Finished chapter 6 and added section on the Bose-Hubbard model to chapter 2

This commit is contained in:
Wojciech Kozlowski 2016-08-14 17:39:47 +01:00
parent 80d4bae49f
commit be066cee9b
14 changed files with 1085 additions and 7 deletions

View File

@ -12,3 +12,4 @@
%********************************** %First Section **************************************

BIN
Chapter2/Figs/1DPhase.png Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 43 KiB

BIN
Chapter2/Figs/BHPhase.png Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 138 KiB

View File

@ -380,6 +380,265 @@ beyond previous treatments.
\section{The Bose-Hubbard Hamiltonian}
\subsection{The Model}
The original Hubbard model was developed as a model for the motion of
electrons within transition metals \cite{hubbard1963}. The key
elements were the confinement of the motion to a lattice and the
approximation of the Coulomb screening interaction as a simple on-site
interaction. Despite these enormous simplifications the model was very
succesful \cite{leggett}. The Bose-Hubbard model is an even simpler
variation where instead of fermions we consider spinless bosons. It
was originally devised as a toy model, but in 1998 it was shown by
Jaksch \emph{et.~al.} that it can be realised with ultracold atoms in
an optical lattice \cite{jaksch1998}. Shortly afterwards it was
obtained in a ground-breaking experiment \cite{greiner2002}. The model
has been the subject of intense research since then, because despite
its simplicity it possesses highly nontrivial properties such as the
superfluid to Mott insulator quantum phase transition. Furthermore, it
is one of the most controllable quantum many-body systems thus
providing a solid basis for new experiments and technologies.
The model we have derived is essentially an extension of the
well-known Bose-Hubbard model that also includes interactions with
quantised light. Therefore, it should come as no surprise that if we
eliminate all the quantized fields from the Hamiltonian we obtain
exactly the Bose-Hubbard model
\begin{equation}
\H_a = -J \sum_{\langle i,j \rangle}^M \bd_i b_j +
\frac{U}{2} \sum_{i}^M \hat{n}_i (\hat{n}_i - 1).
\end{equation}
The first term is the kinetic energy term and denotes the hopping of
atoms between neighbouring sites at a hopping rate given by $J$. The
second term describes the on-site repulsion which acts mainly to
suppress multiple occupancies of any site, but at the same time it
lies at the heart of strongly correlated phenomena in the Bose-Hubbard
model. Unfortunately, despite its simplicity, this system is not
solvable for any finite value $U/J$ using known techniques in finite
dimensions \cite{krauth1991, kolovsky2004, calzetta2006}. However,
exact solutions exist for the ground state in the two limits of
$U = 0$ and $J = 0$ which already provide some insight as they
correspond to the two different phases of the quantum phase transition
that is present in the model. In higher dimensions (e.g.~two or three)
the system is reasonably well described by a mean-field approximation
which provides us with a good understanding of the full quantum phase
diagram.
\subsection{The Superfluid and Mott Insulator Ground States}
Before trying to understand the quantum phase transition itself we
will look at the ground states of the two extremal cases first and for
simplicity we will consider only homogenous systems with periodic
boundary conditions here. We will begin with the $U = 0$ limit where
the Hamiltonian's kinetic term can be diagonalised by transforming to
momentum space defined by the annihilation operator
\begin{equation}
b_\b{k} = \frac{1} {\sqrt{M}} \sum_m b_m e^{i \b{k} \cdot \b{r}_m},
\end{equation}
where $\b{k}$ denotes the wavevector running over the first Brillouin
zone. The Hamiltonian then is given by
\begin{equation}
\H_a = \sum_\b{k} \epsilon_\b{k} \bd_\b{k} b_\b{k},
\end{equation}
where the free particle dispersion is
\begin{equation}
\epsilon_\b{k} = -2 J \sum_{d = 1}^D \cos(k_d a),
\end{equation}
and $d$ denotes the dimension out of a total of $D$ dimensions. It
should now be obvious that the ground state is simply the state with
all the atoms in the $\b{k} = 0$ state and thus it is given by
\begin{equation}
\label{eq:GSSF}
| \Psi_\mathrm{SF} \rangle = \frac{ ( \bd_{\b{k} = 0} )^N } {\sqrt{N!}} | 0
\rangle = \frac{1} {\sqrt{N!}} \left( \frac{1} {\sqrt{M}} \sum_m^M
\bd_m \right)^N | 0 \rangle,
\end{equation}
where $| 0 \rangle$ denotes the vacuum state. This state is called a
superfluid (SF) due to its viscous-free flow properties
\cite{leggett1999}. The macroscopic occupancy of the single-particle
ground state is the signature and defining property of a
Bose-Einstein condensate. Note that the zero momentum states consist of a
superposition of an atom at all the sites of the optical lattice
highlighting the fact that kinetic energy is minimised by delocalising
the particles. This in turn implies that a superfluid state will have
infinite range correlations. It is also important to note that in the
thermodynamic limit the superfluid state is gapless, i.e.~it takes
zero energy to excite the system.
At the other end of the spectrum, i.e.~$J = 0$, the kinetic term
vanishes and the system is dominated by the on-site interactions. This
has the effect of decoupling the Hamiltonian into a sum of identical
single-site Hamiltonians which are already diagonal in the atom number
basis. Therefore, the ground state for $N = gM$, where $g$ is an
integer, is given by
\begin{equation}
\label{eq:GSMI}
| \Psi_\mathrm{MI} \rangle = \prod_m^M \frac{1} {\sqrt{g!}}
(\bd_m)^g | 0 \rangle.
\end{equation}
This state has precisely the same number of bosons, $g$, at each site
and is commonly known as the Mott insulator (MI). Unlike in the
superfluid state, the atoms are all localised to a specific lattice
site and thus not only are there no long-range correlations, there are
actually no two-site correlations in this state at all. Additionally,
this state is gapped, i.e.~it costs a finite amount of energy to
excite the system into its lowest excited state even in the
thermodynamic limit. In fact, the existance of the gap is the defining
property that distinguishes the Mott insulator from the superfluid in
the Bose-Hubbard model. Note that we have only considered a
commensurate case (number of atoms an integer multiple of the number
of sites). If we were to insert another atom on top of the Mott
insulator, the energy of the system would not depend on the lattice
site in which it is inserted. This implies that as soon as $J$ becomes
finite, the ground state would prefer this atom be delocalised and
effectively becomes gapless which means the system is a superfluid and
it will not exhibit a quantum phase transition.
\subsection{Mean-Field Theory}
Now that we have a basic understanding of the two limiting cases we
can now consider the model in between these two extremes. We have
already mentioned that an exact solution is not known, but fotunately
a very good mean-field approximation exists \cite{fisher1989}. In this
approach the interaction term is treated exactly, but the kinetic
energy term is decoupled as
\begin{equation}
\bd_m b_n = \langle \bd_m \rangle b_n + \bd_m \langle b_n \rangle -
\langle \bd_m \rangle \langle b_n \rangle = \Phi^* b_n + \Phi \bd_m
- | \Phi |^2,
\end{equation}
where the expectation values are for the $T = 0$ ground state. Note
that we have assumed that $\Phi = \langle b_m \rangle $ is non-zero
and homogenous. $\Phi$ describes the influence of hopping between
