Research Article Vol. 2, No.

5 / 25 October 2024 / Optica Quantum 371

Multiqubit quantum state preparation enabled

by topology optimization
A. Miguel-Torcal,1,2 A. González-Tudela,3 F. J. García-Vidal,1,2,4
1,2,∗
AND A. I. Fernández-Domínguez
1
Departamento de Física Teórica de la Materia Condensada, Universidad Autónoma de Madrid, E-28049 Madrid, Spain
2
Condensed Matter Physics Center (IFIMAC), Universidad Autónoma de Madrid, E-28049 Madrid, Spain
3
Institute of Fundamental Physics IFF-CSIC, Calle Serrano 113b, E-28006 Madrid, Spain
4
Institute of High Performance Computing, Agency for Science, Technology, and Research (A*STAR), Connexis, 138632, Singapore
*[email protected]

Received 24 May 2024; revised 30 August 2024; accepted 25 September 2024; published 22 October 2024

Using topology optimization, we inverse-design nanophotonic cavities enabling the preparation of pure states of
pairs and triples of quantum emitters. Our devices involve moderate values of the dielectric constant, operate under
continuous laser driving, and yield fidelities to the target (Bell and W) states approaching unity for distant qubits
(several natural wavelengths apart). In the fidelity optimization procedure, our algorithm generates entanglement
by maximizing the dissipative coupling between the emitters, which allows the formation of multipartite pure
steady states in the driven-dissipative dynamics of the system. Our findings open the way toward the efficient and
fast preparation of multiqubit quantum states with engineered features, with potential applications for nonclassical
light generation, and quantum sensing and metrology.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
https://doi.org/10.1364/OPTICAQ.530865

1. INTRODUCTION near-field coupling and the shaping of radiation and absorp-

The preparation and manipulation of highly entangled multi- tion is required. Thus, inverse-designed quantum nanophotonic
qubit states is at the core of all quantum technologies. This devices have shown a notable performance in the context of qubit
makes the degree of control over qubit–qubit interactions a key entanglement formation [25,26] and single-photon generation
aspect in the assessment of material systems for their imple- [27,28].
mentation. In this context, nanophotonic devices offer a wide Here, we employ topology optimization (TO) [29,30] to
range of strategies to tailor photon propagation and the photonic inverse-design dielectric cavities that enable the preparation of
density of states at different length scales [1,2]. This makes Bell states [31] of QE pairs. They are obtained by using the
them stand out among other candidates for quantum hardware fidelity of the density matrix of the system to these target states as
in terms of scalability, integration, and speed of operation [3,4]. the optimization function, in conditions of continuous coherent
For this reason, in recent years, photonic architectures struc- pumping and inter-emitter distances of a few natural wavelengths
tured at the nanoscale to harness photon-assisted interactions (comparable to the cavity dimensions). By setting the laser
among quantum emitters (QEs, acting as qubits) have been pro- fields driving the emitters in phase (anti-phase), antisymmet-
posed as the platform for applications such as quantum light ric (symmetric) Bell-like states are obtained. Our analysis of the
sources [5–7] and detectors [8], quantum networks [9,10] and driven-dissipative dynamics of the QEs reveals that the small
circuits [11,12], memories [13], sensors [14], or simulators differences between the state implemented in the TO cavities
[15,16]. and the target one translates into a slight reduction in its prepa-
Lately, inverse design techniques have revealed unexpected ration time [32]. Finally, we prove the versatility of our design
and counter-intuitive optimization pathways for nanophotonic strategy extending our investigation to QE triples and present-
systems [17,18]. These numerical tools have proven to be par- ing a dielectric device that generates highly entangled tripartite
ticularly successful in functionalities that require a complex states with fidelities to the symmetric W state [33] comparable
trade-off between conflicting mechanisms [19–21], and their with those obtained from the ad hoc optimization of the mas-
impact on the field has even opened the way toward the explo- ter equation parameters. Our results prove that, through a (fully
ration of the fundamental limits of photonic performance [22,23] classical) laser driving, inverse-designed nanophotonic devices
and discovery [24]. More recently, these numerical tools have can be used to produce purely quantum states of QE ensem-
also been exploited in the realm of quantum nano-optics, where bles, showcasing them as a promising resource for quantum
a delicate balance between the engineering of light–matter technology.

2837-6714/24/050371-08 Journal © 2024 Optica Publishing Group

 372 Vol. 2, No. 5 / 25 October 2024 / Optica Quantum Research Article

2. THEORETICAL MODEL
Our nanophotonic cavities are designed to host pairs and triples
of distant QEs, modeled as two-level systems (with perfect quan-
tum yield) under laser driving. The dynamics of the density
matrix for the system, ρ, is described by a master equation of
the form [34]
[︂ ]︂ ∑︂ (︃ 1 {︁ †
)︃
dρ
L ρ ≡ ı ρ, H + γij σj ρσi† − σi σj , ρ =
}︁
, (1)
i,j
2 dt

under the assumption that the QEs are weakly coupled to their
electromagnetic (EM) environment, an approximation that will
be revisited below. In Eq. (1), σi (σi† ) is the annihilation (cre-
ation) operator for the QE labeled as i (ranging from 1 to 2–3),
fulfilling {σi† , σj } = δij . The Hamiltonian above can be written
in the laser frame as
∑︂ ∑︂ ∑︂
H= δi σi† σi + gij σi† σj + Ωi (σi + σi† ), (2)
i i≠j i

