2 December 2019

Integrated Circuits

To celebrate the 60th anniversary of the invention of the integrated circuit, Rob Cork looks back on the pioneering work of two US inventors, Jack Kilby and Robert Noyce, whose research paved the way for virtually all modern electronic devices.

To be able to appreciate the significance of Kilby and Noyce’s individual contributions to the development of the modern integrated circuit (IC), it may be helpful to begin with a short history lesson. In the early 20th century the invention of the vacuum tube had helped to launch the fledgling electronics industry. Due to their physical size, devices which relied on vacuum tubes were extremely bulky and consumed large amounts of power. For example, in 1945 the US Army developed the ENIAC computer for calculating artillery firing tables, which could calculate a trajectory in 30 seconds that would normally take a human 20 hours to complete. However, the ENIAC contained some 20,000 vacuum tubes, approximately 5,000,000 hand-soldered joints, weighed 60,000 pounds and consumed around 150 kW of electricity. It was rumoured at the time that lights across Philadelphia would dim whenever the computer was switched on. Clearly, if electronic devices were to achieve widespread adoption, the size and power requirements would have to be significantly reduced.

In this context, there was great interest in the late 1940s and early 1950s in alternatives to vacuum-tube technology that would allow electronic components and circuits to be miniaturised. One of the most promising solutions was the transistor, which had been developed by Bell Labs in 1948. However, the earliest transistor designs featured electrodes on different surfaces (i.e. different layers). As a result, it was not straightforward to form interconnections between multiple transistors on a single silicon wafer. Instead, transistors were packaged as separate components which would then have to be attached to a circuit board via soldering, making the fabrication of complete circuits both time-consuming and prone to defects.

In the early 1950s, an Englishman called Geoff Dummer was the first person to propose a theoretical design for a monolithic integrated circuit, in which multiple circuit elements would be fabricated closely together on a single block of silicon and interconnected to form a functioning circuit. By reducing the length of the connections between components, Dummer’s novel ‘integrated’ circuit would be capable of operating faster and at lower power than could be achieved with existing technologies, and clearly had the potential to revolutionise the fledgling electronics industry. However, since Dummer did not describe how to actually fabricate such a circuit, let alone on commercial scales, the integrated circuit remained a merely theoretical concept – a solution that promised to deliver so much, yet remained tantalisingly out of reach. Until, that is, Kilby and Noyce arrived on the scene.

On 6 February 1959, Jack Kilby and his employer, Texas Instruments, filed a US patent application titled “Miniaturized Electronic Circuits” (serial no. 791,602). The application described a circuit in which components such as transistors, resistors and capacitors could be fabricated on the same semiconductor wafer and connected via thin loops of wire, the ends of which were thermally bonded to contacts on the surface of the wafer. This provided a technically feasible solution for fabricating integrated circuits. Then, just a few short months later, on 30 June 1959, Robert Noyce and Fairchild Semiconductor Corporation filed an application for a “Semiconductor device-and-lead structure” (serial no. 830,507). Noyce’s application described an alternative approach to manufacturing integrated circuits, in which the electrical connections between different components took the form of metal strips deposited directly onto the surface of the wafer.

Both Kilby’s and Noyce’s circuits shared the common concept of forming a number of components in a single semiconductor wafer by selectively doping parts of the wafer. It was already known that the conductivity of a semiconductor could be increased by the addition of a suitable dopant. If a dopant is used which adds electrons to the semiconductor, the resulting region is referred to as ‘n-type’. On the other hand, the addition of a dopant which removes electrons from the semiconductor, leaving positively-charged ‘holes’, results in a ‘p-type’ region. Crucially, both Kilby’s and Noyce’s solutions were compatible with known semiconductor doping processes, and so could allow miniaturised components having both n-type and p-type regions to be formed on one semiconductor wafer, as a single device.

As an interesting aside, following the grant of Noyce’s application on 25 April 1961 (US 2,981,877), Texas Instruments appealed the decision to grant Noyce’s patent on the grounds that Kilby had made the invention before Noyce. At the time, US patent law operated according to the ‘first-to-invent’ principle, and Texas Instruments produced evidence in the form of Kilby’s lab notebook. This, it was argued, showed that Kilby had made his invention on 24 July 1958, being the date on which he had drawn the first sketch of what later became known as “The Monolithic Idea”. Noyce’s own records, on the other hand, showed that he had arrived at the invention later, in January 1959. The USPTO then granted Kilby’s application on 23 June 1964 (US 3,138,743), which was in turn appealed by Fairchild. Texas Instruments and Fairchild eventually settled their differences via a cross-licensing agreement.

In Kilby’s design for an integrated circuit, components were connected by means of so-called ‘flying wires’ formed from gold. In the below drawing taken from US 3,138,743, these are the wires labelled ‘70’ which span the gaps between components on the surface of the wafer. As well as connecting components on top of the wafer to each other, the gold wires could also be used for connections to other parts of a device, such as the input and output leads beneath the wafer.

Noyce’s design on the other hand, provided connections between various components in the form of surface metallisation. Noyce proposed using metal strips on top of an insulating oxide layer to make the electrical connections between different regions of the device, with the insulating oxide ensuring that the metal layer would not short out the semiconductor junctions. In the below drawing of a transistor taken from US 2,981,877, the metal strips labelled ‘9’ and ‘7’ provide the electrical connections to the p- and n-doped regions, respectively, whilst the underlying oxide ‘tongues’ (5′ and 5′′) prevent the metal strips from shorting out other regions that lie beneath each strip.

For example, the tongue (5′′) on the right-hand side that stretches across the ring of p-type material (4) allows a layer of surface metallisation (7) to be deposited across the insulator to reach the contact (6) to the central n-type region (3), without the metallisation coming into contact with the surrounding ring of p-type material. Noyce formed the insulating layer by oxidising the surface of the semiconductor, followed by etching away selected portions of the oxide to reveal the underlying semiconductor.

As with many of the great inventions, with hindsight Noyce’s idea might seem deceptively obvious, particularly as it is now such a routine feature of semiconductor devices. However, at the time it represented a technological breakthrough since it enabled electrical connections to be formed solely by patterning a layer of surface metallisation, and in doing so elegantly overcame the main obstacle to mass-producing integrated circuits.

Initially, it was not clear which out of Kilby’s and Noyce’s alternative approaches would eventually gain widespread adoption, if at all. As Kilby recalled in his Nobel acceptance speech many years later (8 December 2000), after filing their respective patent applications in 1959: “Noyce, a few others and myself provided the technical entertainment at professional meetings for the next five years as we described and debated the merit of the various miniaturization systems”. Ultimately, it was Noyce’s design that proved more commercially successful, not least because it offered a much simpler fabrication process than Kilby’s ‘flying-wire’ design, and all modern integrated circuits can trace their roots back to Noyce’s original design. Noyce went on to found Intel jointly with Gordon Moore, who is perhaps best known for devising the eponymous ‘Moore’s Law’.

Although Noyce’s design won out commercially, the decision of the Nobel committee to recognise the invention of the integrated circuit in the Nobel Prize in Physics 2000 sadly came too late for Noyce. The prize was awarded to Kilby, but could not be shared posthumously with Noyce since he had passed away a decade earlier, and Nobel prizes can only be awarded while the recipient is still alive. However, in his Nobel acceptance speech Kilby paid tribute to his one-time rival, remarking that “[w]hile Robert and I followed our own paths, we worked hard together to achieve commercial acceptance for integrated circuits. If he were still living, I have no doubt we would have shared this prize”.