Prof. Clifton G Fonstad, Jr

Heterostructure devices, such as laser diodes and high electron mobility transistors, play an increasingly important role in our lives, and are key, enabling components of such common items as compact disk players, cellular telephones, fiber communication links, and direct broadcast television receivers. Their impact would be even greater, however, if a technology existed which could integrate heterostructure devices with silicon VLSI circuitry using the same monolithic wafer-scale batch processing techniques that are largely responsible for the continuing Moore’s Law growth of integrated circuit performance and functionality. Addressing this bottleneck, our research group in the MIT Microsystems Technology Laboratory and Center for Materials Science and Engineering has made significant advances in integrating complex compound semiconductor heterostructure devices with commercial VLSI (very large-scale integration) electronic circuits.

We refer to the general integration methodology we are pursuing as Recess Mounting with Monolithic Metallization, or RM3 (i.e., “RM-cubed”) integration for short. This name embodies the common unifying themes of the several integration techniques we are studying: the mounting of heterostructure devices (or device material) in shallow recesses formed in the dielectric surface layers covering processed integrated circuit wafers, followed by replanarization of the surface and continued wafer-level batch processing to complete the device fabrication and to interconnect the devices electrically with the pre-existing circuitry using monolithic metallization.

Current focus is on an RM3 integration technique we call OptoPill Assembly (OPA), and we are exploring two OPA technologies. In both cases the heterostructures to be integrated are patterned into the proper shape and etched totally free of their original substrate creating thousands of tiny chips or “optopills” ready to be put in place in recesses. The first OPA technology, Manual Pick and Place (MPAP), is a labor-intensive approach that has worked very effectively in a research environment and has allowed us to make rapid progress in developing the post-assembly processes needed complete integration after the recesses are filled. The second OPA technology, Magnetically Assisted Statistical Assembly (MASA) involves automatic assembly whereby the recesses on an IC wafer surface are filled with optopills automatically and in parallel through a batch process exploiting statistics and magnetism.

We are currently involved in three collaborative research programs in which we are using our RM3 technologies to integrate optoelectronic devices on custom designed silicon CMOS ICs for three quite different applications: optical clock distribution on CMOS chips, subsurface imaging in living tissues using diffuse optical tomography, and free-space parallel optical signal processing.

This past year we completed a quantitative evaluation of CMP-processed silicon IC wafers as substrates for rectangular dielectric optical waveguides. We find that they are sufficiently flat that it is feasible to add optical interconnect layers to present back end processes. We are now studying how best to couple such waveguides to RM3-integrated in-plane laser diodes and photodiodes, providing a total optical interconnect fabric for the electronics and optoelectronics communities.

F inally, the research program on microscale thermophotovoltaic (MTPV) energy conversion has moved beyond simply demonstrating the proximity enhancement effect, a goal which was achieved for the first time late in 2000, to analyzing more thoroughly the impact this effect will have on the overall efficiency of the thermal-to-electrical conversion process and to optimizing the corresponding TPV cell design. In addition we have begun analysis of several proposed enhancements to the basic energy-conversion process to further increase efficiency.