Transporting Cold Atoms towards a GaN-on-Sapphire Chip via an Optical Conveyor Belt

By introducing neutral atoms to integrated photonic devices, hybrid photonic-atomic chips (PACs) have attracted extensive research in recent years.^[1–5] Benefiting from the strongly enhanced light-matter interactions due to the tightly optical field confinement at the wavelength and even subwavelength scale, PACs hold great potential in many quantum-based applications, such as quantum memory,^[6,7] novel quantum light sources,^[8] chiral quantum optics devices,^[9,10] nodes of quantum networks,^[11–13] novel quantum optics phenomena with surface plasmons,^[14,15] many-body physics,^[16,17] and quantum sensing.^[18,19]

Early attempts to trap cold neutral atoms near surface microstructures were initially investigated above current carrying microstructures,^[20,21] which can be tailored to create a variety of potential geometry and guiding schemes for cold atoms. Atoms are manipulated by magnetic field, and distances between atoms and the microstructure surface can be reduced to as close as 0.5 μm.^[20] However, the magnetic field transport configuration limits optical access to the atoms in one or more directions. With the advancement of fabrication techniques in photonic structures, manipulation of single atoms at wavelength and even subwavelength scales is possible with the tightly confined optical field confinement near the photonic structures. Many ground-breaking experimental results in coupling atoms to photonic structures have been achieved in various nanophotonic platforms.^[7,22–32] However, these studies move forward with some potential disadvantages. For instance, the platforms based on nanofibers^[7,30–32] are suspended in vacuum, thus being potentially unstable and having poor thermal conductivity, which imposes limitations on the atom trap lifetime and atom coherence time.^[33,34] In addition, vacuum feedthrough for the coupling of light in and out of the nanofiber brings complexity to fiber alignment and assembly. Other platforms based on solid-state microcavities, such as microtoroid or bottle microresonators,^[9,28,29,35] unable to directly load laser cooled atoms into the evanescent-field trap,^[36] face difficulties in deterministic loading and trapping of cold atoms. Although the reported single atom-photon interaction time has been improved from only a few microseconds^[28] to 2 ms,^[29] the further extension of the system to more photonic structures and more atoms is very challenging.

In order to achieve deterministic atom trapping on integrated photonic devices, important theoretical and experimental milestones have been reached with unsuspended waveguides and microring structures.^{[22–27,37,38]} Atoms are loaded into evanescent field of photonic structures from free space with optical tweezers and optical conveyor belts. These methods exhibit a highly precise control of atomic motion near photonic structures, including photonic crystal waveguides^[22–25] and microring resonators.^[26,27,37] Additionally, these demonstrations are compatible with on-chip integrated devices for cooling, transport, and trapping of cold atoms.^[39–41]

In this Letter, we report on transporting free space cooled ⁸⁷Rb atoms towards a GaN-on-sapphire chip^[38] with an optical conveyor belt.^[22,42–44] Successful atomic transport towards the chip is made possible by our platform’s full optical accessibility and careful control of atomic motion with a conveyor belt. The conveyor belt consists of two focused beams, both of which pass through the sapphire substrate perpendicularly. After careful spatial calibration of the beams and phase stabilization, our conveyor belt directly transports 10⁴ trapped atoms with a temperature around 40 μK towards the chip, without extra aberration of the focus beam from the sapphire substrate. The maximum transport efficiency of atoms is about 50% with a transport distance of 500 μm. It paves the way for the further implementation of stable atom trapping on the GaN-on-sapphire chip, promotes the realization of deterministic loading of atoms into the evanescent-field trap, which is promising for realizing the on-chip single-photon-level optical nonlinearity.

Overview of the Photonic-Atom Chip. Figure 1 illustrates our experimental setup for studying the PAC, where a GaN-on-sapphire chip is placed inside the vacuum cell (25 mm × 25 mm × 50 mm). Figure 1(a) provides a photograph of our PAC platform in a vacuum cell. We use low vapor pressure epoxy (torr seal) to stick the chip onto a metal holder while maintaining a high vacuum and enduring the high temperature during the vacuum baking process. Half of the chip without fabrication is glued onto a 316-L stainless steel holder, and the remaining part of the chip is suspended in vacuum with waveguide and microring resonator structures fabricated on the surface. The metal holder is then connected to a CF35 vacuum cube, providing heat dissipation and stability for the chip. Here, we adopt the GaN-on-sapphire platform for the PAC following our previous theoretical proposal,^[38] as the system is more stable without suspended photonic structures. In addition, both GaN and sapphire are wide-band-gap materials that are transparent to ultra-broadband wavelengths (260–1590 nm),^[45,46] so our chip is compatible with lasers working in the visible and near-visible wavelengths for many alkali and alkaline-earth atoms, allowing full optical access for cold rubidium atom cooling, trapping, transport, and detection.

