Semiconductors and their story (part 1)

7 min readMar 24, 2022

Semiconductors and its story

Joe Jia

Andy grove and the Intel founders beside a photomask for the Intel 4004 CPU. — Andy Grove and Intel founders sitting beside a 4004 photomask.

So… semiconductors… I’m not an expert, but I enjoy talking about these intricate things and their manufacturing. I want to put into perspective how complex the process of making a single transistor is and explain the story of how this amazing technology has evolved to what it is now.

I guess it was brutally clear how important these pieces of silicon were to global industries ranging from crazy powerful computers to your family humidifier. It is crazy how much these things disrupted the entire global supply chain when COVID hit and pulled everything down with it. Today we will only be going into just the tip of the iceberg for semiconductors, this article strives to focus on the current cutting-edge technologies which may be quite mind-boggling. This industry sure isn’t an easy one to deal with, semiconductor yields had always been lower than other industries, and producing these things competitively is much more of a tedious task than keeping up with a better car. This article will split semiconductors into 3 parts: The Silicon, The Machines, and The Process.

Jumping right into “The Silicon”, the basic definition of Silicon as described by the Google Dictionary: “The chemical element of atomic number 14, a nonmetal with semiconducting properties”. Silicon is split into many categories of purity, semiconductors utilize the highest purity grade of silicon which is called EGS or “Electronic Grade Silicon”. This silicon is usually grown with the Czochralski Method, short as CZ method. The CZ method grows silicon by utilizing a seed, which is a rod of silicon that has a crosssection of around 1cm squared. The seed is dipped into a “melt” which is pretty much just really really hot silicon that has been melted inside a crucible. The seed is then slowly pulled up and up until a cylindrical crystal of silicon is formed which is then cut into wafers. The seed and crucible are all rotated at specific variable speeds which allows them to adapt to what the material is reacting. This process takes quite a while and there have been many advancements to it since its original form when Jan Czochralski invented it in 1916. Notably, researchers have added silicon rods to the melt which allow it to support larger and larger crystals allowing for more production per melt. This process is, like many things in the semiconductor industry, a complex and secretive endeavor so I can’t explain it too well. The resulting cylindrical silicon is then cut into wafers (ranging from 200–450mm) using a diamond wire saw. These wafers are then post-processed into the uber flat semiconductor magic they are today.

Onward from the silicon, we move towards the actual manufacturing of the semiconductors used in our computers. The silicon is cut down into wafers which range in size, the largest a current gen machine can process is around 450mm. There are two types of silicon called N-type or P-type. These are combined to create transistors, which range from NPN to PNP (P and N being the types). NPN transistors are used for electronics that require fast use time as they turn on quicker. PNP transistors are favored when a device needs to turn off quickly or devices wish to have a fast fall time. It is important to note that the speed of a transistor inside a semiconductor is directly dictated by how fast electricity can flow in and flow out. The smaller the transistor the smaller the capacitance of it which allows electrons to flow in and out faster. This is why you hear things like 14nm or 5nm, NM means nanometer’s and the smaller the value the smaller size it represents. Current generation EUV machines can produce 5nm or smaller chips at volume. Back to the process, to create a NPN transistor we start with a N type wafer and form an oxide layer on top of it. Then we apply a layer of photo resist, which is usually put on via a spin coater insuring even coverage. We then throw this into a lithography machine which projects your chip onto it, the chips design are on a “mask” which is put into a machine to mask the places you dont want exposure. There are multiple types of masks inside the production process. It usually goes like this, an isolation mask creates the chip isolations then we have a polysilicon mask that produces the “gate definition”, then we have an oxide removal mask which creates the contacts on a chip. We then run it through a “metal” mask that creates the interconnections on a chip. After this we do a overcoat mask which creates bonding pads. This is the basic process of using the lithography machine to project masks onto a chip. Through each step of the process we “etch” away regions that did not get exposed. Once we expose the silicon and develop it we then etch away the silicon oxide in exposed regions. Now we can remove the photoresist layer. We then apply or “diffuse” a p-type layer into the silicon and re-oxidize the entire silicon again. The chip is exposed by another mask, which then we can etch away regions needed for an emitter. Reoxidize the entire chip. The chip is then put back into the machine, this time to etch away regions that we will need for metal contacts. Using sputtering, the chips exposed regions have evaporated aluminum on them, creating the contacts. We then attach connectors to these contacts, commonly used in the early days of semiconductors was copper wires which were directly attached. Nowadays the process is much more complex.

The process is not over though! You still need to package the chip, which is essentially the process of adding the actual interface the chip needs to communicate with larger electronics. Or in the case of a common CPU an “IHS” is soldered on. The IHS is the metal piece that conducts the heat away from the chip into the cooling solution.

So now that we have learned about the process of manufacturing these extremely complex semiconductors, what about the story? You see, these things are so incredibly vital to the functionality of the modern-day world yet they, and the incredible genius of technological pioneers who pursued these in their infancy, are commonly pushed to the side. The current idea of “semiconductors” as a computer chip has traces all the way back to the first AC/DC rectifier which was invented way back in the 1800s. However, in this article we will talk more from the invention of the first transistor in Bell Laboratories in the 1940s. The first transistor was created by William Shockley, John Bardeen, and Walter Brattain in 1947. At the time these 3 people were working at bell laboratories and built their transistor upon the concepts laid out earlier, the first transistor was built out of a simple P and N type piece of germanium, with a piece of plastic holding gold contacts onto the germanium. This invention quickly evolved to the Junction Transistor, which quickly found its way into the first transistor radios. The first transistor made out of silicon was only built in 1954 by Morris Tanenbaum as previously the material silicon was replaced with germanium since silicon was still quite a mind-boggling material for scientists at the time. The first use of transistors as integrated circuits only came to be once the MOSFET was invented again at Bell Labs by Mohamed Atalla and Dawon Kahng in 1959. The first Integrated Circuit quickly followed suit, in the same year Robert Noyce created the first IC at Fairchild Semiconductor, truly triggering the leap forward into the semiconductor age. The first CPU only came when a “small” company named Intel was contracted to design and create a chip for a calculator that can integrate all the functions needed inside a singular IC. This was called the 4004, which had a whopping 2300 transistors packed into a simple 60 USD chip. This marked a full step into the modern-day semiconductor industry. Notably, this is also where the genius of a man Andy Grove comes in. Andy Grove designed the 4004 and during his command as CEO of Intel, the company leaped from a valuation of 4 billion USD to nearly 197 billion USD, cementing itself as one of the most important companies in the world. This is also where our famous friend TSMC comes into the picture as a semiconductor manufacturer. At this point the market (and the story) splits into so many segments it’s just too complex to explain and condense into this article alone, so that marks the end of this one.

Thanks for reading! This was a complex one, but we are just getting started with semiconductors. Part 2 will follow this one and hopefully cover everything from ~1975 to ~2010 which should put us within a stone’s throw away from now, 2022.

Bibliography (currently just URLs):

Semiconductors and their story (part 1)

Written by Opinionated meat berry