Semiconductors are an essential part of almost all modern electronic devices, and the vast majority of semiconductors are made in Taiwan. Growing concerns over Taiwan’s reliance on semiconductors – especially given Taiwan’s tenuous relationship with China – led the US Congress to pass the CHIPS and Science Act in late July 2022. The law provides more than $50 billion in subsidies to boost American semiconductors. production and was widely covered in the news. Trevor Thornton, an electrical engineer who studies semiconductors, explains what these devices are and how they are made.
1. What is a semiconductor?
Broadly speaking, the term semiconductor refers to a material – like silicon – that can conduct electricity much better than an insulator like glass, but not as well as metals like copper or aluminum. But when people talk about semiconductors today, they’re usually referring to semiconductor chips.
These chips are usually made from thin wafers of silicon with complex components arranged on them in specific patterns. These models control the flow of current using electrical switches – called transistors – much the same way you control the electrical current in your home by flipping a switch to turn on a light.
The difference between your house and a semiconductor chip is that semiconductor switches are entirely electrical – no mechanical components to flip – and chips contain tens of billions of switches in an area not much larger than the size of a nail.
2. What do semiconductors do?
Semiconductors are how electronic devices process, store and receive information. For example, memory chips store data and software as binary code, digital chips manipulate data according to software instructions, and wireless chips receive data from high-frequency radio transmitters and convert it into electrical signals. These different chips work together under the control of software. Different software applications perform very different tasks, but they all work by switching transistors that control current.
3. How is a semiconductor chip made?
The starting point for the vast majority of semiconductors is a thin slice of silicon called a wafer. Today’s wafers are the size of dinner plates and are cut from single crystals of silicon. Manufacturers add elements like phosphorus and boron in a thin layer on the surface of the silicon to increase the conductivity of the chip. It is in this superficial layer that the transistor switches are made.
Transistors are built by adding thin layers of conductive metals, insulators, and more silicon to the entire wafer, sketching patterns on those layers using a complicated process called lithography, and then removing selectively these layers using computer-controlled plasmas of highly reactive gases to leave specific patterns and structures. Because transistors are so small, it is much easier to add layered materials and then carefully remove unwanted materials than to place microscopic lines of metal or insulators directly on the chip. By depositing, patterning, and etching dozens of layers of different materials, semiconductor manufacturers can create chips with tens of billions of transistors per square inch.
4. How are today’s chips different from earlier ones?
There are many differences, but probably the most important is the increase in the number of transistors per chip.
Among the first commercial applications of semiconductor chips were pocket calculators, which became widely available in the 1970s. These early chips contained a few thousand transistors. In 1989, Intel introduced the first semiconductors to exceed one million transistors on a single chip. Today, the biggest chips contain more than 50 billion transistors. This trend is described by what is known as Moore’s Law, according to which the number of transistors on a chip will double approximately every 18 months.
Moore’s Law has stood for five decades. But in recent years, the semiconductor industry has had to overcome major challenges — primarily, how to keep transistors downsized — to keep up this pace of advancement.
One solution was to go from two-dimensional flat layers to three-dimensional layers with fin-like silicon ridges protruding from the surface. These 3D chips greatly increased the number of transistors on a chip and are now widely used, but they are also much more difficult to manufacture.
5. Do more complicated chips require more sophisticated factories?
Simply put, yes, the more complicated the chip, the more complicated the factory – and the more expensive.
There was a time when almost every American semiconductor company built and maintained their own factories. But today, a new smelter can cost over $10 billion to build. Only the biggest companies can afford this kind of investment. Instead, the majority of semiconductor companies send their designs to independent foundries for manufacturing. Taiwan Semiconductor Manufacturing Co. and GlobalFoundries, headquartered in New York, are two examples of multinational foundries that manufacture chips for other companies. They have the expertise and economies of scale to invest in the extremely expensive technology needed to produce next-generation semiconductors.
Ironically, while the transistor and the semiconductor chip were invented in the United States, no state-of-the-art semiconductor foundry is currently found on American soil. The United States has been here before in the 1980s, when it was feared that Japan would dominate the global memory market. But with the recently passed CHIPS Act, Congress has provided the incentives and opportunities for next-generation semiconductors to be made in the United States.
Maybe your next iPhone’s chips will be “designed by Apple in California, made in the USA.”