Hollow-core fiber loading of nanoparticles into ultra-high vacuum

Optical levitation and motional control of dielectric objects in vacuum provide a unique platform for experiments in fundamental and applied research.^1,2 Optically levitated nanoparticles have been used to study stochastic thermodynamics in underdamped environments,³ develop novel sensing schemes,^4,5 search for new physics,^6–8 and explore many-body phenomena.^9–11 Owing to the flexible spatial and temporal control of the potential landscape and the excellent isolation from the environment,^12–14 levitated optomechanics is also a promising platform for probing macroscopic quantum mechanics or utilizing quantum enabled sensing applications. Recent experiments have entered the quantum regime by preparing nanoparticles in their motional ground state.^15–20 The next generation of quantum experiments will require extended coherence times and hence ultra-high vacuum (UHV) environments at or below $10 − 10$ mbar to avoid collisional decoherence.^12,14,21 Such vacuum levels per se are readily achieved in the laboratory. However, the available methods for nanoparticle loading to date make room-temperature optical levitation below $10 − 9$ mbar a challenging endeavor.

The most established method for loading nanoparticles into optical traps is based on ultrasonic nebulizers, which spray a dilute nanoparticle solution directly into the vacuum chamber at almost atmospheric pressure. While reliable, disadvantages of this method are contamination of the vacuum system and the in-vacuum optics with solvent and nanoparticles. To reach UHV, nebulizer loading followed by a bakeout is in principle possible, but the long turnaround times together with potential misalignment of optics make this approach impractical in most cases. The method is also probabilistic, which is undesirable when working with multiple traps.

Various other approaches have been explored to address the challenge of loading particles into UHV environments. Loading from vibrating surfaces^22–24 has been demonstrated for particle sizes down to 85 nm in diameter. This method eliminates solvent contamination, but it also ejects particles at high speeds and with a large angular spread, making efficient capture in UHV difficult. Another solvent-free loading technique is laser-induced acoustic desorption (LIAD).^25–27 As for piezo loading, however, the speed and emission angle of particles currently prepared with this method do not allow for direct deterministic loading into optical traps at UHV. Finally, particles have been transferred using load-lock schemes^28,29 in a contamination-free and deterministic way. While this method has demonstrated transfers down to $6 × 10 − 5$ mbar, it still requires additional pump-down time to reach the desired UHV level.

Here, we experimentally demonstrate loading of nanoparticles from atmospheric pressure directly into UHV levels below $10 − 9 mbar$⁠. Our method works on the timescale of minutes, without contamination of the UHV chamber, and is spatially precise, i.e., it delivers particles to specific locations. Our approach is based on an optical conveyor belt³⁰ inside a hollow-core photonic crystal fiber (HCF)³¹ for transporting sub-micron particles.³² Silica particles ranging from 100 to 755 nm in diameter are loaded into the optical conveyor belt via an ultrasonic nebulizer and are transported one by one from a low-vacuum loading chamber into a UHV science chamber [see Fig. 1(a)]. The fiber core diameter of 9 μm and the fiber length of 1.4 m enable pressure differentials between the chambers of up to 12 orders of magnitude (see the supplementary material).^33,34 After transport, the nanoparticles are directly deposited into a standing-wave trap in the science chamber. The fiber tip can then be re-positioned and another particle loaded into a different trapping site if required [see Fig. 1(b)]. Loading directly into UHV is realized via an active trigger mechanism, which enables capture of the particle in the target trap without the aid of gas damping.

Our experiment consists of an optical conveyor belt for particle transport from a high-pressure loading chamber to a target optical trap in a clean UHV chamber [Fig. 1(c)]. Various target trap configurations are compatible with our loading technique; the only requirement is the ability to trigger a rapid switching of the potential depth once the particle enters the trap. In our case, we use a 1550 nm standing wave trap in a Sagnac configuration with approximately 1 W input power. The power switching is provided via an AOM at the Sagnac input, while we perform motional readout of the particle via the dark port of the Sagnac interferometer. We use a pair of 0.6 NA aspheric lenses near the center of the Sagnac to achieve both a tight trap focus and a large collection efficiency for readout.

The conveyor belt used for transporting the particles inside the HCF is a moving optical standing wave formed by two 1.5 W counter-propagating 1064 nm beams (Azurlight ALS-IR-1064 10 W), where one beam is coupled into the high-pressure side of the HCF and the other into the UHV side. Each beam is frequency controlled via an AOD, which allows one to set a relative detuning Δ and thereby create a moving standing wave with velocity $v = λ Δ / 2$⁠. The target-trap-facing end of the HCF is mounted on a UHV-compatible 3D-translation stage (Attocube ANPx101/UHV) to enable precise positioning of the particles. The ability to translate the HCF inside the science chamber is useful for trap alignment as well as for loading multiple particles. To maintain the standing wave quality during translation of the HCF, we actively stabilize the coupling of the counter-propagating beam into the HCF using a 3D-piezo positioner [Fig. 1(c)].

To align the HCF optical conveyor belt to the target trap, the HCF tip is used to map the target trap geometry (see the supplementary material). This allows us to determine the positions of the beam foci of the standing wave target trap with respect to the HCF as well as their respective waists (see Fig. 2). We then position the HCF such that particles are deposited into the center of the target trap, where the trap is deepest.