Fig. Fig. 1.  Schematic of our experimental system, with a GaN-on-sapphire chip (5 mm × 10 mm) platform placed inside the vacuum cell. Two pairs of cooling laser beams with a crossing angle of 60° in the x–y plane go parallel to the chip surface, while the third pair intersects the crossing of the beams in the x–y plane and the chip surface at an angle of 60° to the chip surface. The 781 nm dipole laser beams are split by a polarizing beam splitter (PBS), pass through the acousto-optic modulator (AOM) and focus onto the chip by a pair of lens with a focus length f = 100 mm. (a) Photograph of our PAC platform in a vacuum cell. (b) Picture of the micro-fabricated microring resonator and the bus waveguides on top of the GaN-on-sapphire chip taken by a scanning electron microscope. (c) Photograph of our testing setup for the coupling to PAC. (d) Schematic of an optical conveyor belt for atom transport towards a GaN-on-sapphire chip. Trapped atoms are confined in the lattice antinodes and move with the temporally varying trap potential.

DownLoad: Full-Size Img PowerPoint

Figure 1(b) shows the scanning electron microscope (SEM) image of a fabricated microring resonator and the bus waveguides on top of the GaN-on-sapphire chip. The size of the sapphire substrate is 5 mm × 10 mm. The microring resonator is vacuum-cladded for direct interaction between atoms and the evanescent field of the confined modes, with a major radius of 60 μm and a cross-section of 700 nm × 420 nm, and the optical modes of the microring are coupled to a bus waveguide through the evanescent field. Such microring resonators have been widely studied in photonics applications, due to their easy fabrication, high quality factor and small mode volume.^[47] The realization of the on-chip single-photon-level optical nonlinearity highly depends on the cooperativity parameter $C={\displaystyle \frac{3{\lambda }^2}{4{\pi }^2}}{\displaystyle \frac{Q}{V_{\mathrm{m}}}}$, where λ = 780 nm is the wavelength of the D2 line of rubidium atoms. The cooperativity parameter is proportional to the ratio of the quality factor Q and the mode volume V_m for the microring resonator. For our microring resonator parameter, the currently achieved quality factor Q = 3.75 × 10⁴, mainly limited by the surface roughness.

As shown in Fig. 1, our optical configurations of the experiments could be divided into three parts: (i) coupling to PAC, (ii) magneto-optical trap (MOT), and (iii) optical conveyor belt. First, at both ends the GaN-on-sapphire chip, light is coupled in and out of the photonic chip through a high numerical aperture (N.A.) objectives. Figure 1(c) shows the photograph of our testing setup for the coupling to the PAC. A coupling efficiency of about 20% for the GaN waveguide in Fig. 1(b) for optical signals with 780 nm wavelength can be achieved with commercial N.A. = 0.4 objectives. Laser coupled into the waveguide can be utilized to provide an evanescent field trap or to couple input signal photons close to the transition frequency of the atoms to the transported atoms.

The cold ⁸⁷Rb atoms are then prepared through a standard six-beam magneto-optical trap.^[48] The glass cell is connected to a mini cube and a 30 L/s ion pump, resulting in a pressure of 10⁻⁹ mbar measured by the ion-pump current. Three pairs of cooling laser beams are generated from a 780 nm laser, with the power of each beam being about 150 μW and the beam waist being 1 mm. The cooling laser beam is red detuned by 8 MHz from the |F = 2〉 → |Fʹ = 3〉 D2 cycling transition. Additionally, 80 μW of repump laser beams overlap with one of the cooling laser beams. The beams intersect at one point about 1 mm above the surface of the chip, with additional anti-Helmholtz coils aligned with the point providing a magnetic field gradient up to 10 G/cm. To align with our PAC, two pairs of cooling laser beams with a crossing angle of 60° in the x–y plane go parallel to the chip surface, while the third pair intersects the crossing of the beams in the x–y plane and the chip surface at an angle of 60° to the chip surface. Although MOT beam pairs are not oriented orthogonally to each other, 3D atom confinement is still achieved as components of each beam are projected along the orthogonal axis.