neighbouring sites and is the mean -field order parameter. This
decoupling effect means that we can write the Hamiltonian as $\hat{H}_a
= \sum_m^M \hat{h}_a$, where
\begin{equation}
\hat{h}_a = -z J ( \Phi \bd \Phi^* b) + z J | \Phi |^2 + \frac{U}{2}
\n (\n - 1) - \mu \n,
\end{equation}
and $z$ is the coordination number, i.e. the number of nearest
neighbours of a single site. We have also introduced the chemical
potential $\mu$ since we now consider the grand canonical ensemble as
the Hamiltonian no longer conserves the total atom number. The ground
state of the $| \Psi_0 \rangle$ of the overall system will be a
site-wise product of the individual ground states of
$\hat{h}_a$. These can be found very easily using standard
diagonalisation techniques. $\Phi$ is then self-consistently
determined by minimising the energy of the ground state with respect
to $\Phi$.
The main advantage of the mean-field treatment is that it lets us
study the quantum phase transition between the sueprfluid and Mott
insulator phases discussed in the previous sections. The phase
boundaries can be obtained from second-order perturbation theory as
\begin{equation}
\left( \frac{\mu} {U} \right)_\pm = \frac{1}{2} \left[ 2g - 1 -
\left( \frac{zJ}{U} \right) \pm \sqrt{ 1 - 2 (2g + 1) \left( \frac{zJ}{U}
\right) + \left( \frac{zJ}{U} \right)^2} \right].
\end{equation}
This yields a phase diagram with multiple Mott insulating lobes as
seen in Fig. \ref{fig:BHPhase} where each lobe represents a Mott insualtor with a
different number of atoms per site. The tip of the lobe which
corresponds to the quantum phase transition along a line of fixed
density is obtained by equating the two branches of the above equation
to get
\begin{equation}
\left( \frac{J} {U} \right) = \frac{1} {z (2g + 1 + \sqrt{4g^2 + 4g} )}.
\end{equation}
It is important to note that the fact that this formalism predicts a
phase transition is in contrast to other mean-field theories such as
the Bogoliubov approximation \cite{PitaevskiiStringari}. This
highlights the fact that the interactions, which are treated exactly
here, are the dominant driver leading to this phase transition and
other strongly correlated effects in the Bose-Hubbard model.
\begin{figure}
\centering
\includegraphics[width=0.8\linewidth]{BHPhase}
\caption[Mean-Field Bose-Hubbard Phase Diagram]{Mean-field phase
diagram of the Bose-Hubbard model in 1D, i.e.~ $z = 2$, from
Ref. \cite{StephenThesis}. The shaded regions are the Mott
insulator lobes and each lobe corresponds to a different on-site
filling labeled by $n$. The rest of the space corresponds to the
superfluid phase. The dashed lines are the phase boundaries
obtained from first-order perturbation theory. The solid lines are
are lines of constant density in the superfluid
phase. \label{fig:BHPhase}}
\end{figure}
\subsection{The Bose-Hubbard Model in One Dimension}
The mean-field theory in the previous section is very useful tool for
studying the quantum phase transition in the Bose-Hubbard
model. However, it is effectively an infinite-dimensional theory and
in practice it only works in two dimensions or more. The phase
transition in 1D is poorly described, because it actually belongs to a
different universality class. This is clearly seen from the one
dimensional phase diagram shown in Fig. \ref{fig:1DPhase}.
\begin{figure}
\centering
\includegraphics[width=0.8\linewidth]{1DPhase}
\caption[Exact 1D Bose-Hubbard Phase Diagram]{Exact phase diagram
of the Bose-Hubbard model in 1D, i.e.~ $z = 2$, from
Ref. \cite{StephenThesis}. The shaded region corresponds to the
$n = 1$ Mott insulator lobe. The sharp tip distinguishes it from
the mean-field phase diagram. The red dotted line denotes the
reentrance phase transition which occurs for a fixed $\mu$.The
critical point for the fixed density $n = 1$ transition is denoted
by an 'x'. \label{fig:1DPhase}}
\end{figure}
Some general conclusions can be obtained by looking at Haldane's
prescription for Luttinger liquids \cite{haldane1981,
giamarchi}. Without a periodic potential the low-energy physics of
the system is described by the Hamiltonian
\begin{equation}
\hat{H}_a = \frac{1}{2 \pi} \int \mathrm{d} x \left\{ v K [ \hat{\Pi}(x)
]^2 + \frac{v} {K} [\partial_x \hat{\Phi}(x) ]^2 \right\},
\end{equation}
where we have expressed the bosonic field operators in terms of a
density operator $\hat{\rho}(x)$ and a phase operator $\hat{\Phi}(x)$
as $\hat{\Psi}(x) = \sqrt{\hat{\rho}(x)} e^{i \hat{\hat{\Phi}(x)}}$
and $\hat{\Pi}(x)$ is the density fluctuation operator. Provided the
parameters $v$ and $K$ can be correctly determined this Hamiltonian
gives the correct description of the gapless superfluid phase of the
Bose-Hubbard model. Most importantly it gives an expression for the
spatial correlation functions such as
\begin{equation}
\langle \bd_m b_n \rangle = A \left( \frac{\alpha} {|m - n|} \right)^{K/2},
\end{equation}
where $A$ is some amplitude and $\alpha$ is a necessary cutoff to
regularise the theory at short distances. Unlike the superfluid ground
state in Eq. \eqref{eq:GSSF} this state does not have infinite range
correlations. They decay according to a power law. However, for
non-interacting systems $K = 0$ and long-range order is re-established
as before though it is important to note that in higher dimensions
this long-range order persists in the whole superfluid phase even with
interactions present. In order to describe the phase transition and
the Mott insulating phase it is necessary to introduce a periodic
lattice potential. It can be shown that this system exhibits at
$T = 0$ a Berezinskii-Kosterlitz-Thouless phase transition as the
parameter $K$ is varied with a critical point at $K = \frac{1}{2}$
where $K < \frac{1}{2}$ is a superfluid. Above $K = \frac{1}{2}$ the
value of $K$ jumps discontinuously to $K \rightarrow \infty$ producing
the Mott insulator phase. Unlike the gapless superfluid phase the
spatial correlations decay exponentially as
\begin{equation}
\langle \bd_m b_n \rangle = B e^{ - |m - n|/\xi},
\end{equation}
where $B$ is some constant and the correlation length is given by
$\xi = v / \Delta$ where $\Delta$ is the energy gap.
Using advanced numerical methods such as density matrix
renormalisation group (DMRG) calculations it is possible to identify
the critical point by fitting the power-law decay correlations in
order to obtain $K$. The resulting phase transition is shown in
Fig. \ref{fig:1DPhase} and the critical point was shown to be $(U/zJ) = 1.68$. Note
that unusually the phase diagram exhibits a reentrance phase
transition for a fixed $\mu$.
\section{Scattered light behaviour}
\label{sec:a}
@ -832,6 +1091,7 @@ Therefore, we can now express the quantity $n_{\Phi}$ as
\begin{equation}
n_{\Phi} = \frac{1}{8} \left(\frac{\Omega_0}{\Delta_a}\right)^2 \frac{\Gamma}{2} N_K.
\end{equation}
Estimates of the scattering rate using real experimental parameters
are given in Table \ref{tab:photons}. Rubidium atom data has been
taken from Ref. \cite{steck}. Miyake \emph{et al.} experimental
@ -866,7 +1126,7 @@ $\Gamma_\mathrm{sc} = (\Gamma/2) (s_\mathrm{tot}) /
(1+s_\mathrm{tot}+(2 \Delta / \Gamma)^2)$. A scattering rate of 60 kHz
per atom \cite{weitenberg2011} gives $s_\mathrm{tot} = 2.5$.
\begin{table}[!htbp]
\begin{table}
\centering
\begin{tabular}{l c c}
\toprule

View File

@ -4,6 +4,7 @@
\chapter{Probing Correlations by Global
Nondestructive Addressing} %Title of the Third Chapter
\label{chap:qnd}
\ifpdf
\graphicspath{{Chapter3/Figs/Raster/}{Chapter3/Figs/PDF/}{Chapter3/Figs/}}

