Micro Vapor Cells Sealed by Two-step Bonding for Miniature Atomic Clocks

This paper reports the proposal, fabrication, and evaluation of Rb vapor cells sealed by two-step bonding for miniature atomic clocks. The proposed fabrication method is separating the light path filled with Alkali atoms and Alkali dispenser ’ s room at wafer level process. The first fabricated sample is unexpectedly sealed with some air, and the vapor cell is measured the absorption spectra and CPT resonance. The flow solving the problems is indicated, and the vapor cells are fabricated again with the improved process. The FWHM of CPT resonance fits almost the theoretical value in vacuum-sealed samples, which are filled with Rb atoms carried from one dispenser. These results show that fabricated vapor cells are filled with Rb without degassing or leaks, and the reliability of the fabrication process and the production efficiency can be expected. Rb vapor cell sealed in the controlled pressure with buffer gases N 2 /Ar is also fabricated and evaluated. The spectrum line is broadened, and CPT resonance peak is narrowed 4.0 kHz at 90 °C. These results show the proposal fabrication method of the alkali vapor cells is useful for miniature atomic clocks. which resulted in leak, and a polishing process was added in the proposal fabrication process. By the improved fabrication process, two vacuum-sealed vapor cells had almost theoretical absorption spectra and CPT resonant peaks, and the degassing and leak were found to be minimal. The proposal fabrication process is separating the light path room with Rb and dispenser ’ s room, and additionally the fabricated two vapor cells filled with Rb from one Rb dispenser indicated the lower cost fabrication process. The vapor cell sealed in the controlled pressure with buffer gases N 2 /Ar was measured the absorption spectra and CPT resonance. The FWHM of the CPT resonance was narrowed 4.0 kHz at 90 °C. These results show the proposal fabrication method of the alkali vapor cells is widely useful for miniature atomic clocks.


Introduction
The demand of larger amount of data and higher speed communication are always existed. To meet the demand, carrier waves are employed to a wider range or higher frequency, and precise time synchronization system have been developed. In the time synchronization for communication, the base station is needed to have exact time and have atomic clocks. Recently miniature atomic clocks are expected to put on mobile devices like smartphone in order to have accurate time synchronization between the terminal and base station, or terminal and another terminal [1][2][3][4][5][6][7][8][9][10] .
Miniature atomic clocks have vapor cells, which are including alkali atoms Rb or Cs 1,2 .
Applying the laser light to the vapor cell, the reference signal based on the gap of the energy level of the hyperfine structure of Alkali atoms can be obtained from the transmitted light. The reference signal is used to correct the error of the oscillator. This oscillator can keep precise time, and such feedback system is atomic clocks [11][12][13][14] . In addition, recently, the subpicotesla magnetometry or NMR gyroscope using alkali vapor cells has developed [15][16][17][18][19] .
The vapor cells are formed by the microfabrication methods [20][21][22] . Sealing is used anodic bonding at the wafer-level 23 . There have been developed the novel fabrication methods to seal with Alkali atoms, but still it is challenging to seal Alkali atoms at wafer-level packaging 24 . With the mixture of AlkaliCl and BaN6 25 , Rb can be evaporated each cell from the glass ample, and the anodic bonding with a glass wafer is performed in the chamber. This method is needed complicated machines and the cost of the vapor cell is increased. With the alkali atom dispensers, the fabrication process can be straightforward, but the space of the dispenser is needed 4,26,27 . This method refuses miniaturization, and Al/Zr can absorb N2 gas, which is widely used as buffer gas 28 . Figure 1 (a) shows the coherent population trapping (CPT) system, which is normally chosen for miniature atomic clock operation 11,[29][30][31] . CPT is induced by irradiating two laser fields, which are obtained as sidebands from the modulated laser. When the energy difference of the two laser fields is equal to the energy gap of the two ground states, quantum interference impedes photo-absorption, resulting in a narrow transmission peak, which is known as the CPT resonance. By locking the peak, the oscillator's signal can be corrected and keep stable. The short-term frequency stability of the atomic clocks is often characterized by Allan deviation , which is expressed by

The principle and calculation method for CPT resonance
where is the quality factor of the resonance peak, is averaging time, / is the signal-to-noise ratio in detecting the CPT resonance 32 . Q is defined as the ratio of the resonant frequency and FWHM (Full Width Half Maximum) of the CPT resonant peak Δ , which is express as = . (2) Δ is given by , where 2 is the relaxation rate in the absence of laser radiation, is Rabi angular frequency, and * is the decay rate of the excited state. It is known that 2 / * is proportional to the laser intensity 29 . 2 is described as the sum of the relaxation by wall collision 2 , the relaxation by Rb atoms with buffer-gas molecule 2 , and spinexchange interaction between Rb atoms 2 , as follows 2 = 2 + 2 + 2 .
(4) 2 is given as , where is the radius, is the length of the cylindrical column, 0 is diffusion coefficient, is buffer gas pressure at temperature , 0 is reference pressure at a temperature 0 . The cylindrical column image shows in Fig. 1 (b). 2 is given as , where 0 is Loschmidt's constant, is disorientation cross-section which characterizes rubidium-buffer gas collisions, ̅ is the mean relative velocity of rubidium and buffer gas atoms. 2 is given as , where is nuclear spin quantum number, is disorientation cross-section which characterize rubidium atom collisions.
The buffer gases are used to reduce the relaxation due to the wall collisions. Without the buffer gases, the main relaxation depends on the wall collision. The relation is express as where is the 1/e velocity width of a Maxwell Boltzmann distribution and is the mean free path of Rb atom, and is almost same as the minimum length in the vapor cell 33  kHz with N2/Ar buffer gases 3.6 kPa at 90 °C. Figure 2 show the conventional-typed vapor cells using Alkali-atom dispenser and the proposal vapor cell, which is filled with Rb from Alkali-atom dispenser. What you can see from the viewpoint of the structural feature is that an Alkali-atom dispenser remains in Fig. 2 (a), on the other hand, an Alkali-atom dispenser is detached in Fig. 2 (b).There are many fabrication methods to use glass wafer 34,35 , and in this process, by etching the glass Rb atoms can be moved through the gap and the small gap can be sealed by anodic bonding after first sealing.