where δi = ωi − ωL is the detuning of the frequency of the emit-

ters, ωi , with respect to the laser frequency, ωL . The second
term in Eq. (2) reflects the coherent interaction between the QEs
assisted by off-resonant EM modes with strength given by gij , Fig. 1. TO-designed nanophotonic cavity, of radius R and height
and the last one, their laser driving with amplitudes Ωi . Finally, h, hosting a QE pair. The emitters are aligned and oriented along
and back to Eq. (1), the dissipative interaction between the QEs the z-direction, and driven coherently by laser fields of amplitude
(i ≠ j), as well as their radiative decay (i = j) is also accounted Ωi (wavy green lines). The light gray mesh renders the spatial
for by Lindblad operators weighted by the dissipative matrix discretization in the solution of Maxwell’s equations, and the per-
(with entries γij ), which must be positive semi-definite to ensure mittivity map is coded from white (vacuum) to black (ϵ = ϵmax ).
the physical character of the system dynamics [35]. The bottom-right illustrates the spectral position and Lorentzian-
Düng et al. [34] established the connection between the coher- type line shapes of the QEs, with their linewidths and detunings
ent and dissipative coupling parameters in Eqs. (1), (2) and with respect to the laser frequency ωL .
the EM dyadic Green’s function [36] of the dielectric environ-
ment of the QEs, obtaining gij = ω2 p∗ R{G(ri , rj , ω)}p/ℏε0 c2
and γij = 2ω2 p∗ I{G(ri , rj , ω)}p/ℏε0 c2 , where p is the transition [40–42], it consists of a nested iterative procedure in which
dipole moment of the QEs and ri,j their position. This frame- ϵ(r) is shaped with the precision of the spatial discretization
work links the quantum dynamics of ensembles of identical used to solve Maxwell’s equations. Starting from free space,
emitters and the spatial distribution of the dielectric permittiv- in each iteration, the effect of a small, local increment of the
ity in their vicinity, ϵ(r), which has allowed the investigation permittivity, δϵ, on the target function is assessed for each mesh
of QE entanglement generation in different nanophotonic struc- element. By keeping only those that contribute to enlarge Fφ ,
tures [26,37,38]. In these works, the dyadic Green’s function is dielectric cavities with optimum performance (for a set of given
evaluated at the QE frequency. We employ it here to describe constraints) are attained. The high speed and efficiency of the
emitters with slightly different natural frequencies and detuned algorithm resides in the use of first-order Born scattering series
from the driving laser. We anticipate that the validity of this and the exploitation of Lorentz reciprocity in the evaluation of
approach for the nanophotonic cavities that we obtain from the effect of the local dielectric variations on the dyadic Green’s
the TO algorithm will be demonstrated below. In what fol- functions [36].
lows, we will employ the laser frequency, ωL ≃ 2.067 eV (λ = Figure 1 illustrates an inverse-designed cavity hosting a QE
600 nm) for the evaluation of the gij and γij parameters in our pair. The TO algorithm is interfaced with the finite-element EM
calculations. solver implemented in Comsol MultiphysicsTM , whose spatial
The density matrix obtained from the solution of Eq. (1) discretization is sketched by the dark gray thin mesh. The QEs
allows for the calculation of expectation values of any physi- are separated a distance d12 along the z-direction, with their
cal observable of the system. In our case, we will focus on its dipole moments parallel to it. This emitter configuration allows
fidelity [39], Fφ = ⟨ϕ| ρ|ϕ⟩, to a desired pure state, |ϕ⟩. This us to exploit the azimuthal symmetry of the system to solve
quantity ranges from 0 to 1, and we will use it as a measure of Maxwell’s equations within the rz-plane only. As a result, we
the similarity of the quantum state of the QE pair/triple in our obtain cylindrical cavities with rotational symmetry, radius R,
nanophotonic system to the target one. Exploiting the depen- and height h. The dielectric function varies from 1 (white) to its
dence of the master equation parameters on the permittivity of maximum, ϵmax (black), which varies from one design to another.
the medium hosting the QE through the dyadic Green’s func- We set a threshold, ϵmax ≤ 9, corresponding to semiconductor
tion, we have developed an inverse design algorithm based on materials such as GaP [43] in the visible range, to remain in the
TO that provides the optimum (lossless, real-valued) dielectric typical parameter regime of nanophotonics. In the bottom right
map, ϵ(r), for a given target state |ϕ⟩ by maximizing Fφ . A corner, the line shape of the two emitters is represented, with
detailed description of the numerical method can be found else- natural frequencies ωi and linewidths γi = γii . Both are detuned
where [26], we only sketch it here. Inspired by recent reports from the laser frequency ωL .
 Research Article Vol. 2, No. 5 / 25 October 2024 / Optica Quantum 373

