Atomic Structure & Bonding

Before we can talk about diodes or transistors, we have to talk about what silicon is. Every behaviour you'll meet in this subject — the 0.7 V drop of a diode, the gain of a transistor, the way a thyristor latches — traces back to how the silicon atom arranges its electrons and how those electrons share themselves with their neighbours. Atomic structure dictates conductivity, and conductivity is the whole point of a semiconductor.

The Bohr atom in 60 seconds

Niels Bohr's 1913 picture of the atom is the one most of us were taught at school: a dense central nucleus of protons (positive) and neutrons (neutral), orbited by negatively charged electrons arranged in concentric shells. Modern quantum mechanics says the picture is a cartoon — electrons live in fuzzy probability clouds, not on tracks — but the cartoon is almost good enough for the chemistry of bonding. We'll use it throughout.

Each shell is labelled by a principal quantum number $n$: K (n=1), L (n=2), M (n=3), N (n=4), and so on outward. The maximum number of electrons a shell can hold is given by:

So the K shell holds 2, L holds 8, M holds 18, and N holds 32. Atoms fill the inner shells first and only spill into the next shell once the one below is full (with a couple of fiddly exceptions in the transition metals that don't concern us).

Valence: the only shell that matters for bonding

The outermost occupied shell — the one furthest from the nucleus — is called the valence shell, and the electrons in it are valence electrons. Inner-shell electrons are tightly bound and chemically boring; they don't move between atoms and don't make bonds. All the chemistry, and all the electrical behaviour we care about, happens in the valence shell.

Atoms "want" a complete outer shell. For most of the elements you'll meet, complete means eight valence electrons — the famous octet rule. An atom with one or two valence electrons will happily give them up; an atom with six or seven will happily grab a couple more; and an atom with four sits squarely in the middle, which turns out to be exactly where we want it.

Why silicon

Silicon (atomic number $Z = 14$) has the electron configuration 2-8-4: two in the K shell, eight in the L shell, four in the M shell. The four valence electrons in that outer M shell are silicon's entire electrical personality. With four electrons it has neither a strong urge to give them up (like sodium with one) nor to grab more (like chlorine with seven). Instead it shares them.

In a silicon crystal, every atom places itself so that it has four nearest neighbours, and it shares one valence electron with each. Each shared pair forms a covalent bond. From any one atom's point of view, the four electrons it owns plus the four it shares give it a complete outer shell of eight — octet satisfied. This is the diamond-cubic crystal structure, the same geometry as diamond itself, and it is what holds a silicon wafer together.

Germanium and carbon: same family, same idea

Silicon sits in group IV of the periodic table (sometimes called group 14 in the modern numbering). Two of its cousins matter for us: germanium (Z = 32, shells 2-8-18-4) and carbon (Z = 6, shells 2-4). All three have four valence electrons. All three form four covalent bonds with their neighbours. Carbon does it in diamond, silicon does it in a wafer, germanium does it in early transistors and modern high-speed devices. The shared family trait — four valence electrons — is why they all behave like semiconductors.

Why pure silicon is a poor conductor

Here's the punchline. In pure (intrinsic) silicon at room temperature, every valence electron is locked into a covalent bond with a neighbour. There are no free electrons drifting between atoms, and no empty places (holes) for one to hop into. Without free carriers, current cannot flow. Pure silicon is an insulator — not as good an insulator as glass or rubber, but far from being a conductor like copper. Its room-temperature resistivity is roughly 2.3 × 10⁵ Ω·cm, about a million million times that of copper.

That sounds like bad news. It is, in fact, the entire setup for the trick. A material that is a poor conductor on its own — but whose conductivity we can control by adding a vanishingly small amount of impurity — turns out to be exactly what you need to build a switch.