Following the 3D MOT procedure, the temperature of the atom ensemble around 40 μK is finally achieved by a polarization-gradient cooling (PGC) process. With a duration of 2 ms for the PGC, the cooling laser beams detuning is ramped down to −48 MHz from the cycling transition. Figure 2(a) shows a density contour plot of the atom ensemble, which is deduced from a single-shot free space absorption image.^[49] The cold atoms are about 800 μm away from the chip surface, and the atom number density is about 3 × 10¹⁰ cm⁻³ with an atom cloud radius of about 190 μm. The distance between the atom cloud and the chip surface can be adjusted from 300 μm to 1000 μm by adjusting the offset coil and the alignment of the cooling beams. However, in close proximity to the chip, the density and shape of the atomic cloud are altered due to surface reflections, which is consistent with the previous observations.^[50]

Fig. Fig. 2.  (a) Single-shot absorption images of MOT atoms 800 μm away from the chip surface. The absorption imaging is performed with camera of 2048 × 1080 resolution and each pixel point on the camera corresponds to 15.6 μm on the object plane. The red line denotes the interface between free space and photonic chip. (b) Single-shot absorption images of atoms transported toward the chip surface with the optical conveyor belt.

DownLoad: Full-Size Img PowerPoint

Transporting Atoms towards the Chip. The optical conveyor belt is realized by an optical dipole trap, which consists of two linearly polarized counter-propagating Gaussian beams with beam waist w₀ = 20 μm, and the waist is located in the middle of the MOT atom cloud and the chip surface. The intensities of both beams are equal, and their optical frequencies are different by detuning Δν. For the realization of the moving standing wave, the phase fluctuations between the two beams must be minimized. Fluctuations of the relative phase with an rms value of roughly 2π/1000 between the two beams are directly translated into position fluctuations of the dipole trap potential, which limit the lifetime to 3 s, as reported in Ref. [51]. Thus in our experiment, the two counterpropagating dipole laser beams come from a single laser, with λ = 781 nm corresponding to a frequency 2.3 THz red detuned to the ⁸⁷Rb atom D2 transition. The laser is then split into two paths and passed through a double-pass 80 MHz acousto-optic modulator (AOM), with Δν controllable via the phase-synchronized RF signals applied to the AOMs. By moving atoms back and forth within the Rayleigh length in free space, the phase stability is empirically determined with a retention probability of approximately 90% for a single cycle of movement. Therefore, the two beams generate a spatially and temporally varying trap potential $U(z,t)=U_0{\cos }^2({\displaystyle \frac{2\pi \mathrm{\Delta }\nu }{2}}t-{\displaystyle \frac{2\pi }{\lambda }}z)$, where U₀ is the local trap depth and z is the position of atoms along the beam axis. Trapped atoms are confined in the lattice antinodes and moved with the temporally varying trap potential without significant phase noise heating [see Fig. 1(d)].

Fig. Fig. 3.  (a) Atomic density ρ on chip surface with different holding times τ. We measure the trajectories of moving atoms in atom conveyor belt by taking a sequence of images with incremental holding times and record the atomic density on chip surface (averaged 10 times) with Δν_max = 104 kHz and a maximum transport distance d = 500 μm. We observe a clear density peak at τ = 13 ms. The error bar comes from three continuous measurement sequences. Solid curve is the fit result from our theoretical model of atomic accumulation dynamics on the chip surface (see details in the text). Inset is the laser detuning profile; the rising and descending times of the profile are set to 1 ms, and the holding time is τ. (b) Atomic density distribution along the conveyor belt axis for different holding times.