Evaluation and feedback
The evaluation setup for the Rb absorption spectrum is shown in Fig. 4 (a), and for CPT resonance is shown in Fig. 4 (b). In this measurement, VCSEL of 795 nm wavelength is used. A photodiode was used for the detection of the transmitted laser beam. In order to shield external magnetic fields, the vapor cell was covered with permalloy plates.  Figure 4 (d) shows the CPT resonant peak, which was shifted 4.7 kHz from the transition frequency for 87Rb, 6.83 GHz shown in Fig. 1 (a). This shift indicates there are some gases in the vapor cell, and FWHM of the peak was 4.3 kHz. These results meant unexpected gases remains, but the remained gases were worked as the buffer gases very well. It seemed the gases were from the air through the small holes between the Tempax glass and Si surface after Process 3 in Fig. 3 (a). If the leak occurs after Process 5, Rb in the vapor cell can react with much oxygen in air and Rb atoms would be oxidized and the absorption spectra wouldn't be obtained. Fig. 3 (a), and the hole processed by mechanical drills is observed by SEM (Scanning Electron Microscope) in Fig. 5 (b). Figure 5 (c) and (d) were also observed by SEM, and the scratches were found on the Si surface. Those scratches were possibly formed in the mechanical drill process. Figure 5 (e) and (f) show that the air sealing process and expected phenomenon. Air includes approximately 80% N2 and 20% O2, and O2 can be absorbed by the Rb dispenser. N2 cannot be easily absorbed by the Rb dispenser and can behave as a buffer gas.

Figure 5 (a) shows the Si wafer after Process 1 in
In order to smooth the scratches, a polishing process was added after Process 1 in Fig. 3 (a). Then we fabricated and evaluated three vapor cells; two are sealed in vacuum in Process 3, and the other is sealed in the controlled pressure with buffer gases N2/Ar, which mix ratio is N2 : Ar = 5 : 8. Figure 6 (a) shows the vacuum sealed 20 mm-square wafer after Process 3 in Fig. 3 (a), and only the surrounding was founded to be anodically bonded. Figure 6 (b) shows the wafer after Process 4, and the center part was anodicallybonded and cells were sealed. Deposited Rb was also observed at the glass surface of some cells.   Figure 7 (b) shows the measured CPT resonance, and the peak was shifted by 3.6 kHz from the clock frequency, and N2/Ar buffer gasses pressure was estimated to be 2.76 kPa 11 . The pressure was expected to seal 3.6 kPa at 90 °C in the fabrication process, and this difference can be considered because of the unevenness of the temperature and pressure in the chamber during anodic bonding in Process 3 in Fig. 3 (a). The FWHM of the CPT resonance peak was 4.0 kHz at 90 °C, and FWHM from a theoretical calculation using the cylindrical cavity model from the equation (4), (5), (6), and (7) was 2.1 kHz.
The difference might be caused by the power broadening. Although there were some differences from theoretical values, it can be said that the vapor cell fabricated successfully for miniature atomic clocks.

Summary
In summary, the proposal, fabrication, and evaluation of Rb vapor cells sealed by twostep bonding for miniature atomic clocks were introduced. First fabricated vapor cell had some problems, but the result was better for miniature atomic clocks. There was almost no reproducibility, but it can be said that this is the vapor cell using some atmosphere as buffer gases. Then the problem was expected that some scratches during mechanical drills, which resulted in leak, and a polishing process was added in the proposal fabrication process. By the improved fabrication process, two vacuum-sealed vapor cells had almost theoretical absorption spectra and CPT resonant peaks, and the degassing and leak were         (a) After Process 3 in Fig. 3, only the surrounding was anodically bonded and vacuum sealed in 20 mm square wafer.
(b) After Process 4 in Fig. 3, the center part was also anodically bonded, and some cells were observed deposited Rb on the surface of the glass.
(c) After Process 5 in Fig. 3, the vapor cell size is 3.7×3.7×4.0 mm, which cell name is V-CELL 1. On the center part of the glass surface deposited Rb was found.
(d) Absorption line spectrum of Rb D1 line of the fabricated vapor cells, which were named V-CELL 1 and V-CELL 2. They were fabricated in the same wafer and sealed in vacuum. The measurement temperature was 90 °C and the setup was described in Fig. 4 (a).
(e) The measurement results of the CPT resonance of the fabricated vapor cells. The measurement temperature was 90 °C and the setup was described in Fig. 4 (b). FWHM of the peak were 190 kHz and 220 kHz, respectively. (a) Absorption line spectrum of Rb D1 line of the fabricated vapor cell, and the cell was sealed with N2/Ar buffer gases. The measurement temperature was 90 °C and the setup was described in Fig. 4 (a).
(b) CPT resonance of the fabricated vapor cell, and the cell was sealed with N2/Ar buffer gases. The measurement temperature was 90 °C and the setup was described in Fig. 4 (b).
FWHM of the peak was 4.0 kHz.