View File

@ -3,6 +3,7 @@
%*******************************************************************************
\chapter{Quantum Measurement Backaction}
\label{chap:backaction}
% Title of the Fourth Chapter
\ifpdf
@ -244,6 +245,7 @@ between fields. Therefore, the final form of the Hamiltonian
Eq. \eqref{eq:fullH} that we will be using in the following chapters
is
\begin{equation}
\label{eq:backaction}
\hat{H} = \hat{H}_0 - i \gamma \hat{F}^\dagger \hat{F}
\end{equation}
\begin{equation}
@ -277,7 +279,26 @@ nonlocal way and most importantly not only will it not degrade
long-range coherence, it will in fact lead to such long-range
correlations itself.
\mynote{introduce citations from PRX/PRA above}
In Chapter \ref{chap:qnd} we used highly efficient DMRG methods
\cite{tnt} to calculate the ground state of the Bose-Hubbard
Hamiltonian. Related techniques such as Time-Evolving Block Decimation
(TEBD) or t-DMRG are often used for numerical calculations of time
evolution. However, despite the fact our Hamiltonian in
Eq. \eqref{eq:backaction} is simply the Bose-Hubbard model with a
non-Hermitian term added due to measurement it is actually difficult
to apply these methods to our system. The problem lies in the fact
that Matrix Product methods we mentioned are only efficient for
one-dimensional systems that obey the area law for entanglement
entropy, i.e.~systems with only short-range quantum
correlations. Unfortunately, the global nature of the measurement we
consider violates the assumptions made in deriving the area law and,
as we shall see in the following chapters, leads to long-range
correlations regardless of coupling strength. Therefore, we resort to
using exact methods such as exact diagonalisation which we solve with
well-known ordinary differential equation solvers. This means that we
can at most simulate a few atoms, but as we shall see it is the
geometry of the measurement that matters the most and these effects
are already visible in smaller systems.
\section{The Master Equation}
\label{sec:master}

View File

@ -1692,7 +1692,7 @@ obtained by cooling or projecting from an initial ground state. The
combination of tunnelling with measurement is necessary.
\begin{figure}[hbtp!]
\includegraphics[width=\linewidth]{figure3}
\includegraphics[width=\linewidth]{steady}
\caption[Non-Hermitian Steady State]{A trajectory simulation
for eight atoms in eight sites, initially in
$|1,1,1,1,1,1,1,1 \rangle$, with periodic boundary

Binary file not shown.

BIN
Chapter6/Figs/spaces.pdf Normal file

Binary file not shown.