DownLoad: Full-Size Img PowerPoint

Further investigations of the influence of the chip surface on the transport of atoms are shown in Fig. 3(b), where the atomic density distribution along the atom conveyor belt axis for different holding times τ is plotted. A string of pixel values along the conveyor belt axis above the red line in Fig. 2(b) is used to extract the measured atomic density distribution. The origin of the z-axis is set to the chip surface. The parameters of the conveyor belt are the same as in Fig. 3(a). We find that when atoms are close to the chip surface within 100 μm, the peak height of atomic density distribution along the atom conveyor belt axis decreases, which indicates an increasing atom loss rate. The unevenness of the flat surface, which needs to be super-ground to obtain a super-flat mirror surface with a roughness of a few angstroms, is due to the enhanced atomic loss within 100 μm, which has been studied by other groups.

For future PAC experiments, a high density of atoms on the surface of the GaN-on-sapphire chip is desired. Therefore, we experimentally varied the location of the MOT center, and optimized the Δν_max for atom transportation. The results for maximum transport distance d = 1000 μm, 750 μm, and 500 μm are summarized in Fig. 4, with the atomic density on the chip surface calculated from the fitting results, as shown by the solid line in Fig. 3(a). The transport efficiency η is defined as the ratio of the maximum atomic density on the chip surface to the initial atomic density in the standing wave dipole trap. In comparison of the different Δν_max in Fig. 4(a), the transport efficiency reaches the optimum when Δν_max is 100–160 kHz, and the transport efficiency drops when Δν_max is further increased. In particular, the efficiency dramatically decreases when Δν_max is less than 80 kHz. The dependence of η on Δν_max may be attributed to two different reasons. If Δν_max is too large, the acceleration and deceleration process of the conveyor belt may induce a significant loss of atoms. The trap potential gives transported atoms an acceleration boost. Depending on the phase of the atom’s oscillation, abrupt changes in the potential either enhance or decrease the atom’s kinetic energy. Such phenomena have been studied in many other experimental works in detail.^[51,52] In contrast, if Δν_max is too small, the required τ is too large, and the atomic density is limited by the intrinsic atomic loss in the dipole trap. In practical cases, we aim to obtain higher transport efficiency along with less transport time.

Fig. Fig. 4.  (a) Atom transport efficiency η with maximum frequency difference Δν_max for different transport distance d. The transport efficiency is defined as the ratio of the maximum atomic density on the chip surface to the initial atomic density in the standing dipole trap. The transport efficiency reaches the optimum when Δν_max is 100–160 kHz. (b) Atom transport efficiency η with transport time t for different transport distance d. Higher transport efficiency and shorter transport distance are achieved with shorter transport distance.

DownLoad: Full-Size Img PowerPoint

We also notice that the maximum transport efficiency is currently limited to about 50%. While nearly lossless atom transport was achieved for a transport distance of a few millimeters in free space,^[51] the reflection of the dipole trap beams on the chip surface may destroy the moving lattice antinodes in our cases. On the one hand, the interference of the reflected light with the conveyor belt modulates the depth of the dipole trap and reduces the effective trap depth. On the other hand, the vibrational noise of the chip is transmitted to the atoms with the change in the phase of the reflected light. These issues may be mitigated by choosing an appropriate polarization of the dipole trap beams with the incident angle satisfying the Brewster angle, since the reflection can be greatly suppressed. Finally, the intrinsic heating of atoms in the conveyor belt is tested in a static standing wave dipole trap (Δν_max = 0). As shown in Fig. 5, a heating rate of 12.4 mK/s is extracted from the measurement of the trap lifetime for different trap depths, which explains the severe atomic loss when Δν_max < 80 kHz.

Fig. Fig. 5.  Atom lifetime in the standing wave dipole trap for different trap depths. The solid line is a linear fit of our measured atom lifetime for different trap depth, the fitted total heating rate from the measured data is 12.4 mK/s.

DownLoad: Full-Size Img PowerPoint

In summary, we have demonstrated a hybrid photonic-atomic chip platform and successfully transported an ensemble of cold atoms from free space towards the chip with an optical conveyor belt. Our PAC platform is made of transparent materials, allowing full optical access for free space laser cooling, atom transport, and efficient coupling to on-chip photonic waveguides. The maximum transport efficiency of atoms is about 50% with a transport distance of 500 μm. In the future, by combining on-chip MOT laser cooling, optical transport and evanescent-field trapping of cold atoms, a more compact PAC platform is attainable through the implementation of advanced photonic structure fabrication and design strategies. Our PAC platform holds great potential for research of atom-photon interactions and realization of single-photon-level optical nonlinearity, which could find applications in quantum information science and quantum sensing.