Miniature atomic clock makes its debut

02 Nov 2004

A laser-driven atomic clock the size of grain of rice may soon be improving the precision of GPS receivers and replacing the quartz crystal oscillator found in computers and watches. Oliver Graydon reports.

From Opto & Laser Europe November 2004

Scientists in the US have made the world's smallest atomic clock by combining microelectromechanical systems (MEMS) and vertical-cavity surface-emitting laser (VCSEL) technology. The development means that the precision of atomic clock timekeeping could soon be available in handheld devices such as mobile phones, Global Positioning System (GPS) receivers and even wristwatches.

Measuring about the same size as a grain of rice, the inner workings of the clock are about 100 times smaller than those of current designs, and consume just 73mW of electrical power. The prototype, built by researchers at the National Institute of Standards and Technology (NIST) in Boulder, Colorado, provides timekeeping that is accurate to one part in 10¹⁰ - equivalent to 1s in 300 years. This long-term stability is about 1000 times better than that of temperature-compensated quartz crystal oscillators.

"The real power of our technique is that we're able to run the clock on so little electrical power that it could be battery-operated, and that it's small enough to be easily incorporated into a cell phone or some other kind of handheld device," said John Kitching, a physicist from NIST. "And nothing else like it comes close as far as being mass-producible."

Inner workings

Each clock contains a 4mm-high, 1.5mm-wide vertical stack of miniature optical components surrounding a tiny cell containing caesium atoms. From the base up, the layers of the stack comprise: a raw unpackaged 852nm VCSEL; a glass lens; a neutral density filter; a polarizing waveplate; a cell containing caesium atoms; and a silicon photodetector.

The cell is made by etching a square hole with 0.9mm sides into a 1mm-thick silicon wafer. Two glass wafers are then bonded to the wafer to create a tiny hollow cavity. The cavity is filled with caesium atoms before the glass bonding takes place. The glass is also coated with 30nm-thick layer of indium tin oxide to create two transparent heaters that can precisely control the cell's temperature.

"One of the things we have made sure that our design accounts for is the possibility of wafer-scale fabrication," Kitching told OLE. "We can take a wafer of lasers, a wafer of optics and a wafer of cells, and stack them all up and dice them to create the package."

The ability to mass-produce tiny atomic clocks could transform the applications of the technology. For more than 50 years atomic clocks have set the gold standard for time and frequency measurement, but their uses have been limited by their complexity, size and cost. Current cigarette-packet sized designs are hand-assembled, cost around $1000 (€800) and are limited to use in laboratories, GPS satellites and mobile-phone network equipment. So what makes their smaller counterparts more suitable for mass-production?

"A critical difference is the way in which the cell is made," explained Kitching. "In the current generation of clocks the cells containing the caesium atoms are made using glass-blowing techniques. It's a very long process, so that really adds to the expense - and no two cells are identical. With our devices, because everything is lithographically defined, every cell is identical. So in terms of assembling the devices it is much easier, and much quicker and cheaper than conventional designs."

The method of timekeeping used in the latest NIST design is similar to that in the larger atomic clocks installed at standards institutes all over the world, such as NPL in the UK, PTB in Germany and NIST itself. These state-of-the-art clocks rely on a very small but well-defined gap between two energy levels in a caesium 133 gas atom to calibrate a microwave frequency source.

This gap, known as the D2 hyperfine split, corresponds to a frequency of 9.2GHz (to be precise, 9,192,631,770Hz) and is used to define the length of one second.

In NIST's chip clock the VCSEL's drive current is modulated to create optical sidebands that are used to probe the caesium atoms. The frequency of the modulation is then carefully tuned until it exactly matches the hyperfine split. When the alignment is correct there is a small dip (about 1%) in the absorbed optical power, which is detected by the photodetector at the top of the stack.

For the measurement to be accurate the cell needs to be kept at a stable temperature of 85°C, which is achieved by built-in thin-film heaters. It is the need for heating that accounts for the majority of the clock's power consumption of 73mW. However, the NIST team is confident that it can halve the clock's power consumption and increase its stability by a factor of ten. Calculations suggest that a refined design could consume less than 30mW and offer a stability of one part in 10¹¹ which would allow timing precision at the microsecond level over one day.

However, this level of accuracy is still a long way from that achieved by large atomic clocks. For example, NIST's F1 clock boasts a stability of one part in 10¹⁵ equivalent to 1s in 30 million years. "The device that we have demonstrated is the first microfabricated clock and we were simply aiming to get it working and demonstrate that it is possible to make it," said Kitching. "The long-term instabilities that we are seeing have been studied over the last 50 years in larger clocks, and we are confident that we can apply the techniques that have been used to improve their stability to our small clocks."

Sugar-cube solution

Although to date NIST has focused purely on developing the clock's components, Kitching says that it shouldn't be too difficult to integrate these with the necessary supporting electronics to create a complete clock that is very compact. For example, much of the control circuitry could be integrated into a custom-designed silicon chip. In addition, either SiC nano-resonators or thin-film AlN bulk acoustic wave resonators could act as miniature local oscillators for modulating the laser current.

The result of this could be a time reference that is about the size of a sugar cube, costs a few hundred dollars and yet still offers the accuracy of the cigarette-packet-sized versions. GPS receivers are the application most likely to benefit from having precise onboard clocks. The speed and accuracy of the location fix they provide is largely limited to the timing error of their on-board clock, which must measure how long it takes for signals sent from four satellites to reach the receiver. While software and algorithms can help, having a very accurate on-board clock would improve the process further and mean that line-of-sight would only be required with three satellites rather than four.

"This could be very useful for global positioning in urban environments, where you have buildings and other obstructions that essentially block the view of one of the satellites," said Kitching. "Having a very stable clock on your receiver, especially for military applications, is also important in improving jamming resistance."

The potential applications are almost endless. If the clocks can be made cheaply enough they could start to replace the quartz crystal oscillators found in computers, radios and watches. In addition, the technology could be adapted to construct millimetre-sized atomic magnetic-field sensors and modules for spectroscopy calibration.

"I think this is such a general technology, and timing is such a ubiquitous need, that right now we can't even predict what its most important applications are going to be," concluded Kitching.

As for commercializing the design, NIST plans to license the intellectual property to companies that are interested in bringing it to market. "We're a government lab so we are not actually going to make devices, but we are going to push these developments as far as we can and then hand over to a company," said Kitching. "There's definitely a lot of pull from the private sector to get this stuff commercialized and I'm guessing from where we are now that we will see commercial devices on the market in 2-3 years."