PHYSICAL REVIEW A 111, 013104 (2025)

Simulating feedback cooling of incoherent quantum mixtures

Kaiwen K. Zhu () ,* Zain Mehdi , Joseph J. Hope , and Simon A. Haine
Department of Quantum Science, Research School of Physics and Engineering,
The Australian National University, ACT 2600, Australia

(Received 7 September 2024; accepted 18 December 2024; published 9 January 2025)

We develop an approach for efficient and scalable simulations of measurement and control of quantum systems
built upon existing phase-space methods, namely, the truncated Wigner approximation.. We benchmark against
existing particle-filter methods by simulating measurement-based feedback cooling in a two-mode system, whose
low-dimensional nature permits a computation of an exact solution. The advantage of our method is multimode
scalability, which we demonstrate through the successful simulation of measurement-based feedback cooling of
an incoherent quasi-one-dimensional thermal ensemble to quantum degeneracy. As the underlying principle of
our approach exploits a general correspondence between measurement and coherent feedback, we anticipate it
is also applicable across a broad range of other quantum control scenarios.

DOI: 10.1103/PhysRevA.111.013104

I. INTRODUCTION a double-well potential, and second, feedback cooling of a

quasi-one-dimensional (quasi-1D) thermal Bose gas to quan-
Quantum control protocols deal with the manipulation of
tum degeneracy. The low dimensionality of the former permits
quantum systems in order to engineer desired quantum states
an exact solution via Kraus operators [29] against which we
or to enact real-time dynamical control of the system [1,2].
validate our solution, while the latter is an active field of
Such methods have been employed in proposals for quantum-
research and demonstrates the scalability of our approach.
enhanced metrology schemes [3,4], quantum computing [5,6],
Quantum systems may be mapped to phase-space dis-
as well as studying dynamics in circuit and cavity QED [7,8].
tributions [30,31], which under certain conditions can have
One example is measurement-based feedback control (MF),
their evolution mapped into a set of stochastic differential
where measurements performed on an ancillary system entan-
equations (SDEs) [28,32]. At the cost of requiring multi-
gled with the target system are used to alter the target state
ple trajectories and introducing sampling error, this approach
[2,9,10]. Experimentally, MF schemes have been employed to
scales like mean-field simulations while retaining the ability
realize deterministic spin squeezing and noise suppression in
to include many-body correlations. This enables simulations
collective spin systems [11–14], feedback cooling of single
of correlated quantum systems with nontrivial spatial struc-
atoms and micromechanical resonators [15,16], and, more
ture. In degenerate quantum gases, a common choice is the
recently, in preparing ensembles of ultracold atoms below the
Wigner representation and corresponding truncated Wigner
shot-noise limit [17].
approximation (TWA) [33,34]. The TWA has been shown to
There has been particular interest in extending dynamical
effectively produce a first-order correction to mean-field dy-
feedback cooling protocols to large-scale incoherent systems,
namics [35] and has been extensively used to model quantum
namely, feedback cooling thermal atomic gases to quantum
noise [34,36–38], spontaneous scattering [39,40], optical [41],
degeneracy in order to produce Bose-Einstein condensates
atom-atom [42–48], and atom-light [49–52] entanglement,
(BECs) or degenerate Fermi gases [18–26]. While the fun-
and thermal fluctuations in ultracold atomic gases [27,53–
damental description of such a system is well established,
55]. Quantum systems undergoing MF have conditional evo-
simulations that simultaneously account for the multimode
lution that is correlated with the stochastic measurement result
atomic dynamics, quantum correlations beyond mean-field ef-
[10,29], which can also be modeled using phase-space meth-
fects, and the backaction induced by the measurement process
ods [56], but the measurement induces weights that reduce
have proven elusive. This is due to the inherently large Hilbert
the relevance of trajectories that are not consistent with the
space and difficulties of capturing finite-temperature effects
measurement result. This can lead to significant sampling
associated with thermal fluctuations.
issues. Our approach addresses this problem by using phase-
In this work, we introduce a highly scalable field-theoretic
space methods to simulate the unconditional dynamics of an
approach based upon existing phase-space methods [27,28],
equivalent, higher-dimensional quantum system.
which is well suited for exploring the unconditional dynamics
of a wide range of controlled quantum systems. We ap-
ply our method to investigate MF schemes in two systems: II. MEASUREMENT AND COHERENT
first, damping Josephson oscillations of an atomic gas in FEEDBACK SCHEMES
A general quantum feedback control scheme consists of
the system to be controlled (e.g., atomic ensemble) coupled
*
Contact author: kaiwen.zhu@anu.edu.au to an ancillary system (e.g., light) such that information about

2469-9926/2025/111(1)/013104(9) 013104-1 ©2025 American Physical Society

ZHU, MEHDI, HOPE, AND HAINE PHYSICAL REVIEW A 111, 013104 (2025)

