1. Introduction
Notable progress in quantum computing in recent years has resulted in first demonstrations of quantum supremacy with superconducting qubits and photons [
1,
2,
3]. Ultracold ions and atoms remain promising platforms for a scalable quantum computer [
4]. The advantage of ultracold atoms is the potential to create quantum registers of thousands of identical qubits on a micrometer scale [
5]. A substantial improvement of two-qubit gate fidelity in quantum registers based on single trapped atoms has been recently demonstrated [
6,
7,
8,
9]. Quantum simulations of complex problems of many-body physics and correlated quantum phases of matter can be performed using two-dimensional arrays of hundreds of Rydberg atoms in optical tweezers [
10,
11]. However, the fidelity of two-qubit gates for neutral atoms still remains limited. These gates use temporary excitation of atoms into Rydberg states [
12]. The dipole moments of Rydberg atoms scale as
[
13], where
n is the principal quantum number. Ground-state atoms do not interact at distances of few microns, but Rydberg atoms do. Therefore the dipole-dipole interaction of Rydberg atoms can be used for implementation of two-qubit gates and creation of entanglement [
14]. The experimentally demonstrated schemes of two-qubit gates are based on the effect of Rydberg blockade: when two atoms are located at short interatomic distance, they cannot be excited simultaneously to Rydberg states by narrow-band laser radiation [
7,
15,
16,
17,
18].
Multi-qubit gates with many target qubits can provide a remarkable speed-up of quantum algorithms. Realization of geometric and swap gates with atomic qubits using antiblockade was discussed in Refs. [
19,
20,
21]. Schemes for multi-control and multi-target gates based on microwave dressing were proposed in Refs. [
22,
23,
24]. A single-step implementation of the three-qubit controlled gates with Rydberg atoms was reported [
25]. Also, the realization of a two-qubit controlled-PHASE (C
) gate via single-modulated-pulse off-resonant modulated driving embedded in a two-photon transition for Rb atoms with high-fidelity entanglement to be 0.980(7) was recently reported [
26].
Quantum error correction schemes are of essential importance for quantum information processing with neutral atoms. A scheme for fault-tolerant quantum computing based on surface codes was proposed by Auger et al. [
27]. This scheme requires parallel implementation of multiqubit CNOT
(where
N is a number of target atoms) gates with a single control atom and
N target atoms, which are used as ancillary qubits, as shown in
Figure 1a. Recently, a surface code with an atomic quantum processor was experimentally demonstrated [
28]. The non-local connectivity between qubits was achieved by coherent transport of qubits in two dimensions and between multiple zones.
Multi-qubit gates can be built using Rydberg blockade and electromagnetically-induced transparency (EIT) [
29]. EIT is a quantum interference phenomenon that can be observed by two optical fields (probe and control lasers) tuned to interact with quantum states of atoms [
30]. The transmission of a weak probe field is enhanced in the presence of a strong (near-)resonant coupling field [
31]. Several groups studied EIT with Rydberg states theoretically [
32,
33] and experimentally [
34,
35,
36]. From the original proposal [
29] it is clear that implementation of high-fidelity multi-qubit quantum gates based on EIT requires large coupling Rabi frequencies (of order of GHz) for transition between low excited and Rydberg states [
29], which is difficult to achieve in the experiment. For lower values of the coupling Rabi frequency, the fidelity of parallel CNOT
gates becomes substantially limited by the interaction between the target atoms. It is possible to suppress the interaction between the target atoms while keeping strong interaction between control and target atoms by using the dipole-dipole interactions via Förster resonances and asymmetric excitation of control and target atoms to different Rydberg states. Moreover, creation of heteronuclear atomic arrays [
37] provides both extended control over the interatomic interaction and reduced error rates during readout of qubit states. The heteronuclear atomic species were first entangled in the experiment by Zeng et al. [
6]. A two-dimensional
array of two isotopes of Rb atoms was recently demonstrated [
38]. An experimental implementation of a dual-element atomic array with individual control of single Rb and Cs atoms with negligible crosstalk between the two atomic species was recently reported [
39]. There are also related experimental studies of ultracold dense trapped samples of over 1000 of
Rb
Cs molecules in rovibrational ground state with full nuclear hyperfine state control by protocols of stimulated Raman adiabatic passage (STIRAP) with efficiencies of 90% [
40,
41]. Rydberg blockade between a single Rb atom and a single RbCs molecule was recently demonstrated [
42]. A mixture of heteronuclear
Li and
Cs atomic clouds was studied in Ref. [
43].
