The making of an integrated circuit (IC), widely known as a "chip," is perhaps the most amazing manufacturing process the world has ever conceived. A sliver of silicon smaller than a postage stamp can contain billions of transistors, which, acting like on/off switches, are the fundamental active components on the chip. It all starts with the design of the circuits, which carry electrical pulses from one point to another. To learn about the different types of chips, see chip
Transistors to Gates to Circuits
FROM LOGIC TO PLUMBING
Pulses flow through transistors that open or close when activated. The current flowing through one affects the opening or closing of another and so on. Transistors are wired together in Boolean logic gates (see Boolean logic
). Gates make up circuits, and circuits make up CPUs and other components. See Boolean Logic
Circuits were originally designed by humans. Today, logic functions reside in electronic libraries, and designers pick and choose from a menu. New functions have to be designed by humans, gate by gate.
Computers make computers. The computer converts the logical circuits into a plumber's nightmare of transistors, diodes and resistors. These all wind up as lithographic plates from which the chips are made. One CPU chip links billions of transistors together.
The photomask is the actual size of the chip, replicated many times to fit on a round silicon wafer up to 12" in diameter. The transistors are built by creating subterranean layers in the silicon, and a different photomask is created to isolate each layer to be worked on.
Inspecting the Plumbing
People are always more flexible than computers and can find flaws that might go undetected by software analysis. (Image courtesy of Elxsi Corporation.)
CHIPS ARE JUST ROCKS
The base material of a chip is usually silicon, although materials such as sapphire and gallium arsenide are also used. Silicon is found in quartz rocks and is purified in a molten state. It is then chemically combined (doped) with other materials to alter its electrical properties. The result is a silicon crystal ingot up to 12 inches in diameter that is either positively (p-type) or negatively charged (n-type). Slices of the ingot approximately 1/30th of an inch thick are cut from this "crystal salami." The slices are called "wafers."
Drawing the Ingot
The silicon ingot is being drawn from a scalding furnace containing molten silicon. High-speed saws will slice it into wafers about as thick as a dime, which will then be ground thinner and polished like a mirror. (Image courtesy of Texas Instruments, Inc.)
BUILDING THE LAYERS
Circuit building starts out by adhering a layer of silicon dioxide insulation on the wafer's surface. The insulation is coated with film and exposed to light through the first photomask, hardening the film and insulation below it. The unhardened areas are etched away exposing the silicon base below. By shooting a gas under heat and pressure into the exposed silicon (diffusion), a sublayer with different electrical properties is created beneath the surface.
Through multiple stages of masking, etching, and diffusion, the sublayers on the chip are created. The final stage lays the top metal layer (usually aluminum), which interconnects the transistors to each other and to the outside world.
The lady is wearing a "bunny suit," but is not wearing a mask, because the wafers have already been manufactured. (Image courtesy of Hewlett-Packard Company.)
Each chip is tested on the wafer, and bad chips are marked for elimination. The chips are sliced out of the wafer, and the good ones are placed into packages (DIPs, PQFPs, etc.). The chip is connected to the package with tiny wires, then sealed and tested as a complete unit.
Chip making is extremely precise. Operations are performed in a "clean room," since air particles can mix with the microscopic mixtures and easily render a chip worthless. Depending on the design complexity, more chips can fail than succeed.
Packaging the Chip
This machine bonds the chips to the metal structure that will be connected to the pins of the chip housing and carry the signals to and from the circuit board. (Image courtesy of Texas Instruments, Inc.)
In the early 1980s, the 8088 CPU chip in the first PCs had 25 thousand transistors. Twenty years later, Intel's Itanium 2 contained 220 million.
There is a never-ending thirst to build more and more transistors onto a single chip. In order to etch the photomasks finer and create elements as tiny as 32 nanometers and smaller, ultraviolet light has replaced visible light. See feature size
From 2D to 3D
Just as the chip eliminated cutting apart the transistors only to be reconnected in circuit patterns, increasingly, more circuits are built into the same chip, creating complete systems (see SoC
). As we make the chip wider, we are also trying to make it deeper. Not only are we making the elements smaller and the chip larger, we are building chips in layers (3D chips).
Today's chips are yesterday's science fiction; however, it never ends. In the early 2000s, IBM's Almaden Labs made an experimental circuit of carbon monoxide molecules. Taking up space 260,000 times smaller than the equivalent silicon, the circuit performed a calculation by making the molecules collide with each other. Stay tuned!
Dressing for Work
The fabrication of the tiny transistor is an extremely precise one. The slightest contaminants in the air can render the transistor and chip useless. Putting on the "bunny suit" is an elaborate procedure. (Image courtesy of Intel Corporation.)
No Germs in these Rooms
You won't catch the flu working in a chip fabrication plant, at least not in the clean room. Bunny suits and clean rooms are required to produce high yields of defect-free chips. (Photos from top to bottom courtesy of Texas Instruments, Inc., and Motorola, Inc.)