Preface. This is an introduction to how microchips are made to give you an idea of how difficult and amazing they are. This is a very high-level overview gathered from two textbooks full of mind-boggling processes and complexity. Microchips are also constructed out of finite critical, precious, platinum group elements, and rare earth elements — 90% of them produced in China. And all of them mined with declining fossil fuels.
They are the pinnacle of human achievement, the most complex objects on earth, requiring fabrication plants costing billions of dollars, need the most pure chemicals and air of any product, the most vulnerable to electric outages, and supply chains, and will therefore be the first industry to fail when energy shortages become common.
Most of this post came from Quirk and Serda (2001) and Van Zant (2004).
- 2021 How we manufacture silicon
- Posts in 3) Fast Crash: Critical elements, Platinum group elements, precious elements, and Rare Earth elements
Alice Friedemann www.energyskeptic.com author of “Life After Fossil Fuels: A Reality Check on Alternative Energy”, April 2021, Springer, “When Trucks Stop Running: Energy and the Future of Transportation”, 2015, Springer, Barriers to Making Algal Biofuels, and “Crunch! Whole Grain Artisan Chips and Crackers”. Podcasts: Collapse Chronicles, Derrick Jensen, Practical Prepping, KunstlerCast 253, KunstlerCast278, Peak Prosperity , XX2 report
Microchips are incredibly important to civilization — like energy, there isn’t a single business endeavor, infrastructure, or electronic device that isn’t dependent on them.
New cars can have more than 100 semiconductors in their touch screens, computerized engine controls and transmissions, built-in cellular and Wi-Fi connections, collision avoidance systems, cameras and other sensors (Ewing and Clark 2021).
Nearly all knowledge is being stored in electronically in media that won’t be readable after microprocessors stop being made — and lost to the future generations forever, since it’s unlikely we can get back to this level of technology ever again after a global collapse. Microchips in computers have a very short live-span — Consumer Reports recommends getting a new computer rather than fixing it if yours fails after 4 years, and that you may even want to do so if it fails within 2-4 years.
Moore’s Law is a great tragedy, taking us further and further away from a sustainable microchip in the future.
Books and microfiche have a lifespan of 500 years if they’re stored at optimal dryness and temperature, which is why I encourage material scientists, librarians, and perhaps you, to consider what knowledge we might preserve, and how we could do so in “Peak Oil and the Preservation of Knowledge“.
Microprocessors are essential, they’re in just about everything
Billions of chips are created every year for a myriad of applications: in autos, airplanes, ATMs, air conditioners, calculators, cameras, cell phones, clocks, DVDs, machine tools, medical equipment, microwave ovens, office and industrial equipment, routers, security systems, thermostats, TVs, VCRs, washing machines – nearly all electrical devices.
Creating a chip begins by cutting a thin 12 inch slice, called a wafer, from a 99.9999999% pure silicon crystal, one of the purest materials on earth. Wafers require such a high degree of perfection that even a missing atom can cause unwanted current leakage and other problems in manufacturing later on. This is the platform that about 5000 computer chips will be built on. Each chip will contain millions of transistors, capacitors, diodes, and resistors built by punching and filling in holes in more layers than a Queen’s wedding cake.
Particles 500 times smaller than a human hair can cause defects in microchips. The more particles that get on a wafer, the greater the chance there is of a killer defect. Some particles are worse than others — a single grain of salt could ruin all the chips on a wafer. Sodium can travel through layers even faster than stray bits of metal. Particles that outright kill a chip are caught during the testing phase at the factory. Sometimes only 20% make to the end. The traveling particles are insidious, and can cause a chip to malfunction, perform poorly, or die later on (hopefully before your warranty expires). Consumer reports recommends not even trying to repair a personal computer after four years, and in the two to four year range it’s a tossup whether to repair or buy a new one.
Typical city air has 5 million particles per cubic foot. There are processes that require a maximum of 1 particle per square cubic foot.
People are among the worst offenders, as far as particle generation goes. If you walk at a good clip, you emit 7.5 million particles per minute. Even sitting still, you are still emitting particles. A smoker is a particle-emitting dragon long after the cigarette, and a sneezing worker is even worse, a veritable Krakatoa.
City water is not pure enough to be used — it’s full of bacteria, minerals, particulates, and other junk. To make city water clean enough requires many filters, UV-light, and other water treatments. Some fabrication plants use millions of gallons of water a day, requiring a huge investment in water processing and delivery systems.
Microchip fabrication is primarily a chemical process, requiring ultra-clean 99.9999% chemicals and 99.9999999% gases. About one in five steps use water or chemicals to clean the wafers or prepare their surface for the next layer.
Firemen practically need a chemical engineering degree to inspect and fight fires in a chip fabrication plant. During a fire, they risk being exposed to volatile, flammable, or combustible solvents, and chemicals like arsine, used in chemical warfare.
The chips also require humidity to be just right. If the humidity is too high, the wafers accumulate moisture, and the layers won’t stick. Too dry and static electricity will suck particles out of the air and practically glue them to the surface, they’re so hard to remove.