In the present work we optimize the performance of a parallel CNOT gate based on Rydberg blockade and EIT in order to reduce the required coupling Rabi frequency for transitions between low excited and Rydberg states and improve the overall gate fidelity. With decrease of the coupling Rabi frequency the target-target interaction deteriorates the gate performance [
29]. When both control and target atoms are excited to identical Rydberg states [symmetric homonuclear interaction, as shown in
Figure 1b], it is not possible to tune the control-target and target-target interactions independently. However, if the control and target atoms are excited to different Rydberg states [asymmetric homonuclear or heteronuclear interaction, which is illustrated in
Figure 1c], it is possible to meet the conditions of Förster resonance [
44,
45] for control-target interaction and to keep target-target interaction in the van der Waals regime. That allows substantial suppression of the target-target interaction and obtaining higher fidelity of parallel CNOT gate at moderate coupling Rabi frequencies of 100–200 MHz, which can be readily achieved in modern experiments for tightly focused laser beams.
While working on this manuscript we became aware of a foremost experimental work [
46] demonstrating the considered EIT gate protocol for two-qubits, verifying the ability to perform a native CNOT gate. The authors managed to achieve a loss corrected gate fidelity of
, and prepared an entangled Bell state with
by trapping individually a pair of
Cs atoms separated by
.
The paper is organized as follows: in
Section 2 we describe the scheme of multi-qubit CNOT
gate and the physical model used for our numeric simulation of the gate performance. In
Section 3, we discuss the properties of the asymmetric homonuclear and heteronuclear Förster interactions. In
Section 4, we investigate the influence of the parameters of atomic states and laser fields, Förster interaction channels, and gate duration on the gate fidelity in homonuclear symmetric and heteronuclear asymmetric configurations. In
Section 5, we extend our approach to implement a C
NOT
gate with two control and two target atoms and calculate its fidelity for heteronuclear configuration. Analysis of gate errors due to spontaneous emissions is given in
Section 6. In
Appendix A, the model of the atomic system is described taking into account multiple Rydberg interaction channels. Results of calculations are compared with the single-channel model. In
Appendix B, a model of multi-intermediate hyperfine states of the target atom is formulated.
2. Scheme of Rydberg EIT CNOT Gate
The scheme of multiqubit CNOT
gate, proposed in Ref. [
29], is shown in
Figure 2a. The gate operation can be understood as following: (i) If the control atom is initially in the ground state
, the first
-pulse does not change its quantum state. The Raman transfer between states
and
is inhibited due to the interaction with intensive resonant coupling radiation with Rabi frequency
, as shown in
Figure 2b. The fidelity of blocking the population transfer is determined by the value of coupling Rabi frequency
. (ii) If the control atom is initially prepared in the ground state
, then it will be excited to the Rydberg state
by the first laser pulse
. The interaction between control and target atoms will shift the energy levels of the target atoms by a value of
where
denotes k
th target atom, and will make the coupling radiation
off-resonant for the transition between intermediate excited and Rydberg state of the target atoms, as shown in
Figure 2c. Thus, the conditions for EIT are not met anymore, and the Raman population transfer between the states
and
becomes possible. In the ideal limit of a blockade regime, the Rydberg states of the target atoms are never populated.
It is shown in Ref. [
29] that the following conditions for CNOT
gate are satisfied:
We considered several spatial configurations for
target atoms, which are illustrated in
Figure 3. The control atom is at the origin. The target atoms are equally displaced from the control atom. For
(
), the target atoms are placed on the vertices’ of an isosceles triangle (square) where the distance between the nearest target atoms is
.
The interaction of the control atom with radiation in the rotating wave approximation (RWA) is described by the Hamiltonian in basis of
,
and
as
where
ℏ is Planck’s constant, and
is a sharp
pulse which couples the Rydberg state
with
and is applied for
. Explicitly
is defined as following
The interaction of the target atom with radiation for an inverted Y configuration of the atomic energy levels is described by the Hamiltonian in basis of
,
,
, and
as
where
is a smooth Raman
-pulse that couples the ground states of the Rb target atom
and
to the intermediate state
(For Cs atoms, the long-lived ground states, and intermediate state of the target atoms are
,
and
, respectively.) with
. Here
is the detuning from the resonance with the intermediate state
, as shown in
Figure 2. Explicitly
is defined as following
The value
is a constant Rabi frequency which couples the intermediate state
with Rydberg state
[see
Figure 2b].
