How Microprocessors are Made
Creating the Material
First silicon chips must start out as a base material. This base material is pure silicon in the form of a specially "grown" large single crystal.
The Czochralski Process
In 1916 Polish scientist Jan Czochralski was looking for a way to measure crystallization rates of materials. This means that when a crystalline metal or salt was created, it would crystallize at a particular rate. This rate had to be timed somehow to come up with a quantifiable and reproducible number.
One day, while writing notes, he dipped his ink pen into a crucible of molten tin without looking; he thought he was dipping his pen into an ink-well. He did notice his mistake and drew his pen out slowly. Much to his surprise the pen drew out a thin strand of tin which immediately froze (crystallized) on hitting the air. On testing the strand he verified that he had inadvertently created a single crystal structure. He had stumbled on a way, by measuring the speed of the "draw," to determine a crystallization rate of any material.
Shortly after his discovery he began using glass capillary tubes to form crystals from various materials such as tin, lead and zinc.
Though he never imagined that his method would be used to manufacture electronics devices, his method of forming crystals is the most efficient way of forming the pure silicon crystal structures needed for micro-chips.
When Bell Laboratories began creating semiconductor material, Gordon K. Teal and J.B. Little determined that the Czochralski process was exactly the method needed to create pure single crystal silicon wafers.
From Crystal to Wafer
Wafers from the Boule
Once the wafer is drawn it is allowed to anneal (cool) and then it is inspected. The boule is sawn into wafers with a wire saw. At this point the wafers are roughly 2602 millimeters in diameter.
During processing that figure is reduced to about 250mm. As you may notice from the pictures at right wafers sawn from the boule are not going to be exactly the same diameter. So the wafer material is grown to slightly larger than the 250mm standard and then trimmed during processing.
Trimming and Polishing
The wafers are trimmed along the rim to remove any sharp edges.
Because creating the circuit on the wafer is an exacting photographic process the wafers have to be as flat as it is possible to make them. To get that degree of flatness the wafers are put into a polishing machine. They come out looking like mirrors, but what is good to the human eye is not as flat as they need to be so the wafers are subjected to another, chemical polishing step.
The wafers are thoroughly cleaned to remove any polishing grit or chemical residue.
Photoresist & Photographic Circuit Transfer
The wafer is then coated with what is called photo-resist which is a chemical film which reacts to light. When photo-resist is exposed to light it can be washed away with other chemicals. Photo-resist that has not been so exposed will resist the chemical wash.
The depth of the photoresist layer is critical. Too thick and it won't all wash away. Too thin and it may wash away too soon. As you may imagine this layer has to be exact across the entire surface of the wafer. To do this the resist is applied while the wafer is spinning. An automated arm with a spigot fixed to its end is passed over the wafer as it spins. The arm moves from the center of the wafer out at a precise rate. This method insures an even coating of photoresist all the way across the wafer.
Ready for Transfer
Now that the wafer has it's resistive layer on it the next step involves a photographic technique much like that used a few decade ago to create printed photographic images on paper. In a way a negative of the circuit to be created is projected on to the wafer to create a pattern in the photoresist. This pattern is then washed clean in a cheimcal bath. This is then followed by another cleaning step to remove any of the chemicals used to wash away the exposed photoresist.
The process of transferring the circuit pattern onto a wafer is called photolithography though it is really not lithography1 at all. The process takes the wafer, which has been coated with a photoresist, and shoots, photographically, an image of the first layer of the circuit to be created onto the wafer.
Typically the image projected onto the wafer is one one-hundredth the size of the image on the "photographic" plate if not smaller. Ultraviolet light is used because of it's high resolution compared to other wavelengths. By using a computer controlled plate (on which the wafer is mounted) a hundreds of images are "shot" onto the wafer.
Depending on the circuit being created this process can shoot the image of hundreds or up to thousands of circuits on a single wafer. This is where the effects of mass production (economies of scale) begin to take effect.
Each circuit formed this way will eventually be cut out of the wafer. The cut out circuit is called a "die."
Etching or Deposition
Once the photoresist has been washed away and the wafer thoroughly cleaned the wafer is now ready for either etching or deposition.
Etcahing will be performed if channels need to be formed on the dies on the wafer. Etching is typically performed on parts of the circuit that need to have a thin film of metal formed on in these channels. These thin films, following the pattern of the etched channel, will form wires or conductive pathways.
It may be necessary to expose the dies to elemental gas to create semi-conducting areas on each chip. Semi-conducting areas are created by impregnating the silicon with aluminum (most recent development), gallium, gallium-arsenide, boron, selenium, or another element that has conducting properties. Of course, because these elemental gases will only impregnate areas that no longer have photoresist on the, patterns will be formed in the silicon that are semiconducting.
Formed in Layers
Formed Like a Layer Cake
The steps described above may be performed dozens to a score of times to create a complete integrated circuit. In a way they are built up like a layer-cake from the bottom up.
Naturally this takes a lot of planning because the features on each layer must communicate with each other and ultimately all layers must communicate with the outside world.