View File

@ -14,10 +14,616 @@
\section{Introduction}
Light scatters due to its interaction with the dipole moment of the
atoms which for off-resonant light results in an effective coupling
with atomic density, not the matter-wave amplitude. Therefore, it is
challenging to couple light to the phase of the matter-field, as is
typical in quantum optics for optical fields. In the previous chapter
we only considered measurement that couples directly to atomic density
operators just most of the existing work \cite{LP2009, rogers2014,
mekhov2012, ashida2015, ashida2015a}. However, we have shown in
section \ref{sec:B} that it is possible to couple to the the relative
phase differences between sites in an optical lattice by illuminating
the bonds between them. Furthermore, we have also shown how it can be
applied to probe the Bose Hubbard order parameter or even matter-field
quadratures in Chapter \ref{chap:qnd}. This concept has also been
applied to the study of quantum optical potentials formed in a cavity
and shown to lead to a host of interesting quantum phase diagrams
\cite{caballero2015, caballero2015njp, caballero2016,
caballero2016a}. This is a multi-site generalisation of previous
double-well schemes \cite{cirac1996, castin1997, ruostekoski1997,
ruostekoski1998, rist2012}, although the physical mechanism is
fundametally different as it involves direct coupling to the
interference terms caused by atoms tunnelling rather than combining
light scattered from different sources.
We have already mentioned that there are three primary avenues in
which the field of quantum optics with ultracold gases can be taken:
nondestructive measurement, quantum measurement backaction, and
quantum optical potentials. All three have been covered in the context
of density-based measurement either here or in other works. However,
coupling to phase observables in lattices has only been proposed and
considered in the context of nondestructive measurements (see Chapter
\ref{chap:qnd}) and quantum optical potentials. \cite{caballero2015,
caballero2015njp, caballero2016, caballero2016a}. In this chapter,
we go in a new direction by considering the effect of measurement
backaction on the atomic gas that results from such coupling. We
investigate this mechanism using light scattered from these
phase-related observables. The novel combination of measurement
backaction as the physical mechanism driving the dynamics and phase
coherence as the observable to which the optical fields couple to
provides a completely new opportunity to affect and manipulate the
quantum state. We first show how this scheme enables us to prepare
energy eigenstates of the lattice Hamiltonian. Furthermore, in the
second part we also demonstrate a novel type of a projection due to
measurement which occurs even when there is significant competition
with the Hamiltonian dynamics. This projection is fundamentally
different to the standard formulation of the Copenhagen postulate
projection or the quantum Zeno effect \cite{misra1977, facchi2008,
raimond2010, raimond2012, signoles2014} thus providing an extension
of the measurement postulate to dynamical systems subject to weak
measurement.
\section{Diffraction Maximum and Energy Eigenstates}
In this chapter we will only consider measurement backaction when
$\a_1 = C \hat{B}$ and $\c = \sqrt{2 \kappa} \a_1$ as introduced in
section \ref{sec:B}. We have seen that there are two ways of
engineering the spatial profile of the measurement such that the
density contribution from $\hat{D}$ is suppressed. We will first
consider the case when the profile is uniform, i.e.~the diffraction
maximum of scattered light, when our measurement operator is given by
\begin{equation}
\B = \Bmax = J^B_\mathrm{max} \sum_j^K \left( \bd_j b_{j+1} b_j
\bd_{j+1} \right)
= 2 J^B_\mathrm{max} \sum_k \bd_k b_k \cos(ka),
\end{equation}
where the second equality follows from converting to momentum space,
denoted by index $k$, via
$b_m = \frac{1}{\sqrt{M}} \sum_k e^{-ikma} b_k$ and $b_k$ annihilates
an atom with momentum $k$ in the range
$\{ \frac{(M-1) \pi} {M}, \frac{(M-2) \pi} {M}, ..., \pi \}$. Note
that this operator is diagonal in momentum space which means that its
eigenstates are simply Fock momentum Fock states. We have also seen in
Chapter \ref{chap:backaction} how the global nature of the jump
operators introduces a nonlocal quadratic term to the Hamiltonian,
$\hat{H} = \hat{H}_0 - i \cd \c / 2$. In order to focus on the
competition between tunnelling and measurement backaction we again
consider non-interacting atoms, $U = 0$. Therefore, $\B$ is
proportional to the Hamiltonian and both operators have the same
eigenstates.
The combined Hamiltonian for the non-interacting gas subject to
measurement of the from $\a_1 = C \B$ is thus
\begin{equation}
\hat{H} = -J \sum_j \left( \bd_j b_{j+1} + b_j \bd_{j+1} \right) - i
\gamma \Bmax^\dagger \Bmax.
\end{equation}
Furthermore, whenever a photon is detected, the operator
$\c = \sqrt{2 \kappa} C \Bmax$ is applied to the wavefunction. We can
rewrite the above equation as
\begin{equation}
\hat{H} = -\frac{J}{J^B_\mathrm{max}} \Bmax - i \gamma \Bmax^\dagger \Bmax.
\end{equation}
The eigenstates of this operator will be exactly the same as of the
isolated Hamiltonian which we will label as $| h_l \rangle$, where
$h_l$ denotes the corresponding eigenvalue. Therefore, for an initial
state
\begin{equation}
| \Psi \rangle = \sum_l z_l^0 | h_l \rangle
\end{equation}
it is easy to show that the state after $m$ photocounts and time $t$
is given by
\begin{equation}
| \Psi (m,t) \rangle = \frac{1}{\sqrt{F(t)}} \sum_l h_l^m z_l^0 e^{[i h_l -
\gamma(J^B_\mathrm{max}/J)^2 h_l^2]t} | h_l \rangle,
\end{equation}
where $\sqrt{F(t)}$ is the normalisation factor. Therefore, the probability
of being in state $| h_l \rangle$ at time $t$ after $m$ photocounts is
given by
\begin{equation}
p(h_l, m, t) = \frac{h_l^{2m}} {F(t)} \exp\left[ - 2 \gamma \left(
\frac{J^B_\mathrm{max}} {J} \right)^2 h_l^2 t \right] |z_l^0|^2.
\end{equation}
As a lot of eigenstates are degenerate, we are actually more
interested in the probability of being in eigenspace with the
eigenvalue $h_l = (J/J^B_\mathrm{max})B_\mathrm{max}$. This probability
is given by
\begin{equation}
\label{eq:bmax}
p(B_\mathrm{max}, m, t) = \frac{B_\mathrm{max}^{2m}} {F(t)} \exp\left[ - 2
\gamma B_\mathrm{max}^2 t \right] p_0 (B_\mathrm{max}),
\end{equation}
where
$p_0(B_\mathrm{max}) = \sum_{J_\mathrm{max} h_l = J B_\mathrm{max}}
|z_l^0|^2$. This distribution will have two distinct peaks at
$B_\mathrm{max} = \pm \sqrt{m/2\kappa |C|^2 t}$ and an initially broad
distribution will narrow down around these two peaks with successive
photocounts. The final state is in a superposition, because we measure
the photon number, $\ad_1 \a_1$ and not field amplitude. Therefore,
the measurement is insensitive to the phase of $\a_1 = C \B$ and we
get a superposition of $\pm B_\mathrm{max}$. This is exactly the same
situation that we saw for the macroscopic oscillations of two distinct
components when the atom number difference between two modes is
measured as seen in Fig. \ref{fig:oscillations}(b). However, this
means that the matter is still entangled with the light as the two
states scatter light with different phase which the photocount
detector cannot distinguish. Fortunately, this is easily mitigated at
the end of the experiment by switching off the probe beam and allowing
the cavity to empty out or by measuring the light phase (quadrature)
to isolate one of the components \cite{mekhov2009pra, mekhov2012,
atoms2015}.
Unusually, we do not have to worry about the timing of the quantum
jumps, because the measurement operator commutes with the
Hamiltonian. This highlights an important feature of this measurement
- it does not compete with atomic tunnelling, and represents a quantum
nondemolition (QND) measurement of the phase-related observable
\cite{brune1992}. Eq. \eqref{eq:bmax} shows that regardless of the
initial state or the photocount trajectory the system will project
onto a superposition of eigenstates of the $\Bmax$ operator. In fact,
the final state probability distribution would be exactly the same if
we were to simply employ a projective measurement of the same