The model Hamiltonian of the combined system with single control atom and
N target atoms can be written as
where
with dimensions
and
with dimensions
are the identity matrices acting on the control atom, and on the ensemble of target atoms, respectively.
is the Hamiltonian describing the ensemble of target atoms.
can be written in the following form
where
The third and fourth terms of Equation (
7) are the terms describing the interaction between control and target atoms, and the interaction between the target atoms, respectively. In
Section 3, we show how to calculate the interaction energies in homonuclear and heteronuclear architectures.
In our simulations, we choose the maximum value of the Raman pulse
and the detuning
. The Raman pulse is applied for a duration
. Also, we adopted an approach of a non-Hermitian Hamiltonian to consider the finite lifetime of Rydberg state
of control atom by adding the term
to Equation (
3) and the finite lifetime of the intermediate states
of target atoms by adding the term
to Equation (
5) where
and
are the decay rates of Rydberg and intermediate excited states, respectively.
Following the original work [
29], in
Figure 4 we illustrate the dependence of the probability of blocking the population transfer
as a function of the ratio between
and
. In this case, the control atom is not excited to Rydberg state. Therefore, there is no interaction between control and target atoms. As clearly seen in
Figure 4a, for
in the EIT regime [
30] the transfer between ground states of the target atom is blocked. The time dependence of the population transfer is shown in
Figure 4b–d. At low value of
the population transfer
becomes allowed [
Figure 4b]. In the intermediate case of relatively small
, the state
is temporarily depopulated, but finally the system returns to the initial state [see
Figure 4c]. At very high values
the system mostly remains in the state
[see
Figure 4d]. Although the regime of strong coupling is advantageous for maximum gate probability transfer, it requires high coupling Rabi frequencies of order of hundreds of MHz or GHz which are difficult to achieve experimentally for highly excited Rydberg states due to the drop of transition matrix elements for ground-Rydberg laser excitation as
. Therefore, the intermediate values of
are of interest for experimental implementation. Similar behavior of probability transfer is observed for larger number of target atoms.
Müller et al. [
29] analyzed the effect of target-target interactions in the three-qubit GHZ state for three target atoms while considering the control-target interaction to be constant and varying the target-target interaction. This two-dimensional configuration is not relevant to the typical experimental conditions, since changing the values of the target-target interaction requires varying the interatomic distance between target atoms, which also results in change of the distance between the control and target atoms in all possible spatial configurations of atomic ensembles. In our following simulations, we varied the interatomic distances for several spatial configurations, which are illustrated in
Figure 3. In
Appendix A, we developed an analogous model for the one described in this section, where the coupling between many Rydberg states in considered.
3. Homonuclear and Heteronuclear Interaction Energy
Our approach is based on the additional control of the energy of interatomic dipole-dipole interactions using Förster resonances between two distinguishable atoms, which were studied in Ref. [
37]. A Förster resonance [
47] means that the energies of two collective states of two Rydberg atoms coupled by dipole-dipole interaction are equal. This enhances the probability of the population transfer between the collective two-body states which is equivalent to energy transfer between interacting atoms [
48,
49,
50]. Förster resonance is employed to compromise the choice of selected Rydberg states. The interaction between atoms lies in two different regimes: dipole-dipole regime (d-d) where interaction energy can be described as
or van der Waals regime (vdW) where
[
51]. Recently, software packages [
52,
53] facilitated the calculations of interaction energies between alkali and alkaline earth atoms. We used Alkali Rydberg Calculator (ARC) [
52] to calculate the interaction energy of two alkali atoms in
Rydberg states (here
n is the principal quantum number,
ℓ is orbital angular momentum number,
j is the total angular momentum number, and
is the projection of the total angular momentum on the quantization
z-axis) for homonuclear and heteronuclear configurations. We also calculated the Le Roy radius
[
54] which is the internuclear distance between two interacting atoms at which the theory of Le Roy-Bernstein is satisfied, and the interaction potential can be approximated by charge independent atomic distributions. This radius sets the minimum limit of the interatomic distance where our calculations of the interaction energies are valid. As shown in
Figure 3, we considered the quantization
z-axis perpendicular to the interatomic axis, which is the general case for all spatial configurations which are studied in this work. As shown in
Section 2, we investigated the performance of CNOT gate based on EIT with a single control atom for different number of target atoms for homonuclear and heteronuclear configurations. The target atoms are identical in all cases.
