Tiny machines, so small that they're invisible to the human eye, are tackling a wide array of challenges in the manufacturing of automobiles and airplanes, in chemistry and biology, and in medicine and sensing. In optical communications, these silicon micromachinesalso known as micro-electro-mechanical systems (MEMS)--are showing up in optical switches, add/drop multiplexers, optical crossconnects, and other critical components. The devices have the potential to revolutionize lightwave systems or put the power of a laboratory in the palm of your hand.
Manufacturers make MEMS devices by using integrated circuit (IC) batch-processing techniques. Even though the fabrication process is complex, the volumes involved make it economically viable. A typical fabrication cycle begins with a silicon wafer on which multiple layers are deposited and lithographically patterned. Some of the layers are acid-soluble sacrificial oxides included as spacers. Once the deposition process is complete, sacrificial oxide layers are etched away to release elements that may be hinged, spring-loaded, or otherwise suspended (see figure 1). The microfabrication techniques used allow designers to build devices that are tens to hundreds of microns in size.
Figure 1. After deposition and patterning (top), the sacrificial oxide layers are etched away to release the device (bottom) so that individual parts can move under electrical actuation.
MEMS manufacturing benefits from the considerable technological resources of the IC fabrication industry. At the same time, the scale of the devices allows them to be manufactured using previous-generation equipment. In an era in which an IC fab costs $1 billion and becomes obsolete in less than five years, the ability to reuse the equipment for a new class of cutting-edge products is enormously appealing.
IC fabrication techniques also allow designers to integrate micromechanical, analog, and digital microelectronic devices on the same chip, which produces multifunctional integrated systems. Despite their size, MEMS devices have proven to be robust and long lasting, especially those whose parts flex without microscopic wear points.
One of the interesting things about MEMS is that the physics of devices at the microscale can differ completely from physics at the macroscale. Richard Feynman predicted this behavior in a paper he wrote in 1959, well before MEMS became commonplace in the photonics industry. At the macroscale level, electromagnetic forces, for example, are strong and electrostatic ones are weak. At the microscale level, however, the situation is reversed. Another example of this role reversal is the relative importance of friction and inertia. For objects the size of human beings, inertia is relatively important, while friction is less so. At the microscale, friction becomes the dominant effect, while inertia is unimportant. A paramecium, for example, never would be able to discover inertia and Newton's second law.
At Bell Laboratories (Murray Hill, NJ), our group has focused on telecom devices such as optical modulators, variable attenuators, switches, add/drop multiplexers, active equalizers, and optical crossconnects. We see applicationsand opportunitiesfor MEMS components throughout lightwave systems, in the core networks, in passive optical networks, and in metro rings. Indeed, in these places a MEMS-based solution may be the best choice.13 mirror, mirror, in the switch
Our team has developed a 1 X 2 MEMS optical switch that achieves a response time of less than 70 µs. The design features a mirror connected to an electrostatically activated see-saw that is driven by a flat plate. Applying a voltage between the suspended plate and the substrate generates a force that pulls the plate down and displaces the mirror. In the "on" position, the see-saw positions the mirror to deflect light from one fiber to another (see figure 2); in the "off" position, the mirror is pulled out of the way to allow the optical signal to pass undeflected. Depending on the design, the actuation requires from 1.24 V to 20 V, with negligible steady-state power consumption. It has loss of less than 1.5 dB with passive alignment and less than 1 dB with active alignment.
Figure 2. An electrostatically activated MEMS switch consists of a mirror connected to a spring-suspended capacitor plate. A reflector drops into the optical path as necessary, to control the route of the light beam.
At each node in a fiber-optic network, some wavelengths must be added, dropped, or passed. An add/drop multiplexer can use an array of micromirrors to perform this function. A grating spectrally demultiplexes the signal and sends each channel to a separate mirror in the array that routes it to either the output port or the drop port. The array can switch up to 16 wavelengths at speeds of 20 µs with better than 30 dB switching contrast.