3. RESULTS and targeting even (in orange dots) and odd (in blue dots) Bell
states. In Fig. 2(a), the fidelities (used as optimization func-
First, we design dielectric cavities to prepare distant QE pairs
tions) are shown in connected triangles. We obtain F+± >0.95 for
into maximally entangled pure states in the steady-state regime,
all emitter–emitter distances and both symmetries. The small
L ρ = 0 in Eq. (1). We set the dimensions of the structures to
deviation from unity is caused by the finite size of the devices,
R = 6.25λ and h = 16.67λ (λ = 600 nm), and the dipole moment
which we restrict to the micron scale to remain in the domain
of the emitters to |p| = 1 e·nm, which yields a free-space decay
of nanophotonics technology. This is manifested in the slight
rate γ0 = ω3 |p| 2 /3πℏϵ0 c3 = 2.3 µeV. We choose the even and
decreasing trend of the fidelity as a function of d12 for both sets
odd Bell states,
of data, and the fact that, in some cases, the TO algorithm was ter-
1 [︁ minated because the maximum permittivity condition, ϵmax = 9,
|+±⟩ = √ |ge⟩ ± |eg⟩ ,
]︁
(3) was reached in some position within the device. Importantly, all
2
our cavities include a cylindrical ϵ = 1 void around each of the
as our target, with g (e) indicating the ground (excited) state QEs (of height 5 nm and radius 2.5 nm) that the TO algorithm
of each QE, and generate two different sets of devices, result- does not explore, which makes their performance robust against
ing from the maximization of the corresponding fidelities, F+± . emitter misplacement within a few-nanometer range.
In accordance with recent literature [44], we make the detun- To clarify the nature of the quantum states sustained by the TO
ings of the QE frequencies symmetric with respect to the cavities, their purity is plotted in connected circles in Fig. 2(a).
laser frequency, and significantly smaller than their linewidth, This is calculated as P+± = Tr{ρ2+± }, where ρ+± is the system
|δi |/γ0 = 0.2 (i = 1, 2). The laser pumping strengths are set to density matrix (the subscripts indicate the target Bell state). We
|Ωi |/γ0 = 0.7, and their parity is given by the symmetry of the can observe that it is above 0.98 in all cases, indicating the
target state, having Ω1 = ∓Ω2 for |+±⟩. The diffraction limit of pure character of the states formed in the devices [45]. Note
classical optics imposes a lower bound for the inter-emitter dis- as well that the purities present a decreasing slope very simi-
tance, d12 ≳ λ/2, to make a reliable control over the two laser lar to F+± . The deviation of ρ+± from the Bell states in Eq. (3)
fields possible. becomes clearer by computing the ground state populations,
Figure 2 analyzes the performance of the inverse-designed G = ⟨gg| ρ+± |gg⟩. They are rendered in rhombuses in Fig. 2(a),
n+±
cavities obtained for inter-emitter distances between λ and 4λ, obtaining n+± G ∼ 0.04 and a positive slope with increasing dis-
tance. Again, this indicates that the radiation losses experienced
by the TO cavities due to their micron-sized dimensions are
behind the failure to obtain |+±⟩ with higher accuracy.
Once we have verified the capability of the cavities to produce
highly entangled steady states for QEs several laser wavelengths
apart, we explore the physical mechanism behind their opera-
tion. For this purpose, we plot in Fig. 2(b) the coherent and
dissipative coupling strengths versus d12 . Both are normalized
√
to the collective decay rate of the system, defined as γ = γ1 γ2
[46]. We can observe that the dissipative coupling is maximized
[47], reaching absolute values equal to this collective decay,
+±
|γ12 |/γ +± = 1, and its sign is positive (negative) for odd (even)
target Bell states. Note that we are using the ++ and +− super-
scripts to indicate that the parameters correspond to cavities
implementing even and odd Bell states, respectively. However,
12 acquires vanishing values, with opposite sign to γ12 . These
g+± +±

values tend to reproduce the master equation parameters previ-

ously reported in theoretical proposals for emitter entanglement
through cavity or waveguide dissipation [44,48–54]. In all these
proposals, however, the free-space emission was only consid-
ered phenomenologically as an extra parameter, but no realistic
calculation was performed. In the proposal with plasmonic
waveguides [38], such a factor was taken into account showing
how their ability to suppress far-field emission offers a feasible
realization for this dissipative entanglement mechanism. How-
ever, this came at the expense of high absorption losses, which
effectively restricts the QE–QE distances to the sub-wavelength
regime. To our knowledge, our TO dielectric cavities are the
Fig. 2. (a) Fidelity to even and odd Bell states, F+± , versus inter-
first platforms that implement efficiently this phenomenon for
emitter distance for TO cavities with R = 6.25λ and h = 16.67λ. distant emitters thanks to the ability to optimize both the gen-
Purity, F+± , and ground state population, n+± eration of long-range interactions and the suppression of the
G , of the attained
states. The detuning and pumping parameters are |δi |/γ0 = 0.2 far-field emission at the same time.
and |Ωi /|γ0 = 0.7. (b) Dissipative, γ12 , and coherent, g12 , coupling Our TO algorithm is able to produce dielectric cavities that
√ operate efficiently within a wide range of inter-emitter distances.
strengths normalized to the QEs decay rate, γ = γ1 γ2 , as a func-
tion of d12 for same devices as panel (a). (c) Purcell factor, γ/γ0 , To keep the ratio γ12 +±
/γ +± at its maximum value for all d12 , they
versus emitter–emitter distance for all the cases above. The blue must modulate the QE radiation into free-space in different ways.
arrow indicates the configuration considered in Figs. 3 and 4. This is shown in Fig. 2(c), which plots the Purcell factor, γ +± /γ0 ,
 374 Vol. 2, No. 5 / 25 October 2024 / Optica Quantum Research Article