3.1. Heteronuclear Architecture
We consider a Cs atom excited from the ground state
to Rydberg state
as control qubit (as described in
Section 2), and spatially ordered Rb atoms as target qubits. The intermediate state
of target atoms is coupled to Rydberg state
. Due to the interaction between these two atoms the Rydberg states in both atoms will be coupled to other neighboring Rydberg state(s)
[see
Figure 2c]. The most dominant interaction channel, as shown in
Appendix A, is
The asymmetric interaction between these two atoms lies in the regime of dipole-dipole interaction. The interaction Hamiltonian between control and target atoms
can be written as [
55]
where
is a dipole-dipole interaction coefficient for the target atom
j separated from the control atom by distance
,
is the angle between the quantization axis and the interatomic axis. The energy defect
is the energy difference between the collective two-atom Rydberg states for the dominant interaction channel. The collective state
is defined as
where
is the excited Rydberg state of the target atom, and
is the set of all basis states of any target atom with dimensions
.
In this architecture, all target atoms are identical, and they interact in the vdW regime. The Hamiltonian describing their interaction can be written as
where
is van der Waals coefficient calculated by fitting the model function with the calculated energy level using ARC function
getC6fromLevelDiagram for
rStart =
,
rStop = 20
, and
minStateContribution = 0. The interatomic distance between different target atoms
. The set of all basis states of the control atom
with
. The collective state
where
3.2. Homonuclear Architecture
The case when all interacting atoms are the same atomic species (Rb), all interactions are in the vdW regime, since all atoms all excited to the same Rydberg state
and the interaction is in the vdW regime. The control-target Hamiltonian in this case will be in the following form
and the target-target interaction Hamiltonian will be the same as given in Equation (
13) for Rb atoms. For Cs homonuclear interaction with all atoms are excited to the same Rydberg state
, the van der Waals coefficient
.
In
Figure 5a, we show the dependence of the dipole-dipole interaction coefficient
for Cs and Rb atoms excited to Rydberg states, on the angle
between quantization
z-axis and the interatomic axis for two different cases of the projection of the total angular momentum on the quantization
z-axis. We have
which does not meet the maximum value of interaction energies but for using negative projection of the Rydberg state of Cs atom strengths the interaction. The interaction between the two atoms vanish at the magic angle
, and
.
In
Figure 5b,c, we show the evolution of blockade radius (dotted-blue curve) and Le Roy radius (dotted-red curve) as a function of the principal quantum number
n of the excited Rydberg state for homonuclear interaction between two Rb atoms excited symmetrically to Rydberg states
and the heteronuclear interaction between Rb and Cs atom excited asymmetrically to Rydberg states
, and
, respectively. It is noted that for homonuclear interactions with symmetric Rydberg states, the evolution of blockade radius is steady while being fluctuant for the asymmetric heteronuclear (or homonuclear) interactions. For Rb atom excited to Rydberg state
interacting with Cs atom excited to Rydberg state
, the value of blockade radius reaches local maxima. Local maxima points are repeated also at
, and 122 (the principal quantum number corresponding to the Rydberg state of Cs atom). Local minima points (
, and 94) can be imputed to principal quantum numbers where the dipole-dipole couplings of these pair states counteract each other. The evolution of Le Roy radius
is steady in both cases.
4. Fidelity of Multiqubit Entangled States
Fidelity is a measure of the closeness of two arbitrary quantum states. We employed the standard definition [
56] of fidelity between arbitrary states
and
of a quantum system as
where we considered
as the calculated density matrix after performing a partial trace over the subspace of computational states of control and target atoms, and
is the density matrix of the multi-qubit entangled state
. We numerically calculated the density operator of the system when it was initially prepared in the superposition of the computational ground states of control atom, which results in simultaneous blocking and transferring operations
where
. The case when
corresponds to a two-qubit Bell state, while for
we end in a GHZ-state, which is a useful resource in quantum computing and cryptography [
57]. For
, we have four computational states
and
. Generally, the total number of basic computational states in Rydberg blockade of CNOT gate is equal to
from the total
states, where
is the number of states in control (target) atom.
