Semiconductors are at the heart of most electronics, but have you ever wondered how one works? In this article, we explain what semiconductors are; how they work, with an eye to N-type vs. P-type semiconductors; what semiconductors are made of and how they're built; and where they are found and useful for. We'll wrap up with a word on Moore's Law, and what happens when semiconductor manufacturing bumps up against the laws of physics.
What Is a Semiconductor?
A semiconductor is a material or device that only conducts electricity under certain conditions. Some elements in the periodic table, such as carbon, silicon, and germanium, are intrinsic semiconductors due to the number of electrons in their outermost electron shells. By comparison, extrinsic semiconductors are composite materials created by treating or infusing intrinsic semiconductors with other elements from the periodic table (more on this below).
Conductor, Semiconductor, Insulator
On the periodic table, semiconductors (also known as metalloids) crop up along a diagonal line between conductors and insulators. Noble gases such as helium and neon are poor conductors of electricity, so we call them electrical insulators. Copper and other metals are conductors, which allow the free movement of electrons throughout the material. Transition metals are also excellent thermal conductors. But if you heat up a copper wire, it becomes less conductive and acts more like an electrical insulator. Heating up semiconductors makes them more conductive.
Semiconductors are defined by their behavior with respect to electricity, and they require electricity to do what they do. When connected to a voltage source, a semiconductor will transmit energy under certain conditions; change the conditions and the amount of energy flowing through the semiconductor also changes. Without power, even a space-age semiconductor is just another pretty rock.
What Are Semiconductors Made Of?
How semiconductors work is a function of what they're made of. Many semiconductors are based on silicon, with atoms of other elements (such as phosphorus or boron, which are themselves semiconductors) interspersed throughout the silicon crystal lattice in a process called "doping." The choice of dopant determines what properties the finished semiconductor will have.
There exist a huge number of exotic semiconductor materials, almost as many as there are material science projects using them. Still, modern manufacturing is dominated by two basic types of semiconductors: N-type and P-type.
N-type vs. P-type Semiconductors
Like carbon, silicon has four electrons in its outermost electron shell, so it can form four bonds. N-type semiconductors use dopant atoms from elements in group V, with five or more electrons in their valence shell. Readily available phosphorus is a common choice. The "spare" electron from each dopant atom acts like a negative charge, so semiconductors of this type are called negative-type or N-type.
In contrast, P-type semiconductors use a dopant element from group III on the periodic table (such as boron, indium, or gallium). These elements have only three electrons in their valence shell, so where N-type semiconductors are doped with electron donors, P-type semiconductors use dopants that act as electron acceptors.
Boron fits into the crystal lattice just like silicon atoms, but since boron has only three electrons to use, a kind of hole is created where an electron might have been. The absence of a negative charge behaves like a positive charge, so these semiconductors are called positive-type or P-type.
When a semiconductor includes both N-type and P-type materials, it is called a complementary metal-oxide semiconductor (CMOS, pronounced "sea-moss").
How Are Semiconductors Made?
Many semiconductors start their lives as massive cylindrical ingots of monocrystalline silicon called boules. Boules are formed by dipping a seed crystal into pure molten silicon. Extrinsic semiconductors (those that use dopant) can also be created through vapor deposition inside an electromagnetic field. These gigantic crystals take many hours to grow.
Ingots are sliced into wafers just a few millimeters thick, lapped and polished to a flawless mirror shine, and then etched with functional patterns (such as resistors, transistors, and capacitors) using high-intensity light in a process called photolithography. The result is called an integrated circuit.
Many wafers contain multiple copies of a single type of integrated circuit, which are eventually cut apart and packaged individually.
Since they're thin and fragile, integrated circuits are often enclosed in a housing of plastic or metal, which is then connected to a printed circuit board using metal wires. You'll see this general form factor used in surface-mounted technology (SMT) components used by manufacturers, and it's also used in consumer-oriented computer chips, like the CPUs sold by companies such as Intel and AMD. However, "wafer-scale computing" uses processors that are made of an entire wafer, for massive throughput at very low latency.
How Do Semiconductors Work?
When voltage is applied across a material such as a semiconductor, the negative side of the voltage pushes electrons (since like charges repel), and the positive side pulls electrons toward itself. Organizing the motion of electrons into one direction creates an electrical current. Fluctuations in the flow of electricity can encode data, such as analog sensor readings or binary code.
A semiconductor device controlled by fluctuations in an electric field is called a field-effect transistor or FET. The term "CMOS" often specifically refers to the layered physical structure of devices such as metal-oxide-semiconductor field-effect transistors, or MOSFETs. MOSFETs have a metal gate electrode atop an oxide insulator, which is itself affixed to a slice of semiconductor material. Because of the synergy between N-type and P-type materials, CMOS logical systems are frequently more energy efficient than purely NMOS or PMOS logic.
What Is a Band Gap?
Semiconductors rely on quantum effects such as band gaps, so we do have to discuss a bit of quantum stuff—we'll stick to plain language as much as possible.