Fig. 4. (a) Population dynamics of the QE pair within the TO

cavity in Fig. 3(b) under even coherent driving and initially in their
ground state, nG (0) = 1. The transients obtained from Eqs. (1) and
(5) are rendered in color solid and black dotted lines, respectively.
Vertical gray lines indicate the preparation times of the TO steady
Fig. 3. (a) Fidelity, normalized coupling strengths, and Purcell state, τ, and the phenomenological pure dark state, τDS . Panels (b)
factor versus iteration step for the TO cavity targeting the odd Bell and (c) show the spectral density at the emitters position (yellow,
state for d12 = 3.4λ. (b) Permittivity map obtained as a result of the purple) and the cross-spectral density (light blue). Dotted lines
TO procedure in panel (a). The dielectric constant is represented in render their single-mode fitting at the laser frequency (vertical line).
gray scale between 1 (white) and 5 (black). The red and blue arrows
indicate the position and orientation of the QEs.
previously reported for the formation of dissipatively entangled,
dark states: γ12 = γ, g12 = 0.
for all the cavities in the panels above. At small distances, the The dielectric map, ϵ(r), of the nanophotonic cavity obtained
dissipative coupling is strong and the emission rate of the QEs is at the end of the optimization procedure described above is
Purcell-enhanced (by a factor 1.5) to achieve a maximal fidelity. shown in Fig. 3(b). The permittivity is fully characterized within
In contrast, for large d12 , the QE interactions are weak, requiring the rz-plane, but to facilitate its visibility, it is displayed within
a strong reduction in their decay rate and yielding γ +± /γ0 ≃ 0.1. xz- and yz-planes. The QEs are sketched as blue and red arrows
As anticipated, the devices showcase a 15-fold difference in the along the z-axis. The gray scale codes the dielectric constant
Purcell factor, which is key for the formation of dissipatively linearly from 1 (white) to ϵmax = 5 (black). Two different regions
entangled quantum states with similar F+± for QEs separated by can be distinguished in ϵ(r). Few wavelengths apart from the
very different distances. QEs and near the edges of the cavity, elliptical-shaped, high-
Having assessed the performance of the TO cavities for Bell contrast periodic reflectors are apparent over a smooth ϵ ≃ 3
state preparation, we focus next on different aspects of the background. These we can link to the reduction of the collective
inverse-design procedure. To do so, we select a particular device, Purcell factor that minimizes the radiation decay experienced by
indicated by the blue vertical arrow in Fig. 2(c) (d12 = 3.4λ, odd the emitters. In their near-field and centered around them, two
symmetry), and use it as a test bed for the inspection of the oper- sets of lower-contrast, spherical-shaped shells can be observed.
ation of our optimization algorithm. Figure 3(a) plots the fidelity These are embedded into an ϵ ≃ 1 background. These contribute
to the target state (blue line), the coupling strengths (red and yel- to the tailoring of the emitter–emitter interactions, maximizing
low lines), and the collective Purcell factor (green) versus the (vanishing) its dissipative (coherent) coupling. Small, deeply
iteration step k. Each step corresponds to a complete scan within subwavelength, and isolated rings of high permittivity are dis-
the rz-section of the cavity, evaluating the Green’s function and tributed along the radial direction in between these two regions.
consolidating or discarding a δϵ = 0.003 increment in each mesh Our TO algorithm conforms these two structural elements (far-
point. Three different regimes can be distinguished in the evolu- field reflectors, and near-field shells and rings) to optimize the
tion of these quantities along the iterative procedure toward the balance between the two mechanisms, photon-assisted interac-
condition F+− = 1. First, a sharp drop in the Purcell factor takes tions and photon leakage, that govern the quantum state of the
place, while both couplings grow slowly in absolute value and QE dimer. Note that the maximum permittivity in the device
different sign. Between the steps 100 and 300, γ/γ0 varies very acquires only a moderate value, well below the threshold set for
little, while γ12 /γ increases quickly, approaching its maximum, the TO algorithm.
and g12 reaches its minimum value. For k>300, both the Purcell Figure 4(a) explores the preparation time (from the onset of
factor and coherent coupling tend smoothly to 0, while the dis- the laser driving) required for the emergence of the steady state
sipative coupling, already very close to the condition γ12 /γ = 1, in the cavity in Fig. 3(b). It plots the population dynamics in
converges toward it. In all this process, the fidelity grows almost the first excitation manifold and in the Bell state basis, n+± (t) =
linearly from 0.2 at k = 0 to F+− = 0.96 at k = 800, when the cav- ⟨+±| ρ(t)|+±⟩. The system is initially in its ground state, nG (0) =
ity implements master equation parameters very similar to those 1. We can clearly observe that once the QE pair is pumped, the
 Research Article Vol. 2, No. 5 / 25 October 2024 / Optica Quantum 375

