g explains the physical changes that turn silicon into a CPU.
First Steps
The fabrication process most commonly used today is CMOS (complementary metal oxide semiconductor). In CMOS, chip vendors design the processor logic so that it uses pairs of transistors. The logic uses only one transistor at a t
ime to conserve power. These transistor pairs are made of complementary materials, as shown in the figure
"The Goal: CMOS Transistor Pair."
Today's processors use silicon because it is both plentiful and relatively cheap: Beach sand is made of pulverized silicon dioxide. Another advantage to silicon is that its oxide makes an excellent insulator, which simplifies the fabrication process. As the name semiconductor implies, silicon is neither a good conductor nor a good insulator. But by adding trace elements to silicon, you can change its electrical properties so that it becomes conductive. More on that in a moment.
The silicon used in the fabrication process comes from the wafer itself. The wafers are made by sawing slices from a silicon ingot grown under controlled conditions. The precision of this growth is such that each ingot, which is 6 to 8 inches in diameter and up to several feet long, is literally a giant crystal. One of the initial steps is to add a fine layer of silicon
dioxide, called the field oxide, to the wafer surface. Next, workers add a photoresist to the surface. Later, an etching step removes sections of the field oxide. The desired trace elements are added to the exposed areas of silicon using the ion implantation, as shown in the figure
"Starting Point: Photoresist for Arsenic Implantation."
Crystals with a Charge
The trace elements, called dopants, establish the silicon's conductivity by placing charge carriers in the material's crystalline lattice. The addition of arsenic to the silicon creates a crystal with an electron surplus. The extra electrons can migrate about and carry a current. Thus, the material is known as negative, or n-type material. Adding boron to the mix creates a crystal with an electron shortage, making it a positive, or p-type material. The latter material conducts because electrons can migrate between the positive vacancies (known as holes) in the lattice. These two types of material are necessa
ry because they're used to build junctions that will selectively conduct when the proper voltage or current is applied. A transistor consists of several such junctions made of these dissimilar materials.
Bipoloar transistors are made by sandwiching a p-type layer between two n-type layers (making an n-p-n transistor), or sandwiching an n-type layer between two p-layers (making a p-n-p transistor). The sandwich layer, known as a gate, controls the flow of current through the device. The other layers are the source, where the current enters the device, and the drain, where the current exits. Note that we're using the term sandwich loosely here: As you can see in the figures, the various layers are fabricated adjacent to one another. Normally the charge carriers are attracted to one another at the boundaries between the n- and-p-type materials, and to the connections at the source and drain. This means that no current flows through the transistor. If you apply a current of the proper polarity to the gate, th
e standoff between the charge carriers disappears and current flows through the transistor. A transistor thus acts as a valve or switch in a digital circuit.
Most processor circuits are made of field-effect transistors (FETs). In this design, the gate material has an insulated electrode attached to it. Applying the proper voltage (not current) to this electrode creates an electrical field that arranges the charge carriers in the gate material so that a temporary conductive channel appears between the source and drain electrodes, and current flows through the transistor.
CMOS Pairs
CMOS designs use complementary pairs of FET transistors. The p-channel FET is made in a large well of n-type material implanted into the wafer. The n-channel FET uses the doped wafer substrate itself to make the gate channel. Many successive steps of exposures, etchings, and implantations are required before the CMOS pair of transistor are fully assembled.
Now that the transistors are made, they mus
t be connected together into useful circuits. More processing steps add tungsten plugs that establish electrical connections to the transistor's sources, gates, and drains. Then fab workers deposit a layer of aluminum onto the wafer, as shown in the figure
"Next Step: Aluminum Becomes the First Metal Layer."
Another etching step strips away most of the aluminum, except for specific connections between transistors.
Typically, an insulating diaelectric is placed over the first aluminum layer, and then more exposures, metal deposits, and etchings build successive aluminum interconnection layers, as shown in
"Final Assembly."
The various layers of aluminum wire all the transistors together into adders, multipliers, bus interfaces, instruction decoder, and other logic units that make up the processor.
illustration_lin
k (18 Kbytes)
illustration_link (17 Kbytes)
illustration_link (14 Kbytes)
illustration_link (18 Kbytes)
Building Transistors, Layer by Layer