operator. What is unusual in our case is that this has been achieved
with weak measurement in the presence of significant atomic
dynamics. The conditions for such a measurement have only recently
been rigorously derived in Ref. \cite{weinberg2016} and here we
provide a practical realisation using in ultracold bosonic gases.
This projective behaviour is in contrast to conventional density based
measurements which squeeze the atom number in the illuminated region
and thus are in direct competition with the atom dynamics (which
spreads the atoms), effectively requiring strong couplings for a
projection as seen in the previous chapter. Here a projection is
achieved at any measurement strength which allows for a weaker probe
and thus effectively less heating and a longer experimental
lifetime. Furthermore. This is in contrast to the quantum Zeno effect
which requires a very strong probe to compete effectively with the
Hamiltonian \cite{misra1977, facchi2008, raimond2010, raimond2012,
signoles2014}. Interestingly, the $\Bmax$ measurement will even
establish phase coherence across the lattice,
$\langle \bd_j b_j \rangle \ne 0$, in contrast to density based
measurements where the opposite is true: Fock states with no
coherences are favoured.
\section{General Model for Weak Measurement Projection}
\section{Determining the Projection Subspace}
\subsection{Projections for Incompatible Dynamics and Measurement}
\section{Conclusions}
In section \ref{sec:B} we have also shown that it is possible to
achieve a more complicated spatial profile of the
$\B$-measurement. The optical geometry can be adjusted such that each
bond scatters light in anti-phase with its neighbours leading to a
diffraction minimum where the expectation value of the amplitude is
zero. In this case the $\B$ operator is given by
\begin{equation}
\B = \Bmin = J^B_\mathrm{min} \sum_m^K (-1)^m \hat{B}_m
= 2 i J^B_\mathrm{min} \sum_k \bd_k b_{k - \pi/a} \sin(ka).
\end{equation}
Note how the measurement operator now couples the momentum mode $k$
with mode $k - \pi/a$. This measurement operator no longer commutes
with the Hamiltonian so it is no longer QND and we do not expect there
to be a steady state as before. In order to understand the measurement
it will be easier to work in a basis in which it is diagonal. We
perform the transformation
$\beta_k = \frac{1}{\sqrt{2}} \left( b_k + i b_{k - \pi/a} \right)$,
$\tilde{\beta}_k = \frac{1}{\sqrt{2}} \left( b_k - i b_{k - \pi/a}
\right)$, which yields the following forms of the measurement operator
and the Hamiltonian:
\begin{equation}
\label{eq:BminBeta}
\Bmin = 2 J^B_\mathrm{min} \sum_{\mathrm{RBZ}} \sin(ka) \left( \beta^\dagger_k
\beta_k - \tilde{\beta}_k^\dagger \tilde{\beta}_k \right),
\end{equation}
\begin{equation}
\label{eq:H0Beta}
\hat{H}_0 = - 2 J \sum_{\mathrm{RBZ}} \cos(ka) \left( \beta_k^\dagger
\tilde{\beta}_k + \tilde{\beta}^\dagger_k \beta_k \right),
\end{equation}
where the summations are performed over the reduced Brilluoin Zone
(RBZ), $0 < k \le \pi/a$, to ensure the transformation is
canonical. We see that the measurement operator now consists of two
types of modes, $\beta_k$ and $\tilde{\beta_k}$, which are
superpositions of two momentum states, $k$ and $k - \pi/a$. Note how a
spatial pattern with a period of two sites leads to a basis with two
modes whilst a uniform pattern had only one mode, $b_k$. Furthermore,
note the similarities to
$\D = \Delta \hat{N} = \hat{N}_\mathrm{even} - \hat{N}_\mathrm{odd}$
which is the density measurement operator obtained by illuminated the
alttice such that neighbouring sites scatter light in anti-phase. This
further highlights the importance of geometry for global measurement.
Trajectory simulations confirm that there is no steady state. However,
unexpectedly, for each trajectory we observe that the dynamics always
ends up confined to some subspace as seen in
Fig. \ref{fig:projections}. In general, this subspace is neither an
eigenspace of the measurement operator or the Hamiltonian. In
Fig. \ref{fig:projections}(b) it in fact clearly consists of multiple
measurement eigenspaces. This clearly distinguishes it from the
fundamental projections predicted by the Copenhagen postulates. It is
also not the quantum Zeno effect which predicts that strong
measurement can confine the evolution of a system as this subspace
must be an eigenspace of the measurement operator \cite{misra1977,
facchi2008, raimond2010, raimond2012, signoles2014}. Furthermore,
the projection we see in Fig. \ref{fig:projections} occurs for even
weak measurement strengths compared to the Hamiltonian's own
evolution, a regime in which the quantum Zeno effect does not
happen. It is also possible to dissipatively prepare quantum states in
an eigenstate of a Hamiltonian provided it is also a dark state of the
jump operator, $\c | \Psi \rangle = 0$~\cite{diehl2008}. However,
this is also clearly not the case as the final state in
Fig. \ref{fig:projections}(c) is not only not confined to a single
measurement operator eigenspace, it also spans multiple Hamiltonian
eigenspaces. Therefore, the dynamics induced by $\a_1 = C\Bmin$ project
the system into some subspace, but since this does not happen via any
of the mechanisms described above it is not immediately obvious what
this subspace is.
\begin{figure}[hbtp!]
\includegraphics[width=\linewidth]{Projections}
\caption[Projections for Non-Commuting Observable and
Hamiltonian] {Subspace projections. Projection to a
$\mathcal{P}_M$ space for four atoms on eight sites with
periodic boundary conditions. The parameters used are $J=1$,
$U=0$, $\gamma = 0.1$, and the initial state was
$| 0,0,1,1,1,1,0,0 \rangle$. (a) The
$\langle \hat{O}_k \rangle = \langle \n_k + \n_{k - \pi/a}
\rangle$ distribution reaches a steady state at
$Jt \approx 8$ indicating the system has been projected. (b)
Populations of the $\Bmin$ eigenspaces. (c) Population of
the $\hat{H}_0$ eigenspaces. Once the projection is achieved
at $Jt\approx8$ we can see from (b-c) that the system is not
in an eigenspace of either $\Bmin$ or $\hat{H}_0$, but it
becomes confined to some subspace. The system has been
projected onto a subspace, but it is neither that of the
measurement operator or the
Hamiltonian. \label{fig:projections}}
\end{figure}
To understand this dynamics we will look at the master equation for
open systems described by the density matrix, $\hat{\rho}$,
\begin{equation}
\dot{\hat{\rho}} = -i \left[\hat{H}_0, \hat{\rho} \right] +
\left[ \c \hat{\rho} \cd - \frac{1}{2} \left( \cd \c \hat{\rho} + \hat{\rho}
\cd \c \right) \right].
\end{equation}
The following results will not depend on the nature or exact form of
the jump operator $\c$. However, whenever we refer to our simulations
or our model we will be considering $\c = \sqrt{2 \kappa} C(\D + \B)$
as before, but the results are more general and can be applied to ther
setups. This equation describes the state of the system if we discard
all knowledge of the outcome. The commutator describes coherent
dynamics due to the isolated Hamiltonian and the remaining terms are
due to measurement. This is a convenient way to find features of the
dynamics common to every measurement trajectory.
Just like in the preceding chapter we define the projectors of the
measurement eigenspaces, $P_m$, which have no effect on any of the
(possibly degenerate) eigenstates of $\c$ with eigenvalue $c_m$, but
annihilate everything else, thus
$P_m = \sum_{c_n = c_m} | c_n \rangle \langle c_n |,$ where
$| c_n \rangle$ is an eigenstate of $\c$ with eigenvalue $c_n$. Note
that in the specific case of our quantum gas model
$\c = \sqrt{2 \kappa} C(\D + \B)$ so these projectors act on the
matter state. Recall from the previous chapter that this allows us to
decompose the master equation in terms of the measurement basis as a
series of equations $P_m \dot{\hat{\rho}} P_n$. We have seen that for
$m = n$ we obtain decoherence free subspaces,
$P_m \dot{\hat{\rho}} P_m = -i P_m \left[\hat{H}_0, \hat{\rho} \right]
P_m$, where the measurement terms disappear which shows that a state
in a single eigenspace is unaffected by observation. On the other
hand, for $m \ne n$ the Hamiltonian evolution actively competes