population is transferred first into the even Bell state, and n++ (t) to any emitter ensemble. As discussed above, the limiting factor
(orange) develops a plateau (preceded by significant oscillations) is the calculation of the EM dyadic Green’s function, which
that showcases a meta-stable regime [32] that extends up to t ≃ is greatly simplified if the QEs are aligned. Thus, to prove the
40γ0−1 . In this time window, n+− (t) is negligible. At longer times, versatility of our method, we consider now QE triples, located
the population of the odd Bell states grows quickly, reaching along the z-axis and with dipole moments parallel to it. There
n+− ≃ 0.96 at t ≳ 103 γ0−1 , which sets the preparation time for exist multiple ways to generate entanglement in tripartite sys-
the cavity steady state. To obtain another estimation of this tems, which have been the object of much research in recent years
time, we compute the inverse of the Liouvillian gap of Eq. (1) [61,62]. We focus on a well-known class of three-qubit states,
[55,56]. The value obtained, τ = 350γ0−1 , is indicated by the the so-called W states [63,64]. These are a class of pure states
vertical gray dotted line in Fig. 4(a). We can benchmark this that present high robustness against noise and losses, which
estimation against the inverse of the Liouvillian gap for the means that they retain the maximum possible amount of bipar-
phenomenological master equation that yields pure dark Bell tite entanglement when any one of the three QEs is lost (traced
states in Ref. [32], τDS = 700γ0−1 (see vertical gray solid line). out) [33]. In particular, we take the symmetric W state
Thus, we can conclude that the small deviation from the target
1 [︁
|+++⟩ = √ |gge⟩ + |geg⟩ + |egg⟩ ,
]︁
state in the TO cavities allows for a shorter preparation time than (6)
the exact Bell state generated under an ad hoc theoretical model. 3
To gain insight into the population dynamics in Fig. 4(a), we as the target for our TO algorithm. We keep the same cav-
investigate next the role of the cavity fields as intermediaries of ity dimensions (R = 6.25λ and h = 16.67λ) as before which,
the QE–QE interactions behind them. To do so, we must refine to accommodate a third QE, requires a reduction of the inter-
our model of the system to account for the photonic degrees emitter distances to d12 = d23 = d = 1.17λ (where the extremal
of freedom in its quantum density matrix, ϱ. Thus, we perform QEs are labeled as 1 and 3, and the central one as 2).
EM calculations for the spectral densities, Jij (ω) = γij (ω)/2π To proceed with the cavity design maximizing F+++ , the
[57] for each emitter and between them, shown in Figs. 4(b) fidelity to the state in Eq. (6), we must first set the exter-
and 4(c), respectively. In these panels, the solid lines render the nal parameters (driving amplitudes and emitter detunings) of
EM simulations for our TO device, and dotted lines their single the Liouvillian. In the case of QE pairs, these were set in
Lorentzian fittings [58] accordance with recent literature [44]. We do not have such
analytical insight in QE triples, and need to use a different
Gi Gj Γa /2
Jij (ω) = (4) approach. Operating at the master equation level, we perform
π (ω − ωa )2 + (Γa /2)2 a particle-swarm-optimization (PSO) [65] of its parameters,
in the vicinity of the laser frequency (indicated by vertical gray taking F+++ as the objective function. Imposing invariance
lines). The fitting parameters, G1 ≃ G2 = 7.75γ0 = 17.1 µeV and under the exchange of QEs 1 and 3, ten quantities remain to
Γa = 2459γ0 = 5.4 meV, yield very good agreement with the full be optimized: six internal ones, describing the emitter–emitter
spectra within a 0.03-eV window around ωL (much larger than interactions and decay rates, and four external ones. To limit
the QE detunings). the range of parameter values, we used the EM dyadic Green’s
Equation (4) allows for the direct parameterization of a master function for a bulk medium with ϵmax = 9 (threshold permit-
equation of the form [57] tivity in the TO algorithm) to estimate the spatial variation
they can experience within the inter-emitter distance, d. In this
[︂ ]︂ (︃
1 {︁ †
)︃ way, we found that the conditions sign{g12,23 } = −sign{g13 } and
L ϱ ≡ ı ρ, H + Γa aϱa − a a, ϱ
′ ′ †
}︁
2 sign{γ12,23 } = −sign{γ13 } had to be fulfilled. Our PSO computa-
(5) tion in the constrained 10-dimensional parameter space involved
dϱ
(︃ )︃
∑︂ 1 {︁ † 2 × 103 particles, up to 5 × 103 iteration steps, and 1000 runs
+ γ0 σi ϱσi −†
σ σi , ϱ =
}︁
,
i
2 i dt under different initializations. The quantum steady state obtained
this way presents a fidelity F+++ = 0.91 and a purity P = 0.99,
with H ′ = (ωa − ωL )a† a + i δi σi† σi + Gi σi† a + Ωi σi† + h.c. It
∑︁
indicating that it corresponds to a pure state of the QE triple
describes the coupling of both emitters to their EM environ- slightly different from Eq. (6). This PSO procedure does not only
ment, with strengths Gi , approximated by a single cavity mode allow us to benchmark the performance of our TO algorithm, it
with frequency ωa and linewidth Γa [59]. Black dotted lines in also provides us with the four external parameters that it requires:
Fig. 4(a) plot the QE populations obtained with this refined the QE-laser detunings, δ1,3 /γ0 = 0.55 and δ2 /γ0 = −0.3, and
model. The excellent agreement with the transients in solid laser amplitudes, Ω1,3 /γ0 = 0.33 and Ω2 /γ0 = −0.73.
lines prove the accuracy of the original description in the weak- Figure 5(a) shows the map ϵ(r) maximizing F+++ . In con-
coupling, quasi-degenerate regime (|δi | = 0.2γ0 ≃ 2γ) in which trast to the PSO described above, no QE symmetries are
the inverse-designed cavity operates [60]. Moreover, the solu- imposed in the TO algorithm and, hence, in the TO cav-
tion to Eq. (5) enables us to compute the population transient ity itself. Its structure resembles very much that in Fig. 3(b),
for the cavity mode, na (t) = tr{a† aϱ(t)}. This is rendered by the presenting periodic reflectors with high permittivity contrast
green dotted line in Fig. 4(a), and shows that the plateau in n++ (t) and a larger, ϵ ∼ 2, background near its boundaries than at
coincides with the time window in which the cavity population is the central region, where ϵ ∼ 1. Remarkably, ϵmax = 3.6, lower
non-negligible. Thus, the cavity mode sustains this meta-stable than that in Fig. 3(b). This originates from the shorter inter-
regime in the quantum dynamics, beyond which, its population emitter distance, as the cavities for Bell state preparation at
decays and the two QEs, already in the target Bell state, become d12 ≃ d = 1.17λ (not shown) present a permittivity range very
effectively decoupled from their EM environment. similar to that in Fig. 5(a). The main panel of Fig. 5(b) shows
Up to here, we have designed nanophotonic cavities to host the tomography of the steady-state density matrix hosted by
quantum states of QE pairs, but our TO algorithm can be applied the dielectric cavity in panel (a). It is restricted to the ground
 376 Vol. 2, No. 5 / 25 October 2024 / Optica Quantum Research Article

⟨geg| ρ|geg⟩ = 0.527, and ⟨egg| ρ|egg⟩ = 0.204, revealing that

the weight of QE 2 is larger than the two extremal ones. Notice
the small discrepancy between the populations of the latter, a
direct consequence of the slightly asymmetric character of ϵ(r).
As discussed above, the degree of entanglement of W states
exhibit a strong robustness against the disposal of one of the
qubits. Indeed, the bipartite states that result from the trac-
ing out of one of the QEs, ρ(k) = Trk {ρ} (where k = 1, 2, 3),
in Eq. (6), present a significant Wootters concurrence [66],
C(ρ(k) ) = 2/3 = 0.667 for all k. Here, C = 1 for a maximally
entangled two-qubit state, such as the Bell states in Eq. (3).
The concurrence calculation for the partial traces of the den-
sity matrix in Fig. 5(b) yields C(ρ(1) ) = 0.624, C(ρ(2) ) = 0.344,
C(ρ(3) ) = 0.632. Thus, the degree of bipartite entanglement
obtained by tracing out the extremal QEs is similar to that in
the perfectly symmetric W state. However, it is slightly lower
if the intermediate QE is lost. This evidences the higher sensi-
tivity to decoherence effects in the intermediate emitter of the
quantum state in the inverse-designed TO cavity.