From a carrier viewpoint, the ideal network would allow flexible provisioning and rapid restoration after service interruptions. Large MEMS-based optical crossconnects offer a viable way to accomplish this. A typical design consists of an array of two-axis micromirrors facing a fixed mirror or a second MEMS array (see figure 3). Light from an input fiber is focused onto one of the micromirrors, which routes it to the appropriate output micromirror by bouncing it off the fixed mirror at a specific angle. The output micromirror then sends the signal to a designated output fiber. Such fabrics are low loss, require minimal power, and work independently of data rate and data format. They operate at both 1.3 µm and 1.5 µm and can scale to very large port counts.
Figure 3. In MEMS-based optical crossconnects, light from input fibers passes through imaging lenses to an array of mirrors that can tilt on two axes. A mirror deflects a given beam to the flat reflector, which returns the beam to the mirror array to be reflected to the appropriate output fiber.
Our optical crossconnect design uses an array of two-axis micromirrors. We have used the design to develop crossconnects with a petabit capacity. The switches have been deployed commercially since July 2000.
Optical MEMS technology can allow designers to build a variety of novel lightwave subsystems. It does entail some interesting technical challenges, however. On one hand, the use of IC fabrication methods means that engineers must face the same issues that are involved in building very-large-scale-integration (VLSI) chips. On the other hand, optical MEMS are also lightwave devices. Coupling light onto and off of the chiptaking an optical signal from the macroscale to the microscaleactually becomes a larger challenge than that posed by fabrication concerns. Solving the coupling issue requires the development of hermetically sealed packages with windows; flat mirrors that maintain planarity over a wide temperature range; and low loss, high reflectivity coatings. The teams that will compete in this space will be multidisciplinary, combining capabilities in physics, chemistry, VLSI processing, mechanical engineering, optical engineering, and reliability testing.
The challenges are great, but so are the payoffs. Market analysts expect the optical MEMS industry to become a multibillion-dollar market in five years. That means that people can expect this strategic technology to provide numerous engineering challenges and business opportunities for many years to come. oe
1. Randy Giles et al., Optical Fiber Communications (OFC) Conference, postdeadline paper #PD2, (1998).
2. A.G. Dentai et al., Electronics Letters, 33, 718, 1997.
3. J. E. Ford et al., J. Lightwave Technology, 17, 904, 1999.
Dare to experiment
It pays to experiment, says David Bishop, director of micromechanical research at Lucent Technologies (Murray Hill, NJ), and he should know. When Bishop and some associates set out to build a MEMS optical switch, they weren't even sure it was possible. Today, MEMS switching is a hot market sector with dozens of competitors.
"About three years ago, our business unit suggested that electronic switches were going to run out of gas," Bishop says. Intrigued by MEMS devices, the group asked if Bishop could build an optical switch using micro-electro-mechanical methods. "I certainly didn't know how at the time," he says, "but they planted a seed in my mind that germinated for a while."
Two years ago, the group managed to demonstrate an array of two-axis mirrors, but they had doubts that the devices were stable enough for optical switching. "We easily could have talked ourselves out of building these mirrors because we could see potential problems," Bishop says. "But we were experimenting with everything, so we went ahead. We were shocked at how well they worked."
When the group showed their successful device to the management at Bell Laboratories, they got the green light. "It was a 'gulp' kind of moment," Bishop says. "They gave us everything we asked for and gave us no excuse to fail. I was reminded of the saying, 'Be careful what you wish for, you might get it.'"
Management had only one conditionwhatever Bishop did, he had to do it quickly. "The marketplace wasn't going to give us the luxury of spending years to get this switch to work," he says. The group met the challenge, which went from concept to product in 18 months. "This was record time for something this complicated," says Bishop. "It has been a wild and interesting ride."
Bishop believes that MEMS techniques allow researchers to build the most amazing devices, such as micro-microphones, micromotors, micro-microscopes, and optical switches. "The people that work in this field, including myself, think that this technology is just so incredibly cool that we just enjoy working in it," he says.
Laurie Ann Toupin
David Bishop, Randy Giles, Vladimir Aksyuk
David Bishop, Randy Giles, and Vladimir Aksyuk are with Lucent Technologies' Bell Laboratories, Murray Hill, NJ.