In our simulations, we consider different configurations of the control and target atoms to be either Rb (
Rb) or Cs (
Cs). We have taken decay rates
and
of the Rydberg state of the control atom
and the intermediate state of the target atoms
, respectively, from the data of ARC [
52]. The lifetime of Rb
state (first excited state of Rb) is
while for Rb
state the lifetime is
s. The lifetime of Cs
(first excited state of Cs) of the target atom is
, while for Cs
state we have
. Since lifetimes of higher excited states of target atoms are much longer, in the following calculations we consider Rb
and Cs
states as the intermediate state of the target atom. The Rydberg excitation schemes through these intermediate states were experimentally demonstrated in Refs. [
5,
7]. In
Section 6, we justify this choice of the intermediate state
.
4.1. Homonuclear Architecture
In this section, we study the gate performance for the case when control and target atoms are same atomic species, which can be either Rb or Cs. There are two different scenarios of homonuclear interaction: (i) symmetric interaction
, when both control and target atoms are excited to the same Rydberg state, and (ii) asymmetric interaction
, when the control atom is excited to one Rydberg state with principal quantum number
n, and the target atoms are excited to a different Rydberg state with principal quantum number
. Note that in the latter case, the interaction between the target atoms is also symmetric (the quantum states
of target atoms are identical)—
Symmetric interaction channel .
Figure 6 is a contour plot of fidelity of entangled states as a function of interatomic distance between control and target atoms
(
) and the ratio
where we considered
Rb Rydberg states of control and target atoms with lifetime
. In the case of only one target atom it is possible to achieve high fidelity
for a wide range of interatomic distances
and moderate values of
, since the target-target interaction does not exist in this case. Considering schemes with more target atoms
, the fidelity for
drops at small interatomic distances because of the increase of the influence of strong target-target interactions. The optimum interatomic distance is found to be around
. For
target atoms, the fidelity
for very high values of Rabi frequency
at
as in
Figure 6d.
In principle, it is also possible to consider asymmetric homonuclear interactions () to achieve high fidelities by reducing target-target interaction compared to control-target interaction. This case will be considered in a future work.
4.2. Heteronuclear Architecture
The contour plot of the fidelity
of entangled states in heteronuclear configuration is shown in
Figure 7 as a function of interatomic distance
R (
) and the ratio
. This case corresponds to asymmetric heteronuclear interaction between control and target atoms, while the target atoms interact in the vdW regime. The control Cs atom is excited to Rydberg state
with lifetime
and Rb target atoms are excited to Rydberg state
. Heteronuclear configuration is clearly advantageous in terms of fidelity comparing to the symmetric homonuclear configuration, shown in
Figure 6. The regime of a CNOT gate with one target atom for the selected range of interatomic distances
, as shown in
Figure 7a, is governed purely by dipole-dipole interaction and allows to achieve fidelity
for
. With increase of the number of target atoms
, the maximum obtained fidelity slightly drops, similarly to the homonuclear case. The fidelity
within the region
and
. This justifies the advantage of heteronuclear configuration for implementation of multiqubit CNOT
gates. Moreover, the two-species architecture is useful for improvement of readout without cross-talk when the state of a Rb data qubit is not affected by measurements performed by resonant light scattering by ancillary Cs atoms [
37], which was recently demonstrated experimentally for arbitrary two-dimensional arrays of Rb and Cs atoms [
39].
