Enabling atom-clad waveguide operation in a microfabricated alkali vapor-photonic integrated circuit

TL;DR

This work addresses the challenge of integrating rubidium vapor with photonic integrated circuits in a scalable, hermetically sealed platform by anodically bonding silicon-nitride PICs to borosilicate vapor cells housing rubidium pill dispensers. The authors show that standard high-power pill activation degrades waveguides, and develop a low-power pulsed activation strategy together with a counter-propagating desorption laser to suppress rubidium-induced losses, enabling waveguide-based spectroscopy in a 3 mm air-clad region. They demonstrate controllable transient and quasi-steady rubidium densities by adjusting activation pulse length, duty cycle, and device temperature, with a logistic model and Beer-Lambert relation guiding the density dynamics. The results establish a compact, manufacturable vapor–PIC platform with potential for cavity QED, quantum nonlinear optics, and chip-scale atomic sensing, and outline paths to further robustness via surface passivation.

Abstract

Integrating alkali atomic vapors with nanophotonic devices offers a scalable route to quantum technologies that leverage strong atom-photon interactions. While there have been many approaches to such integration, the general reliance on traditional glass vapor cells, distilled alkali metals, and epoxy sealing limits reproducibility and scalability. Moreover, mitigating adverse Rb-photonics interactions is essential, particularly as devices become more compact and the alkali source lies in close proximity to the photonic elements. Here, we demonstrate the successful operation of compact and fully integrated devices that combine silicon nitride photonic integrated circuits (PICs) with microfabricated borosilicate vapor cells and pill-type rubidium (Rb) dispensers through hermetic seals via anodic bonding. We show how successful operation hinges on optically activating the dispenser in a low-power pulsed mode, releasing controlled amounts of Rb vapor on demand while mitigating photonic degradation. Simultaneously, a counter-propagating desorption laser completely suppresses Rb-induced losses and enables waveguide-based atomic vapor spectroscopy. Using this approach, we demonstrate repeatable control of vapor density by tuning activation pulse length, duty cycle, and device temperature. These results establish a compact, manufacturable, and scalable vapor-PIC device, and set the stage for future demonstrations in cavity quantum electrodynamics, quantum nonlinear optics, and chip-scale atomic sensors.

Figures (10)

