|gg⟩ + |ge⟩ + |eg⟩ + |ee⟩) .
2

3. Apply CZ. (a) Ideal case, no detuning):

CZ
|ψin ⟩ −−→ · · ·

(b) With dynamical decoupling):

(DD)
U
−−→ e−iHT /(2ℏ) (X ⊗ X) e−iHT /(2ℏ) |ψin ⟩ .
|ψin ⟩ −−CZ

4. Finish CNOT: apply Hadamard on qubit 2 again

H 2 |gg⟩ + |ee⟩
· · · −−→ √ ,
2
which is the desired Bell state.

Timeline Diagram of the DD-Protected CZ Sequence

DD-protected CZ (total time T = πℏ/V0 )

q1 H X

e−iHT /2 e−iHT /2 |gg⟩ + |ee⟩

|gg⟩ √ time
2

q2 H X H

Explanation. The first Hadamard prepares a superposition on qubit q1 , and the Hadamard
on q2 implements the standard CNOT construction. During the evolution segments of duration
T /2, the laser is turned off, so the system evolves only under detuning and the interaction term
V0 |ee⟩ ⟨ee|. A global π pulse reverses the sign of the detuning contributions while preserving the
entangling term, thereby cancelling dephasing from inhomogeneous broadening.

Summary
Inhomogeneous broadening produces random phases that degrade the Bell-state coherence. Ram-
sey interferometry illustrates this dephasing, and two-qubit spin echo (or more general dynamical
decoupling) removes unwanted single-qubit phases while preserving the entangling interaction.

11