Optical microelectromechanical systems (MEMS) devices are driving the trend toward all-optical telecommunication networks. With the ability to directly manipulate an optical signal, MEMS have several applications that eliminate unnecessary optical-electrical-optical (O-E-O) conversions. In the recent past, tilting-mirror MEMS have received a great deal of attention as they entered the marketplace as optical cross-connect switches.

Grating light valve (GLV) diffractive optical MEMs are an alternative approach. Basically, the GLV device is a spatial light modulator with applications in areas such as dynamic gain equalizers and dynamic wavelength selective products.

In an ideal world, all new products would use well-tested technologies, provide economies of scale through mass production, and can be sourced from multiple suppliers. Semiconductors are an excellent example of products that benefit from these attributes. However, most MEMS devices are based on commercially unproven core technologies and require specialized manufacturing processes not easily integrated into a semiconductor fab — including bulk micromachining, heavy metals, or self-assembled lubricant films.

Over the last several years, researchers at Silicon Light Machines have overcome these limitations and developed methods to build optical MEMS device in a CMOS manufacturing facility. The compatibility between GLV MEMS and CMOS fabrication provides highly optimized and tested processes, rapid wafer turnaround time and device development, high yield, tight process tolerances and automated process monitoring.

The GLV optical MEMS device is a dynamic diffraction grating that can act as a seamless one-dimensional array of attenuator elements or switches. When used in an optical network module, the GLV device can selectively shape the optical power spectrum or redirect channels to different ports. Characteristic to the GLV device is fine analog attenuation control, high extinction ratio, high-speed actuation capability, and spatially continuous adjustment. In addition, GLV devices do not experience the fatigue, wear and stiction failure modes typical in other MEMS devices.

The GLV device consists of an array of parallel micro-ribbons suspended above an air gap, and is configured such that alternate ribbons can be dynamically actuated. The ribbons are under high tension of about 800 Mpa so that they remain taut when not actuated. The top layer of the ribbon is aluminum, which serves as both the reflective layer and the top electrode for electrostatic actuation. When a voltage is applied to the ribbon, electrostatic attraction deflects the ribbon downward. The sub-layers of the ribbon are a carefully designed sandwich of stoichiometric Si3N4 and SiO2 films that provides the spring-like restoration force that counterbalances the electrostatic actuation force, and which provides stiffness and stress balance so the ribbon remains flat across its width. Ribbons are about 500 mm long, 10 mm wide, 300 nm thick and closely spaced. The gap between ribbons is less than 0.5mm.

To build the GLV device, the ribbon layers are deposited on a sacrificial layer, which is isotropically etched from underneath the ribbon layers to release the ribbons and form the air gap. The thickness of the sacrificial layer must be carefully optimized for operation in the 1550 nm spectral region. Below the sacrificial layer is an etch-stop layer that protects the underlying bottom electrode during the release process.

The GLV device has alternate "active" ribbons and "bias" ribbons. The bias ribbons have a single common control connection and are generally held at ground potential — the same as the common bottom electrode. Individual electrical connections to each active ribbon electrode provide for independent actuation.

When the voltage of the active ribbons is set to ground potential, all ribbons are undeflected, and the device acts as a mirror. As the voltage to an active ribbon is increased, this region of the array begins to diffract light, thus attenuating the light that is reflected specularly. Maximum attenuation occurs when the bias ribbons reaches one-fourth of the wavelength of the incident light. The GLV device is designed so that one-fourth wavelength deflection is obtained when 10 to 20 volts are applied to the active ribbons.

To form a complete GLV device, the ribbons are replicated several thousand times to form a one-dimensional array of diffracting elements. Again, the compatibility with CMOS processing makes this an easy task. A key feature of the GLV array is that the individual diffraction elements are seamless in that there are no physical boundaries, or dark spaces between elements. This feature means that for certain applications in optical communications the GLV device is agnostic to dense wavelength division multiplexing (DWDM) channel count.

The key metrics for a GLV device are uniformity, insertion loss, extinction ratio, attenuation accuracy, and PLD. And of course, the spectral region is the 1500-1600 nm band used by DWDM optical networks. To obtain high performance of these parameters, many factors of the GLV device must be optimized including the ribbon reflectance, size of the ribbon gaps, parasitic diffraction, and substrate reflectance.

Insertion loss is the excess, or unwanted, optical loss of signal when the device is set to minimum attenuation. Extensive characterization shows that typical insertion loss across the entire GLV device array is about 1 dB, with a uniformity of 0.3 dB.

Extinction ratio is the maximum contrast between the ON and OFF states (specular and diffractive) of the GLV device. We routinely see contrast of greater than 30 dB over the entire array, and have recently developed improvement that will allow us to increase this to 35-40 dB in the future.

Polarization dependent loss is the difference in insertion loss data (in dB) for the two principal axes of polarization, as defined by directions parallel and perpendicular to the length of the ribbon. Results show that the PDL does not exceed 0.16 dB over 10 dB of programmed attenuation and wavelengths that span the c-band.

When used in applications for optical networks, the spectrum of DWDM signals (channels) is dispersed and the discrete wavelengths are then projected onto the GLV — each channel landing on a different place on the light valve array. The DWDM signal is then manipulated (attenuated, equalized, and switched) by applying programming voltages to the active ribbons. The diffracted light diverges from the specular light, and the specular portion is collected, recombined and focused back into the original fiber.

The GLV optical MEMS device is a proven means to switch, modulate and attenuate light for optical communications applications. The fundamental GLV technology surpasses most other light-modulation technologies in terms of speed, accuracy, reliability and manufacturability — making it an optimal solution for large telecom system developers and subsystem OEMs.