All new semiconductor technology begins in the research lab, where success is measured by the ability to produce a handful of devices that set new standards for performance. Before adoption of any technology into higher volume consumer applications, huge challenges must be overcome. The fourth wave of semiconductor technology, InP, faces the same barriers to adoption that all new technologies encounter. Attempting to force a semiconductor technology into the commercial market will inevitably lead to failure — but where there are compelling advantages for the customer to adopt a new technology, a successful market launch of a new technology is virtually unstoppable.

Market pull for InP
For InP-based devices, customer pull in several target markets is undeniable. In the fiber optic arena, InP is the only semiconductor technology that allows photodetectors and lasers to be integrated on the same substrate with other analog and mixed signal functionality, providing advances in integration and cost reduction that could spur significant paradigm shifts in the fiber optic components. In the wireless industry, InP-based amplifiers provide significant improvements in both higher performance and lower power consumption, high linearity and low temperature sensitivity that significantly enhance the battery life and reception in current handset designs. In millimeter wave applications that are beyond the capabilities of GaAs or Si, InP devices can be easily fabricated for passive imaging and other applications that are just emerging in the marketplace.

New technologies need coordinated accomplishments on a number of fronts to make the leap from the research lab to the production line. This was true for Si, and it was also true for GaAs. Fortunately, most of the equipment and process innovations that took GaAs into production just a decade ago can also be applied to InP production. Although there are specific alloy differences, advances made in epitaxial equipment, including multiwafer MOCVD and MBE systems, are used in both InP and GaAs production, as are dry etching and substrate thinning. Since the manufacturing infrastructure is the same for both technologies, there is no large investment for new fabrication line equipment that must be made to switch to InP. Furthermore, substrate vendors are now making great strides in large- diameter InP substrate development, just as they did for GaAs a decade ago.

When GaAs technology was initially funded for defense and aerospace applications, it took more than a decade to transition into commercial applications. InP is making that same transition today, going down the same path initiated for GaAs in the early 90s. In 1993, TRW announced its plans to produce GaAs HBTs for the commercial marketplace. This plan was initially received with skepticism, as HBT was perceived to be neither reliable nor manufacturable in high production quantities. Today, GaAs-based HBTs are used in many high-volume consumer applications and is the technology of choice for several commercial telecommunications applications, including those from an industry leader, which licensed the technology and holds 19 percent of the worldwide GaAs MMIC market for wireless applications. Today, we are seeing InP follow the same ramp up in terms of substrate size and quality generating a concomitant reduction in price, and the same acceptance by customers who see a performance differentiator.

Still, the advantages of InP process technology are often overshadowed by concerns regarding cost, yield and reliability of any process technology in its formative years. The lack of InP processing experience rife in the market today is perhaps the greatest barrier as companies attempt to move the InP process from the research lab into the production line. In other words, regardless of InP's superior performance, a cost-effective manufacturing process is required to allow this technology to compete effectively in the marketplace.

One approach has been to model the InP process development on proven, high-volume GaAs HBT production processes. Lessons learned from the GaAs transition are applicable to the InP scenario. For example, the process flow in the GaAs HBT process is directly transferable to the InP HBT process.

Challenges for volume production
For any new technology to gain acceptance and long-term success in the market, it must have not only the ability to be produced in volume, but must have proven reliability — the key to the initial application. In this area, InP HEMT MMICs are already proving themselves in challenging in-orbit applications. In this way, the stringent reliability requirements of satellite components are actually paving the way for commercial, terrestrial applications of InP.

A key facet of technology adoption, the ability to produce any technology in volume, must be designed into the semiconductor process during the development phase. However, it only comes into its own during the volume production phase when the fabrication line is fully exercised. As volume production increases, the cost for any semiconductor device — including those made using InP — will come down. As processing experience improves and yields rise, prices for a wide range of components decrease by factor of two or more through the life cycle of typical device.

Initial production snags are to be expected when new technologies transition from R&D to prototype to volume production. Ge, Si and GaAs each experienced growing pains, and InP can benefit by examining the types of problems GaAs encountered when making its transition to the commercial market.

When initial attempts to commercialize GaAs were made several years ago, availability, cost and reproducibility of GaAs substrates were a challenge, as was establishing a reproducible GaAs HBT production process. A backside process for thinning and fabricating GaAs substrates was also critical. The result was a high-volume, cost-efficient GaAs production line. When applied to the InP production process, these lessons expedite volume production.

Production challenges aside, cost of materials is another factor, and one already addressed through GaAs experience. Wafer processing requires some of the most expensive materials, and HEMT and HBT devices are only as good as those materials with which they are built. Extensive material characterization and complete understanding of the material quality before processing is the key to success for InP devices. Establishing correlations between material characteristics and device performance for both GaAs and InP is a key element to a stable and successful fabrication. Fundamental physical modeling enables theoretical correlation to actual results.

Still behind Si
InP lithography today is several generations behind Si, yet InP still yields performance that far exceeds that of any production Si process. As InP moves to smaller geometries, this performance differentiation will increase significantly. SiGe HBT devices with 0.13µ emitters cannot compete with InP HBT devices with emitters 10 times that size. Imagine what will happen to circuit performance as InP HBTs are reduced to sizes approaching that of Si today! InP HEMT has the highest performance of any transistor, with published results for integrated circuits operating at 220 GHz, far in excess of the published results of any other substrates.

New technologies such as InP first generate excitement in university and government laboratories, as well as corporate research and development centers. Reputations made as "firsts" are published and presented at key conferences, but, as the technology matures, researchers often lose interest and move on to other "firsts."

However, this phase of technology development is actually the most important, and in many ways, the most difficult. The companies that are successful in transitioning these technologies into production are not necessarily the first to publish. Success will be found by the company that can transition an exciting technology into a viable commercial application, long after researchers have moved on to more exciting projects.

A final key to success is ensuring that the research engineers — the people who developed the technology — are participating in the transition from research to prototype to production, moving on to new research endeavors only after they have achieved full production.

Rather than merely throwing the technology over the fence to its production people, incorporate a team effort among all who participate in the birth and nurturing of the technology.