IEEE Communications Magazine • March 200275The Lucent LambdaRouter: MEMSTechnology of the Future Here Today0163-6804/02/$17.00 © 2002 IEEEABSTRACTMEMS devices are beginning to impactalmost every area of science and technology. Infields as disparate as wireless communications,automotive design, entertainment, and lightwavesystems MEMS is increasingly becoming a keytechnology. In this article we discuss MEMSdevices in general, show how and where they willbe used in lightwave systems, and then show indetail how they are allowing a billion dollar busi-ness to be born, that of large all-optical cross-connects. In particular we will highlight oneparticular device, the Lucent LambdaRouter,and show how it is built from the chip on up anddiscuss its performance and applications.INTRODUCTIONDespite the recent slowdown in the telecomequipment market, the underlying fundamentalsof our industry remain strong. The demand forbandwidth, the driver for everything we do, isstill growing at a compound annual growth rate(CAGR) of 100 percent. This then drives theneed for ultra long-haul transmission systems,optical crossconnects at the nodes, metro ringsto feed these core networks, and access networksto drive bandwidth into all of the above. All ofthis will be done under the control of MPLS sig-naling protocols. In this article we focus on therole of the large optical crossconnect for coreswitching and show how MEMS is the only tech-nology currently capable of allowing high-capaci-ty scalable architectures.The need for crossconnects in core networksis driven by the desire of service providers tooffer a rich new class of wavelength services.Designing networks around wavelengths insteadof packets allows the construction of networksthat combine the attributes of ultra high capaci-ty, scalability, flexibility, self-healing, and autoprovisioning, with time of day services and band-width by the minute, hour, week, or year. Anexample of such a network is shown in Fig. 1.The core of the network routes wavelengths toand from destination cities. The standard elec-tronic packet routers load the wavelengths at theedge of the network, but the core of the networkroutes wavelengths, not packets.The inevitability of such a mesh-based net-work as opposed to legacy ring-based architec-tures is driven by the rate of growth of capacityin optical fiber systems. With recent demonstrat-ed capacities of more than 10 Tb/s in a singlefiber, core packet routers are clearly incapableof keeping up. As shown in Fig. 2, electronics ofany kind doing the switching function at thenodes in high-capacity networks is quicklybecoming impractical since the rate of increaseof speed and performance of electronics is muchslower than optics. Optics for the switching func-tion allows network elements to be built that aresmaller, faster, and cheaper with much loweroperating power than their electronic counter-parts. This is why the all-optical crossconnect isthe idea that has launched 100 startups. Thesenetwork elements are universally agreed on asDavid J. Bishop, C. Randy Giles, and Gary P. Austin, Lucent TechnologiesOPTICAL SWITCHING■ Figure 1. Example of a mesh-based network architecture that uses largeregions of optical transparency of the type allowed by ultra-long-haul transportsystems and all-optical crossconnects. This type of network allows for rapidprovisioning, auto-restoration, and virtual private networks (VPNs).AZO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBAZO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBO L BO L BOLBOLBIEEE Communications Magazine • March 200276being the key devices for the next generation ofterabit networks.WHAT ARE MEMS DEVICES?The technology that allows high-port-countdata-rate-independent switches is micro-elec-tro-mechanical-systems (MEMS). As shown inFig. 3, these are silicon micromachines builtjust the same way as a silicon integrated circuit.Starting with a silicon wafer, one deposits andpatterns materials such as polysilicon, siliconnitride, silicon dioxide, and gold in a sequenceof steps, producing a complicated three-dimen-sional structure. However, unlike an integratedcircuit, at the end one releases the device oretches parts of it away, leaving pieces free tomove. A typical example uses HF to removethe oxides leaving the nitrides, polysilicon, andmetals. These devices offer a number of advan-tages to systems designers. Because they arebuilt using IC batch-processing techniques,these devices, albeit complicated, are inexpen-sive to produce because many are fabricated inparallel. Also, they get ever better with time,driven forward by the $200 billion/year verylarge-scale integration (VLSI) business and itsrelentless development of new tools, tech-niques, and processes.VLSI fabrication techniques also allowdesigners to integrate micromechanical, analog,and digital microelectronic devices on the samechip, producing multifunctional integrated sys-tems. Contrary to intuition, MEMS devices haveproven to be robust and long-lived, especiallyones whose parts flex without microscopic wearpoints. Research in this area has been extremelyactive over the last decade, producing micro-scopic versions of most macro-machines. MEMSdevices have a number of desirable attributes tooffer to the systems architect such as small size,high speed, low power, and a high degree offunctionality. In particular, many of us believethat the size scale at which these machines workwell make them a particularly good match tooptics problems where the devices, structures,and relevant wavelengths range in size from oneto several hundred microns.Another feature that can be designed intothese devices is something called self-assembly.This technology allows microscopic springs toassemble complex devices during the releasestep. An example of this is the optical crosscon-nect mirror shown on the right of Fig. 4. Springsat the sides, shown as the gold lifting members,assemble the devices during the release step,leaving the device as shown, raised up off of thesilicon substrate.WHERE DO WE SEEMEMS BEING USED INLIGHTWAVE NETWORKS?Shown in Fig. 5 is a schematic lightwave systemwith places shown in red where these devicesmight find application [1–3]. The possible appli-cation areas range from data modulators, vari-able attenuators, active remote nodes, active■ Figure 2. This diagram shows the design space for electronic