First, a bit of background. Electrons don't act like objects on the macroscale; mostly ignoring Newton's laws of kinematics, electrons cleave instead to laws describing the probability of where they'll turn up. "Like charges repel" is also the rule that governs the places electrons choose to hang out around their host atoms. Pinned to their positively charged nucleus by the electromagnetic force, electrons will still try to separate themselves equally from one another in 3D space, leading to elaborate (and sometimes delicious!) distribution patterns like tiny Rorschach blots.
Collectively, electrons that are not tightly attached to just one nucleus are called the conduction band. By absorbing energy in the form of a phonon (heat) or a photon (light), electrons can gain enough energy to jump to the conduction band. The band gap is the energy difference between the highest energy level or "top" of the valence band (that is, the energy embodied in the highest-energy electron that won't bounce out of a valence shell) and the bottom of the conduction band.
Insulators have a wide band gap that is difficult for electrons to cross without absorbing a great deal of energy, if they can cross it at all. By contrast, conductors have a band gap that is narrow to nonexistent. Semiconductors are sometimes considered to be insulators with narrow band gaps.
For a visual explanation of N-type versus P-type semiconductors, including a clear visual explanation of the band gap, check out the following video, starting at about four minutes in:
Since it's a measure of energy, the band gap is expressed in electron volts, abbreviated eV.
What Are Semiconductors Used For?
Most electronic devices rely on semiconductors, assembled into transistors and integrated circuits. Because of their ability to transmit electricity (but only sometimes), semiconductors can serve as sensors, emitters, and/or their combined form: the transistor. There are too many applications to name, but here are three of the most important, each of them showing off the key properties of a semiconductor in its own way.
Solar Cells
Solar cells use semiconductors to harvest renewable energy from the sun. In a solar panel, a layer of semiconducting material absorbs energy from the sun as photons and emits it in the form of electrons, producing an electrical current.
Solar panel efficiency has climbed significantly over the last 20 years, due to developments like layering different semiconductors, each with its own distinct band gap, to catch as broad a range of the Sun's energy as possible.
Light-Emitting Diodes (LEDs)
LEDs are another semiconductor application that relies on the band gap, but for LEDs the band gap plays a role both functional and aesthetic. When a particle or material absorbs energy, it enters an excited state. Releasing energy equivalent to the breadth of its band gap returns it to its previous state.
Sometimes, when a particle "relaxes," it releases a photon instead of an electron. Each photon has its own wavelength, corresponding to a color. Fine-tuning semiconductor devices using this effect allows for a wide range of colors, from low-energy and long wavelength (infrared) through the visible spectrum into the high-energy ultraviolet band. But as wavelength decreases, it gets tougher to surmount the engineering challenges imposed by components getting ever smaller, which brings us to...
Transistors
Possibly the most important use of semiconducting materials is in transistors, which enable computing and radio communications. Transistors are electronic components, devices which can function as switches or amplifiers. Usually, transistors have at least three connections and often four, such as in a generic MOSFET. Electrical voltage or current applied to one pair of terminals controls the current output through the other pair. One way to understand how transistors work is by comparing them with a garden hose with a kink in it. Positively charging the transistor acts like un-bending the hose, permitting the flow of electrons.
Because they're useful as switches, transistors also play a role as logic gates, physical structures that encode Boolean logic. The CPUs and GPUs at the heart of laptops and PCs rely on semiconductors etched with logical structures that can move data without moving parts.
Moore's Law: The End of Semiconductor Scaling (Maybe)
Software has exploded in complexity since the days of vacuum-tube computation. The devices required to do all that data crunching at speed have mostly kept pace in complexity, both in their internal logical structure and in the absolute size of features in an integrated circuit. Every new process node allows fabs and foundries to pack more features onto a single chip. Today, billions of transistors, so tiny as to be invisible to the unaided eye, take up the same footprint as a single transistor from the 1950s. Integrated circuits get smaller, faster, and more efficient over time—but there are limits.
As feature size shrinks, so does the amount of power it takes to activate each individual transistor. The wavelengths of light that can etch those tiny patterns into a wafer get smaller, too, so higher-energy light is required to do the etching. Meanwhile, as features pack in together, the chip has to work harder to dissipate heat. Together, these phenomena are often referred to as Moore's Law (although Moore was more focused on the physical size of transistors, while the frequency and thermal/electrical properties of shrinking circuits were described by Dennard scaling). All the while, as feature size approaches a single atom wide, it gets easier and easier for quantum effects like electron tunneling to produce errors in the manufacturing process and the finished product.
Constrained by the laws of physics, semiconductor companies are exploring different strategies to wring each incremental performance improvement out of their next process node. We may not need circuits that fire at 10 gigahertz, if we can get throughput by other means than raw speed. Chips don't have to reach internal temperatures approaching that of the surface of the Sun.
Similarly, traces don't need to be a single atom wide. If a chip designer can't cram another transistor per square millimeter into their next-gen chip design, maybe it's time to look for new types of semiconductor materials. Another promising option is a "3D" form factor, with functional layers stacked up like the strata in sedimentary rock. Academics have even proposed volumetric crystalline semiconductors that would look right at home in an episode of Stargate: SG-1. You can guess which we favor.