some system observable is encoded onto the ancilla. In MF, separated by interval τ , where the atoms  are entangled
an arbitrary positive operator-valued measure measurement with light pulses via Ĥent (t ) = h̄λ j ent ˆ †
j (t, t p )Jz b̂ j b̂ j ,
can always be mapped to projective measurements made on where λ is an atom-light coupling constant, b̂ j is the
the ancillary system, and the measurement result y is subse- annihilation operator corresponding to the jth pulse of
quently input to a controller whichalters the target system light and obeys [b̂ j , b̂†k ] = δ jk , and ent j (t, t p ) = 1/t p for
through a Hamiltonian term Ĥfb = j f j (y)Ôsj for some set t ∈ [t j , t j + t p ) is the unity normalized temporal mode
of system operators {Ôsj }. In coherent feedback (CF), the of each pulse beginning at t j = jτ with width t p  τ .
explicit measurement step is absent and control of the target This Hamiltonian generates correlations between the
system is mediated entirely through coherent interactions be- atomic number difference observable Jˆz with the phase
tween the target and ancillary, achieved by engineering Ĥfb quadrature of each light pulse defined as Ŷ j = i(b̂ j − b̂†j ).
such that the system-ancilla correlations affect the system in A projective measurement of Ŷ j yields a result y j which is
the desired way [1,57,58]. In both cases, imperfect measure- used to construct the estimate Jzest (t j ) = y j , where is
ment strength of the system observable, and measurement a scale factor chosen to minimize (Jˆz (t j ) − Jzest (t j ))2 ;
backaction are accounted for through the unitary dynamics for Glauber coherent states with (real) amplitude β0 ,
that produces the system-ancilla correlations. = (2λβ0 )−1 and b̂†j b̂ j  = |β0 |2 . We approximate ∂t Jzest
Importantly, it has been shown that for any given MF via differencing consecutive measurements, which produces
scheme, we may always find an equivalent CF scheme which 
the MF Hamiltonian Ĥfb = h̄ kfb j fbj (t, τ )(y j − y j−1 ),
generates the same dynamics for the controlled system [57].
where fbj (t, τ ) = 1/τ for t ∈ (t j + t p , t j+1 + t p ] is the
This is true only for unconditional dynamics, as the concept
temporal mode of the jth feedback operation with width τ
of a conditional state does not exist in CF, but this lack
starting after the jth entangling pulse.
of conditioning also means that phase-space simulations of
To derive TW equations for the system, we first obtain
the CF system are free of the sampling issues that constrain
the equivalent CF Hamiltonian via the replacement y j → Ŷ j
conditional simulations. This is precisely our approach: for a
quantum system undergoing MF, we use phase-space methods in Ĥfb . Applying operator correspondences and the standard
to model an equivalent CF scheme, recovering the uncon- truncation procedures [27,32] now produces the following set
ditional dynamics in the presence of feedback. We provide of SDEs for the Wigner phase-space variables [α1 , α2 , β j ]:
χ 
a detailed proof of the equivalency in Appendix C. We can
i∂t αm = (2|αm |2 − 1) + lm [λM(t ) − u(t )] αm + καn ,
obtain an equivalent CF scheme by retaining and coherently 2
coupling the ancillary
 state to the system through a Hamil- i∂t β j = −λent j (t, t p )Jz β j + i kfb  j (t, τ )Jz ,
fb
(1)
tonian Ĥfb → j f j (ŷ) ⊗ Ôsj , where ŷ is the operator of the
 ent lm = (−1) , the backaction term is M(t ) =
m
measured ancilla observable. Phase-space equations for the where
combined system may now be obtained by applying standard  (t, t p )( 1
− |β j | ),
2
and we define the feed-
j j 2 
operator correspondences, and the result is a set of equa- back signal u(t ) = kfb j fbj (t, τ )(y j − y j−1 ) where
tions for both system and ancilla phase-space variables, where y j = i(β j − β ∗j )|t=t j +t p . When t p is sufficiently short
feedback is implemented on each trajectory individually. This compared to the timescale set by the natural atomic
provides a stochastic simulation of the feedback-controlled dynamics, the latter remains approximately constant,
system that is valid in all regimes where the phase-space and the entangling dynamics may be solved analytically
simulations are valid, and can efficiently scale to systems with as β j (t j + t p ) = β j (t j ) exp(−iλJz (t j )) and αm (t j + t p ) =
many modes. αm (t j ) exp{−ilm λ(|β j (t j )|2 − 21 )}. From these solutions, it is
clear that measurement backaction arises from the quantum
intensity fluctuations in the light coupling into the relative
III. PEDAGOGICAL EXAMPLE phase of the atomic system.
IN A DOUBLE-WELL POTENTIAL
We demonstrate and validate our approach using a A. Example: Feedback cooling of a CSS
two-mode atomic ensemble described by the double-well We first simulate MF cooling of a spin-coherent state
Hamiltonian Ĥs = h̄χ [(â1† â1 )2 + (â2† â2 )2 ]/2 + h̄κ (â1† â2 + (CSS) [61,62] using three methods: our unconditional CF
†
â2 â1 ), where â j are the annihilation operators corresponding approach, an exact integration using Kraus operators, and
to each mode (well) for j ∈ {1, 2}, and χ and κ are the the number phase Wigner (NPW) particle filter, a leading
nonlinear interaction and tunneling constants, respectively. candidate for simulations of controlled quantum systems sub-
The system furnishes a representation of the su(2) algebra ject to numberlike measurements [56,63]. The latter two
and the feedback aims to extract energy by damping methods are conditional in nature and require an additional
out oscillations in the population difference between round of averaging over independent stochastic trajectories.
the wells represented by the pseudospin observable Briefly, the atomic state is conditioned following each mea-
Jˆz = (â1† â1 − â2† â2 )/2 [59,60]. This may be achieved by surement as ρ̂ = K̂ (y)ρ̂s K̂ † (y), where K̂ (y) = exp{−(y −
performing a measurement to estimate this quantity and 2β0 sin(λJˆz ))2 /4}/(2π )1/4 is the Kraus operator corresponding
subsequently applying the Hamiltonian Ĥfb = h̄u(t )Jˆz , where to the quadrature measurement result y. In the NPW solu-
u(t ) is the feedback signal u(t ) = kfb ∂t Jzest conditioned on tion, pulses of continuous stochastic weighted simulation are
the time derivative of the estimate (Jzest ) with gain parameter spaced between deterministic evolution. We provide details of
kfb . We model stroboscopic homodyne measurements the implementation for both methods in Appendix A.

013104-2
SIMULATING FEEDBACK COOLING OF INCOHERENT … PHYSICAL REVIEW A 111, 013104 (2025)

FIG. 1. Unconditional dynamics of the first two pseudospin moments for a CSS with pseudospin triplet √ [0, 0.4, −0.3] subject to MF
damping. Simulation parameters are χ = 0.01, κ = 0.09, λ = 0.8 × 10−4 , kfb = 0.1, N = 100, and β0 = 107 , and 62 measurements are
made over the MF interval. Our method, the exact solution, and NPW are plotted using solid lines, dashed lines, and markers, respectively; Jˆx ,
Jˆy , and Jˆz moments are plotted as red squares, blue triangles, and green circles, respectively. Shaded regions are bootstrapped 2σ -confidence
intervals in our method. (ai),(bi) Pseudospin means. The dashed colored lines are dynamics in the absence of MF, and the SU(2) Wigner
function at different times is also displayed in (bi). The feedback potential could be realized by varying the depth of each potential well, as
illustrated in the inset of (ai). (aii),(bii) Pseudospin variances. The purple line denotes the variance of a CSS on the Bloch sphere equator.

