NEMS: All you need is feedback.

A lmost all electronic devices — from wristwatches to radios to computers — rely on electrical oscillators in some way. However, all oscillators are prone to losses, so it is necessary to keep adding energy to keep them oscillating. The most intuitive way to do this is to use feedback: take a small fraction of the output of the oscillator, amplify it and then, somehow, feed this amplified signal back into the oscillator circuit without disturbing its overall balance too much. This approach to building self-sustaining oscillator circuits is also used in other kinds of devices such as lasers. On page 342 of this issue, Philip Feng, Christopher White, Ali Hajimiri and Michael Roukes at Caltech report the first example of a self-sustaining nanoelectromechanical oscillator1. Such devices could have applications in sensing, precision timekeeping and communications. The Caltech device is basically a nanoscale bridge made of silicon carbide that resonates at roughly 428 million cycles per second, which is in the ultrahigh frequency (UHF) band. The device is similar to a simple mass-and-spring system: an a.c. voltage from an external oscillator applies a sinusoidal force to the system, causing energy to move back and forth between the potential energy of the spring and the kinetic energy of the mass. In every cycle, however, a small fraction of energy is lost to friction or by radiation to the surroundings, with the size of this loss depending on the quality factor of the resonator. The external oscillator is needed to replenish these losses, and the oscillations start to fade away if the external oscillator is switched off (Fig. 1). Previously, most nanoelectromechanical resonators relied on external oscillators. In some cases, the frequency of the external oscillator was swept around the mechanical resonance, which allowed information about the mass of the resonator2, the energy losses3 and so forth to be extracted from the frequency response. In other cases, phase-lock techniques were used to build frequency-tracking circuits that could measure very small masses and forces4. In the Caltech experiment, the external oscillator is no longer needed because the energy that sustains the oscillations comes solely from a d.c. power supply (Fig. 1). Oscillations can be started by a small disturbance in the circuit, such as a thermal ‘kick’, and if the feedback system is well tuned, only oscillations at the desired frequency are sustained. When the resonator is running in a steady state, its mechanical oscillations are first converted into electrical oscillations by an electromechanical transducer and then amplified. These electrical oscillations are converted back into an oscillating mechanical force by another transducer, which is applied to the resonator to compensate for all the energy losses. The key to sustaining the oscillations is to tune the phase and amplitude around the feedback loop very accurately so that the mechanical force is applied at the correct time and with the right phase. The Caltech group had to surmount a number of technological challenges to get its self-sustaining nanoelectromechanical oscillator to work. The nanoscale bridge at the centre of the resonator moves by only ~1 nm at most, so an ultrasensitive transducer is needed to convert the motion into electrical signals. They use a magnetomotive transducer, which relies upon Faraday’s law to generate an electromotive force on the moving bridge in a magnetic field, but even then the signal is buried in a huge background. Roukes and co-workers have spent over a decade painstakingly developing sophisticated techniques to overcome such challenges, and all this work has paid off in the new device. Even at its development stage, the nanoelectromechanical oscillator can compete with state-of-the-art oscillators in terms of stability and spectral purity. The frequency of a perfect oscillator does not change, and all its oscillatory energy is focused at a single frequency. The frequency of a real oscillator, however, fluctuates in short time intervals, and can drift over longer periods of time. These imperfections are usually caused by external agents, such as amplifiers and/or fundamental physics5. Electrical engineers have developed intuitive measures for quantifying the performance of oscillators In the past, nanoelectromechanical resonators have been passive devices that required external oscillators to keep them working, so the development of a self-sustaining resonator powered only by a d.c. voltage is a major advance. NEMS