You are on page 1of 18

Quantum Engineering - Eksamen

Sebastian Yde Madsen


June 6, 2023

Maybe set up figure refs.

1 Explain the properties of Fock and coherent states and their implications
for beamsplitters and interferometers.
• Formally, we define Fock states as occupation number states of multiple identical particles. These are
elements of the Fock space F. For Bosonic (integer spin particles - like photons) we define the Fock space as the
symmetrized (and normalized) direct sum over tensor product spaces of n single-Boson Hilbert-spaces:
s P  
∞ X O k
M ni ! 1
F≡ Q i P  P̂α H (1)
k=0 i (ni !) i ni ! α m=1
| {z }
Ŝ+

allowing for a variable number of particles in a given state - i.e. - Second Quantization. Specifically, we call the
following occupation number representation:
(
{0, 1, 2, 3, . . .}, for Bosons
|n1 , n2 , . . .⟩ , ni ∈ (2)
{0, 1}, for Fermions

a Fock state, with ni denoting the number of particles in the ni ’th single particle basis, present in the state.

In Second Quantization, these states are commonly defined through action of the Bosonic creation operator
on the vacuum state:
N different basis states ni   ni 
â†i â†i
z }| { N
Y YN
|n1 , n2 , . . .⟩ = √ |0, 0, . . .⟩ ≡ √ |0⟩ (3)
i=1
ni ! i=1
ni !

1
Specifically the creation (annihilation) operator obey the bosonic commutation relations [â†i , â†j ] = [âi , âj ] = 0
and [âi , â†j ] = δij , and we can define the occupation number operator n̂i ≡ a†i ai , that counts the number of
particles in basis state ni 2 :
n̂i |n1 , n2 , . . . ni , . . .⟩ = ni |n1 , n2 , . . . ni , . . .⟩ (4)
Now, work with single-mode Fock states |n⟩. By quantizing the EM-field - one can even demonstrate vacuum-
fluctuations as non-zero variance of E-field even for vacuum-state:
2
(∆Eˆx (z, t))2 n
= ⟨n| Eˆx (z, t)2 |n⟩ − ⟨n| Eˆx (z, t) |n⟩ = ε20 sin(kz)2 (2n + 1) − |{z}
0 (5)
zero mean field

which is also evident from [n̂, Eˆx (z, t)] ∝ â − ↠̸= 0.3
• Now, lets Introduce the Coherent state - eigenstate of the annihilation operator:

â |α⟩ = α |α⟩ (6)

1 Note that in general we define Âi ≡ 1⊗(i−1) ⊗ Â ⊗ 1 ⊗ . . .


2 Can be seen by utilizing the annihilation and creation operators - corresponds to resulting coefficient.
3 Think generalized uncertainty relation: ⟨(∆Â)2 ⟩⟨(∆B̂)2 ⟩ ≥ 1 ⟨[Â, B̂]⟩ 2

2i

1
As occupation number operator is hermitian - it’s eigenstates (the Fock states) constitute a complete basis - we can
perform the expansion:
X∞
|α⟩ = cn |n⟩ (7)
n=0

and from normalization req. and action of â, specifically:



1
X αn
|α⟩ = e− 2 |α| √ |n⟩ (8)
n=0 n!

These states are ”special” in the sense that they are the ”most classical-like states”. Specifically we observe:
– Classical-like expectation of E-field: ⟨Eˆx (z, t)⟩α = ⟨α| Eˆx (z, t) |α⟩ ∝ |α|sin(ωt − k · r − θ) - note |α| is related to
amplitude.

Figure 1: Vertical width is (∆Eˆx (z, t))2 α

– Given avg. photon number as n̄ = ⟨α| n̂ |α⟩ = |α|2 , the fractional uncertainty in photon number decreases, when
avg. photon number increases:
p
∆n ⟨n̂2 ⟩α − ⟨n̂⟩2α 1
= = √ (Decreases for increasing n̄) (9)
n̄ n̄ n̄

– The phase-distribution becomes increasingly localized for an increasing avg. photon number (peak of Gaussian
becomes smaller):
 12
2|α|2

1 2
(ϕ−θ)2
P(ϕ) = | ⟨ϕ|α⟩ |2 ≈ e−2|α| , (Large n̄ = |α|2 approx.) (10)
2π π

Furthermore, as the annihilation and creation operator isn’t hermitian (not observable quantities), one oftentimes
defines the hermitian quadrature operators4 :
1 1
X̂1 ≡ (â + ↠), X̂2 ≡ (â − ↠) (11)
2 2i
In fact - one can equivalently define the coherent states as the ones minimizing the uncertainty of these quadrature
operators, s.t.:
1
(∆X̂1 )2 α = (∆X̂2 )2 α = (12)
4
and we might picture the coherent state in the ”phase space” formed by these hermitian operators as:

4 These correspond to a dimensionless scaling of canonical position and canonical momentum - they obey [X̂1 , X̂2 ] = i/2.

