Reading time: 3 minutes
Joachim Burghartz is the director of the Institut für Mikroelektronik Stuttgart (IMS Chips) and the former director of Dimes (now called EKL) at Delft University of Technology.
In 1997, I was fortunate to be a member of the executive committee of the IEEE Bipolar Circuits and Technology Meeting (BCTM) in Minneapolis-Saint Paul, when we organized a special event on the occasion of the 50th anniversary of the invention of the transistor at Bell Labs. We had invited semiconductor pioneers, including John Moll, Ray Warner, Jim Early, Tak Ning, Jan Slotboom and, most prominently, Jack Kilby.
In 2000, just three years after our BCTM event, Kilby received the Nobel Prize in Physics, jointly with Herbert Kroemer and Zhores Alferof, “for his part in the invention of the integrated circuit.” The other part of that invention is attributed to Bob Noyce, co-founder of Fairchild and Intel and nicknamed “mayor of Silicon Valley.” Both Kilby and Noyce developed, independently, the integrated circuit (IC).
While Kilby was first only by a couple of months, Noyce’s idea of the IC leveraged Hoerni’s planar process to define a truly monolithic circuit integration, which became the backbone of driving Moore’s Law for more than fifty years. Kilby’s IC demonstration looked more like what we now call a hybrid IC or a system-in-a-package (SiP), as it combined three bipolar transistors on a common piece of germanium by means of wire bonding. Many tried to downplay Kilby’s invention. Nevertheless, Jack Kilby received the Nobel Prize in Physics while Noyce never did.
Looking at today’s challenges in advancing semiconductor technology, however, I believe that Kilby also invented, or at least stimulated, something that was fifty years ahead of its time – heterogeneous integration. There, separately manufactured components are integrated into a higher-level assembly providing enhanced functionality. Such SiPs are not only feasible as rigid components but also as hybrid systems-in-foil (HySiF), in which large-area electronics and ultra-thin chiplets are assembled and interconnected onto a foil carrier, enabling high-performance flexible electronics.
In general, hybrid implementations of circuits and systems have advantages and disadvantages compared to monolithic ones. They allow for combining different material systems, they make it possible to build up systems with tested known-good dies, but they’re large in size.
They can be made smaller by realizing them as SiPs. If, for instance, you would open up a common sensor IC from Bosch, you’d find two dies inside the plastic package, a sensor chip and a processor chip, interconnected through wire bonding. The interconnect ability of such SiP chiplets, however, is limited by the large pad size and pitch, which hampers system design.
Those large pads are also cost adders because they occupy valuable CMOS chip real estate. One of the solutions to those shortcomings is the chipfilm patch concept, introduced by IMS Chips. It exploits wafer-level processing to embed ultra-thin chips into a foil patch carrier and adds interconnects to those chiplets, covering both the chip and peripheral areas. Therefore, pads can be omitted or shifted to the periphery, so that chip sizes are minimal and the interconnect ability of chips is sufficiently high.
In HySiFs, only the interconnect benefit of chipfilm is leveraged because the targeted flexible electronic components are intentionally large in size. If, however, the HySiF concept is transferred to a rigid assembly of densely packed chiplets, the full potential of the chipfilm concept can be exploited. Such micro-hybrid dies, thus, feature the form factor of a common CMOS chip, yet offer the benefits of a hybrid. This allows for applying packaging solutions readily available for common chips.
The emerging integrated photonic and quantum technologies call for just this kind of flexibility in the integration of chips made in different technologies, such as semiconductor lasers, chips with waveguide systems for optical signal processing, optical detector dies, MEMS, and more. Those technologies are intended for high-volume markets, so that they would greatly benefit from the small form factor and wafer-level processing.
Without any doubt, Kilby had never aimed at heterogeneous integration when he stitched together his three bipolar transistors on that germanium die. But, certainly, he has stimulated semiconductor engineers to rethink monolithic integration from time to time, with several great inventions not directly linked to Moore’s Law. For my part, I was definitely stimulated by Jack Kilby during my professional career, both by his invention and by having had the honour to meet him in person.