The procedure for loading nanoparticles via the HCF into the target trap is as follows: Nanoparticles are diluted in isopropanol and loaded via a nebulizer into the loading chamber at ∼1 bar. A particle falling into one of these trap sites (as confirmed by an infrared camera) is transported along the moving antinodes from the loading chamber to the science chamber. We first describe the high-pressure case (⁠ $> 1 mbar$ in the science chamber), where the particle can be stably levitated in front of the HCF. When the particle arrives near the exit of the HCF, the detuning is reduced to decrease the particle velocity for its transfer to the target trap [see Fig. 3 (Multimedia view)]. As the particle enters the target trapping region its mechanical motion becomes visible in the dark-port detection signal. At the trap center (where the trapping frequencies are maximized), the conveyor belt is turned off, thereby completing the transfer. Figure 3 shows a video sequence of a complete transfer of a 100 nm diameter particle at a pressure of 8 mbar together with the power spectral density of the dark-port signal recorded during this procedure. Particle transfers into a UHV environment require an additional trigger mechanism.

A significant benefit of the HCF-loading technique is the capability to deterministically deliver and retrieve particles with micrometer-level spatial resolution. We demonstrate this by loading and manipulating several particles controllably into different standing wave trap sites at 5 mbar as shown schematically in Fig. 1(b). Specifically, we show how a single particle can be deposited, retrieved, and redeposited into successive antinodes of our target trap. The measured frequencies are in agreement with a theoretical model based on the measured trap geometry [Fig. 4(a)]. Discrepancies with the model are attributed to an observed slipping behavior of the particle into a neighboring antinode, likely due to the tilt of the HCF with respect to the target trap axis. Finally, we demonstrate how multiple particles can be loaded into desired trap sites simultaneously and also retrieved back into the HCF [see Fig. 4(b)].

The above-described particle transfer procedure entails depositing particles with slow detuning into the target trap. This requires stable levitation in front of the hollow-core fiber, which we have observed to work reliably only down to approximately 1 mbar (without feedback). To load particles into UHV, we instead ballistically eject the particle from the HCF at a constant detuning toward the target trap. Without gas damping, the conservative nature of the trapping potential means that the particle cannot be trapped without additional deceleration. Therefore, to transfer particles directly into a UHV environment, we employ a trigger that deepens the target trap as the nanoparticle passes its center²⁶ (see inset Fig. 5).

We monitor the particle signal in the dark port of the Sagnac interferometer to identify the moment the particle enters the target trap. The resulting detector signal is sent to a three-stage trigger circuit consisting of an amplifier, a rectifier, and a comparator. If the signal exceeds the comparator threshold, the trigger sends a TTL pulse to an FPGA (see the supplementary material), which, after a certain delay, simultaneously increases the target trap depth and turns off one conveyor belt beam. By leaving the counter-propagating HCF beam on, the nanoparticle is additionally decelerated by radiation pressure, which assists the loading process.

For the trigger to work effectively, it needs to increase the target trap power when the particle is near the trap center. There is an experimental delay between the detection of the particle in the target trap and the moment when the trap power is increased. This delay is at least 125 μs, but in practice will be longer depending on the specific settings of the trigger circuit (see the supplementary material). To make the time the particle crosses the trap center coincide with the trap activation, we choose the particle velocity accordingly (via the conveyor belt velocity). The minimum delay sets an upper limit on the particle velocity (see Fig. 5): above 33 mm/s, the particle will have passed the trap center before the potential deepens. A lower limit is given by the stability of the conveyor belt trap: we observed that at science chamber pressures below $∼ 10 − 2$ mbar, particles moving slower than 16 mm/s escape the conveyor belt before reaching the target trap. Quickly traversing the low-pressure region thereby avoids particle loss. Alternatively, the conveyor belt trap stability could be improved by active feedback cooling.

The procedure for UHV transfers is then similar to that of the high-pressure case described earlier, except for a higher particle velocity when leaving the HCF and the addition of the triggering mechanism. Once the trigger increases the target trap power and turns off the conveyor belt beam, the particle is fully trapped and the remaining conveyor belt beam can be shut off. As an example, Fig. 6 shows the loading of a 365 nm silica nanoparticle at a science-chamber pressure of $2 × 10 − 9 mbar$⁠. We used this method to load particles in a diameter range between 143 and 365 nm into a standing wave trap at UHV. We have demonstrated transfers down to a pressure of $8 × 10 − 10 mbar$⁠, limited only by the loading chamber pressure during these experimental runs (see the supplementary material).

A UHV transfer from start (particle trapped in the conveyor belt) to finish (particle in the target trap at UHV) takes less than 3 min, despite the manual trigger setup. Note that the initial capture time from the nebulizer varies, but is typically on the order of seconds to minutes. We expect that the transfer from the loading chamber into the UHV trap could be reduced to less than 10 s by automation.

Our hollow-core fiber loading technique enables fast, clean, and deterministic delivery of nanoparticles to optical traps in ultra-high vacuum. It can also be applied to different trapping schemes and particle sizes and is particularly suited for compact on-chip levitation platforms.³⁵ One could also use other schemes such as piezo loading or LIAD in the loading chamber in lieu of a nebulizer, which would completely eliminate the introduction of solvent and could also improve the reliability of the initial loading process.

Our technique opens the door to room-temperature levitated optomechanics experiments at unprecedented vacuum levels, a prerequisite for a number of proposed experiments aiming to utilize the quantum regime of levitated solids for both fundamental science and sensing applications. As this method enables high-resolution positioning of nanoparticles, it is also relevant for the growing community exploring architectures with multiple traps.