2
• Now, in few photon limit we need quantum mechanical treatment of (50:50) beamsplitter (BS). Specifically, just
thinking of the setup as:

with â2 = tâ1 and â3 = râ1 is in contradiction to commutation relations. We need to include extra port for vacuum
state, s.t.:

Specifically we connect the input and output operators by a unitary transformation:


    h π
â2 â0 i
= U† U, U ≡ exp i (â†0 â1 + â0 â†1 ) (13)
â3 â1 4

now considering single-photon input-state â†1 |0⟩0 |0⟩1 = |0⟩0 |1⟩1 , and utilizing the transformation, yielding:

1
â†1 = √ (iâ†2 + â†3 ) (14)
2
we observe action of BS:
BS 1
|0⟩0 |0⟩1 −−→ √ (i |1⟩2 |0⟩3 + |0⟩2 |1⟩3 ) (15)
2
π
and recover expected result → reflected part phase-shift by 2, but equal prob. (Note - is entangled state5 )

• Now, if we instead send in the classical-like coherent state D̂(α)1 |0⟩0 |0⟩1 = |0⟩0 |α⟩1 , and utilize the transformation to
express the annihilation and creation operators in the Displacement operator, we see the action of the beamsplitter
is:  
BS iα α
|0⟩0 |α⟩1 −−→ √ √ (16)
2 2 2 3
yields more classical result - no entanglement and equal amplitude6 - will always measure something in both ”arms”.
(also still reflected part phase-shifted by π2 )

• Finally we might consider sending in 2 photons, i.e. â†1 â†0 |0⟩0 |0⟩1 = |1⟩0 |1⟩1 , from which we observe:

BS i
|0⟩0 |0⟩1 −−→ √ (|2⟩2 |0⟩3 + |0⟩2 |2⟩3 ) (17)
2
rather peculiar - both photons are always observed in the same arm (with equal prob). Can be explained by destructive
interference (wave-property). Specifically Recall Feynman’s rule for obtaining the probability for an outcome that can
occur by several indistinguishable processes: add the probability amplitudes of all the processes and then calculate
the square of the modulus:
2
1 1 i i
P[|1⟩0 |1⟩1 −BS
−→ |1⟩2 |1⟩3 ] = |AT · AT + AR · AR |2 = √ √ + √ √ =0 (18)
2 2 2 2

Add interferometry example

5 Define entangled state here


6 Makes sense with equal amplitude - coherent state is superposition of infinitely many single-mode Fock states

3
2 Explain what squeezed states of light are and how they can be produced
and measured
• As the annihilation and creation operators are not hermitian - not observable - one oftentimes defines the hermitian
quadrature operators7 :
1 1 i
X̂1 ≡ (â + ↠), X̂2 ≡ (â − ↠), with [X̂1 , X̂2 ] = (19)
2 2i 2
and from the generalized uncertainty relation8 we can can view the coherent states |α⟩ as the ones minimizing the
uncertainty of these quadrature operators, s.t.:
1
(∆X̂1 )2 α
= (∆X̂2 )2 α
= (20)
4
However, this doesn’t inherently prevent the uncertainty being smaller in one of the quadrature operators (as long as
the other one is correspondingly larger), specifically, consider:

Figure 2: Squeezed states corresponds to shaded area.The solid-line border is a hyperbola determined from the equalized uncertainty
relation.

As such we define the squeezed states as the ones where:


1 1
(∆X̂1 )2 α
< or (∆X̂2 )2 α
< (21)
4 4

• Mathematically, we generate the squeezed states using the Squeeze operator:


h1 i
Ŝ(ξ) ≡ exp (ξ ∗ â2 − ξ(↠)2 ) , ξ ∈ C s.t. ξ = reiθ (22)
2
with 0 ≤ r ≤ ∞ being squeeze parameter and 0 ≤ θ ≤ 2π. Think of this operator as two-photon generalization of
the displacement operator - the operator acting on the vacuum would create some sort of “two-photon coherent state”
as it is clear that photons will be created or destroyed in pairs by the action of this operator.
Moreover, considering the expansion of the squeezed vacuum state in Fock states, one can observe that:
2m
tanh(r)
P[Odd nr. photons] = | ⟨2m + 1|ξ⟩ |2 = 0, P[Even nr. photons] = | ⟨2m|ξ⟩ |2 ∝ (23)
cosh(r)

only an even number of photons may be exist - squeezing creates an even number of photons on the vacuum state.

• Now, defining squeezed vacuum state |ξ⟩ ≡ Ŝ(ξ) |0⟩, we see that for θ = 0, the squeezing is in X̂1 , as:
1 −2r 1 2r
(∆X̂1 )2 ξ
= e , (∆X̂2 )2 ξ
= e (24)
2 2
which can be seen in the phase-space, as:

7 see footnote 4
8 see footnote 3

4
Figure 3: θ = 0 → squeezing in canonical position.

and squeezing along an arbitrary axis in phase-space plane might be achieved by considering the rotated quadratures:
    
Ŷ1 cos(θ/2) −sin(θ/2) X̂1
= (25)
Ŷ2 sin(θ/2) cos(θ/2) X̂2
corresponding to:

• Furthermore, one might both displace and squeeze the vacuum state, i.e.:
|α, ξ⟩ ≡ D̂(α)Ŝ(ξ) |0⟩ (26)
which (for θ = 0 → X̂1 squeezing) in phase space might look like:

An interesting fact about these states are that depending on whether we squeeze in X̂1 or X̂2 , we either create:
– An amplitude-squeezed state, where the field’s intensity is known with greater precision (below vacuum
fluctuations), but the phase is less precise OR
– A phase-squeezed state, where the field’s phase is known with greater precision (below vacuum fluctuations),
but the intensity is less precise.
see below:

Figure 4: (a): Phase-squeezed state, (b): Amplitude-squeezed state.