The evolution of the first two pseudospin moments (where below by 1/M, where M is the number of modes [65]. In two
Jˆx = (â1† â2 + â2† â1 )/2 and Jˆy = i(â1† â2 − â2† â1 )/2 encode co- modes, the matrix is Gnm 2M
= ân† âm  for n, m ∈ {1, 2}, so the
herences between the wells) is displayed in Fig. 1, where our TMCF is bounded below by 1/2.
approach exhibits quantitative agreement with both the exact The evolution of the TMCF (blue) and system energy
solution and NPW across both orders. We chose Hamiltonian (red) is plotted in Fig. 2(bi) for three different measurement
parameters that produced nontrivial dynamics while remain- strengths (different transparencies) which we parameterize
ing in the regime of validity for TWA where the truncation in terms of the single-shot measurement uncertainty Jˆz =
error is minimal; the dynamics in the absence of MF are 1/2λ|β0 |. For the optimal strength of Jˆz = 2.6, the TMCF
plotted using dashed lines. In the inset of Fig. 1(aii), the vari- grows rapidly from its minimal value to a steady-state value
ances of the conjugate observables (Jˆx , Jˆy ) increase following 2M
of ffrac = 0.960 ± 0.001, indicative of a highly pure state.
each measurement due to measurement backaction, though This growth is accompanied by a decrease in system en-
this behavior is absent in the variance of Jˆz (Fig. 1(bii)), which ergy, which approaches a steady-state energy close to that
approaches the steady state smoothly. This is consistent with of the ground state (Eground ) and corresponds to the scenario
a quantum nondemolition measurement, where there is no where the opposing mechanisms of measurement backaction,
backaction in the measured observable [9,64]. and feedback, which act to increase and extract energy from
the system, respectively, are balanced. In comparison, the
strengths of Jˆz = 12 and Jˆz = 0.3 are too weak and too
B. Example: Feedback cooling of a two-mode thermal state strong, respectively, and both lead to suboptimal cooling ef-
We now show that our approach is fully capable of working ficacy. The quality of the feedback signal is degraded in the
across incoherent regimes by simulating feedback cooling former, while the measurement backaction (and hence heat-
of a two-mode thermal state (described by ρ̂s = Î/(N + 1)), ing) is too large in the latter.
which loosely describes a mixture of particle distributions
across the double well. The pseudospin moments are shown
IV. FEEDBACK COOLING OF A BOSE GAS
in Fig. 2, where we again observe quantitative agreement
between our approach and the exact solution. A visual rep- Finally, we demonstrate the scalability of our CF approach
resentation of the state in terms of its SU(2) Wigner function by simulating MF cooling of a quasi-1D incoherent thermal
and sampled trajectories at different instances is displayed in mixture to a BEC via nondestructive phase-contrast imag-
Figs. 3(ai)–3(aiii), while Figs. 3(bi)–3(bii) show the evolution ing, which yield measurements of the spatial atomic density
of the Jˆy and Jˆz distributions. [19,26]. The feedback potential is proportional to the time
We quantify coherence by defining a two-mode condensate derivative of the spatial density estimates, and extracts energy
fraction (TMCF), borrowing from the Penrose-Onsager defi- from the system by damping out density fluctuations. Previous
2M
nition of the fraction ( ffrac ) defined as the largest eigenvalue studies have employed mean-field approximations [23,66] or
of the one-body density matrix ρ(x, x ) = x|ψ̂ † (x)ψ̂ (x )|x the NPW particle filter, which either assumes an unphysical
normalized by the trace, such that the fraction is bounded zero-temperature coherent ensemble as the initial state or is

013104-3
ZHU, MEHDI, HOPE, AND HAINE PHYSICAL REVIEW A 111, 013104 (2025)

FIG. 2. (ai),(aii),(bii),(cii) Pseudospin moments for a maximally mixed two-mode thermal state subject to MF cooling. The parameters and
plotting convention are the same as those used in Fig. 1 and the dynamics in the absence of MF are also plotted. (bi) Decrease and growth of
the system energy (red) and TMCF (blue), respectively, for three different measurement strengths parameterized in terms of the single-shot
measurement uncertainty Jˆz = 1/2λ|β0 |.

limited to extremely small particle numbers (N ≈ 1000) due form

to prohibitively large sampling requirements [20,21]. In all  
h̄λ
cases, a quantitative, dynamical study of the coherence growth ĤPC = − dx n̂(x) dx b̂† (x ) f ∗ (x − x )
across the BEC phase transition [67] which accounts for quan- tp
tum correlations is lacking, which we demonstrate is possible 
using our approach. × dx b̂(x ) f (x − x ), (2)
Representing the multimode atomic state and optical field
using the bosonic field operators ψ̂ (x) and b̂(x), respec- where n̂(x) = ψ̂ † (x)ψ̂ (x) is the atomic density operator and
tively, the Hamiltonian describing phase-contrast imaging f (x) = F −1 { f˜(k)} is the diffraction-limited blurring kernel
for a single measurement pulse of duration t p is of the characterized in Fourier space as f (k) ∝ exp(−rd4 |k|4 /16),
where rd is the diffraction limit determined by the ex-
perimental apparatus [19,60,68,69]. Applying our CF map-
ping, the fields (in terms of their Wigner phase-space
functionals ψ̂ (x) → ψ (x) and b̂(x) → β(x)) transform un-
der a single measurement as ψout (x) = ψin (x)e−iλβ0 G(x) and
βout (x) = βin (x)e−iλninf (x) , where G(x) ∝ dx (x ) f (x − x )
is the phase-scrambling stochastic backaction term aris-
ing from quantum fluctuations in the optical intensity
βin (x) = β0 + (x), where (x) is a complex Gaussian
∗
noise with  correlations (x) ( x) = δ(x − x )/2, and
ninf (x) ∝ dx f (x − x )nin (x ) is the diffraction-blurred den-
sity such that the quadrature signal Y (x) = i[βout (x) −
∗
βout (x)] ≈ 2λβ0 ninf (x) − 2Im[(x)] encodes a noisy estimate
of the spatial density given by nest (x) = Y (x)/2λβ0 . In prac-
tice, nest (x) is further convolved with a Gaussian kernel in
order to capture the finite imaging and control resolution,
which we denote by ñest (x). In the absence of measurement,
FIG. 3. (ai)–(aiii) The SU(2) Wigner function and sampled CF the dynamics of the atomic field are described by the cold-
trajectories represented on the Bloch sphere at the initial, intermedi- atom Hamiltonian [67] which generates the truncated-Wigner
ate, and final times. For clarity, only 2000 out of the 5000 simulated equation of motion [27],
trajectories are shown. (bi),(bii) Time evolution of the Jˆy and Jˆz
distributions computed from Wigner trajectories. The dashed lines i h̄∂t ψ (x) = [h(x) + g|ψ (x)|2 ]ψ (x), (3)
correspond to the intermediate time in (aii), and the broadening of the
P(Jˆy ) distribution at measurement times due to backaction is visible where h(x) = (−h̄2 ∂x2 /m + mω02 x 2 )/2 + Vfb (x) is the single-
in (bi). particle Hamiltonian describing harmonic trapping with

013104-4
SIMULATING FEEDBACK COOLING OF INCOHERENT … PHYSICAL REVIEW A 111, 013104 (2025)

