A new nanomechanical resonator inspired by the structure of a spider’s web could be used in quantum sensors to sense ultra-small forces such as gravity. The resonator, which was designed using machine learning, works at room temperature – a first for a device of this kind.
Nanomechanical resonators are tiny vibrating beams that oscillate at very high resonant frequencies – often in the megahertz or gigahertz range. They are employed in a range of applications, including telecommunications, and can also be used to detect and determine the mass of tiny objects such as single DNA molecules or viruses. They work on the principle that whenever a small particle is absorbed onto the beam, the frequency at which the beam vibrates changes in a way that can be monitored and used to calculate the particle’s mass.
The ultra-sensitivity of these devices can work against them, however, as it makes them extremely sensitive to ambient thermal noise. As a result, such resonators need to be kept at temperatures near absolute zero (–273.15 °C) to counter these unwanted vibrations.
A team of researchers from TU Delft in the Netherlands has now made a resonator that works at room temperature thanks to its excellent isolation from external noise. The new device’s design was inspired by the structure of a spiderweb, which is one of nature’s best vibration detectors.
Spider silk threads are very tough and stiff, boasting strength-to-weight ratios on a par with steel. They are thus able to withstand large impacts while remaining sensitive enough to detect and trap small flying insects. Crucially, they are also most sensitive to vibrations emanating from inside the web, rather than from vibrational disturbances in the surrounding environment, such as wind.
This unique behaviour is the result of millions of years of evolution and thus represents a good starting point for machine-learning algorithms to design nanomechanical sensors, says Richard Norte, who co-led the new study together with Miguel Bessa.
Making use of Bayesian optimization
The researchers chose a popular mechanical-resonator material, silicon nitride (Si3N4), for their sensor. They used an algorithm known as Bayesian optimization to find a good spiderweb-type design quickly and efficiently, having first specified that the machine-learning programme should consider devices made from a slab of 20 nm-thick Si3N4 freely suspended over a length of several millimetres.
To their surprise, they found that the algorithm proposed a relatively straightforward web, one that consists of just six strings assembled in an apparently simple way. Better still, the team’s computer simulations showed that the device would work at room temperature due to its high mechanical quality factor, which is the ratio of energy stored in a resonator over the energy dissipated over one oscillation cycle. This quality factor, denoted Qm, exceeds one billion in this temperature range thanks to a novel “torsional soft-clamping” mechanism that isolates the device’s vibration modes from the ambient thermal environment and was discovered by the data-driven optimization algorithm.
“What is fascinating is that the machine-learning algorithm independently homes in on torsional vibration mechanisms, which are actually used by spiderwebs in nature when detecting prey, although the algorithm does not have any prior knowledge of how a spiderweb functions,” they explain.
The researchers then made a real-world sensor based on this optimized design, set it vibrating with piezoelectric stages and used an optical interferometer to measure the time it took for the vibrations to stop. These “ringdown” measurements provide information about the rate of decay of the resonator’s amplitude and so the rate at which it dissipates energy – values that are then used to calculate its Qm.
The researchers report that almost no energy is lost outside of their microchip-based “spiderweb”. “The vibrations move in a circle on the inside and don’t touch the outside,” Norte explains. “This is somewhat like giving someone a single push on a swing and having them swing on for nearly a century without stopping.”
Extending to other geometries
The new device could be used to search for dark matter or probe ultra-tiny forces such as gravity that are notoriously difficult to measure, the researchers say. While their initial design uses silicon nitride, they believe that their approach could be extended to other materials such as diamond, gallium arsenide, silicon carbide, indium gallium phosphide, fused silica glass, silicon, phosphorus carbide and even superconducting films. Using machine learning to design such devices is just a first step toward developing the next generation of nanomechanical resonators, the researchers add, and it might also be extended to geometries other than spiderweb-like designs.
Nanomechanical qubit could be highly stable
Spurred on by their results, the team now plans to develop new machine-learning algorithms to design optical nanostructures rather than just mechanical ones. “These structures will be used to solve other physics problems such as creating lightsails that travel at a quarter of the speed of light,” Bessa and Norta tell Physics World. “They will require a careful design that allows them to be ultra-lightweight and highly reflective – characteristics that are difficult to achieve simultaneously.”
The work is detailed in Advanced Materials.