5. Scheme of Rydberg EIT CNOT
The proposals for implementation of C
NOT
/C
Z
gates with many control and many target atoms have been limited to gates with either many control atoms and single target atoms, or to single control atom and many target atoms. The most general case for arbitrary number of control and many target atoms has not been studied extensively. Recently, such schemes were proposed in several theoretical approaches [
23,
24,
58]. Young et al. [
23] designed a protocol which uses microwave dressing to implement multi-qubit gates with many control and many target atoms. This protocol reduces intraspecies interaction energies and maximizes the interspecies interaction energies, leading to asymmetric blockade, which simplifies the state preparation and enhances the speed of quantum algorithms and reduces the need for fault-tolerant error correction schemes. In this section, we modify the previously studied CNOT scheme based on EIT in order to implement a four-qubit gate with two control and two target atoms (C
NOT
gate) [see
Figure 8a] simultaneously by proposing an asymmetric sequence of laser pulses acting on control atoms. We consider the following sequence [see the scheme in
Figure 8b]:
We apply -pulses to excite the control atoms from ground state to highly excited Rydberg state in sequence.
Then we apply smooth Raman laser pulse to couple the ground states of the target atoms and , simultaneously, to the intermediate dark state .
Finally, we apply pulses to return the control atoms from highly excited Rydberg state to ground state in reversed sequence applied in step 1.
By applying this sequence of laser pulses, it is possible to efficiently implement the following gates by properly tuning the system parameters:
The system dynamics for the case of only one control atom excited to the Rydberg state
and the second control atom remaining in the ground state
[i.e.,
or
] corresponds to the previously considered cases for homonuclear symmetric and for heteronuclear configurations in
Figure 6b and
Figure 7b, respectively.
The architecture of this gate becomes possible if the qubits satisfy the following conditions: (1) Cs control atoms are excited to Rydberg state where their dominant interaction regime is vdW i.e., , (2) Rb target atoms are excited to Rydberg state and similarly their dominant interaction regime is vdW i.e., , (3) The regime of interaction between control Cs and target Rb atoms, is dipole-dipole interaction i.e., .
In
Figure 9a,b, we plot the contours of fidelity when the system is initially prepared in the superposition of ground states on control atoms, while the target atoms are in states
:
with gate duration of
. The total number of computational basis states for a scheme with
k control atoms interacting with
N target atoms is equal to
. In
Figure 9a, we consider all possible interaction between control and target atoms. The maximum achieved fidelity is
for a high value of Rabi frequency
for
. It is also noticed a sharp drop in fidelity for
which can be a result of the interaction between control atoms, since in this regime interaction between target atoms almost vanishes as seen in
Figure 9c. In
Figure 9b, we study a non-realistic case where we neglected the interaction between control atoms, which can be compared with the case of CNOT
in
Figure 7b, considering the different spatial arrangements. The maximum value of fidelity becomes possible for a wider range of intratomic distances. This case proves that the destructive pattern in system dynamics is a direct result of
.
6. Gate Errors
In this section we discuss the effect of finite lifetimes, and the role of the excitation through different intermediate states and of the spatial arrangement of target atoms on the fidelity for of CNOT and CNOT gates.
CNOT—In
Figure 10a,b, we plot the fidelity as a function of interatomic distance between Cs control and Rb target atoms for CNOT
gate for
, and
, respectively, with
,
, and total gate time
. Solid (dashed) curve represents the case of excitation of the target atoms through the second (first) resonance level of the intermediate state
(
). It is clear that using the second resonance level enhanced the obtained fidelity for CNOT
(CNOT
) to be
(
) at
, compared with
(
) for excitation through the first resonance level.
CNOT—In
Figure 10c, we plot the fidelity of multi-control and multi-target C
NOT
gate as a function of interatomic distance
between Cs control and Rb target atoms with total gate time
. Solid (Dashed) curve represents the case of excitation of the two target atoms through the second (first) resonance level of the intermediate state
(
) which shows a possible fidelity
(
) at
.
In our model, we considered the time gap between excitation and de-excitation of Rydberg state on the control atom to be
(
s) caused by the pulse sequence
which is typically required by a Rydberg blockade gate can cause an expected atom loss, not only for control atom but also for the ensemble of target atoms, which can also a source of errors for implementing the physical system in experiment. In [
59], the authors reported their findings in constructing a native CNOT gate based on optimizing smooth Gaussian-shaped pulses.
In
Figure 11, we show the effect of spatial arrangement and the number of target atoms on the obtained fidelity of CNOT
gate for Rabi frequency
for two different values of
. It is clear that the spatial arrangement of target atoms around the control atom slightly changes the fidelity according to their trapping positions. We also show that the heteronuclear architecture can be advantageous in terms of fidelity as a value for arrays with
.