FIG. 4. (ai) Decrease of the momentum variance and growth of the condensate fraction for a quasi-1D Bose gas with N ≈ 106 atoms
initialized in a thermal mixture subject to MF cooling over 60 trap cycles at a rate of 150 measurements per cycle with strength λβ0 = 3.7 ×
10−5 . Inset: The density profiles of the real, estimate, and smoothed estimates as well as the feedback potential constructed at an intermediate
time. (aii) Unconditional 1D column densities of the initial thermal state (red), an intermediate state (green) at a time corresponding to the inset
in (ai), and final condensed state (blue), which agrees well with the Thomas-Fermi profile (shaded) given by nTF (x) = [μ − V (x)]/N, where
μ = 15.13 h̄ω0 is the chemical potential. Inset: A single Wigner trajectory for the initial (thermal) and final (cooled) states. Shaded regions
represent 3σ -confidence intervals. (bi),(bii) Time evolution of the spatial density (n(x)) and first-order correlation function (g1 (x, 0)) across
the ensemble, respectively.

feedback potential Vfb (x) = h̄kfb ∂t ñest (x), and g is the inter- field-theoretic approach starting with an incoherent thermal
atomic interaction strength. ensemble. This was only possible using our unconditional CF
In order to initialize trajectories of ψ (x), samples of a ther- approach, which we have shown to be scalable to multimode,
mal state are obtained by propagating the stochastic projected large-scale atomic systems where both traditional open-
Gross-Pitaevskii equation (SPGPE) to equilibrium, which are quantum system methods and particle-filter methods fail. Our
the phase-space equations corresponding to a grand-canonical approach offers a significant reduction in computational com-
description accounting for energy and number-exchanging plexity; the two-mode benchmark using 500 fictitious and
interactions between the atomic gas and a thermal reservoir real NPW trajectories [56,63] exhausted 100 hours of CPU
[53,70,71]; we provide further details in Appendix B. We time, while our method using 5000 unconditional trajectories
simulate feedback cooling in 1D with stroboscopic density completed in five minutes, which is a 1200-fold speedup.
measurements for an ensemble of N ≈ 1.1 × 106 atoms with It may seem counterintuitive that a computational speedup
interaction strength g = 0.0001 √ h̄ω0 x0 and diffraction limit is achieved despite simulating a higher-dimensional quan-
rd = 0.52 x0 , where x0 = h̄/mω0 is the natural oscillator tum system. However, as phase-space representations retain
length scale. These values of rd and g are consistent with the mean-field scalability, the increase in complexity due to the
use of 87 Rb atoms with trapping frequency ω0 = 2π × 13 Hz. introduction of the ancillary field is negligible compared to
Figure 4(ai) shows the growth in condensate fraction from the speedup obtained from avoiding stochastic integration al-
ffrac = 5.8 ± 0.5% to ffrac = 98.77 ± 0.03% as well as a 70- gorithms and resampling algorithms needed to address rapid
fold reduction in the momentum variance over 60 trap periods, sample impoverishment inherent to conditional particle-filter
signaling the formation of an extremely pure condensate. The techniques [72,73]. As the latter is exponentially exasperated
evolution of the first-order correlation function g1 (x, 0) is also for large fields, particle-filter methods are fundamentally un-
displayed in Fig. 4(bii), where it approaches unity across the scalable to high-dimensional systems.
spatial extent of the BEC in the steady state, which is an The standard limitations common to all phase-space meth-
excellent demonstration of off-diagonal long-range order [67]. ods still apply. The Wigner function must remain positive
Figure 4(aii) shows the density profiles of the initial (thermal) under time evolution to ensure the quantum state is well
and final (condensed) states, the latter agreeing well with the represented in terms of trajectories, which is guaranteed for
approximate Thomas-Fermi (T = 0 K) ground-state profile, initially positive states evolving under truncated equations of
as expected for a strongly interacting atomic BEC. motion [27,35,74].
Our findings represent an instance where feedback cool- The natural extension of this work is to extend the MF
ing of atomic gases has been modeled using a full quantum scheme for the quasi-1D Bose gas to higher-dimensional

013104-5
ZHU, MEHDI, HOPE, AND HAINE PHYSICAL REVIEW A 111, 013104 (2025)

geometries; though our results are promising and may of- is obtained by integrating over the continuous measurement
√
fer insight towards optimal parameter regimes for efficient current dy = Jˆz dt + dW/(2 γ ) in order to produce
cooling, extensive variables such as the final BEC size or  t +t NPW
JzNPW
est
= ( t00 p dy)/t pNPW , where γ is the measurement
condensate fraction may only be interpreted as physical ther-
rate parameter which satisfies the equation γ t pNPW = λ2 β02 .
modynamic quantities in 3D [67,75]. Our approach may also
This equality may be obtained by either equating the CF and
be adapted to incorporate measurement-induced spontaneous
NPW backaction terms (i.e., Eqs. (1) and (A1)) and enforcing
emission losses [51], which is the primary limiting factor in
that the noise terms ◦ dV and  j obey the same distribution,
the efficacy of feedback cooling from high-energy thermal
where  j is a complex Gaussian noise characterized by
states [26]. Finally, as our approach fundamentally exploits
(t ) j (t )∗k  = (t − t )δ jk /2 which appears in the initial
a general correspondence between measurement and coherent
sampling of the light variables β j = β0 +  j , or by equating
feedback, its validity does not rely on details of any particular
the signal-to-noise ratio of a single-shot measurement.
feedback scheme. Combined with the built-in scalability and
In practice, we evolve a swarm of 500 fictitious trajectories
the ability to generalize our approach to model continuous
(each with their own noise dV j ) under the same measurement
measurements, we anticipate it may be used in other physical
noise dW to simulate a single conditional trajectory. Over the
systems, such as quantum optomechanics [76], as well as
measurement period, trajectory weights decay exponentially
across the broader field of quantum control scenarios, such as
with respect to the difference between their Jz values and the
reservoir engineering [77–79], state preparation [80,81], and
measurement record. This rapidly leads to a single trajectory
active stabilization of quantum systems [82,83].
dominating, and so the state is severely undersampled and
not well represented. This is referred to in the literature as
ACKNOWLEDGMENTS particle impoverishment [85] and we employ the Kitigawa
We thank Australian Research Council Project No. sequential importance resampling algorithm to address this
DP190101709, the National Computational Infrastructure for problem [72].
access to computing resources, Stuart Szigeti for insightful Kraus operator implementation. Let the system (ρ̂s ) and
discussions, and Isabelle Savill-Brown for careful proofread- ancilla (ρ̂a ) be the atomic ensemble and light, respectively. An
arbitrary atom-light state (ρ̂0 = ρ̂s ⊗ ρ̂a ) is expressed in the
ing of this manuscript. We acknowledge the Ngunnawal 
people as the traditional custodians of the land upon which Dicke basis as ρ̂0 = N/2 n,m=−N/2 ρnm |nm| ⊗ |β0 β0 |, where
this research was conducted, and recognize that sovereignty ρnm are coefficients for the atomic state and we have chosen
was never ceded. S.A.H. acknowledges support through an the light as a coherent state with amplitude β0 . The posten-
Australian Research Council Future Fellowship Grant No. tanglement state is given by ρ̂ent = Ûent ρ̂0Ûent
†
, where Ûent =
FT210100809. K.Z. is thankful for the support provided by exp(−iλJz ⊗ b̂ b̂), and is equal to
ˆ †

