On this page
A modern chip holds billions of parts, but in 1958 every resistor, capacitor and transistor was a separate piece wired to the others by hand. Jack Kilby asked whether all of them could be made from one block of material. This post covers his life, the summer he built the first integrated circuit, a worked example on why the idea looked risky, and what a student can learn from how he worked.
Who was Jack Kilby?
In his Nobel autobiography, Kilby describes growing up in Great Bend, Kansas, where his father ran a small electric company. During a huge ice storm in his teens, his father worked with amateur radio operators to reach customers whose power and phone lines were down. That sparked Kilby's interest in electronics.
He studied electrical engineering at the University of Illinois and graduated in 1947, a year before Bell Labs announced the transistor. He then joined Centralab, a Milwaukee electronics manufacturer, and took evening classes for a master's degree. In 1958 he moved to Texas Instruments in Dallas, which he says was the only company that agreed to let him work on miniaturisation more or less full time.
What problem was Kilby trying to solve?
His Nobel lecture begins with vacuum tubes. The ENIAC computer of 1946 contained more than 17,000 tubes, weighed 60,000 pounds and used 174 kilowatts. Transistors were smaller, but complex systems still had to be assembled from tens or hundreds of thousands of separate parts. Kilby writes that such systems would have been unreliable and unaffordable.
Several groups were working on miniaturisation. Some tried making all parts the same size so assembly could be automated. Others used thin films. Kilby noticed that every approach used different materials and processes for different parts, and he looked for a way to reduce both.
How did Kilby build the first integrated circuit?
Kilby joined TI in May 1958 and had no vacation due that year, so he worked through the company's "mass vacation", when most staff were away. He used the time to think. His reasoning was simple. TI could already make semiconductors cheaply, and resistors and capacitors could also be made from semiconductor. They would not perform as well as the usual parts, but if everything could be made from one material, it could be made in a single piece.
The lecture gives the sequence:
24 July 1958: In his lab notebook he described what became known as "The Monolithic Idea": resistors, capacitors and transistors, all made of the same material, in a single chip. Resistors would use the bulk resistance of silicon; capacitors would use p-n junctions.
28 August 1958: His supervisor, Willis Adcock, was enthusiastic but sceptical. Kilby first built the circuit from separate silicon parts to prove each element could be made from semiconductor, and demonstrated it.
12 September 1958: Using germanium wafers that already held transistor structures, technicians cut bars about 0.12 by 0.4 inches. The first three phase-shift oscillators were completed, and the first oscillated at about 1.3 megahertz.
19 September 1958: A flip-flop circuit, showing digital circuits could be built the same way, was completed.
TI announced the "Solid Circuit" in March 1959. Kilby's lecture credits Robert Noyce of Fairchild Semiconductor with showing the value of the planar process with metal leads over the oxide. The 2000 Nobel Prize in Physics gave Kilby half the prize "for his part in the invention of the integrated circuit".
Worked example: why did critics doubt the chip?
Kilby lists the objections. The first was that production yields would always be too low to make a profit: at the time, he writes, fewer than 10 per cent of all transistors made actually worked properly. Simple probability shows why this mattered.
Suppose each part on a chip works with probability p, independently of the others. A chip with n parts works only if every part works, so
P(chip works) = pⁿ
With p = 0.9 and n = 10: 0.9¹⁰ ≈ 0.349. About a third of chips work.
With p = 0.99 and n = 10: 0.99¹⁰ ≈ 0.904.
With p = 0.99 and n = 100: 0.99¹⁰⁰ ≈ 0.366.
So adding parts drags the yield down fast, and integration only pays when each part is very reliable. These numbers are an illustration, not TI's data.
Common mistakes:
Adding instead of multiplying. "Each part fails 1 time in 100, so 10 parts fail 10 times in 100" is only roughly right for small n and fails badly for large n.
Assuming independence without saying so. Real defects cluster on a wafer, so pⁿ is a model, and you should state the assumption.
Unit slips. Kilby's bars were about 0.12 × 25.4 ≈ 3.0 mm by 0.4 × 25.4 ≈ 10.2 mm.
Our post on common probability mistakes covers more traps of this kind.
What can a student take from Kilby's way of working?
Kilby did not invent new physics. He took what his company could already make and asked a different question of it. He wrote the idea down with a date, built a rough version to convince a sceptic, and then built the real thing within weeks. His lecture also notes that the transistor came first, from Bardeen, Brattain and Shockley; you can read about it in our post on John Bardeen and the transistor.
You can use the same steps on a hard problem:
List what you already have. Write the formulas and facts you know apply before reaching for anything new.
Sketch, then test the parts. Check each step separately, for example the units of each term, before combining them.
Date your work. Keep a problem notebook, as Kilby kept a lab notebook, so you can see how an idea developed.
For multi-step probability questions like the yield example, the Fermi Problem Library has worked problems with a step-by-step solver that explains the reasoning rather than printing an answer.
Try this today: work out how reliable each part must be for a 1,000-part chip to work half the time.