• Figure 1: Hybrid device that integrates an atomic vapor cell, Rb pill source, and a photonic integrated circuit.(a) A borosilicate cover glass, micro-machined borosilicate frame, and a photonic integrated circuit (PIC) are anodically bonded together along with a miniature Rb pill to form a compact “vapor-PIC” package. (b) A photograph of a device after bonding and Rb pill activation.
• Figure 2: Experimental setup and schematic for simultaneous Rb pill activation, Rb density assessment, and PIC waveguide absorption spectroscopy.(a)Photograph of a device in setup: The device rests on a heated copper sample mount while the lensed fibers couple light in and out of the device. A D-mirror and lens guide the 980 nm activation beam onto the Rb pill. (b)Side view schematic: the device (green/white) containing the Rb pill (pink) rests on top of a temperature-controlled copper base heater (orange). A 980nm activation laser (black arrow) is amplitude-modulated, as indicated by the square-wave inset, and focused onto the pill under a microscope camera. A 780nm probe laser is split (50:50) with one arm coupled through a lensed fiber into the PIC waveguide, and the other interrogating the free-space Rb vapor. Probe light reflected from the chip exits upward to detector D1. (c)Top view schematic: two lensed fibers couple light in and out of the PIC waveguide (green/white): the 780nm probe (red arrow) and a counter-propagating 801nm desorption laser (magenta arrow). Probe transmission is measured by detector D2, while the desorption beam prevents Rb buildup on the waveguide during pill activation. The region of the device where the waveguides are air-cladded and interact with the Rb vapor is highlighted with dashed circle. (d-f)Illustrations showing three different measurement methods: free-space interrogation of Rb using reflection off the chip (path length of 4mm), free-space interrogation using a fully oxide-clad, waveguide-coupled grating out-coupler (path length of 2mm), and interrogation by the waveguide evanescent field (interaction length of 60µm or 3mm).
• Figure 3: Loss induced by standard activation that results in a stable vapor density. Temperature dependence of the total insertion loss at 780nm, away from any Rb transition, of a 60µm long PIC waveguide before and after Rb is introduced. The Rb vapor is generated using a Rb pill dispenser after heating with a strong 980nm external laser ($\approx 1W$). After Rb introduction, there is an $\approx 10.5dB$ increase in insertion loss of the waveguide. Inset: Doppler-broadened absorption spectrum of the $D_2$ manifold through the waveguide at 240. The error bars represent 1dB measurement uncertainty, and are one standard deviation values associated with coupling variation.
• Figure 4: Rubidium density control via activation laser.(a)Transient Rb density. Transmission of the $5S_{1/2},F=2 \rightarrow 5P_{3/2}$ transition in ^85Rb is monitored for 100 s in free-space (configuration in Fig. \ref{['fig:setup']}(d)) while a 980nm activation laser dispenses Rb from a pill. Four activation lengths -- 2s (dark blue), 3s (red), 4s (green), and 5s (light blue) -- are compared at identical laser power ($\approx 100mW$) and device temperature of 180. Data points (symbols) are joined by dashed lines to guide the eye. Longer activation pulse lengths ($tON$) yield higher Rb vapor densities, seen as deeper transmission dips: the 2 s trace (dark blue) maintains the highest transmission, whereas the 5 s trace (light blue) shows the lowest. Upper right inset: a 0 s to 20 s zoom-in of the data reveals an exponential drop in transmission (linear rise in Rb density) during each ON window and an exponential recovery when the laser is OFF; solid square pulses beneath the traces indicate the programmed 2 s, 3 s, 4 s, and 5 s activation windows. Lower left inset: full Doppler-broadened absorption spectrum of the $D_2$ manifold from a reference cell; the arrow marks the hyperfine line whose transmission is tracked in the main panel. (b)Quasi-steady Rb vapor density via repetitive operation. Transmission of the $5S_{1/2},F=1 \rightarrow 5P_{3/2}$ transition in ^87Rb is monitored for 300s in free-space while a 980nm activation laser periodically releases Rb from a pill. Three programmed duty cycles --- low (light blue), medium (green) and high (red) --- are shown as solid square-wave traces beneath the data (high level = laser ON). During the pulsing window (0 s to 150 s), a nearly constant Rb density is achieved. When pulsing stops at $t\approx150s$, the vapor gradually re-adsorbs onto the cell walls and all traces return toward unit transmission. The transmission during the pulsing period is fitted with a four-parameter logistic function and the transmission after pulsing stops is characterized by an exponential. Further discussion about the fitting is present in the main text. Upper right inset: Doppler-broadened absorption spectrum of the $D_2$ manifold; the arrow marks the hyperfine line tracked in the main panel.
• Figure 5: Temperature-dependent Rb loading and unloading during low power activation.(a) Transmission of the $5S_{1/2},F=1 \rightarrow 5P_{3/2}$ transition in ^87Rb is monitored for 400s in free-space using an on-chip grating coupler (Fig. \ref{['fig:setup']}(e)) while a 980nm activation laser periodically releases Rb from a dispenser pill. The solid square-wave trace beneath the data marks the laser timing (high = ON). The device base is held at 150 (dark blue circles), 160 (red triangles), 170 (green squares), and 180 (light blue diamonds); laser power and duty cycle remain fixed. Scattered points are data; solid curves are fits to the logistic model for the time interval between 0s and 150s) followed by a single-exponential recovery for all times $>150s$. Inset: Doppler-broadened $D_2$ absorption spectrum recorded in a reference cell; the arrow indicates the hyperfine transition tracked in panel a. (b) Reaction rate $k$ extracted from the 0s to 150s fits versus base temperature. (c) Transmission-decay rate constant obtained from fitting the $>150s$ data vs. base temperature. Error bars in panels (b) and (c) denote one standard deviation uncertainties from the fits.