the D.N.F. Dunbar Honours Scholarship and Australian Gov- 

ernment Research Training Program. K.Z. thanks Jun Hua for ρ̂ent = ρnm |nm| ⊗ |β0 eiλm β0 eiλn |. (A2)
n,m
insightful discussion on the scope of this work.
Prior to each measurement, we construct the probability dis-
DATA AVAILABILITY tribution P(y, t ) = Trl [ρ̂ent (t )|yy|], which takes the form
N/2 
The source code and data for all simulations presented in ρnn (t ) 1
this paper may be found at [84]. P(y, t ) = √ exp − [y − 2β0 sin(λn)]2 , (A3)
n=−N/2
π 2

APPENDIX A: SIMULATION IMPLEMENTATION from which we generate a single measurement result y

DETAILS using rejection sampling. To condition ρ̂ent onto result
y, we trace out the light and apply the Kraus oper-
Number-phase Wigner implementation. The atomic NPW
ator K̂ (y) = exp{−(y − 2β0 sin(λJˆz ))2 /4}/(2π )1/4 to obtain
equations closely resemble Eq. (1) once the optical terms (β j )
ρ̂y = K̂ (y)Trl (ρ̂ent )K̂ † (y) with appropriate renormalization,
are discarded. The Hamiltonian terms are identical, while the
which is taken as the new atomic state at the begin-
backaction and weights evolution are as follows:
ning of the next time step. Between measurements, the
√
i γ state is evolved as ρ̂a (t + t ) = Û (t )ρ̂a (t )Û † (t ), where
dαmj = . . . − lm αmj ◦ dV j , Û (t ) = exp(−i(Ĥs + Ĥfb )t/h̄).
2
TWA implementation. The backaction term (M(t )) in
dω j √
= −2γ (Jz − Jˆz )2 dt + 2 γ Jz ◦ dW, (A1) Eq. (1) imprinted onto the relative phase of the atomic system
ωj comprises a deterministic and stochastic rotation about the
where ◦ dV and ◦ dW are real Wiener noises in Jˆz axis due to the coherent amplitude and quantum intensity
the
 jStratonovich  j calculus and  f (â) = Ec [ f (α)] = noise. In practice, we can always cancel out the former by
j ω f (α )/
j
jω are conditional expectation values applying a counter rotation with magnitude λ|β0 |2 after each
over a swarm of fictitious trajectories indexed by j. For entangling pulse.
each measurement, we evolve Eq. (A1) continuously over The SU(2) subspace is number conserving, while TW sam-
a period t pNPW such that t pNPW /dtCF  1, which allows us ples capture number fluctuations. To ensure fair comparison
to update the trajectories instantaneously over a single dtCF between the CF and exact solutions, we manually normalize
time step. The measurement result over a single pulse t pNPW trajectories to N, i.e., |α1j |2 + |α2j |2 = N.

013104-6
SIMULATING FEEDBACK COOLING OF INCOHERENT … PHYSICAL REVIEW A 111, 013104 (2025)

APPENDIX B: SPGPE THERMAL STATE GENERATION Under MF, we apply the conditional feedback Hamiltonian
Ĥfb (y) = j f j (y)Ôsj to ρ̂s (y) for some time t to obtain
Quasi-1D thermal state initialization. The complete
SPGPE accounts for both energy- and number-exchanging the conditional feedback state ρ̂sfb (y) = ÛfbMF (y)ρ̂s (y)ÛfbMF† (y),
processes between a coherent “C field” and thermal reservoir where ÛfbMF (y) = exp(−iĤfb (y)t/h̄). The unconditional
[55,71]. For initializing thermal trajectories, it is sufficient to feedback state is recovered by averaging over all conditional
neglect the energy-damping terms, and the remaining terms states,
form the simple-growth SPGPE, 
dψ|SG = · · · + P{γ (μ − L)ψdt + dWγ (x, t )}, (B1) ρ̂s,fbMF = dy P(y)ÛfbMF (y)ρ̂s (y)ÛfbMF† (y), (C2)

where we have defined the Liouvillian operator Lψ =

(ĤCA + g|ψ|2 ψ )ψ and dWγ (x, t ) is a complex Wiener noise where P(y) = N (y)−1 is the probability distribution of mea-
characterized by the correlations dWγ∗ (x, t )dWγ (x , t ) = surement results.
In CF, no projective measurement is performed on the
2γ T̃ δ(x − x )dt, where γ and T̃ = kb T /h̄ω0 are the growth
ancillary.  Instead, an unconditional feedback Hamiltonian
rate and dimensionless temperature parameters, respectively.
Ĥfb (ŷ) = j f j (ŷ) ⊗ Ôsj which acts on the combined system-
To obtain thermal samples, we evolve Eq. (B1) to equilibrium
ancillary space is directly applied to the postentanglement
in the Hermite-Gauss basis with the projector P{} restricting
state (ρ̂ent ) to obtain the unconditional combined feed-
evolution into n = 100 modes.
back state ρ̂ fb = ÛfbCFÛent (ρ̂s ⊗ ρ̂a )Ûent
†
ÛfbCF† , where ÛfbCF =
In theory, increasing T̃ should lead to thermal states with
near-zero condensate fraction. However, the 1D reduction pre- exp(−it Ĥfb t/h̄). We now show that the unconditional
vents the existence of “true” thermal states, and we find an feedback state for the system is identical to that of the MF
effective lower bound of ffrac ≈ 5–6% regardless of parameter scheme in Eq. (C2). First, expanding ÛfbCF in the spectral basis

choice or the number of modes, n. As the simple-growth of the ancillary yields ÛfbCF = exp(−it j f j (ŷ) ⊗ Ôsj /h̄) =
  
