By: Dr. Andreas Schumacher
Semiconductors have become foundational building blocks of modern societies, economies, militaries, and consumer goods. But public and policy debates about semiconductors are often reduced to “nanometers” as a measure of their merit. The 21st century’s most essential resource deserves a more nuanced debate.
Long before the first transistor (a single electrical switch) was demonstrated in 1947, it was clear that switches could be combined to form logic circuits. By linking enough of those circuits together, computers could be built. After a decades-long path along the exponential shrink roadmap of Moore’s Law (the observation that the number of transistors in a dense integrated circuit doubles about every two years), leading-edge factories can today produce features with the size of three nanometers (nm). Billions of transistors can now be packed on a device the size of a postage stamp with structures much smaller than the wavelength of visible light. Manufacturing at 3 nm, therefore, requires extreme ultraviolet light, and the machines used to process these types of semiconductors are among the most complex machines ever manufactured by mankind.
Manufacturing these type of semiconductors becomes very expensive. For it to be economically feasible, major economies of scale are crucial: manufacturing the smallest nodes requires some of the highest volume applications (e.g. mobile phones) to be economically lucrative. This, in turn, favors ever larger and more specialized producers and has given rise to a division of labour in the industry: foundries for manufacturing and fabless companies specializing in design. The largest and most advanced manufacturing firm is Taiwan Semiconductor Manufacturing Company (TSMC), which dominates a landscape in which 90% of sub-10 nm semiconductors are manufactured in Taiwan. The geopolitical implications are obvious.
Another branch of semiconductor development involves memory chips. Memory semiconductors are conceptually simple: they allow storage and read-out of information in the form of electrical signals (0 and 1 — what makes “digital” digital). Depending on whether storage requires constant electrical power, or whether information is retained also without electrical power, memory is classified either as volatile or non-volatile. Memory semiconductors are shrinking at a rate just marginally slower than microprocessors.
Still other semiconductors evolved towards optimizing the individual, discrete transistor. This can chiefly be done along two dimensions: faster switching (eventually at radio frequencies) and switching of ever higher electrical power. The latter are known as “power semiconductors.” Power semiconductors are typically not highly integrated — just a single switch — but designing and manufacturing high-performance power semiconductors require sophisticated control of the underlying technologies and processes. A power semiconductor that is 3% more efficient can make an electrical car drive 3% further or save 3% of energy. The materials used to manufacture a power semiconductor have evolved. Whereas basic, cost-optimized devices are still based on silicon, more advanced, energy-efficient devices require novel materials such as silicon carbide and gallium nitride.
While much of the public attention over the last years focused on microprocessors of the smallest node, following this shrink path is economically attractive only for highest volumes and if highest performance at the lowest power is a design requirement. Much of the microcontrollers that control our everyday life — for washing machines, cars, airplanes, and industrial control systems — are manufactured at feature sizes of 22 to 40 nm. Microcontrollers and device in this range are optimized for different requirements, e.g. high security and reliability, or low-power consumption. “Analog/mixed signal” is another important type of integrated circuit. These semiconductors are the interface between our real (analog) world and the digital world. Devices are optimized to process both analog and digital electrical signals. At the highest frequencies, those analog electrical signals are electromagnetic waves, giving rise to a class of radiofrequency semiconductors. WiFi, Bluetooth, and radar chips belong to this category.
Sensors, too, are often based on semiconductor technology, e.g. for temperature, magnetic fields, gas, or pressure. The number of physical effects semiconductors can sense and convert into electrical signals is impressive. In a mobile phone, for example, the microphone, the camera, temperature, and positioning are all semiconductor-based. The first “quantum” sensors, which make use of quantum-mechanical effects to detect exceedingly small signals, are entering commercial use.
While the world of semiconductors may be essential for modern societies and economies, it is also vast and diversified: power semiconductors, sensors, memory, microcontrollers, and microprocessors. Some are manufactured at the smallest nodes, others on novel materials, but all require enormous research efforts and knowhow. Advancement cannot be described along only one dimension, and modern economies rely on many different types of semiconductors. This is important for informed public and policy debates and targeted and efficient industrial policies.
Dr. Andreas Schumacher is Executive Vice President for Strategy and Mergers & Acquisitions at Infineon Technologies, currently on sabbatical at the Center for Strategic and International Studies (CSIS) in Washington DC. The views expressed herein are strictly his own.