So it shouldn’t surprise you that it costs over 3 billion dollars to build a clean room. The inside is composed of non-shedding materials, especially stainless steel. Floors have sticky mats to pull dirt off of operators’ shoes. Pens, notebooks, tools, and mops – everything is built of material that sheds as few particles as possible, but even so, equipment particles cause a third of the contamination.
How chips are made
Wafers move from workstation to workstation and have different operations performed on them at each one. Wafer fabrication for a chip might involve 450 processes with operations that overall take several thousand individual steps. The machines that make this all happen include high-temperature diffusion furnaces, wet cleaning stations, dry plasma etchers, ion implanters, rapid thermal processors, vacuum pumps, fast flow controllers, residual gas analyzers, plasma glow dischargers, vertical furnaces, optical pyrometers, etc.
If you were shrunk to chip size and tied to a wafer, you’d go through the car wash from hell. You’ll be moved along by robotic wafer handlers from one machine to the next, where you’d be layered with different materials, centrifuged, electro-polished, dyed, scraped, heated to 1,800 F, ultrasonically agitated, sputtered, doped, hard baked, dipped in toxic chemical baths, irradiated, blasted with ultrasonic energy, spray-cleaned, dry-cleaned, scrubbed, micro-waved, x-rayed, shot with metal, etched, and probed.
At various points, the “Survivor” show comes on. Chips are examined at an atomic level for defects, and their electrical functioning tested. They’re usually thrown out if anything is wrong, since most mistakes can’t be fixed.
There are many problems that can cause a chip to fail besides contamination. The wafer must be perfectly flat in structure and while it goes through the workstations. If the wafer were 10,000 feet high, you’d see bumps or holes no higher than 2 inches – more than that and the layering is thrown off. If the wrong step was performed after 3,841 correctly performed steps, the chip was under or overheated, the layer didn’t fully stick, was improperly aligned before the next layer was added, or a chemical misapplied, the chip is thrown out. It’s amazing any chips make it out the door.
After your makeover, you’d emerge in a designer outfit composed of up to 25 layers embedded with millions of transistors, diodes, and resistors. You’ll find yourself “best in show” at tattoo competitions and irresistible to Terminator fans.
Dependencies of the internet [select to enlarge].
Each box denotes an abstract or concrete component, resource, or function of the Internet or one of its dependencies. Arrows denote dependency; dashed arrows denote optional dependency (a “one of these” relationship). At the top is the use case. We quickly move through the higher layers which represent mostly functional and abstract pieces that we recognize as pieces of the Internet and on to the lower layers which reside in the realm of hardware. Once we move beyond hardware manufacturing we enter the realm of chemical compounds and natural resources that are required for many of the relevant manufacturing processes, from making fiber optics to printing circuit boards. We end with ores or otherwise naturally occurring resources. We decided not to depict operational or deployment dependencies since they tend to involve a small number of processes like power generation and transportation. Source: “What are the Internet’s dependencies?” by barath. 2011. contrapositin.org
I encourage you to read about how pencils are made and Thwaite’s attempt to build a simple toaster from scratch as well, since just about everything is complicated, and we many not even be able to make toasters after a collapse.
You may also want to read about how dependent the internet and microchip manufacture are on electricity that is mostly generated from coal: “The Cloud Begins With Coal. Big data, big networks, big infrastructure, and big power. An overview of the electricity used by the global digital ecosystem”. Aug 2013. Mark P. Mills. Digital Power Group.
There may be multiple “collapses” before stability returns to the pre-fossil fuel population of 1 billion. At that point, everything we take for granted will be rusted, crumbling, or broken apart — all the the roads, bridges, skyscrapers, dams, pipelines, electric grid, drilling and mining equipment, cars, trucks, and so on.
I don’t see how we could ever recover from that, ever build anything sophisticated enough to read the terabytes of information stored on any remaining computers.
You need educated engineers to rebuild with, but they need to be drawn from the 10% of the population who aren’t growing food, and decades of stability are required to gain a PhD education in engineering. You’ll also be short on fossil fuel engineers, who can maybe get at some of the remaining coal, but much of the natural gas and oil are so remote or stranded (since the energy drilling and pipeline infrastructure has by now rusted apart), that metallurgy engineers will be hard pressed to make iron or steel with so little energy (2370 degrees Fahrenheit to smelt iron).
Nor will alternate energy such as wind, solar, dams, geothermal survive oil decline, since their materials, construction, delivery, and maintenance are oil dependent. Wind is too intermittent to provide energy without natural gas peaking plants to keep the amount of electricity in the system steady so the electric grid isn’t fried.
Of course future civilizations can mine skyscrapers for a while, but they’ll be so short on energy it will be hard to do much. And Tree charcoal will be in short supply (read John Perlin’s outstanding book: “A Forest Journey: The Role of Wood in the Development of Civilization” to understand why)
Ewing J, Clark D (2021) Lack of Tiny Parts Disrupts Auto Factories Worldwide. New York Times.
Quirk M, J. Serda J (2001) Semiconductor manufacturing technology. Prentice Hall.
Van Zant, P (2004) Microchip Fabrication, fifth edition. McGraw-Hill.