SPGPE realizes a grand-canonical description, we adjust the dy exp(−it j f j (y)Ôsj /h̄) ⊗ |yy| = dy ÛfbMF (y) ⊗
particle number (N) at equilibrium simply by varying param- |yy|. Tracing over the ancillary noting that [ŷ, Ôsj ] = 0, we
eters μ and T̃ . obtain

APPENDIX C: CF AND MF CORRESPONDENCE ρ̂s, CF = dy1 y1 |ÛfbCFÛent (ρ̂s ⊗ ρ̂a )Ûent
fb †
ÛfbCF† |y1 
Equivalence of MF and CF unconditional state. We 
now show that a single iteration of MF and CF leads = dyy1 |y2 ÛfbMF (y2 )K̂ (y2 )ρ̂s K̂ † (y3 )ÛfbMF† (y3 )y3 |y1 
to identical unconditional states for the system. Gen-

eralizing Eq. (A1), the combined system-ancilla state
postentanglement is described by ρ̂ent = Ûent (ρ̂s ⊗ ρ̂a )Ûent
†
. = dy N (y)−1ÛfbMF (y)[N (y)K̂ (y)ρ̂s K̂ † (y)]ÛfbMF† (y)
Assuming a pure state for the ancillary (i.e., ρ̂a = |ψa ψa |), 
a projective measurement of an ancillary observable (ŷ) con- = dy P(y)ÛfbMF (y)ρ̂s (y)ÛfbMF† (y) ≡ ρ̂s,fbMF . (C3)
ditions the system state (onto the measurement result defined
by ŷ|y = y|y),
Hence, under our mapping Ĥfb (y) → Ĥfb (ŷ), the uncondi-
ρ̂s (y) = N (y)Tra (|yy|ρ̂ent |yy|) tional system state in MF obtained after a single (and, by
extension, a series of) feedback iteration(s) may be directly
= N (y)ψa |Ûent |yρ̂s y|Ûent
†
|ψa . (C1)
recovered from our proposed CF scheme. We emphasize that
We have traced out the ancillary and introduced the nor- the equivalent CF scheme need not be experimentally realis-
malization factor N (y) = Tr(K̂ † (y)K̂ (y)ρ̂s )−1 , where K̂ (y) = tic; as long as the Hamiltonian is Hermitian, it is, in principle,
ψa |Ûent |y is an equivalent expression for the Kraus operator. physical.

[1] J. Zhang, Y.-X. Liu, R.-B. Wu, K. Jacobs, and F. Nori, Quantum [5] H. M. Wiseman and L. Bouten, Optimality of feedback con-
feedback: Theory, experiments, and applications, Phys. Rep. trol strategies for qubit purification, Quantum Inf. Proc. 7, 71
Quantum Feedback: Theor. Expt. Applic. 679, 1 (2017). (2008).
[2] K. Jacobs and A. P. Lund, Feedback control of nonlinear quan- [6] A. Vepsäläinen, R. Winik, A. H. Karamlou, J. Braumüller,
tum systems: A rule of thumb, Phys. Rev. Lett. 99, 020501 A. D. Paolo, Y. Sung, B. Kannan, M. Kjaergaard, D. K.
(2007). Kim, A. J. Melville, B. M. Niedzielski, J. L. Yoder, S.
[3] L. Pezzè, A. Smerzi, M. K. Oberthaler, R. Schmied, and Gustavsson, and W. D. Oliver, Improving qubit coher-
P. Treutlein, Quantum metrology with nonclassical states of ence using closed-loop feedback, Nat. Commun. 13, 1932
atomic ensembles, Rev. Mod. Phys. 90, 035005 (2018). (2022).
[4] S. S. Szigeti, O. Hosten, and S. A. Haine, Improving cold-atom [7] J. E. Reiner, W. P. Smith, L. A. Orozco, H. M. Wiseman, and
sensors with quantum entanglement: Prospects and challenges, J. Gambetta, Quantum feedback in a weakly driven cavity QED
Appl. Phys. Lett. 118, 140501 (2021). system, Phys. Rev. A 70, 023819 (2004).

013104-7
ZHU, MEHDI, HOPE, AND HAINE PHYSICAL REVIEW A 111, 013104 (2025)