Scales Limitations and Breakthroughs
When integrated circuits were first manufactured with this process the typical size of the wafer was twenty-five millimeters or roughly one inch in diameter. Over time and because individual chips are cheaper if more can be made at the same time, that diameter has steadily creeped upward to thirty-eight (38mm), fifty-one (51mm), seventy-five (75mm), one hundred (100mm), up through three hundred (300mm) and now four hundred fifty (450mm) millimeters in diameter. The three hundred millimeter wafer is slightly smaller than twelve inches in diameter and the four hundred-fifty millimeter wafer slightly larger than seventeen (17") inches in diameter.
As the wafers have increased in size the actual features on the integrated circuit have shrunk. When Intel created the first integrated circuit microprocessor (the Intel 4004) the typical size of each part of the transistor (wire, junction, or connector) in that circuit were in the ten (10µm3) micrometer range; that was in 1971. By 1975 the scale was in the three (3µm) micrometer range, by 1982 the one (1µm) micrometer scale. By 1990 circuit features were measured in the nine hundred (900nm) nanometer range and continue to be in the billionth of a meter ranges today.
This year, 2011, circuit features will break the thirty-two nanometer barrier and continue to get smaller. It is projected that by 2012 circuit features will hit the sixteen (16nm4) nanometer mark and by 2015 the eleven (11nm) nanometer mark.
As circuit features have shrunk problems have arisen. With less space between features electrons tend to "leak" from one device to an adjacent device on the same circuit. This causes transistors that are supposed to be "off" to suddenly be "on."
This was caused by a "leak" of electrons from one transistor to a nearby transistor. This has caused manufacturers to use different materials (aluminum instead of gallium-arsenide) or to add additional circuit features to each transistor in order to electrically isolate them.
Because circuit features continue to shrink the photographic process of transferring the circuit image to the wafer has also met with problems. Though the ultraviolet wavelength of light is very fine and provides a high resolution at nanometer scales these images still tend to get fuzzy as features get smaller. There was thought to be no way to improve that resolution if the image is to be projected from the mask to the chip.
To get around this problem photolithography has been done in a complete vacuum, but even this step had its limitations. Now "immersion" lithographically is used because purified water tends to keep the light beam tight all the way down to the surface of the wafer.
Immersion photolithography has it's own problems; the water must be very pure and have no trapped gasses (such as nitrogen or oxygen) in it. This, in turn, has added another layer of complexity to the process.
As the circuit sizes have shrunk materials scientists working with physicists have explored using different elemental gases, using finer photographic techniques, and even to enhance the materials making up the lens optics.
Circuit Creation Steps
The video immediately above demonstrates seventy-two distinct steps of photoresist, deposition, etching, etc. to create a transistor.
The processes described here are very general by necessity. Many of the manufacturing techniques used today are the trade secrets of the manufacturers involved.
There have been many benefits to electronics manufactured with ever smaller circuit sizes.
Less electrical energy is required to run smaller transistors. Because less energy is required devices that used to be plugged into a wall socket can now run on battery power alone.They can also be powered for longer with less energy at the source. Additionally, they tend to run cooler with the reduction in power requirements.
Another benefit of smaller circuitry is the reduction in the cost to make each individual chip. As chip sizes have shrunk and the wafer size enlarged, the cost to manufacture them has also fallen. This makes each integrated circuit cost mere pennies rather than dollars to produce.
Naturally smaller integrated circuits mean that the device it goes into can also be smaller.
Finally, smaller circuit design means that ever more powerful computers can be manufactured that are much smaller than the version that came immediately before.
The net results is that modern chips are smaller, cheaper, more powerful, and use less electricity than a version produced just two years before. And this will be true two years later.
Intel co-founder Gordon E. Moore predicted in 1965 that roughly every two years the number of circuits packed into a given space would double.
"The future of integrated electronics is the future of electronics itself. The advantages of integration will bring about a proliferation of electronics, pushing this science into many new areas.
"Integrated circuits will lead to such wonders as home computers or at least terminals connected to a central computer, automatic controls for automobiles, and personal portable communications equipment. The electronic wristwatch needs only a display to be feasible today." - Gorden E. Moore (Electronics, Volume 38, Number 8, April 19, 1965)
1 Lithography is really a printing process using a stone (or metal plate) to transfer ink to paper. The use of this term in electronics is necessarily inaccurate, but as with many new technologies the terms had to be borrowed from somewhere.
2 A millimeter is 0.0393 inches. milli in millmeter means thousandth
3 A micrometer is 0.000393 inches. micro in micrometer means millionth
4 A nanometer is 0.000000393 inches. nano in nanometer means billionth
The author was not compensated in any way, either monetarily, with discounts, or freebies by any of the companies mentioned.
Though the author does make a small profit for the word count of this article none of that comes directly from the manufacturers mentioned. The author also stands to make a small profit from advertising attached to this article.
The author has no control over either the advertising or the contents of those ads.
More by this Author
A discussion of fibers and resins used in Fiber Reinforced Plastic. Includes human-made and natural fibers as well as various types of resins. Also discussed are the advantages and disadvantages of each material; fiber...
How to prepare and paint fiberglass-reinforced plastic. Includes discussions on immersion service paints, outdoor paints, and the effects of exposure to the elements.
How to use fiber and resin to repair and create objects.