5
• In practice, one can utilize Parametric Down Conversion to generate squeezed states. This is a process where a
high-energy pump beam/field ωp are directed into a non-linear medium (here of second-order nonlinear susceptibility
χ2 ) such that some photons of that field are converted into pairs of photons, s.t. ωp = ωi + ωs 9 :

Specifically for producing squeezed states we consider Degenerate Parametric Down Conversion, where the
created pair of photons are identical, s.t. ω ≡ ωs = ωi .
In this setup (assuming single-mode) the hamiltonian of the system is given by:

H ≡ ℏωp b̂† b̂ + ℏω ↠â} + iℏχ2 (â2 b̂† − (↠)2 b̂)


| {z (27)
| {z } | {z }
Pump field Signal field Interaction (squeezing)

Now, utilizing the Parametric approximation10 , we find that in the interaction picture, this hamiltonian becomes:

HI ≡ iℏ(η ∗ â2 − η(â)2 ), η = βχ2 (28)

which in turn means that the associated time-development operator has the form of the squeeze operator, i.e.:
h i h i
Ui (t, t0 = 0) = exp − iHI t/ℏ = exp (η ∗ â2 − η(â)2 )t (29)

i.e. - time-evolving a state with this system corresponds to squeezing the state! Can be tuned by changing t - time
inside crystal corresponding to length of crystal.
Include how to measure squeezed light - balanced homodyne detection 7.3 in book

3 Discuss the optical Bloch equations and the Bloch sphere representation
of a two-level atom
• Starting w. position repr. of electron bound to atom (no external fields):

P̂ 2
H= + V (r), V (r) = Coulomb potential (30)
2m
Now taking external field into account + working in Coulomb Gauge (∇ · A = 0), and neglecting rel. effects:
1  2
H(r, t) = P̂ + eA(r, t) + V (r) (31)
2m
where A(r, t) is determined by solving wave equation11 .
Now, utilizing dipole approximation12 , we assume spatially uniform vector potential A(r, t) ∼ A(t), and finally
get:
P̂ 2
H= + V (r) + er · E(t) (32)
2m | {z }
Interaction w. field

now defining dipole moment d = −er, we arrive at final form:

H = H0 − d · E(t) (33)
| {z }
Perturbation

• Considering a classical (not quantized) sinusoidal field:


(
E0 cos(ωt), t ≥ 0
E(t) = (34)
0 ,t < 0

9 energy conservation
10 Assume that the pump field is in a strong coherent classical field, which is strong enough to remain undepleted of photons over the relevant
time scale.
11 The wave equation for the vector potential is ∇2 A = 1 ∂ 2 A
c2 ∂t2
12 For ||r|| of typical atomic dims. (a few Ångströms), and typical optical λ (few hundred nm), we have k · r << 1

6
and assuming that the electron starts out in an energy eigenstate, s.t. H0 |i⟩ = Ei |i⟩ and by expanding a general
state ket in this basis, utilizing normalization and inserting in SE, yields system of coupled 1. order diffs (for time
dependent. evolution coeffs):
d X
iℏ cn (t) = cm (t) ⟨m| (−d · E(t)) |i⟩ eiωnm t (35)
dt m

which we usually handle by performing power-series expansion of the coeffs. in the perturbation strength (typically to
second order) cn (t) = cn (t)(0) + λcn (t)(1) + λ2 cn (t)(2) + . . . . By inserting in system of diffs above, we get a recursive
relation with c(n) being dependent on c(n−1) , and under weak-field assumption13 this can be solved and we can define:
h i
P |i⟩ → |f ⟩ (t) ≈ |cf (t)(0) + cf (t)(1) + cf (t)(2) + . . . |2 (36)

and in case the field has frequency close to resonance of 2-level system14 , we can make Rotating Wave Approxi-
mation and obtain Fermi’s Golden Rule (in steady state limit t → ∞).
• Now, in the case that we have a strong field (still w. small detuning), making the weak-field approximation inappro-
priate, we turn to the Rabi Model. Here we again consider the hamiltonian on the form (33) with field on form (34),
but we instead denote ground state and excited state of 2 level system as |g⟩ and |e⟩, and consider a state of the form:

|ψ(t)⟩ = cg (t)e−iEg t/ℏ |g⟩ + ce (t)e−iEe t/ℏ |e⟩ (37)

utilizing this in SE and solving the resulting set of 2 coupled diffs (still 1. order diffs), we find probability of exciting
to |e⟩, as:
h i Ω2
P |g⟩ → |e⟩ (t) = |ce (t)|2 = 2 sin2 (ΩR t/2), Ω = − ⟨e| − d̂ · E0 |g⟩ /ℏ (38)
ΩR

where ΩR is the Rabi frequency, defined as ∆2 + Ω2 and ∆ is the detuning, e.g. consider:

i.e. the excitation probability oscillates. For no detuning, where ∆ = 0, we define the π-pulse as one that sends all to
excited state, and π/2-pulse as one that sends to coherent superposition.
• Now, all the dynamics of the 2-level system may also be graphically represented. Specifically, denoting a general
coherent superposition as:
|ψ⟩ = cg |g⟩ + ce |e⟩ (39)
we consider the density operator formalism:

|cg |2 cg ce ∗
   
ρ11 ρ12
|ψ⟩ ⟨ψ| ≡ ρ = = (40)
cg ∗ ce |ce |2 ρ21 ρ22

and we can actually find the diff eqs. describing the dynamics of the 2-level system interacting with the external field
(if we disregard spontaneous emission) using Von Neumann Equation:

d
iℏ ρ = [H, ρ] (41)
dt

13 c (0) = 1, c (0) = 0 s.t. for t > 0 ci (t) ≈ 1, cf (t) << 1 - very little excitation.
i f
14 We mean a small detuning, i.e. ∆ = ω − ωf i is small