[8] W. Cui and F. Nori, Feedback control of Rabi oscillations in [27] P. Blakie†, A. Bradley†, M. Davis, R. Ballagh, and C.W.
circuit QED, Phys. Rev. A 88, 063823 (2013). Gardiner, Dynamics and statistical mechanics of ultra-cold
[9] H. M. Wiseman and G. J. Milburn, Quantum theory of field- Bose gases using c-field techniques, Adv. Phys. 57, 363
quadrature measurements, Phys. Rev. A 47, 642 (1993). (2008).
[10] K. Jacobs, Quantum Measurement Theory and Its Applications [28] C. W. Gardiner and P. Zoller, Quantum Noise, Springer Series
(Cambridge University Press, Cambridge, 2014). in Synergetics (Springer, Berlin, 2010).
[11] K. C. Cox, G. P. Greve, J. M. Weiner, and J. K. Thompson, [29] H. M. Wiseman and G. J. Milburn, Quantum Measurement and
Deterministic squeezed states with collective measurements and Control (Cambridge University Press, Cambridge, 2009).
feedback, Phys. Rev. Lett. 116, 093602 (2016). [30] R. P. Rundle and M. J. Everitt, Overview of the phase space
[12] L. K. Thomsen, S. Mancini, and H. M. Wiseman, Spin squeez- formulation of quantum mechanics with application to quantum
ing via quantum feedback, Phys. Rev. A 65, 061801(R) (2002). technologies, Adv. Quantum Technol. 4, 2100016 (2021).
[13] R. Inoue, Shin-Ichi-Ro Tanaka, R. Namiki, T. Sagawa, and [31] H.-W. Lee, Theory and application of the quantum phase-space
Y. Takahashi, Unconditional quantum-noise suppression via distribution functions, Phys. Rep. 259, 147 (1995).
measurement-based quantum feedback, Phys. Rev. Lett. 110, [32] C. W. Gardiner, Handbook of Stochastic Methods for Physics,
163602 (2013). Chemistry, and the Natural Sciences (Springer-Verlag, Berlin,
[14] O. Hosten, N. J. Engelsen, R. Krishnakumar, and M. A. Ka- 1985).
sevich, Measurement noise 100 times lower than the quantum- [33] E. Wigner, On the quantum correction for thermodynamic equi-
projection limit using entangled atoms, Nature (London) 529, librium, Phys. Rev. 40, 749 (1932).
505 (2016). [34] P. D. Drummond and B. Opanchuk, Truncated Wigner dynam-
[15] J. Guo, R. Norte, and S. Gröblacher, Feedback cooling of a ics and conservation laws, Phys. Rev. A 96, 043616 (2017).
room temperature mechanical oscillator close to its motional [35] A. Polkovnikov, Phase space representation of quantum dynam-
ground state, Phys. Rev. Lett. 123, 223602 (2019). ics, Ann. Phys. 325, 1790 (2010).
[16] C. Schäfermeier, H. Kerdoncuff, U. B. Hoff, H. Fu, A. Huck, [36] M. J. Steel, M. K. Olsen, L. I. Plimak, P. D. Drummond, S. M.
J. Bilek, G. I. Harris, W. P. Bowen, T. Gehring, and U. L. Tan, M. J. Collett, D. F. Walls, and R. Graham, Dynamical quan-
Andersen, Quantum enhanced feedback cooling of a mechan- tum noise in trapped Bose-Einstein condensates, Phys. Rev. A
ical oscillator using nonclassical light, Nat. Commun. 7, 13628 58, 4824 (1998).
(2016). [37] A. Sinatra, C. Lobo, and Y. Castin, The truncated Wigner
[17] M. Gajdacz, A. J. Hilliard, M. A. Kristensen, P. L. Pedersen, method for Bose-condensed gases: Limits of validity and ap-
C. Klempt, J. J. Arlt, and J. F. Sherson, Preparation of ultracold plications, J. Phys. B: At., Mol. Opt. Phys. 35, 3599 (2002).
atom clouds at the shot noise level, Phys. Rev. Lett. 117, 073604 [38] S. A. Haine, Quantum noise in bright soliton matter-wave inter-
(2016). ferometry, New J. Phys. 20, 033009 (2018).
[18] S. A. Haine, A. J. Ferris, J. D. Close, and J. J. Hope, Control of [39] A. A. Norrie, R. J. Ballagh, and C. W. Gardiner, Quantum turbu-
an atom laser using feedback, Phys. Rev. A 69, 013605 (2004). lence and correlations in Bose-Einstein condensate collisions,
[19] S. S. Szigeti, M. R. Hush, A. R. R. Carvalho, and J. J. Hope, Phys. Rev. A 73, 043617 (2006).
Continuous measurement feedback control of a Bose-Einstein [40] S. A. Haine and A. J. Ferris, Surpassing the standard quantum
condensate using phase-contrast imaging, Phys. Rev. A 80, limit in an atom interferometer with four-mode entanglement
013614 (2009). produced from four-wave mixing, Phys. Rev. A 84, 043624
[20] M. R. Hush, S. S. Szigeti, A. R. R. Carvalho, and J. J. Hope, (2011).
Controlling spontaneous-emission noise in measurement-based [41] P. D. Drummond and A. D. Hardman, Simulation of Quantum
feedback cooling of a Bose-Einstein condensate, New J. Phys. Effects in Raman-Active Waveguides, Europhys. Lett. 21, 279
15, 113060 (2013). (1993).
[21] M. L. Goh, Z. Mehdi, R. L. Taylor, R. J. Thomas, A. S. Bradley, [42] S. A. Haine and M. T. Johnsson, Dynamic scheme for
M. R. Hush, J. J. Hope, and S. S. Szigeti, Feedback cooling generating number squeezing in Bose-Einstein condensates
Bose gases to quantum degeneracy, arXiv:2206.05069. through nonlinear interactions, Phys. Rev. A 80, 023611
[22] H. M. Hurst, S. Guo, and I. B. Spielman, Feedback induced (2009).
magnetic phases in binary Bose-Einstein condensates, Phys. [43] B. Opanchuk, M. Egorov, S. Hoffmann, A. I. Sidorov, and
Rev. Res. 2, 043325 (2020). P. D. Drummond, Quantum noise in three-dimensional BEC
[23] E. P. Yamaguchi, H. M. Hurst, and I. B. Spielman, Feedback- interferometry, Europhys. Lett. 97, 50003 (2012).
cooled Bose-Einstein condensation: Near and far from equilib- [44] L. Slodička, G. Hétet, N. Röck, P. Schindler, M. Hennrich, and
rium, Phys. Rev. A 107, 063306 (2023). R. Blatt, Atom-atom entanglement by single-photon detection,
[24] J. T. Young, A. V. Gorshkov, and I. B. Spielman, Feedback- Phys. Rev. Lett. 110, 083603 (2013).
stabilized dynamical steady states in the Bose-Hubbard model, [45] R. J. Lewis-Swan and K. V. Kheruntsyan, Sensitivity to thermal
Phys. Rev. Res. 3, 043075 (2021). noise of atomic Einstein-Podolsky-Rosen entanglement, Phys.
[25] M. H. Muñoz-Arias, P. M. Poggi, P. S. Jessen, and I. H. Deutsch, Rev. A 87, 063635 (2013).
Simulating nonlinear dynamics of collective spins via quan- [46] S. A. Haine, J. Lau, R. P. Anderson, and M. T. Johnsson,
tum measurement and feedback, Phys. Rev. Lett. 124, 110503 Self-induced spatial dynamics to enhance spin squeezing via
(2020). one-axis twisting in a two-component Bose-Einstein conden-
[26] Z. Mehdi, S. A. Haine, J. J. Hope, and S. S. Szigeti, Fundamen- sate, Phys. Rev. A 90, 023613 (2014).
tal limits of feedback cooling ultracold atomic gases, Phys. Rev. [47] S. P. Nolan, J. Sabbatini, M. W. J. Bromley, M. J. Davis, and
Lett. 133, 073401 (2024). S. A. Haine, Quantum enhanced measurement of rotations with

013104-8
SIMULATING FEEDBACK COOLING OF INCOHERENT … PHYSICAL REVIEW A 111, 013104 (2025)

