Jack Kilby, Bob Noyce and the 3D Integrated Circuit

The first (2D) Integrated Circuits were not monolithic. In fact, they had hand-soldered interconnects between circuit components. Let's fast-forward 50 years to the time of the 3D Integrated Circuit... the first 3D-ICs are not monolithic either. Will history repeat itself? Will the monolithic idea win? 

In the 1950s, circuits were built by connecting together discrete components such as transistors, resistors and capacitors. The industry, however, realized it was running into an interconnect problem...

• As the number of components for a circuit grew, the number of interconnects increased exponentially. Not only was soldering thousands of leads and connections difficult and costly, but also it reduced reliability and caused errors in computation.
• Connecting together discrete components to form a circuit increased size of the built circuit.
  

The solution, as we know now, was the integrated circuit. Jack Kilby of Texas Instruments came up with the idea that you could build transistors, capacitors and resistors on the same semiconductor substrate. You can see a picture of his first prototype above: it was a 1.3MHz oscillator with 1 transistor, 1 capacitor and a few resistors, all built on the same piece of germanium. Kilby used hand-soldered wires to connect together these components [Click for more details].

Bob Noyce of Fairchild Semiconductor, on the other hand, integrated connections between components. His first prototype, a flip-flop, is shown above. With the monolithic integrated circuit concept developed by Noyce, one could get very high connection density in circuits. Interconnects between components could be made much smaller! [Read Computer History Museum's article on the invention of the Integrated Circuit]

Interestingly, 50 years from the time of Noyce and Kilby, we are seeing a similar scenario. The first 3D Integrated Circuits are coming out, and they're not monolithic. Different components of the 3D-IC are built separately, then stacked using through-silicon via (TSV) technology or wire-bonding. Please see the figure above for more details. Just like the first prototypes from Kilby, we are finding it hard to get high connection density, since connections between vertically stacked components are not integrated. To give some numbers for TSV density, the Intl. Technology Roadmap for Semiconductors predicts that minimum TSV diameter will not change much between 2009 and 2015, and will be around 1um [Click for more details]. This compares with sub-50nm dimensions for on-chip interconnects, making it hard to move many of these on-chip interconnects into the third dimension.

Going monolithic could be an option, and my company is exploring it, as are others such as Stanford University and BeSang. With monolithic 3D and integrated vertical connections, one can get very high connection density between components of the 3D-IC. The monolithic idea took the (2D) IC community by storm... will history repeat itself for the 3D-IC?

- Post by Deepak Sekar