4. CONCLUSION
To conclude, we have generated pure quantum steady states of
QE pairs and triples under coherent driving conditions through
the inverse-design of their dielectric environment. By means
Fig. 5. (a) Permittivity map for the inverse-designed cavity max- of a topology-optimization algorithm that acts at the level of
imizing the fidelity to the W state in Eq. (6) for an inter-emitter the electromagnetic dyadic Green’s function, we have obtained
distance d = 1.17λ, and the same dimensions as the device in nanophotonic cavities that engineer simultaneously the coher-
Fig. 3(b). The dielectric constant is coded in white-to-black lin- ent and dissipative interactions between the emitters and their
ear scale. The position and orientation of the emitters along the radiative decay. First, we have performed a thorough study of
z-direction is indicated by red and blue arrows. (b) Quantum tomog- the capability of our devices to prepare even and odd Bell states,
raphy of the steady-state density matrix (real part) for a QE triple in showing that they exploit a dissipation-driven mechanism to
the cavity in panel (a). Only the entries within the ground and single- entangle pairs of quantum emitters separated several natural
excitation manifolds are shown. The colors represent the absolute wavelengths apart. Analyzing the population dynamics in the
value of the real part of the density matrix in logarithmic scale. system, we have shown that the small discrepancy between the
Inset shows the difference between the TO density matrix and a pure states hosted by the cavities and exact Bell states trans-
phenomenological one, resulting from the direct PSO of the master
lates into shorter preparation times. Finally, we have tested the
equation parameters maximizing F+++ (same color scale as in the
versatility of our approach by applying it to a triple of quan-
main panel).
tum emitters, successfully realizing a highly entangled tripartite
state akin to a symmetric W state. We believe that our results
and the single excitation manifolds, where the basis is formed prove that inverse design is a powerful tool for the conception,
by |+ + +⟩ and the states |α⟩ = √13 [|gge⟩ − a|geg⟩ − b|egg⟩] implementation and refinement of quantum hardware based on
√ nanophotonic platforms, with applications in areas such as for
and | β⟩ = √13 [|gge⟩ − b|geg⟩ − a|egg⟩], with a = (1 + 3)/2 and quantum sensing or metrology [67,68], and with potential advan-
√
b = (1 − 3)/2. The vertical axis displays the real part of pop- tages in terms of scalability and speed of operation with respect
ulation and coherence, and the color scale codes its absolute to other enabling technologies.
value in logarithmic scale. By simple inspection, we can extract Funding. HORIZON EUROPE Framework Programme (101070700
F+++ = ⟨+ + +| ρ|+ + +⟩ = 0.87, whose discrepancy (∼ 0.04) (MIRAQLS)); Fundación BBVA; Comunidad de Madrid (Y2020/TCS-
with that for the PSO Liouvillian is very similar to the devi- 6545); Ministerio de Ciencia, Innovación y Universidades (CEX2023-
ation of F+± from unity in Fig. 2(a). Thus, we can conclude 001316-M, PID2021-126964OB-I00, PID2021-127968NB-I00, TED2021-
that the performance of the cavities is similar in both cases. To 130552B-C21, TED2021-130552B-C22).
further verify the similarity between the quantum state in the Acknowledgments. We thank Alejandro Vivas-Viaña and Carlos
TO device and its PSO counterpart, in the inset of Fig. 5(a), we Sánchez Muñoz for fruitful discussions.
present the tomography of the difference between the two den- Disclosures. The authors declare no conflicts of interest.
sity matrices. The deviations are most apparent in the W-state
population and its coherence with the ground state, and these Data availability. Data and code underlying the results presented in this
paper may be obtained from the authors upon reasonable request.
remain in the same range, below 0.05.
Finally, we analyze in more detail the nature of the QE-triple
state implemented by the nanophotonic cavity. It presents a REFERENCES
large purity, P = 0.94, and the single excitation section of the 1. V. Giannini, A. I. Fernández-Domínguez, S. C. Heck, et al., “Plas-
tomography in Fig. 5(b) indicates that the state amplitudes in the monic nanoantennas: fundamentals and their use in controlling the
bare basis are not equal, in contrast to |+ + +⟩. The calculation radiative properties of nanoemitters,” Chem. Rev. 111, 3888–3912
of the populations in this basis yield ⟨gge| ρ|gge⟩ = 0.210, (2011).
 Research Article Vol. 2, No. 5 / 25 October 2024 / Optica Quantum 377

