1. Introduction

The latest atomic clocks are expected to be installed in mobile devices, such as smartphones, to achieve accurate time synchronization between the terminal and base stations or between terminals [1–6]. These clocks are generally based on the CPT (Coherent Population Trapping) system and use alkali atom vapor cells [7]. The vapor cells are also used for subpicotesla magnetometry or nuclear magnetic resonance (NMR) gyroscopes [8–11]. With the increasing applications of vapor cells in various fields, there is a growing demand for the fabrication of vapor cells at low cost and in large quantities.

Alkali atom vapor cells are formed using microfabrication methods [12,13]. Although a novel fabrication method has been developed, it is still challenging to seal alkali atoms in wafer-level packaging [14]. Using a mixture of RbCl and BaN₆, Rb can be deposited in each cell from a glass ampoule; anodic bonding with a glass wafer is then performed in a chamber [15]. This method requires complex machineries and increases the cost of the vapor cell. Using alkali azide compounds, alkali atom can be sealed with N₂ as one of the buffer gasses. However, long-term instability was reported in this method; moreover, and it is challenging to treat with an accurate amount of powder source, which contaminates the interface in the wafer bonding process [16,17]. With alkali atom dispensers, the fabrication process can be straightforward, but the dispenser takes considerable space, which inhibits miniaturization of this type of vapor cell [2,6,18,19]. In addition, the dispenser includes Zr/Al, which can absorb N₂ buffer gas. A decrease in the amount of buffer gases in the vapor cell causes drift of the CPT resonance peak and induces long-term frequency instability [6,20].

2. Proposed vapor cell

Figure 1 shows the conventional vapor cells using the alkali-atom dispenser and the proposed vapor cell, which is filled with Rb atoms from the alkali-atom dispenser. The conventional vapor cell has additional space to hold the dispenser (Fig. 1(a)). The alkali-atom dispenser is cut away in our proposed vapor cell, as shown in Fig. 1(b). There are many fabrication methods that use glass wafers [21,22]. In this process, by creating a small gap and dividing the bonding process into two processes, the distillation of alkali atoms at the wafer-level can be performed. By distillation, other contents such as Zr/Al do not exist in the vapor cell, and the buffer gases cannot be absorbed.

 

Fig. 1. (a) Vapor cell fabricated using the conventional fabrication method, which requires an alkali atom dispenser. (b) Vapor cell fabricated using our proposed fabrication method. This method also uses an alkali atom dispenser, but the dispenser is cut away in the fabrication process.

Download Full Size | PDF

3. Fabrication

3.1 Process chart

Figure 2 shows a process chart of the proposed vapor cell. (Process 1) A 2-mm thick Si wafer is processed by mechanical drills. (Process 2) Anodic bonding is performed with a 1-mm thick Tempax borosilicate glass wafer. The temperature is 450 °C, and the applied voltage –600 V under atmospheric pressure for 5 min. (Process 3) After filling the Rb dispenser (SAES Getters RB/AMAX/PILL/1-0.6), anodic bonding is performed with 1-mm thick Tempax glass, where the center part is etched 2.0 µm deep by diluted hydrogen fluorine. The bonding temperature is 400 °C, and the applied voltage –450 V for 7 min in vacuum or under controlled pressure with N₂/Ar buffer gases, with a mix ratio of N₂:Ar = 5:8. The activation of the Rb dispenser is performed using a YAG laser. Increasing the temperature for anodic bonding, Rb atoms exposed to the YAG laser get vaporized and fill the connected cavities. The density of Rb atoms in each cell depends on the number of connected cells, Rb atoms exposed to the YAG laser, and the process conditions. (Process 4) The center of the wafer is anodically bonded. The temperature is 400 °C, and the applied voltage –900 V under atmospheric pressure for 5 min. At 400 °C, the clusters of Rb atoms are movable. The top and bottom side of the wafer are almost the same temperature, and electrical reaction occurs at the anodic bonding interface. These conditions eventually prevent Rb atoms from remaining on the interface. The Rb atoms are distilled after cooling to room temperature. (Process 5) Subsequently, dicing is performed.

 

Fig. 2. Process chart of the proposed vapor cell using the Rb dispenser.

Download Full Size | PDF

3.2 Fabrication result

Figure 3(a) shows the 20-mm square wafers after Process 3; at this stage, only the surroundings were anodically bonded. There was a change in color of the bonded area; moreover, the center part of the wafer was curved because of the difference in the pressure between the inside and outside of the cell and the difference in the coefficient of thermal expansion between the Si and Tempax glasses [23,24]. Figure 3(b) shows the wafer after Process 4; at this stage, the center part was anodically bonded, and the cells were successfully sealed with Rb atoms. Deposited Rb was observed on the glass surfaces of some cells. Figure 3(c) shows the schematics of a 3.7${\times} $3.7${\times} $4.0 mm vapor cell after Process 5.