against measurement. In general, if $\c$ does not commute with the
Hamiltonian then a projection to a single eigenspace $P_m$ is
impossible unless the measurement is strong enough for the quantum
Zeno effect to occur.
We now go beyond what we previously did and define a new type of
projector $\mathcal{P}_M = \sum_{m \in M} P_m$, such that
$\mathcal{P}_M \mathcal{P}_N = \delta_{M,N} \mathcal{P}_M$ and
$\sum_M \mathcal{P}_M = \hat{1}$ where $M$ denotes some arbitrary
subspace. The first equation implies that the subspaces can be built
from $P_m$ whilst the second and third equation are properties of
projectors and specify that these projectors do not overlap and that
they cover the whole Hilbert space. Furthermore, we will also require
that $[\mathcal{P}_M, \hat{H}_0 ] = 0$ and $[\mathcal{P}_M, \c] =
0$. The second commutator simply follows from the definition
$\mathcal{P}_M = \sum_{m \in M} P_m$, but the first one is
non-trivial. However, if we can show that
$\mathcal{P}_M = \sum_{m \in M} | h_m \rangle \langle h_m |$, where
$| h_m \rangle$ is an eigenstate of $\hat{H}_0$ then the commutator is
guaranteed to be zero. This is a complex set of requirements and it is
unclear if it is possible to satisfy all of them at once. However, we
note that there exists a trivial case where all these conditions are
satisfied and that is when there is only one such projector
$\mathcal{P}_M = \hat{1}$.
Assuming that it is possible to have non-trivial cases where
$\mathcal{P}_M \ne \hat{1}$ we can write the master equation as
\begin{equation}
\label{eq:bigP}
\mathcal{P}_M \dot{\hat{\rho}} \mathcal{P}_N = -i
\left[\hat{H}_0, \mathcal{P}_M \hat{\rho} \mathcal{P}_N \right] +
\left[ \c \mathcal{P}_M \hat{\rho}
\mathcal{P}_N \cd - \frac{1}{2} \left( \cd \c \mathcal{P}_M \hat{\rho}
\mathcal{P}_N + \mathcal{P}_M \hat{\rho} \mathcal{P}_N \cd \c \right)
\right].
\end{equation}
Crucially, thanks to the commutation relations we were able to divide
the density matrix in such a way that each submatrix's time evolution
depends only on itself. Every submatrix
$\mathcal{P}_M \dot{\hat{\rho}} \mathcal{P}_N$ depends only on the
current state of itself and its evolution is ignorant of anything else
in the total density matrix. This is in contrast to the partitioning
we achieved with $P_m$. Previously we only identitified subspaces that
were decoherence free, i.e.~unaffected by measurement. However, those
submatrices could still couple with the rest of the density matrix via
the coherent term $P_m [\hat{H}_0, \hat{\rho}] P_n$.
We note that when $M = N$ the equations for
$\mathcal{P}_M \hat{\rho} \mathcal{P}_M$ will include decoherence free
subspaces, i.e.~$P_m \hat{\rho} P_m$. Therefore, parts of the
$\mathcal{P}_M \hat{\rho} \mathcal{P}_M$ submatrices will also remain
unaffected by measurement. However, the submatrices
$\mathcal{P}_M \hat{\rho} \mathcal{P}_N$, for which $M \ne N$, are
guaranteed to not contain any of these measurement free subspaces
thanks to the orthogonality of $\mathcal{P}_M$. Therefore, for
$M \ne N$ all elements of $\mathcal{P}_M \hat{\rho} \mathcal{P}_N$
will experience a non-zero measurement term whose effect is always
dissipative/lossy. Furthermore, these coherence submatrices
$\mathcal{P}_M \hat{\rho} \mathcal{P}_N$ are not coupled to any other
part of the density matrix as seen from Eq. \eqref{eq:bigP} and so
they can never increase in magnitude; the remaining coherent evolution
is unable to counteract the dissipative term without an `external
pump' from other parts of the density matrix. The combined effect is
such that all $\mathcal{P}_M \hat{\rho} \mathcal{P}_N$ for which
$M \ne N$ will always go to zero due to dissipation.
When all these cross-terms vanish, we are left with a density matrix
that is a mixed state of the form
$\hat{\rho} = \sum_M \mathcal{P}_M \hat{\rho} \mathcal{P}_M$. Since
there are no coherences, $\mathcal{P}_M \hat{\rho} \mathcal{P}_N$,
this state contains only classical uncertainty about which subspace,
$\mathcal{P}_M$, is occupied - there are no quantum superpositions
between different $\mathcal{P}_M$ spaces. Therefore, in a single
measurement run we are guaranteed to have a state that lies entirely
within a subspace defined by $\mathcal{P}_M$. This is once again
analogous to the qubit example from section \ref{sec:master}.
Such a non-trivial case is indeed possible for our $\hat{H}_0$ and
$\a_1 = C\Bmin$ and we can see the effect in
Fig. \ref{fig:projections}. We can clearly see how a state that was
initially a superposition of a large number of eigenstates of both
operators becomes confined to some small subspace that is neither an
eigenspace of $\a_1$ or $\hat{H}_0$. In this case the projective spaces,
$\mathcal{P}_M$, are defined by the parities (odd or even) of the
combined number of atoms in the $\beta_k$ and $\tilde{\beta}_k$ modes
for different momenta $0 < k < \pi/a$ that are distinguishable to
$\Bmin$. The explanation requires careful consideration of where the
eigenstates of the two operators overlap.
\subsection{Determining the Projection Subspace}
To find $\mathcal{P}_M$ we need to identify the subspaces $M$ which
satisfy the following relation
$\sum_{m \in M} P_m = \sum_{m \in M} | h_m \rangle \langle h_m
|$. This can be done iteratively by (i) selecting some $P_m$, (ii)
identifying the $| h_m \rangle$ which overlap with this subspace,
(iii) identifying any other $P_m$ which also overlap with these
$| h_m \rangle$ from step (ii). We repeat (ii)-(iii) for all the $P_m$
found in (iii) until we have identified all the subspaces $P_m$ linked
in this way and they will form one of our $\mathcal{P}_M$
projectors. If $\mathcal{P}_M \ne 1$ then there will be other
subspaces $P_m$ which we have not included so far and thus we repeat
this procedure on the unused projectors until we identify all
$\mathcal{P}_M$. Computationally this can be straightforwardly solved
with some basic algorithm that can compute the connected components of
a graph.
The above procedure, whilst mathematically correct and always
guarantees to generate the projectors $\mathcal{P}_M$, is very
unintuitive and gives poor insight into the nature or physical meaning
of $\mathcal{P}_M$. In order to get a better understanding of these
subspaces we need to define a new operator $\hat{O}$, with eigenspace
projectors $R_m$, which commutes with both $\hat{H}_0$ and
$\c$. Physically this means that $\hat{O}$ is a compatible observable
with $\c$ and corresponds to a quantity conserved by the
Hamiltonian. The fact that $\hat{O}$ commutes with the Hamiltonian
implies that the projectors can be written as a sum of Hamiltonian
eigenstates $R_m = \sum_{h_i = h_m} | h_i \rangle \langle h_i |$ and
thus a projector $\mathcal{P}_M = \sum_{m \in M} R_m$ is guaranteed to
commute with the Hamiltonian and similarly since $[\hat{O}, \c] = 0$
$\mathcal{P}_M$ will also commute with $\c$ as required. Therefore,
$\mathcal{P}_M = \sum_{m \in M} R_m = \sum_{m \in M} P_m$ will satisfy
all the necessary prerequisites. This is illustrated in
Fig. \ref{fig:spaces}.
\begin{figure}[hbtp!]
\includegraphics[width=\linewidth]{spaces}
\caption[A Visual Representation of the Projection Spaces of
the Measurement]{A visual representation of the projection
spaces of the measurement. The light blue areas (bottom
layer) are $R_m$, the eigenspaces of $\hat{O}$. The green
areas are measurement eigenspaces, $P_m$, and they overlap
non-trivially with the $R_m$ subspaces. The $\mathcal{P}_M$
projection space boundary (dashed line) runs through the
Hilbert space, $\mathcal{H}$, where there is no overlap
between $P_m$ and $R_m$. \label{fig:spaces}}
\end{figure}
In the simplest case the projectors $\mathcal{P}_M$ can consist of
only single eigenspaces of $\hat{O}$, $\mathcal{P}_M = R_m$. The
interpretation is straightforward - measurement projects the system
onto a eigenspace of an observable $\hat{O}$ which is a compatible
observable with $\a$ and corresponds to a quantity conserved by the
coherent Hamiltonian evolution. However, this may not be possible and
we have the more general case when
$\mathcal{P}_M = \sum_{m \in M} R_m$. In this case, one can simply
think of all $R_{m \in M}$ as degenerate just like eigenstates of the
measurement operator, $\a$, that are degenerate, can form a single