2. D. E. Chang, J. S. Douglas, A. González-Tudela, et al., “Colloquium: 28. J. Yang, M. A. Guidry, D. M. Lukin, et al., “Inverse-designed silicon
quantum matter built from nanoscopic lattices of atoms and photons,” carbide quantum and nonlinear photonics,” Light: Sci. Appl. 12, 201
Rev. Mod. Phys. 90, 031002 (2018). (2023).
3. A. W. Elshaari, W. Pernice, K. Srinivasan, et al., “Hybrid integrated 29. M. P. Bendsoe and O. Sigmund, Topology Optimization: Theory,
quantum photonic circuits,” Nat. Photonics 14, 285–298 (2020). Methods, and Applications (Springer Science & Business Media,
4. A. González-Tudela, A. Reiserer, J. J. García-Ripoll, et al., 2003).
“Light–matter interactions in quantum nanophotonic devices,” Nat. 30. J. Jensen and O. Sigmund, “Topology optimization for nano-
Rev. Phys. 6, 166 (2024). photonics,” Laser Photonics Rev. 5, 308–321 (2011).
5. R. Mitsch, C. Sayrin, B. Albrecht, et al., “Quantum state-controlled 31. D. Sych and G. Leuchs, “A complete basis of generalized Bell states,”
directional spontaneous emission of photons into a nanophotonic New J. Phys. 11, 013006 (2009).
waveguide,” Nat. Commun. 5, 5713 (2014). 32. A. Vivas-Viaña, A. González-Tudela, and C. S. Muñoz, “Unconven-
6. M. K. Bhaskar, D. D. Sukachev, A. Sipahigil, et al., “Quantum non- tional mechanism of virtual-state population through dissipation,”
linear optics with a germanium-vacancy color center in a nanoscale Phys. Rev. A 106, 012217 (2022).
diamond waveguide,” Phys. Rev. Lett. 118, 223603 (2017). 33. W. Dür, G. Vidal, and J. I. Cirac, “Three qubits can be entangled in
7. C. Groiseau, A. I. Fernández-Domínguez, D. Martín-Cano, et al., two inequivalent ways,” Phys. Rev. A 62, 062314 (2000).
“Single-photon source over the terahertz regime,” PRX Quantum 5, 34. H. T. Dung, L. Knöll, and D.-G. Welsch, “Resonant dipole-dipole inter-
010312 (2024). action in the presence of dispersing and absorbing surroundings,”
8. F. Najafi, J. Mower, N. C. Harris, et al., “On-chip detection of non- Phys. Rev. A 66, 063810 (2002).
classical light by scalable integration of single-photon detectors,” Nat. 35. H.-P. Breuer and F. Petruccione, The Theory of Open Quantum
Commun. 6, 5873 (2015). Systems (Oxford University Press, 2007).
9. A. Sipahigil, R. E. Evans, D. D. Sukachev, et al., “An integrated 36. L. Novotny and B. Hecht, Principles of Nano-Optics, 2nd ed. (Cam-
diamond nanophotonics platform for quantum-optical networks,” Sci- bridge University Press, 2012).
ence 354, 847–850 (2016).
37. D. Dzsotjan, A. S. Sorensen, and M. Fleischhauer, “Quantum emit-
10. J. Q. Grim, A. S. Bracker, M. Zalalutdinov, et al., “Scalable in ters coupled to surface plasmons of a nanowire: a Green’s function
operando strain tuning in nanophotonic waveguides enabling three- approach,” Phys. Rev. B 82, 075427 (2010).
quantum-dot superradiance,” Nat. Mater. 18, 963–969 (2019).
38. A. Gonzalez-Tudela, D. Martin-Cano, E. Moreno, et al., “Entangle-
11. P. Lodahl, S. Mahmoodian, S. Stobbe, et al., “Chiral quantum optics,”
ment of two qubits mediated by one-dimensional plasmonic waveg-
Nature 541, 473–480 (2017).
uides,” Phys. Rev. Lett. 106, 020501 (2011).
12. A. Blanco-Redondo, “Topological nanophotonics: toward robust
39. R. Jozsa, “Fidelity for mixed quantum states,” J. Mod. Opt. 41,
quantum circuits,” Proc. IEEE 108, 837–849 (2020).
2315–2323 (1994).
13. S. L. Mouradian, T. Schröder, C. B. Poitras, et al., “Scalable integra-
40. S. Mignuzzi, S. Vezzoli, S. A. R. Horsley, et al., “Nanoscale design
tion of long-lived quantum memories into a photonic circuit,” Phys.
of the local density of optical states,” Nano Lett. 19, 1613–1617
Rev. X 5, 031009 (2015).
(2019).
14. L. Ostermann, D. Plankensteiner, H. Ritsch, et al., “Protected sub-
41. R. Bennett and S. Y. Buhmann, “Inverse design of light–matter
space Ramsey spectroscopy,” Phys. Rev. A 90, 053823 (2014).
interactions in macroscopic QED,” New J. Phys. 22, 093014 (2020).
15. A. González-Tudela, C.-L. Hung, D. E. Chang, et al., “Subwavelength
42. R. Bennett, “Inverse design of environment-induced coherence,”
vacuum lattices and atom–atom interactions in two-dimensional pho-
Phys. Rev. A 103, 013706 (2021).
tonic crystals,” Nat. Photonics 9, 320–325 (2015).
43. J. Cambiasso, G. Grinblat, Y. Li, et al., “Bridging the gap between
16. C. Tabares, A. Muñoz de las Heras, L. Tagliacozzo, et al., “Variational
dielectric nanophotonics and the visible regime with effectively loss-
quantum simulators based on waveguide QED,” Phys. Rev. Lett. 131,
less gallium phosphide antennas,” Nano Lett. 17, 1219–1225 (2017).
073602 (2023).
17. S. Molesky, Z. Lin, A. Y. Piggott, et al., “Inverse design in nanopho- 44. H. Pichler, T. Ramos, A. J. Daley, et al., “Quantum optics of chiral spin
tonics,” Nat. Photonics 12, 659–670 (2018). networks,” Phys. Rev. A 91, 042116 (2015).
18. S. So, T. Badloe, J. Noh, et al., “Deep learning enabled inverse design 45. M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum
in nanophotonics,” Nanophotonics 9, 1041–1057 (2020). Information: 10th Anniversary Edition (Cambridge University Press,
19. A. Y. Piggott, J. Lu, K. G. Lagoudakis, et al., “Inverse design and 2010).
demonstration of a compact and broadband on-chip wavelength 46. Z. Ficek and R. Tanaś, “Entangled states and collective nonclassical
demultiplexer,” Nat. Photonics 9, 374–377 (2015). effects in two-atom systems,” Phys. Rep. 372, 369–443 (2002).
20. Z. Lin, X. Liang, M. Lončar, et al., “Cavity-enhanced second-harmonic 47. A. Canaguier-Durand, R. Pierrat, and R. Carminati, “Cross density of
generation via nonlinear-overlap optimization,” Optica 3, 233–238 states and mode connectivity: probing wave localization in complex
(2016). media,” Phys. Rev. A 99, 013835 (2019).
21. Z. Lin, A. Pick, M. Lončar, et al., “Enhanced spontaneous emission 48. A. F. Alharbi and Z. Ficek, “Deterministic creation of stationary
at third-order Dirac exceptional points in inverse-designed photonic entangled states by dissipation,” Phys. Rev. A 82, 054103 (2010).
crystals,” Phys. Rev. Lett. 117, 107402 (2016). 49. M. J. Kastoryano, F. Reiter, and A. S. Sørensen, “Dissipative prepara-
22. Z. Kuang and O. D. Miller, “Computational bounds to light–matter tion of entanglement in optical cavities,” Phys. Rev. Lett. 106, 090502
interactions via local conservation laws,” Phys. Rev. Lett. 125, (2011).
263607 (2020). 50. H. Zheng and H. U. Baranger, “Persistent quantum beats and long-
23. P. Chao, B. Strekha, R. Kuate Defo, et al., “Physical limits in electro- distance entanglement from waveguide-mediated interactions,” Phys.
magnetism,” Nat. Rev. Phys. 4, 543–559 (2022). Rev. Lett. 110, 113601 (2013).
24. P. R. Wiecha, A. Arbouet, C. Girard, et al., “Deep learning in 51. A. Vivas-Viaña, D. Martín-Cano, and C. S. Muñoz, “Dissipative stabi-
nano-photonics: inverse design and beyond,” Photonics Res. 9, lization of maximal entanglement between non-identical emitters via
B182–B200 (2021). two-photon excitation,” arXiv, (2023).
25. G.-X. Liu, J.-F. Liu, W.-J. Zhou, et al., “Inverse design in quantum 52. T. Ramos, H. Pichler, A. J. Daley, et al., “Quantum spin dimers from
nanophotonics: combining local-density-of-states and deep learn- chiral dissipation in cold-atom chains,” Phys. Rev. Lett. 113, 237203
ing,” Nanophotonics 12, 1943–1955 (2023). (2014).
26. A. Miguel-Torcal, J. Abad-Arredondo, F. J. García-Vidal, et al., 53. T. Ramos, B. Vermersch, P. Hauke, et al., “Non-Markovian dynamics
“Inverse-designed dielectric cloaks for entanglement generation,” in chiral quantum networks with spins and photons,” Phys. Rev. A 93,
Nanophotonics 11, 4387–4395 (2022). 062104 (2016).
27. E. G. Melo, W. Eshbaugh, E. B. Flagg, et al., “Multiobjective inverse 54. M. B. Plenio, S. F. Huelga, A. Beige, et al., “Cavity-loss-induced
design of solid-state quantum emitter single-photon sources,” ACS generation of entangled atoms,” Phys. Rev. A 59, 2468–2475
Photonics 10, 959–967 (2023). (1999).
 378 Vol. 2, No. 5 / 25 October 2024 / Optica Quantum Research Article