7
Specifically. Denoting the rotating frame coords. as:
ρ̃11 = ρ11 , ρ̃22 = ρ22 , ρ̃12 = ρ12 ei∆t , ρ̃21 = ρ21 e−i∆t (42)
we define the Bloch Vector as:

u = ρ̃12 + ρ̃21 = 2Re(ρ̃12 )

R ≡ uêx + vêy + wêz , v = i(ρ̃21 − ρ̃21 ) = 2Im(ρ̃12 ) (43)

w = ρ̃22 − ρ̃11 , (Population inversion.)

as ||R||2 = 1 → corresponds to point on unit-sphere:

 
and by considering expectation value of dipole operator ⟨d̂⟩ = Tr ρd̂ we see that we can interpret:
– u: atomic dipole component in phase with the field.
– v: atomic dipole component in quadrature with the field.

Find out how this makes sense.


Finally - the diff eqs. for the rotating frame coords, can be expressed in terms of the coordinates of the Bloch Vector
- we call these the Bloch Equations:

u̇ = −∆v

v̇ = ∆u − Ωw ⇐⇒ Ṙ = R × W , W = Ωêx + ∆êz (44)

ẇ = Ωv

Note that as Ṙ · W = 015 , we know that R · W = RW cos(θ) is a constant which means that for excitation with a fixed
Rabi frequency and detuning the magnitude W is constant, and since R is also fixed, the Bloch vector moves around
a cone around W with θ constant - however if detuning is exactly 0 (not just constant) - s.t. W is exactly along êx
axis - the R vector rotates in a circle about êx axis (in êy , êz plane) between w = −1 (corresponding to being in |e⟩ as
ρ22 = 1) and w = 1 (corresponding to being in |g⟩ as ρ11 = 1). We can view the 0 detuning case as Rabi oscillations
- and the π and π/2 pulses therefore have a visually interpretation in this case - See figure below, (N.B. x and y axis
are rotated 180 degrees compared to fig. above):

• Now - include spontaneous emission by introduce dampening term dt


d
ρ22 = −Γρ22 and get Optical Bloch equa-
tions:  
u/2
Ṙ = R × W − Γ  v/2  (45)
w+1

Interpret dampening term - Bloch vector rotates and shrinks

Discuss steady state solution

15 The cross product of 2 vectors produces one that is perpendicular to both of the vectors (and all vectors in their plane)

8
4 Discuss the significance of the Jaynes-Cummings model and its applica-
tions in experiments with Rydberg atoms in microwave cavities or with
trapped ions
• We now consider the fully QM treatment of interaction w. EM-field and atomic 2-level system. The Jaynes–Cummings
model (JCM) is a theoretical model in quantum optics. It describes the system of a two-level atom interacting with
a single quantized mode of an optical cavity, with or without the presence of light.
• We consider an atom, with levels |g⟩ and |e⟩, interacting with a (quantized) single-mode cavity field of the form:
 1/2
ℏω
Ê = e (â + ↠)sin(kz), e = polarization vector. (46)
ε0 V
Now, defining raising/lowering operators as â+ ≡ |e⟩ ⟨g| , â− ≡ |g⟩ ⟨e|, and the inversion operator as σ̂3 ≡ |g⟩ ⟨g| +
|e⟩ ⟨e|16 , setting the zero-point half way between |g⟩ and |e⟩:

and disregarding the constant off-set term in the field Hamiltonian + utilizing Rotating Wave Approximation
yields a total system hamiltonian of the form:
1
H= ℏω0 σ̂3 + ℏω ↠â} + ℏλ(σ̂+ â + σ̂i ↠) = ĤA + ĤF + ĤI (47)
2 | {z | {z }
| {z } Field Interaction
Atom

(N.B. λ = dg/ℏ)
• Now, the case with ∆ = ω0 − ω = 0 with system starting in state |i⟩ ≡ |e⟩ ⊗ |n⟩ (s.t. |f ⟩ ≡ |g⟩ ⊗ |n + 1⟩) (N.B. field
part is non-classical Fock state) is called the Resonant JMC - and we consider the state-ket:
|ψ(t)⟩ = ci (t) |i⟩ + cf (t) |f ⟩ (48)
Now, by inserting the interaction part of the Hamiltonian in SE, with init. conds. ci (0) = 1, cf (0) = 0, one obtains:
√ √
ci (t) = cos(λt n + 1) =⇒ P |i⟩ = |ci (t)|2 = cos2 (λt n + 1)
 
(49)
17
which is non-zero even for n = 0 → even in vacuum we observe oscillations between ground and excited state - Rabi
Vacuum Oscillations - the result of the atom spontaneously emitting a photon then re-absorbing it, re-emitting it,
etc.: an example of reversible spontaneous emission:

Figure 5: Here the plots is over T = λt

16 Resembles pauli-x
17 Can

find cf (t) = −isin(λt n + 1) in similar way

9
• Now, lets consider case where field part is initially in the (”most classical”) coherent state, s.t.:

2 X αn
|ψ(0)⟩Field = e−|α| /2
√ |n⟩ , |ψ(0)⟩Atom = cg |g⟩ + ce |e⟩ ⇒ |ψ(0)⟩ = |ψ(0)⟩Atom ⊗ |ψ(0)⟩Field (50)
n=0 n!

and for ce = 1, cg = 0, solving SE, one finds:


( P∞ √
|ψg (t)⟩ = n=0 −icn sin(λt n + 1) |n + 1⟩
|ψ(t)⟩ = |ψg (t)⟩ ⊗ |g⟩ + |ψe (t)⟩ ⊗ |e⟩ , with P∞ √ (51)
|ψe (t)⟩ = n=0 cn cos(λt n + 1) |n⟩

and it actually turns out the the Rabi oscillations here initially gets damped, and then after some time ”revives”
however with smaller amplitude - specifically, consider:

Figure 6: Here the plot is over T = λt

We contribute √this phenomenon


√ to the fact that we have a dominant frequency around the average photon number,
i.e. Q(N̄ ) = λ n̄ + 1 ≈ λ n̄, n̄ >> 1, which gets ”de-phased” with the other (slightly dominating) frequences in
the range n̄ ± ∆n, and by utilizing ”time-frequency uncertainty relation” one gets a characteristic collapse time of:
1
tc ∼ (52)
λ

Write about dressed states
Write about application in experiments w. rydberg atoms in microwaye cavities or with trapped atoms

5 Discuss the basics of laser cooling for ions, neutral atoms or mechanical
resonators
• Lets consider the laser cooling of ions in a Linear Paul trap. From Earnshaw’s Theorem it is given that spatial
confinement in 3D is not possible via static fields - need dynamic fields. The Linear Paul trap confines ions on a string,
by applying a DC-field along the string-axis (z-axis) and an AC field in the radial directions (x- and y-axis; an AC
Quadrupole field):

Now, based on the following assumptions:


– Approximation of the trap’s shape: To obtain a true quadrupole field - surface of the electrodes should be
hyperbolic - however in practice - much easier to construct a trap with cylindrical electrodes as seen above -
resulting potential will only deviate very little if dimensions are chosen correctly.

10
– Small oscillations: The ions’ oscillations in the trap are small, that is, their motion can be approximated as
harmonic.
– Quasi-static approximation: The frequency of the oscillating electric field ω is assumed to be significantly
higher than the ion’s natural oscillation frequency ωz . This allows the ion to effectively ”see” a time-averaged
potential, also known as the pseudo-potential.
The ions experience an effective pseudo-potential of the form of harmonic oscillator:
1 1
Φ(x, y, z) = mωz2 z 2 + mωr (x2 + y 2 ) (53)
2 2
q
1 q2
− a2 ω.
p
with the radial frequency ωr = 2 a+ 2 ω and axial/natural frequency ωz =

• Now generally as T ∝ v 2 - we cool ions by slowing them down - specifically cooling in linear Paul trap is done trough
a combination of Doppler cooling and Resolved side-band cooling.
• Initially we consider case with:
ωw << Γ (54)
i.e. When the ion’s oscillation frequency in the trap ωz is smaller than the decay rate Γ of the transition used for
cooling - ion behaves as a free particle18 with a changing Doppler shift.
Here we cool w. Doppler cooling - relies on Doppler effect:

±
 kObs vObs   kz vz 
ωObs = 1± ωSrc =⇒ ω± = 1 ± ωL (55)
c c
specifically - red-tune laser s.t. ion moving towards, due to Doppler effect - photons (w. p = ℏk) gets absorbed and
slows ion in that direction:

- when photon is re-emitted via. spontaneous emission gives momentum kick to ion. Re-emission in random
isotropic direction yields avg. no momentum over many cycles:

Now even though ⟨p⟩ = 0, each kick moves ion in random dir - gives random walk - motional ground state not reached
- Doppler cooling limit:
ℏΓ
TDoppler = (56)
2kb
• To cool further we consider case w. :
ωz >> Γ (57)
i.e. Resolved Side-band cooling the decay rate of the transition is smaller than the separation of the mo-
tional/vibrational levels, and distinct side-bands appear at integer-multiples ω ± nωz (”the side-bands are resolved”
or. ”we can resolve phonon modes”):

Figure 7: Left is red-sideband and right is blue-sideband

18 The timescale of the oscillation is much smaller than the timescale at which absorption and re-emission cycles occur.

11
and we can tune the laser s.t. the energy of the absorbed photons is less than the energy of spontaneously emitted
photons, s.t. the energy of the ion will be reduced.
• QM description by considering single ion in trap as (tensor) product state between its internal energy states |e⟩ , |g⟩
and its external vibrational/motional state:

and we get hamiltonian on the form:  


H = H0 + dE(ẑ(t), t)σ̂− + H.c (58)
where E(ẑ(t), t) is a classical electrical field, and we instead quantize the COM vibrational modes of the ion as:

ẑ ∝ (âe−iωz t + H.c) (59)

s.t.:
1 1  1 †
H= ℏω0 σ̂z + ℏωz ↠â + + dE0 (eiω0 t e−iη(â +â) + H.c.) (60)
|2 {z } |2 {z 2} | {z
External field
}
Internal Atom Harmonic oscillator trap

and now by assuming to be in so-called Lamb-Dicke regime (η << 1), where the vibrational amplitude of the ion is
much smaller than the wavelength of the laser light field, and going into the rotating (interaction picture) frame and
performing the RWA, and letting the the laser frequency be red-detuned by ωz (i.e. ω = ω0 − ωz ), we then get for
the 1-st red sideband:
dE0
H ∝ ℏηΩ â+ â + σ̂− ↠, Ω ≡
 
(Rabi oscillation) (61)

which drives transitions on the form:
|g⟩ |n⟩ → |e⟩ |n − 1⟩ (62)
which then - via. spontaneous emission will perform transition:

|e⟩ |n − 1⟩ → |g⟩ |n − 1⟩ (63)

leaving the atom in the ground state of a lower vibrational mode - consider:

6 Describe the basic trapping techniques for ions and neutral atoms
7 Describe the treatment of open quantum systems by the quantum master
equation and stochastic wave functions
8 Discuss quantum cryptography as realized by the BB84 protocol
• Due, to quantum properties it is possible to perform provably secure symmetric-key19 cryptography.
• Specifically, we consider the BB84 protocol, which can be described as a protocol for establishing a secure shared
key between two parties, Alice and Bob, that wishes to communicate afterwards.

19 Means that the same key is used for both the encryption and decryption processes.

12
1. Alice and Bob initially agree on a set of 2 non-orthogonal basis’20 - lets say z− and x− basis:

z-basis: {|0⟩ , |1⟩}


1 1
x-basis: {|+⟩ , |−⟩} = { √ (|0⟩ + |1⟩), √ (|0⟩ − |1⟩)}
2 2

2. Alice generates two n-length bit-strings; one denoting the bits she wants share {a1 , a2 , . . . an } and a randomly
generated one specifying which basis she is to use for encoding {b1 , b2 , . . . , bn }
3. Bob generates a random n-length bit string {b′1 , b′2 , . . . , b′n }.
4. Alice sends her sequence of bits {ai } encoded in quantum states, to Bob, via public quantum channel, as follows:
– For i = 1, 2, . . . , n

if bi = 0 - utilize z-basis: if bi = 1 - utilize x-basis:


∗ If ai = 0 prepare and send |0⟩. ∗ If ai = 0 prepare and send |−⟩.
∗ If ai = 1 prepare and send |1⟩. ∗ If ai = 1 prepare and send |+⟩.

5. Bob performs measurement on each of the incoming quantum states, as follows:


– For i = 1, 2, . . . , n

∗ If b′i = 0 measure incoming quantum state i in z-basis and store result a′i .

∗ If b′i = 1 measure incoming quantum state i in x-basis and store result a′i .
6. Bob and Alice now share and compare their randomly generated ”basis denoting” bit strings {bi } and {b′i } via a
classical public channel.
7. Finally Alice and Bob can determine where their basis choices coincided. For each quantum state, if Bob chose
the same basis as Alice to measure the qubit, then, barring noise or eavesdropping, his measurement outcome
should match the state that Alice sent.
The ”key” is then composed of the bits corresponding to these qubits — where Alice and Bob used the same
basis. Alice knows what state she sent (and thus what Bob should have measured), and Bob knows what state he
measured. Since their basis choices were the same, these states (and the corresponding bits) should be the same.
For instance, consider example:

Figure 8: 5 qubit example of BB84 protocol.

in which the shared secret key becomes 101.


• Now, lets consider the case in which a malicious third-party, Eve, tries to eavesdrop.

Due to the No-Cloning Theorem it is not possible for Eve to simply copy the quantum states, and measure
them for her self, whilst passing an undisturbed copy along to Bob.
As such, she would have to simply try to guess the basis used by Alice, measure in that basis, and pass along a similar
state to bob, encoded in the basis she used to measure.
Now, lets say she had somehow become privy to the set of non-orthogonal basis’ used by Alice and Bob - and lets
consider the probability of Bob and Alice detecting her eavesdropping.

Specifically, If Eve intercepts and measures each qubit in a randomly guessed basis, she will guess wrong 1/2 of
the time, and then when Bob measures the qubit in the correct basis, he will get a wrong result 1/2 of the time. So,

20 To ensure that they cannot be simultaneously measured - same as eigenstates of two non-commuting observables.

13
Eve introduces an error in about 1/4 of the qubits she intercepts - probability that any given qubit has been tampered
by Eve is PE = 14 . Now if Alice and Bob then compare some m-length subset of the keys ai and a′i where their
measurement bases matched (subset of the key), via a classical public channel, the probability PD that they detect
Eve’s presence (i.e., find at least one error) is equal to 1 minus the probability that they find no errors in these m
qubits.
Given the probability for a single qubit, PE , the probability that they don’t detect Eve (find no errors) in m qubits is
(1 − PE )m , and therefore, the probability that they detect Eve’s presence is:
 m  m
m 1 3
PD = 1 − PN D = 1 − (1 − PE ) = 1 − 1 − =1− (64)
4 4
25
As such, already at m = 25 they have probability 1 − 1 − 41 = 0.9992... of detecting Eve.
• Note that if Alice and Bob didn’t use non-orthogonal basis’ (i.e. eigenstates of two non-commuting observables), then
Eve could eavesdrop and measure without disturbing.