eigenspace $P_m$. However, these subspaces correspond to different
eigenvalues of $\hat{O}$ distinguishing it from conventional
projections. Instead, the degeneracies are identified by some other
feature.
In our case, it is apparent from the form of $\Bmin$ and $\hat{H}_0$
in Eqs. \eqref{eq:BminBeta} and \eqref{eq:H0Beta} that
$\hat{O}_k = \beta_k^\dagger \beta_k + \tilde{\beta}_k^\dagger
\tilde{\beta_k} = \n_k + \n_{k - \pi/a}$ commutes with both operators
for all $k$. Thus, we can easily construct
$\hat{O} = \sum_\mathrm{RBZ} g_k \hat{O}_k$ for any arbitrary
$g_k$. Its eigenspaces, $R_m$, can then be easily constructed and
their relationship with $P_m$ and $\mathcal{P}_M$ is illustrated in
Fig. \ref{fig:spaces} whilst the time evolution of
$\langle \hat{O}_k \rangle$ for a sample trajectory is shown in
Fig. \ref{fig:projections}(a). Note that unlike the $\c$ or $\H_0$ we
can actually see that this observable's distribution does indeed
freeze. These eigenspaces are composed of Fock states in momentum
space that have the same number of atoms within each pair of $k$ and
$k - \pi/a$ modes.
Having identified and appropriate $\hat{O}$ operator we proceed to
identifying $\mathcal{P}_M$ subspaces for the operator
$\Bmin$. Firstly, since $\Bmin$ contains $\sin(ka)$ coefficients atoms
in different $k$ modes that have the same $\sin(ka)$ value are
indistinguishable to the measurement and will lie in the same $P_m$
eigenspaces. This will happen for the pairs ($k$, $\pi/a -
k$). Therefore, the $R_m$ spaces that have the same
$\hat{O}_k + \hat{O}_{\pi/a - k}$ eigenvalues must belong to the same
$\mathcal{P}_M$.
Secondly, if we re-write $\hat{O}$ in terms of the $\beta_k$ and
$\tilde{\beta}_k$ modes we get
\begin{equation}
\hat{O} = \sum_{\mathrm{RBZ}} g_k \left(
\beta^\dagger_k \beta_k + \tilde{\beta}_k^\dagger \tilde{\beta}_k
\right),
\end{equation}
and so it's not hard to see that
$\B_{\mathrm{min},k} = (\beta^\dagger_k \beta_k -
\tilde{\beta}_k^\dagger \tilde{\beta}_k)$ can have the same
eigenvalues for different values of
$\hat{O}_k = \beta^\dagger_k \beta_k + \tilde{\beta}_k^\dagger
\tilde{\beta}_k$. Specifically, if a given subspace $R_m$ corresponds
to the eigenvalue $O_k$ of $\hat{O}_k$ then the possible values of
$B_{\mathrm{min},k}$ will be $\{-O_k, -O_k + 2, ..., O_k - 2, O_k\}$
just like $\Delta N = N_\mathrm{even} - N_\mathrm{odd}$ can have
multiple possible values for a fixed
$N = N_\mathrm{even} + N_\mathrm{odd}$. Therefore, we can see that all
$R_m$ with even values of $O_k$ will share $B_{\mathrm{min},k}$
eigenvalues and thus they will overlap with the same $P_m$
subspaces. The same is true for odd values of $O_k$. However, $R_m$
with an even value of $O_k$ will never have the same value of
$B_{\mathrm{min},k}$ as a subspace with an odd value of
$O_k$. Therefore, a single $\mathcal{P}_M$ will contain all $R_m$ that
have the same parities of $O_k$ for all $k$, e.g.~if it includes the
$R_m$ with $O_k = 6$, it will also include the $R_m$ for which
$O_k = 0, 2, 4, 6, ..., N$, but not any of $O_k = 1, 3, 5, 7, ..., N$,
where $N$ is the total number of atoms.
Finally, the $k = \pi/a$ mode is special, because $\sin(\pi) = 0$
which means that $B_{\mathrm{min},k=\pi/a} = 0$ always. This in turn
implies that all possible values of $O_{\pi/a}$ are degenerate to the
measurement. Therefore, we exclude this mode when matching the
parities of the other modes.
To illustrate the above let us consider a specific example. Let us
consider two atoms, $N=2$, on eight sites $M=8$. This particular
configuration has eight momentum modes
$ka = \{-\frac{3\pi}{4}, -\frac{\pi}{2}, -\frac{\pi}{4}, 0,
\frac{\pi}{4}, \frac{\pi}{2}, \frac{3\pi}{4}, \pi\}$ and so the RBZ
has only four modes,
$\mathrm{RBZ} := \{\frac{\pi}{4}, \frac{\pi}{2}, \frac{3\pi}{4},
\pi\}$. There are 10 different ways of splitting two atoms into these
four modes and thus we have 10 different
$R_m = \{O_{\pi/4a}, O_{\pi/2a} ,O_{3\pi/4a} ,O_{\pi/a}\}$ eigenspaces
of $\hat{O}$
\begin{table}[!htbp]
\centering
\begin{tabular}{l c c}
\toprule
$m$ & $R_m$ & Possible values of $B_\mathrm{min}$ \\ \midrule
0 & $\{2,0,0,0\}$ & $ -\sqrt{2}, 0, \sqrt{2} $ \\
1 & $\{1,1,0,0\}$ & $ -\frac{1 + \sqrt{2}}{\sqrt{2}}, -\frac{1
- \sqrt{2}}{\sqrt{2}}, \frac{1
- \sqrt{2}}{\sqrt{2}}, \frac{1 +
\sqrt{2}}{\sqrt{2}} $ \\
2 & $\{1,0,1,0\}$ & $ -\sqrt{2}, 0, \sqrt{2} $ \\
3 & $\{1,0,0,1\}$ & $ -\frac{1}{\sqrt{2}},
\frac{1}{\sqrt{2}} $ \\
4 & $\{0,2,0,0\}$ & $ -2, 0, 2 $ \\
5 & $\{0,1,1,0\}$ & $ -\frac{1 + \sqrt{2}}{\sqrt{2}}, -\frac{1
- \sqrt{2}}{\sqrt{2}}, \frac{1
- \sqrt{2}}{\sqrt{2}}, \frac{1 +
\sqrt{2}}{\sqrt{2}} $ \\
6 & $\{0,1,0,1\}$ & $ -1, 1 $ \\
7 & $\{0,0,2,0\}$ & $ -\sqrt{2}, 0, \sqrt{2} $ \\
8 & $\{0,0,1,1\}$ & $ -\frac{1}{\sqrt{2}},
\frac{1}{\sqrt{2}} $ \\
9 & $\{0,0,0,2\}$ & $0$ \\
\bottomrule
\end{tabular}
\caption[Eigenspace Overlaps]{A list of all $R_m$ eigenspaces for $N
= 2$ atoms at $M = 8$ sites. The third column displays the eigenvalues of
all the eigenstates of $\Bmin$ that lie in the given $R_m$.}
\label{tab:Rm}
\end{table}
In the third column we have also listed the eigenvalues of the $\Bmin$
eigenstates that lie within the given $R_m$.
We note that $ka = \pi/4$ will be degenerate with $ka = 3\pi/4$ since
$\sin(ka)$ is the same for both. Therefore, we already know that we
can combine $(R_0, R_2, R_7)$, $(R_1, R_5)$, and $(R_3, R_8)$, because
those combinations have the same $O_{\pi/4a} + O_{3\pi/4a}$
values. This is very clear in the table as these subspaces span
exactly the same values of $B_\mathrm{min}$, i.e.~the have exactly the
same values in the third column.
Now we have to match the parities. Subspaces that have the same parity
combination for the pair $(O_{\pi/4a} + O_{3\pi/4a}, O_{\pi/2a})$ will
be degenerate in $\mathcal{P}_M$. Note that we excluded $O_{\pi/a}$,
because as we discussed earlier they are all degenerate due to
$\sin(\pi) = 0$. Therefore, the (even,even) subspace will include
$(R_0, R_2, R_4, R_7, R_9)$, the (odd,even) will contain $(R_3, R_8)$,
the (even, odd) will contain $(R_6)$ only, and the (odd, odd) contains
$(R_1, R_5)$. These overlaps should be evident from the table as we
can see that these combinations combine all $R_m$ that share any
eigenstates of $\Bmin$ with the same eigenvalues.
Therefore, we have ended up with four distinct $\mathcal{P}_M$ subspaces
\begin{align}
\mathcal{P}_\mathrm{even,even} = & R_0 + R_2 + R_4 + R_7 + R_9
\nonumber \\
\mathcal{P}_\mathrm{odd,even} = & R_3 + R_8 \nonumber \\
\mathcal{P}_\mathrm{even,odd} = & R_6 \nonumber \\
\mathcal{P}_\mathrm{odd,odd} = & R_1 + R_5 \nonumber.
\end{align}
At this point it should be clear that these projectors satisfy all our
requirement. The conditions $\sum_M \mathcal{P}_M = 1$ and
$\mathcal{P}_M \mathcal{P}_N = \delta_{M,N} \mathcal{P}_M$ should be
evident from the form above. The commutator requirements are also
easily satisfied since the subspaces $R_m$ are of an operator that
commutes with both the Hamiltonian and the measurement operator. And
finally, one can also verify using the table that all of these
projectors are built from complete subspaces of $\Bmin$ (i.e.~each
subspace $P_m$ belongs to only one $\mathcal{P}_M$) and thus
$\mathcal{P}_M = \sum_{m \in M} P_m$.
\section{Conclusions}
In summary we have investigated measurement backaction resulting from
coupling light to an ultracold gas's phase-related observables. We
demonstrated how this can be used to prepare the Hamiltonian
eigenstates even if significant tunnelling is occuring as the
measurement can be engineered to not compete with the system's
dynamics. Furthermore, we have shown that when the observable of the
phase-related quantities does not commute with the Hamiltonian we
still project to a specific subspace of the system that is neither an
eigenspace of the Hamiltonian or the measurement operator. This is in
contrast to quantum Zeno dynamics \cite{misra1977, facchi2008,
raimond2010, raimond2012, signoles2014} or dissipative state
preparation \cite{diehl2008}. We showed how this projection is
essentially an extension of the measurement postulate to weak
measurement on dynamical systems where the competition between the two
processes is significant.

