We take it for granted now, but the fact that you can flip your phone from portrait to landscape mode depends on accelerometers. As everyone knows, though, the damn things often get it wrong, leaving you staring at a screen that refuses to reorient until you give it a good shake. One of the reasons for the screen refusing to orient correctly is that accelerometers have to balance sensitivity to small changes with the speed of response—a slow accelerometer is a sensitive accelerometer.
This compromise, however, is also due to fabrication limitations. A recent paper in Nature Photonics shows that clever fabrication can result in an accelerometer that is both fast and sensitive.
An accelerometer works by sensing the motion of a suspended test mass. If suspended test mass doesn't mean a lot to you, think of a device containing a bit of hardware consisting of a spring attached with a weight at the end of it. Every time the device moves, the acceleration will set the mass in motion: the direction and vigor of the acceleration are reflected in the direction and amplitude of the spring's oscillations.
Three factors come into play in determining the performance of an accelerometer. The size of the test mass—the more massive it is, the smaller its oscillations are for a given acceleration. For maximum sensitivity, you want it as light as possible.
The second factor is called the Q of the oscillator. Q is, put simply, a measure of how long it takes the test mass to return to rest after it has been given a shake. A high Q oscillator will oscillate for longer than a low Q oscillator, which makes things easier to detect.
The final factor is the resonant frequency of oscillation, which is where the accelerometer is most sensitive—if we were to shake the accelerometer at just that frequency, we would get a giant oscillation. This, however, is not terribly useful, because we want to detect a lot of different frequencies.
The way to do this is to set the resonance frequency above the maximum acceleration that you want to measure. The response of the test mass to accelerations below that frequency is fairly even across the board. Unfortunately, the response is also substantially less sensitive. In any case, the resonant frequency defines the range of frequency components in an acceleration that you can measure, which determines the speed of the accelerometer's response.
So, the perfect accelerometer has a very tiny test mass that doesn't have a strong resonant response, and that response should be at high frequencies so that it's also fast. Of course, such an accelerometer would pick up every vibration, including those due to thermal noise inside the case of the device; generally not a desirable feature.
Obviously, a grand compromise plus some innovative thinking is required to break this trifecta of embuggerance. This is exactly what a team of researchers from Cal Tech and the University of Rochester have done. They recognized that if you take a very tiny mass and make it part of a very high Q resonator, then the mass will be very insensitive to excitation due to temperature. But if you could measure the amplitude of the oscillations, you could measure very tiny accelerations. The key, then, is in how you measure things.
Researchers created their mass by etching away the surrounds of a silicon nitride membrane (one of the springiest materials known to man), leaving a block with a mass of just 10pg (10-12g) suspended by the same material. That material choice means that the resonance frequency is just under 30kHz, allowing a researcher to potentially sense accelerations at a rate of something like 15kHz (which is fast). So, we have an extremely sensitive accelerometer—now we just need to give it a high Q.
As it turns out, the springy material (silicon nitride) is a great material for making tiny optical components. The researchers fabricated a tiny zipper-like structure next to the mass. The zipper acts like a pair of mirrors that reflects light back and forth—which means the researchers placed an optical resonator right next to their mechanical resonator. As the test mass moves back and forth, it changes the length of the optical resonator ever so slightly, and that changes the frequency of the light that it will resonate.
The upshot is that if you shine a laser light into the optical resonator, then as the mass passes back and forth near the resonator, it causes the amplitude of the light exiting the other side of the resonator to change. The higher the Q of the optical resonator, the more sensitive it becomes. So, even though the test mass is only making the tiniest of motions, the sensing device picks it up with ease.
The researchers show that their accelerometer is very close to the absolute limit of what is allowed by quantum mechanics (in terms of the trade, it is called shot-noise limited).
I should note that this is unlikely to turn up in your phone anytime soon. The problem lies in the detection system: that laser was a relatively expensive device, and if you want your accelerometer for a few pennies, this isn't going to do it for you. I can see this being used in research labs for monitoring vibrational noise, and in motion capture suites, where accelerometers are combined with clever algorithms to calculate absolute motion from measured accelerations.
Nature Photonics, 2012, DOI: 10.1038/nphoton.2012.245