 

Fig. 3. (a) After Process 3 in Fig. 2, only the surroundings was anodically bonded and vacuum sealed in a 20-mm-square wafer. (b) After Process 4 in Fig. 2, the center part was anodically bonded, and some cells were observed with deposited Rb on the surface of the glass. (c) After Process 5 in Fig. 2, the vapor cell size is 3.7${\times} $3.7${\times} $4.0 mm.

Download Full Size | PDF

4. Evaluation

4.1 Vapor cells fabricated to seal only Rb atoms in vacuum

In this measurement, a VCSEL (Vertical Cavity Surface Emitting Laser) of 795 nm wavelength was used. A photodiode was used to detect the transmitted laser beam. The vapor cell was covered with permalloy plates to shield it from external magnetic fields [6,25].

The Rb absorption spectra of two vacuum-sealed Rb vapor cells are shown in Fig. 4(a). The vapor cell was heated to 90 °C; the two samples showed almost the same spectral lines. The absorption lines had the same appropriate line shapes compared to the theoretical Rb spectrum in a vacuum-sealed vapor cell [26]. The CPT resonance spectra of the two vapor cells are shown in Fig. 4(b). A weak magnetic field was applied to the Zeeman shift. The FWHM/2 values of the peaks were 95 and 110 kHz, respectively. The theoretical value is estimated to be 131 kHz, and compared with the experimental values, the small amount of gases possibly remain and behave as buffer gases [27]. The estimations of the pressure of the remaining gases were approximately several dozen Pa in both vapor cells. These are relatively low values compared to that in conventional cavities sealed by anodic bonding [28]. It is considered that the dispenser works as a getter in the connected cavities in the Process 3 in Fig. 2, and the dispenser is separated, and each cell doesn’t have the getter in the Process 4 in Fig. 2. The remaining gases would be the emitted gases from the surface of bonding interface or closed cavity and the non-reactive gases remaining in the Process 3 in Fig. 2. In addition, two vapor cells were fabricated with one dispenser. This indicates that several vapor cells can be fabricated with one dispenser, and this method is more cost-effective than conventional method.

 

Fig. 4. (a) Absorption line spectrum of Rb D1 line of the fabricated vapor cells sealed in vacuum. The measurement temperature was 90 °C. (b) Measurement results of the CPT resonance of the fabricated vapor cells. The measurement temperature was 90 °C. The FWHM of the peak were 190 and 220 kHz, respectively.

Download Full Size | PDF

4.2 Vapor cell fabricated to seal Rb atoms with buffer gases

Figure 5(a) shows the measured CPT resonance of the vapor cell with buffer gases. The peak was shifted by 3.6 kHz from the clock frequency, and the pressure of the buffer gases was estimated to be 2.76 kPa [27]. The pressure was expected to be 3.6 kPa at 90 °C in the fabrication process. This difference was could have caused by the fluctuations of the temperature and pressure in the chamber during anodic bonding in Process 3 in Fig. 2. The FWHM (Full Width at Half Maximum) of the CPT resonance peak was 4.0 kHz at 90 °C; that from a theoretical calculation was 2.1 kHz [27]. This difference could have been caused by power broadening. Figure 5(b) shows the Allan deviation of the frequency stability of the vapor cell. The short-term stability was 6.0×10⁻¹⁰ τ^-1/2; this is represented by the dotted line. The high short-term noise was possibly due to the power broadening. The Allan deviation at an averaging time of 3000 s was 1.7 × 10⁻¹¹. Under the condition controlling the cell temperature and laser frequency/intensity, these results possibly indicate that getter materials such as Al/Zr do not exist and almost no continuous degassing and leakage occurs, because of the lower amounts or lack of drift compared to that in previous studies [4,6]. Furthermore, long-term stability and miniaturization were improved using our method compared to that in cells fabricated using the conventional method of using the alkali-atom dispenser [4,6,19].

 

Fig. 5. (a) CPT resonance of the fabricated vapor cell. The cell was sealed with N₂/Ar buffer gases. The measurement temperature was 90 °C, and FWHM of the peak was 4.0 kHz. (b) Allan deviation of the vapor cell.

Download Full Size | PDF

5. Summary

This study proposed a new method to fabricate and evaluate Rb vapor cells sealed by wafer-level distillation and two-step bonding for miniature atomic clocks. The vacuum-sealed vapor cells had almost the same absorption spectra and CPT resonant peaks as the theoretical absorption spectra and CPT resonant peaks. It was shown that two vapor cells were fabricated with one dispenser, and this method implied that several vapor cells could be fabricated with one dispenser. The vapor cell, sealed under controlled pressure with buffer gases N₂/Ar, was then evaluated in miniature atomic clocks. Vapor cells fabricated using our method showed substantially improved long-term stability compared to the vapor cell fabricated by the conventional method of using the alkali-atom dispenser. It is concluded that the proposed vapor cell can resolve existing problems of traditional cell fabrication; moreover, it can be widely used, and is cost-effective.