9 Discuss different experimental implementations of quantum bits and de-


scribe how gates are performed
• Commonly, DiVencenzo’s Criteria is referred to as the defining requirements for what might constitute a suitable
platform for perform quantum computing, and as such, implementing quantum bits. The criteria:
1. Scale-able system: The platform must support the implementation of an increasing number of qubits to be
use-full.
2. High Fidelity Initialization: The ability to initialize qubits in a well-defined state.
3. Long relative Decoherence times: An implementation of qubits with long decoherence times compared to
the characteristic gate-operation times.
4. Universal Gate-set: The ability to implement any universal set of gates.
5. Qubit-Specific Measurement Capability: The ability to repeatedly measure individual qubits to get output.
• Generally speaking the qubit can be considered as a 2-level system, and the pauli-z eigenstates, denoted {|0⟩ , |1⟩} are
commonly used as the computational basis.
• One possibility is Optical Quantum Computing, where one utilizes the quantum features of photons to perform
quantum computing. Specifically, single qubits states can be encoded in polarization directions, e.g.:

|0⟩ ∼ |H⟩ , |1⟩ ∼ |V ⟩ (65)

λ λ λ
And a general superposition can be created through a 4 → 2 → 4 wave-plate21 sequence:

and as such, one might in general implement single-qubit gates by means ofLinear Optical Elements, e.g. consider
the action of the Hadamard gate as seen on fig. (10), which can be realized by a half wave-plate oriented at 22.5
degrees:

21 Quarter-Wave Plate (λ/4): A quarter-wave plate is a device that alters the polarization state of a photon. It introduces a phase shift of π/2

(or a quarter of a wavelength, λ/4) between the orthogonal polarization components of the light wave

14
Generally speaking, the single-qubit gates are implemented using Linear Optical Elements.

However, to implement at complete set of gates, we need to able to entangle, and as such, the ability to implement a
2-qubit gate is required. Specifically, we will consider the CNOT gate, which can be decomposed as:

problem is that no known or foreseen material has an optical non-linearity strong enough to implement this conditional
phase shift - i.e. no material exhibits strong enough Kerr effect22 .

In practice, one can only implement a probabilistic/non-deterministic CNOT on optical platforms - however
1
in its standard form only success 16 of the time - problem as success probability for n CNOT’S:

1
(66)
16n
decrease exponentially.

However, by utilizing quantum teleportation the success probability can be boosted. Problem is that this proce-
dure required an increasing number of photons to realise - just for single CNOT one originally needed >10,000 pairs of
entangled photons to achieve a success probability of > 95%. Today there exists methods for alleviating this problem.
• Another possibility for realizing 2-level system for qubits is to trap a chain or lattice of atoms using using either ion
traps (e.g. a linear Paul trap) or MOTs for neutral atoms, respectively. Here, the 2-level system is realized by e.g.
different Zeeman sub-levels23 , or hyperfine structure levels24 .
Both atoms and ions, require very low temperatures to operate - to avoid decoherence from thermal noise of surround-
ings, and become spatially well defined.
• Lets consider the Linear Paul trap. From Earnshaw’s Theorem it is given that spatial confinement in 3D is not
possible via static fields - need dynamic fields. The Linear Paul trap confines ions on a string, by applying a DC-field
along the string-axis (z-axis) and an AC field in the radial directions (x- and y-axis; an AC Quadrupole field):

Explain single qubit gates on ion trap and Cirac-Zoller (or maybe Mølmer Sørensen gate) for CNOT

10 Describe the basic idea of quantum computing and the operation of


a specific quantum computing algorithm (Deutsch-Jozsa algorithm or
Grover)
• Basic idea of quantum computing is to utilize physical quantum mechanical properties to enable more effecient com-
putation.

22 A phenomenon in optics where the refractive index of a material changes in response to the intensity of light passing through it. This is a
type of optical non-linearity.
23 The external magnetic field couples with the magnetic moment of the electron. This interaction causes the energy levels to split, which can

be observed as a splitting of the spectral lines.


24 Results from the coupling between the magnetic moment of the nucleus and the magnetic field produced by the electrons orbiting around it.

15
• The basic unit of information in classical computation, is the bit which is discrete in the sense that it is binary and
can only be 0 or 1:
Classical bit: b = 0 ∨ b = 1 (67)
In quantum computation, the basic unit of information is the qubit, and convention dictates that the eigenstates
of the pauli-z operator is taken as computational basis. In contrary to the classical bit, the qubit |q⟩ isn’t discrete in
the sense that it, due to its QM nature, may be put in an arbitrary superposition of its basis states:
   
1 0
Quantum bit: |q⟩ ≡ α |0⟩ + β |1⟩ , with α, β ∈ C and |0⟩ = , |1⟩ = (68)
0 1

corresponding to arbitrary point on Bloch sphere:

• Specifically, in a classical computer, a bit string of length N can represent 2N different states (nr. of perms.), and
one can only perform operations on 1 of these at a time, however, on a quantum computer (QC) a tensor-product
multi-qubit-state of N qubits |q⟩N , may be in a superposition of 2N basis states (the size of the Hilbert space), and
as such, performing a single operation on the state may correspond to performing an operation on 2N basis states at
once - this is commonly dubbed Quantum Parallelism.
• On a classical computer, logical gates, such as AND, or OR are utilized to perform operations and a classical com-
puter is mostly useful if it is an Universal computer - it should implement a universal set of logical gates25 - take any
bit-string to all possible bit-strings.
Likewise, for a QC to be use-full it must also be able to take any multi-qubit state to all other multi-qubit states in
the pertaining state space. On a quantum computer operations are performed by unitary operators, and there exists
multiple sets of universal gates, though a common one is {Rx (θ), Ry (θ), Rz (θ), Phase shift, CNOT}.

