Last year Dartmouth made headlines around the globe for a tiny innovation: the world’s smallest untethered mobile robot. A joint project between Dartmouth computer scientists and Thayer School engineers, the microrobot was created in the microengineering lab in Cummings Hall. We asked microengineering lab director Professor Christopher Levey, co-PI and lead creator of the breakthrough robot’s electromechanical design and fabrication, to tell us how he and his colleagues took microrobots to a whole new level.

The entire robot is under 10 micrometers thick — one-tenth the thickness of a human hair — and consists of a locomotion drive that is 60 micrometers by 120 micrometers and a steering arm that is 8 micrometers by 130 micrometers. When placed on a penny and viewed through a microscope it looks like a hair on Lincoln’s chin.

Whereas earlier mobile devices this size were powered via wires or rails, our robot moves freely on its surface. It is the tiniest robot without tethers that is engaged in its own locomotion and steering. Power and instructions are broadcast from a grid of electrodes buried under the surface it walks on — there is no direct electrical contact.

The robot contains two independent actuators — the “scratch drive” for forward locomotion and the steering arm for turning. The scratch drive moves like an inchworm, moving forward an average of 10 nanometers each step. We have demonstrated this motion for more than 40 million such steps with no sign of fatigue, covering a distance of about one foot over the course of half an hour. If a person took that many steps, she would travel halfway around the world. The robot turns by snapping down a paddle on the end of its steering arm, then pivoting around it. An operator can control forward motion and turning with a sequence of commands through the wireless transmission that also powers the motion.

About eight years ago, I was having a discussion with Bruce Donald and Daniela Rus, both computer science professors at Dartmouth at the time (Bruce is now at Duke, and Daniela is at MIT), about their work involving large-scale (brick-sized) self-reconfiguring robots. We started thinking about how to develop similar robot modules but on a much, much smaller scale. Such robots would have several advantages: they would be on the right scale to interact with other very small things and we could make large numbers of them at very low cost by using the microfabrication techniques first developed for the microelectronics industry. The complex interactions of thousands of robots could lead to effects that are qualitatively very different from those of just a few robots, in the same way that the behavior of a computer is much richer than that of a few switches.

Micromachining research spans a broad range of disciplines, including electrical, mechanical, and chemical engineering, physics, materials science, computer science, and robotics. The success of this microrobot is a reflection of Thayer’s non-departmental focus and a strong interdisciplinary collaboration that spanned across Dartmouth. It started with the brainstorming session I had with Bruce Donald and Daniela Rus and grew to include many others, particularly recent computer science doctoral student Craig McGray Adv’05 and current doctoral candidate Igor Paprotny, who did most of the work. Thayer School Professors Ursula Gibson ’76 and Francis Kennedy and Ellen Pettigrew ’08, an intern with Dartmouth’s Women in Science Project, also helped.

Given the minute size of the robot and the delicate environment in which it operates, we faced many challenges. The broadest was how to pack the functionality of locomotion, steering, memory, power delivery, and reception and decoding of control commands into this small a package. We met this challenge by keeping it simple; the entire robot is a single piece of silicon. There are no parts to assemble and no batteries or microprocessors to wire up. Memory is stored in a simple spring-loaded electromechanical switch, the steering arm, which also serves mechanically to make the robot turn.

We also needed to develop a surface environment that would supply power to the robot without using wires, tracks, springs, or other physical connectors. I came up with the idea of using a grid of interdigitated electrodes buried in an insulator to induce a charge polarization in the robot and a resulting vertical force. The scratch-drive mechanism then translates this force into lateral locomotion. After considerable effort by the graduate students in the Thayer microengineering lab fabricating this environment, we found that this scheme works well. It is a “one-wire” system, that is, there is a single signal wire plus one ground wire attached to the environment. All commands are also sent on the signal wire, and distributed uniformly over the environmental surface, to be decoded by the robot.

Another challenge was how to address various steering arms and scratch drives individually. Our analysis showed that we could not do this using the standard microfabrication techniques, which allow a designer control of only in-plane shapes; we needed to curl our devices up out of the plane. We took advantage of a previous Thayer microengineering lab invention to do this: Stress Engineered Micro-Structures (SEMS). Engineers working on micromachines typically try to minimize stress gradients in their materials so objects won’t curl. In previous work with former Thayer School Professor Al Henning ’77 Adv’79 and graduate student Chia-Lun Tsai Th’98, we controlled stress gradients to create usefully curved micromachine parts. The trick is to add a “stressor” layer under tension to the top of the device and to control the extent of the resulting curl through in-plane patterning of the size of the stressor. We now use this SEMS technique to add just the desired curl to our robot steering arms, and hence to allow control of individual arms: Arms curled a long way from the base are thereby programmed to respond only to high signal voltages; arms with little curl will snap down in response to lower voltages.

An obstacle we needed to overcome was that of charging. We found that the scratch drive built up an electrostatic charge, causing it to occasionally stick to the surface. We got around this by alternating positive and negative charges. If the environment charges up positively, a fraction of a second later we switch to a negative waveform to reverse that process. We keep reversing the polarity in order to avoid building up any one charge. The robot responds equally to either polarity.

Moving robots this size anywhere we want in two dimensions is a tremendous accomplishment. We managed to free scratch drives from rails and tethers and then added a tiny silicon paddle (coated on the upside with chromium — the SEMS stressor layer) so the robot can turn on command. We can control the paddle’s motion separately from the scratch drive; the paddle responds to higher voltage levels. And by varying the size and shape of the paddle, we can create robots that respond to various voltages.