55. D. Manzano and P. Hurtado, “Harnessing symmetry to control quan- 62. N. Brunner, J. Sharam, and T. Vértesi, “Testing the structure of mul-
tum transport,” Adv Phys 67, 1–67 (2018). tipartite entanglement with Bell inequalities,” Phys. Rev. Lett. 108,
56. V. V. Albert and L. Jiang, “Symmetries and conserved quantities in 110501 (2012).
Lindblad master equations,” Phys. Rev. A 89, 022118 (2014). 63. A. Acín, D. Bruß, M. Lewenstein, et al., “Classification of mixed three-
57. I. Medina, F. J. García-Vidal, A. I. Fernández-Domínguez, et al., qubit states,” Phys. Rev. Lett. 87, 040401 (2001).
“Few-mode field quantization of arbitrary electromagnetic spectral 64. L. Borsten, D. Dahanayake, M. J. Duff, et al., “Freudenthal triple
densities,” Phys. Rev. Lett. 126, 093601 (2021). classification of three-qubit entanglement,” Phys. Rev. A 80, 032326
58. M. Sánchez-Barquilla, F. J. García-Vidal, A. I. Fernández- (2009).
Domínguez, et al., “Few-mode field quantization for multiple emit- 65. M. R. Bonyadi and Z. Michalewicz, “Particle swarm optimization for
ters,” Nanophotonics 11, 4363–4374 (2022). single objective continuous space problems: a review,” Evol. Comput.
59. R.-Q. Li, D. Hernángomez-Pérez, F. García-Vidal, et al., “Trans- 25, 1–54 (2017).
formation optics approach to plasmon-exciton strong coupling in 66. W. K. Wootters, “Entanglement of formation of an arbitrary state of
nanocavities,” Phys. Rev. Lett. 117, 107401 (2016). two qubits,” Phys. Rev. Lett. 80, 2245–2248 (1998).
60. G. McCauley, B. Cruikshank, D. I. Bondar, et al., “Accurate Lindblad- 67. J. Franke, S. R. Muleady, R. Kaubruegger, et al., “Quantum-
form master equation for weakly damped quantum systems across enhanced sensing on optical transitions through finite-range interac-
all regimes,” npj Quantum Inf. 6, 74 (2020). tions,” Nature 621, 740–745 (2023).
61. A. R. R. Carvalho, F. Mintert, and A. Buchleitner, “Decoher- 68. W. J. Eckner, N. Darkwah Oppong, A. Cao, et al., “Realizing spin
ence and multipartite entanglement,” Phys. Rev. Lett. 93, 230501 squeezing with Rydberg interactions in an optical clock,” Nature 621,
(2004). 734–739 (2023).