a spin-1 Bose-Einstein condensate in a ring trap, Phys. Rev. A [67] C. J. Pethick and H. Smith, Bose–Einstein Condensation in
93, 023616 (2016). Dilute Gases, 2nd ed. (Cambridge University Press, Cambridge,
[48] S. S. Szigeti, S. P. Nolan, J. D. Close, and S. A. Haine, High- New York, 2008).
precision quantum-enhanced gravimetry with a Bose-Einstein [68] Z. Mehdi, Superfluid dissipation and feedback cooling in
condensate, Phys. Rev. Lett. 125, 100402 (2020). ultracold atomic gases, Ph.D. thesis, Australian National Uni-
[49] S. A. Haine, Information-recycling beam splitters for quantum versity, 2024.
enhanced atom interferometry, Phys. Rev. Lett. 110, 053002 [69] K. Zhu, Feedback cooling of degenerate quantum gases, Thesis,
(2013). Australian National University, 2023.
[50] S. A. Haine and Wing Yung Sarah Lau, Generation of atom- [70] S. J. Rooney, P. B. Blakie, and A. S. Bradley, Numerical method
light entanglement in an optical cavity for quantum enhanced for the stochastic projected Gross-Pitaevskii equation, Phys.
atom interferometry, Phys. Rev. A 93, 023607 (2016). Rev. E 89, 013302 (2014).
[51] M. Kritsotakis, J. A. Dunningham, and S. A. Haine, Spin [71] A. S. Bradley, S. J. Rooney, and R. G. McDonald, Low-
squeezing of a Bose-Einstein condensate via a quantum dimensional stochastic projected Gross-Pitaevskii equation,
nondemolition measurement for quantum-enhanced atom inter- Phys. Rev. A 92, 033631 (2015).
ferometry, Phys. Rev. A 103, 023318 (2021). [72] G. Kitagawa, Monte Carlo filter and smoother for non-gaussian
[52] L. A. Fuderer, J. J. Hope, and S. A. Haine, Hybrid method nonlinear state space models, J. Comput. Graph. Stat. 5, 1
of generating spin-squeezed states for quantum-enhanced atom (1996).
interferometry, Phys. Rev. A 108, 043722 (2023). [73] S. Kiesewetter and P. D. Drummond, Coherent Ising machine
[53] P. B. Blakie and M. J. Davis, Projected Gross-Pitaevskii equa- with quantum feedback: The total and conditional master equa-
tion for harmonically confined Bose gases at finite temperature, tion methods, Phys. Rev. A 106, 022409 (2022).
Phys. Rev. A 72, 063608 (2005). [74] M. K. Olsen and A. S. Bradley, Numerical representation of
[54] C. N. Weiler, T. W. Neely, D. R. Scherer, A. S. Bradley, M. J. quantum states in the positive-P and Wigner representations,
Davis, and B. P. Anderson, Spontaneous vortices in the forma- Opt. Commun. 282, 3924 (2009).
tion of Bose–Einstein condensates, Nature (London) 455, 948 [75] L. P. Pitaevskii, S. Stringari, L. P. Pitaevskii, and S. Stringari,
(2008). Bose-Einstein Condensation, International series of mono-
[55] S. J. Rooney, P. B. Blakie, and A. S. Bradley, Stochastic graphs on physics (Oxford University Press, Oxford, 2003).
projected Gross-Pitaevskii equation, Phys. Rev. A 86, 053634 [76] M. Aspelmeyer, T. J. Kippenberg, and F. Marquardt, Cavity
(2012). optomechanics, Rev. Mod. Phys. 86, 1391 (2014).
[56] M. R. Hush, Efficient simulation of controlled large quantum [77] J. F. Poyatos, J. I. Cirac, and P. Zoller, Quantum reservoir
systems, Ph.D. thesis, Australian National University, 2012. engineering with laser cooled trapped ions, Phys. Rev. Lett. 77,
[57] K. Jacobs, X. Wang, and H. M. Wiseman, Coherent feedback 4728 (1996).
that beats all measurement-based feedback protocols, New J. [78] D. Basilewitsch, F. Cosco, N. L. Gullo, M. Möttönen, T. Ala-
Phys. 16, 073036 (2014). Nissilä, C. P. Koch, and S. Maniscalco, Reservoir engineering
[58] S. Lloyd, Coherent quantum feedback, Phys. Rev. A 62, 022108 using quantum optimal control for qubit reset, New J. Phys. 21,
(2000). 093054 (2019).
[59] J. F. Corney and G. J. Milburn, Homodyne measurements [79] B. Tissot, H. Ribeiro, and F. Marquardt, Reservoir engineering
on a Bose-Einstein condensate, Phys. Rev. A 58, 2399 for classical nonlinear fields, Phys. Rev. Res. 6, 023015 (2024).
(1998). [80] R. van Handel, J. K. Stockton, and H. Mabuchi, Modelling and
[60] D. A. R. Dalvit, J. Dziarmaga, and R. Onofrio, Continu- feedback control design for quantum state preparation, J. Opt.
ous quantum measurement of a Bose-Einstein condensate: A B: Quantum Semiclass. Opt. 7, S179 (2005).
stochastic Gross-Pitaevskii equation, Phys. Rev. A 65, 053604 [81] A. R. Mills, C. R. Guinn, M. M. Feldman, A. J. Sigillito, M. J.
(2002). Gullans, M. T. Rakher, J. Kerckhoff, C. A. C. Jackson, and
[61] C. Gross, Spin squeezing, entanglement and quantum metrol- J. R. Petta, High-fidelity state preparation, quantum control, and
ogy with Bose-Einstein condensates, J. Phys. B: At., Mol. Opt. readout of an isotopically enriched silicon spin qubit, Phys. Rev.
Phys. 45, 103001 (2012). Appl. 18, 064028 (2022).
[62] F. T. Arecchi, E. Courtens, R. Gilmore, and H. Thomas, Atomic [82] C. Sayrin, I. Dotsenko, X. Zhou, B. Peaudecerf, T. Rybarczyk,
Coherent States in Quantum Optics, Phys. Rev. A 6, 2211 S. Gleyzes, P. Rouchon, M. Mirrahimi, H. Amini, M. Brune,
(1972). J.-M. Raimond, and S. Haroche, Real-time quantum feedback
[63] M. R. Hush, A. R. R. Carvalho, and J. J. Hope, Number-phase prepares and stabilizes photon number states, Nature (London)
Wigner representation for efficient stochastic simulations, Phys. 477, 73 (2011).
Rev. A 81, 033852 (2010). [83] J. Guo, J. Chang, X. Yao, and S. Gröblacher, Active-feedback
[64] E. O. Ilo-Okeke and T. Byrnes, Theory of single-shot phase quantum control of an integrated low-frequency mechanical
contrast imaging in spinor Bose-Einstein condensates, Phys. resonator, Nat. Commun. 14, 4721 (2023).
Rev. Lett. 112, 233602 (2014). [84] K. Zhu, Feedback Cooling Code, https://github.
[65] O. Penrose and L. Onsager, Bose-einstein condensation and com/kozse/Feedback-Cooling-of-Incoherent-Quantum-
liquid helium, Phys. Rev. 104, 576 (1956). Mixtures/tree/main (unpublished).
[66] S. S. Szigeti, Controlled Bose-condensed sources for atom [85] T. Li, T. P. Sattar, and S. Sun, Deterministic resampling: Un-
interferometry, Ph.D. thesis, Australian National University, biased sampling to avoid sample impoverishment in particle
2013. filters, Signal Proc. 92, 1637 (2012).

013104-9