View File

@ -214,4 +214,8 @@
\renewcommand{\b}[1]{\mathbf{#1}}
\newcommand{\op}{\hat{o}}
\newcommand{\h}{\hat{h}}
\newcommand{\N}{\hat{N}}
\newcommand{\N}{\hat{N}}
\newcommand{\B}{\hat{B}}
\newcommand{\Bmax}{\hat{B}_\mathrm{max}}
\newcommand{\Bmin}{\hat{B}_\mathrm{min}}
\newcommand{\D}{\hat{D}}

View File

@ -21,6 +21,13 @@
year={2011},
school={LMU}
}
@phdthesis{StephenThesis,
title={Strongly correlated one-dimensional systems of cold atoms in
optical lattices },
author={Clark, Stephen Richard James Franz},
year={2007},
school={University of Oxford}
}
@misc{steck,
title={Rubidium 87 D line data},
author={Steck, Daniel A},
@ -58,6 +65,19 @@ year = {2010}
year={2012},
publisher={Springer Science \& Business Media}
}
@book{leggett,
title={Quantum liquids: Bose condensation and Cooper pairing in condensed-matter systems},
author={Leggett, Anthony J},
year={2006},
publisher={Oxford University Press}
}
@book{PitaevskiiStringari,
title={Bose-Einstein Condensation and Superfluidity},
author={Pitaevskii, Lev and Stringari, Sandro},
volume={164},
year={2016},
publisher={Oxford University Press}
}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Igor's original papers
@ -210,6 +230,18 @@ year = {2010}
year={2015},
publisher={APS}
}
@article{caballero2015njp,
title={Quantum properties of light scattered from structured
many-body phases of ultracold atoms in quantum
optical lattices},
author={Caballero-Benitez, Santiago F and Mekhov, Igor B},
journal={New Journal of Physics},
volume={17},
number={12},
pages={123023},
year={2015},
publisher={IOP Publishing}
}
@article{caballero2016,
title = {Quantum simulators based on the global collective light-matter interaction},
author = {Caballero-Benitez, Santiago F. and Mazzucchi, Gabriel and Mekhov, Igor B.},
@ -224,6 +256,13 @@ year = {2010}
doi = {10.1103/PhysRevA.93.063632},
url = {http://link.aps.org/doi/10.1103/PhysRevA.93.063632}
}
@article{caballero2016a,
title={Bond Order via Light-Induced Synthetic Many-body Interactions
of Ultracold Atoms in Optical Lattices},
author={Caballero-Benitez, Santiago F and Mekhov, Igor B},
journal={arXiv preprint arXiv:1604.02563},
year={2016}
}
@article{mazzucchi2016,
title = {Quantum measurement-induced dynamics of many-body ultracold
bosonic and fermionic systems in optical lattices},
@ -1040,3 +1079,149 @@ doi = {10.1103/PhysRevA.87.043613},
year={2008},
publisher={Nature Publishing Group}
}
@article{douglas2012,
title={Scattering-induced spatial superpositions in multiparticle localization},
author={Douglas, James S and Burnett, Keith},
journal={Physical Review A},
volume={86},
number={5},
pages={052120},
year={2012},
publisher={APS}
}
@article{douglas2013,
title={Scattering distributions in the presence of measurement backaction},
author={Douglas, James S and Burnett, Keith},
journal={Journal of Physics B: Atomic, Molecular and Optical Physics},
volume={46},
number={20},
pages={205301},
year={2013},
publisher={IOP Publishing}
}
@article{lee2014,
title={Classical stochastic measurement trajectories: Bosonic atomic
gases in an optical cavity and quantum measurement
backaction},
author={Lee, Mark D and Ruostekoski, Janne},
journal={Physical Review A},
volume={90},
number={2},
pages={023628},
year={2014},
publisher={APS}
}
@article{weinberg2016,
title={What happens in a measurement?},
author={Weinberg, Steven},
journal={Physical Review A},
volume={93},
number={3},
pages={032124},
year={2016},
publisher={APS}
}
@article{brune1992,
title={Manipulation of photons in a cavity by dispersive atom-field coupling: Quantum-nondemolition measurements and generation of ``Schr{\"o}dinger cat'' states},
author={Brune, M and Haroche, S and Raimond, J M and Davidovich, L and Zagury, N},
journal={Physical Review A},
volume={45},
number={7},
pages={5193},
year={1992},
publisher={APS}
}
@inproceedings{hubbard1963,
title={Electron correlations in narrow energy bands},
author={Hubbard, John},
booktitle={Proceedings of the royal society of london a:
mathematical, physical and engineering sciences},
volume={276},
number={1365},
pages={238--257},
year={1963},
organization={The Royal Society}
}
@article{jaksch1998,
title={Cold bosonic atoms in optical lattices},
author={Jaksch, Dieter and Bruder, Ch and Cirac, Juan Ignacio and Gardiner, Crispin W and Zoller, Peter},
journal={Physical Review Letters},
volume={81},
number={15},
pages={3108},
year={1998},
publisher={APS}
}
@article{greiner2002,
title={Quantum phase transition from a superfluid to a Mott
insulator in a gas of ultracold atoms},
author={Greiner, Markus and Mandel, Olaf and Esslinger, Tilman and
H{\"a}nsch, Theodor W and Bloch, Immanuel},
journal={nature},
volume={415},
number={6867},
pages={39--44},
year={2002},
publisher={Nature Publishing Group}
}
@article{krauth1991,
title={Bethe ansatz for the one-dimensional boson Hubbard model},
author={Krauth, Werner},
journal={Physical Review B},
volume={44},
number={17},
pages={9772},
year={1991},
publisher={APS}
}
@article{kolovsky2004,
title={Quantum chaos in the Bose-Hubbard model},
author={Kolovsky, Andrey R and Buchleitner, Andreas},
journal={EPL (Europhysics Letters)},
volume={68},
number={5},
pages={632},
year={2004},
publisher={IOP Publishing}
}
@article{calzetta2006,
title={Bose-Einstein-condensate superfluid--Mott-insulator transition in an optical lattice},
author={Calzetta, Esteban and Hu, Bei-Lok and Rey, Ana Maria},
journal={Physical Review A},
volume={73},
number={2},
pages={023610},
year={2006},
publisher={APS}
}
@article{leggett1999,
title={Superfluidity},
author={Leggett, Anthony J},
journal={Reviews of Modern Physics},
volume={71},
number={2},
pages={S318},
year={1999},
publisher={APS}
}
@article{fisher1989,
title={Boson localization and the superfluid-insulator transition},
author={Fisher, Matthew PA and Weichman, Peter B and Grinstein, G and Fisher, Daniel S},
journal={Physical Review B},
volume={40},
number={1},
pages={546},
year={1989},
publisher={APS}
}
@article{haldane1981,
title={Effective harmonic-fluid approach to low-energy properties of
one-dimensional quantum fluids},
author={Haldane, FDM},
journal={Physical Review Letters},
volume={47},
number={25},
pages={1840},
year={1981},
publisher={APS}
}

View File

@ -101,7 +101,7 @@
% To use choose `chapter' option in the document class
\ifdefineChapter
\includeonly{Chapter5/chapter5}
\includeonly{Chapter1/chapter1}
\fi
% ******************************** Front Matter ********************************