Opto-electronics -> Monolithic integration requires clever process, packaging schemes

Mixed-materials monolithic integration is the best and perhaps only way to accomplish this goal. In other integration techniques such as hybrid packaging, optical or electrical-power efficiency may reduce the performance or functionality of the integrated platform.

Also, monolithic integration reduces or eliminates the need for pick-and-place, bonding (wafer- and/or ball-bonding) and other assembly processes necessary in hybrid assemblies, including those that place discrete elements on a silicon substrate or "workbench."

A further advantage of monolithic integration on a substrate like silicon, vs. a compound semiconductor, comes from having a much better heat-conducting platform (an order of magnitude or better), which can directly improve the power efficiency of individual components such as power transistors and lasers. This thermal advantage provides a lowering of the operating power requirements or an increase in the output power potential at a given input power level, as well as, perhaps, improving reliability due to reduced long-term operating temperatures. This may also present the possibility of reducing the need for costly and somewhat complicated customized isothermal controls.

At the same time, there are challenges that remain before these advantages can truly become mainstream practice, and before systems of all types can take full advantage of the many capabilities being offered by this technology platform.

There is a need for embedded as well as modeling software development that merges the electronic (both digital an d RF/analog) and photonic functionality. This software development would greatly simplify the overall subsystem design process and maximize the potential benefits.

On the wafer-processing side, monolithic integration of the unique photonic and electronic structures and process requires the development of new and clever process techniques. Management of thermal budgets and surface planarity issues during the manufacturing process flow must be done to eliminate or minimize adverse effects on all of the integrated circuit elements on the chip.

Clever combinations of existing packaging techniques or completely new ones will need to come together to create new package designs that incorporate optical and electrical interconnect functionality.

Developing cost models that guide the partitioning of a system are very challenging. They cannot be developed simply by asking, "What was the previous total discrete device cost vs. the new integrated platform?" The overall cost of the system — manufacturing, installation, operation and maintenance, etc. — must be the guiding set of criteria for partitioning the system appropriately.

There are many fundamental advantages for the integration of generic optoelectronic systems. For one, the costs could be significantly reduced by the use of fewer very difficult, and expensive, optical-fiber connections between components. With fewer discrete fiber connections and the number of individual packages required on subsystem boards, improvements could also be made in reliability. Additionally, higher performance levels and data rates are enabled due to shorter interconnection lengths between functions and components.

Much smaller footprints are becoming more critical as optoelectronic communication subsystems are brought closer to the end users. This reduction in size due to integration will relieve the overcrowding and costly floor space issues already being faced in many central offices and switching centers. Complete photonic and ele ctronic integration will increase subsystem functionality, robustness and design flexibility. Optical interconnects reduce electromagnetic interference and improve signal integrity better than traditional electrical-transport methods.

Reconfigurability, design flexibility and remote programmability are necessary to lower the costs of installation and maintenance of dynamic communication systems. This will significantly improve the profitability of existing services and generate new streams of revenue for the service providers.

Telecommunication's ultimate vision for monolithic integration of electronics and photonics would be a product platform of optical nodes at various sections of the system.

A given node could be a monolithic electro-optic integrated circuit, with input and output fibers secured in silicon V-grooves, coupled to optical waveguides on the chip. The waveguides could be fabricated in electro-optic material to provide several different telecommunication functions such as splitting, routing, demultiplexing and multiplexing, switching, modulation and amplification — all integrated with lasers and detectors. Multiple tunable lasers could provide wavelength reconfiguration. All the functions will have the driver electronics and system-management control integrated with the devices.

This kind of integration would have a significant impact on the telecommunications industry. Any given required node configuration would have the manufacturing process flow defined to suit the thermal manufacturing constraints. The required masks and processing tools could be designed to fabricate the desired node in an ASIC-like manner. Small systems changes to a given node could be made with relatively little impact on the process flow, providing a means for a vendor to quickly and simply adapt to changes in a customer's needs.

Other contributors to this article are Robert W. Bollish, engineering operations manager, Thoughtbeam Inc., A Motorola Company (Austin, Texas), and B arbara Foley Barenburg, manager of integrated microphotonics research, Physical Sciences Research Labs, Motorola Inc. (Tempe).