Actually not quite possible to get any state as set of unitary matrices form a continuous group - however Solo-
vay–Kitaev theorem tells that one can approximate any desired unitary operation to arbitrary accuracy by using
a long enough sequence of gates from this set. Specifically a quantum circuit of m gates can be approximated to ε
precision by a quantum circuit of:
c
O(m log(m/ε) ) (69)
gates from universal set - ”polylog” in error - very good. However - decomposing an arbitrary unitary operation in
these gates is not generally solved - problematic.

• Now, lets consider an example of a quantum algorithm that outperforms best classical algorithm; Deutsche’s algo-
rithm. Consider some hidden function:
f : {0, 1} −→ {0, 1} (70)
which is guaranteed to be either constant, or balanced26 . Now, classically, determining this feature would entail having
to evaluate both f (0) and f (1). However, as demonstrated by Deutsche, this quality can be accurately determined by
a quantum computer through only one evaluation of f , by utilizing superposition and interference according to the
circuit seen in fig (9)

25
26 In this instance, f is considered balanced if f (0) ̸= f (1).

16
Figure 9: Circuit model of Deutsches algorithm

To begin with, a composite state of qubits is initialized in the zero state, followed by a bit flip, on the second qubit:

|ψ0 ⟩ = |0⟩a ⊗ |0⟩b −→ |ψ1 ⟩ = 1a ⊗ Xb |ψ0 ⟩ = |0⟩a ⊗ |1⟩b (71)

Now, by putting the two states in a positive and negative superposition of the basis states, respectively:
   
|0⟩a + |1⟩a |0⟩b − |1⟩b 1   1  
|ψ2 ⟩ = Ha ⊗ Hb |ψ1 ⟩ = √ ⊗ √ = |0⟩a ⊗ |0⟩b − |1⟩b + |1⟩a ⊗ |0⟩b − |1⟩b (72)
2 2 2 2

it is clear that the action of the unitary matrix representing f is:


1   1  
|ψ3 ⟩ = Uf |ψ2 ⟩ = |0⟩a ⊗ |0 ⊕ f (0)⟩b − |1 ⊕ f (0)⟩b + |1⟩a ⊗ |0 ⊕ f (1)⟩b − |1 ⊕ f (1)⟩b (73)
2 2
and by implementing |0 ⊕ x⟩ − |1 ⊕ x⟩ = (−1)x (|0⟩ − |1⟩):

(−1)f (0)   (−1)f (1)  


|ψ3 ⟩ = |0⟩a ⊗ |0⟩b − |1⟩b + |1⟩a ⊗ |0⟩b − |1⟩b (74)
2 2
(−1)f (0) |0⟩a + (−1)f (1) |1⟩a
   
|0⟩b − |1⟩b
= √ ⊗ √ (75)
2 2

and (−1)x⊕y = (−1)−x (−1)y :

|0⟩a + (−1)f (1)⊕f (0) |1⟩a


   
f (0) |0⟩b − |1⟩b
|ψ3 ⟩ = (−1) √ ⊗ √ (76)
2 2
it is seen that interfering the superposition of the first qubit, by applying a final hadamard gate, yields:
  
(−1)f (0) |1⟩a ⊗ |0⟩b√−|1⟩b if f (0) ̸= f (1)
2
|ψ4 ⟩ = 
|0⟩ −|1⟩
 (77)
f (0)
(−1) |0⟩a ⊗ b

2
b
if f (0) = f (1)

Therefore it can readily be concluded that f is constant, if a projective measurement in the computational basis of
the first qubit, yields |0⟩, and vice versa. By means of Deutsches algorithm, ”O(1)” was achieved instead of ”O(2)”.

While this might seem very modest, Deutsch, in collaboration with Richard Jozsa, since generalized the algorithm,
such that it can determine whether some g : {0, 1}n → {0, 1}, is balanced or constant. As any n-length bit string has
2n distinct permutations, and it would take 1 more than half of these27 to determine this feature with certainty in the
worst case, the Deutsche-Josza algorithm astoundingly achieves O(1) for something that classically is O(2n ).

27 Meaning 2n
that one would have to compute f a maximum of 2
+ 1 times.

17
11 Discuss typical sources of errors in quantum computations, their theo-
retical description and approaches for their correction
Appendix

|0⟩ |0⟩

 
√1
1 1
Ĥ = 2 1 −1
|0⟩+i|1⟩ |0⟩+i|1⟩
√ √
2 2

|0⟩+|1⟩ |0⟩+|1⟩
√ √
2 2
|1⟩ |1⟩

Figure 10: On the left Bloch sphere is shown a qubit in the |0⟩ state (with a blue arrow), and on the right is the state resulting from
applying the hadamard gate, putting the qubit in an equal superposition of |0⟩ and |1⟩, or alternatively the P auli − X basis.

18

You might also like