Metalloids

Every other family on this table has a definition someone could check an element against. The metalloids do not. IUPAC has never issued one, no two reference works agree on the membership, and the number of elements included ranges from about four to about a dozen depending on which criteria are used and how strictly. Boron, silicon, germanium, arsenic, antimony and tellurium appear on almost every list; polonium and astatine appear on many; selenium, aluminum, carbon and even bismuth appear on some.

This is not sloppiness. It is a category attempting to name a gradient, and gradients do not have edges. Everything worth understanding about metalloids follows from accepting that, rather than from trying to fix the boundary.

What the label is actually measuring

Several distinct properties are bundled into the word, and they do not switch over at the same element.

• Electrical behaviour. A metal's conductivity falls as it warms; a semiconductor's rises, because heat promotes electrons across an energy gap. This is a qualitative, unambiguous test — and it is the one criterion that behaves cleanly.
• Mechanical behaviour. Metalloids are brittle. They fracture rather than deform, because their bonding is directional and covalent rather than a delocalised electron sea.
• Electronegativity. They cluster around 1.9 to 2.2 on the Pauling scale, between the metals below and the nonmetals above.
• Oxide character. Their oxides are typically amphoteric or weakly acidic, rather than the clearly basic oxides of metals or the clearly acidic ones of the reactive nonmetals.
• Appearance. Most look like metals. Silicon, germanium, antimony and tellurium are all grey and lustrous, which is why they were mistaken for metals historically and why antimony was used as a metal for centuries.

Run those five tests over the periodic table and they mark out overlapping but different sets. That is why the staircase drawn on most tables — the diagonal line from boron down to astatine — is a convention rather than a boundary, and why some tables draw it in a different place.

The band gap is the sharpest thing here

If one number captures the metal-to-nonmetal transition, it is the energy gap between the highest filled band of electron states and the lowest empty one.

In a metal the two overlap, so an arbitrarily small electric field moves electrons and conduction costs nothing. In an insulator the gap is several electronvolts and thermal energy at room temperature — about 0.025 eV — cannot bridge it. Between those extremes sits a gap small enough that a useful number of electrons cross it, and the population that crosses depends steeply on temperature, on illumination, and on deliberately added impurities.

Walking down group 14 shows the transition as a set of numbers rather than a classification. Diamond's gap is around 5.5 eV, which makes carbon in that form a transparent insulator. Silicon's is 1.12 eV. Germanium's is 0.67 eV. Grey tin's is essentially zero, and white tin, the ordinary form, is a plain metal. The same column, the same bonding geometry at the top, and the character changes because larger atoms overlap their orbitals more diffusely and the bands broaden until they touch.

Boron is the outlier at the top. Its gap is wide — over an electronvolt and a half — but it is structurally unlike its neighbours, building icosahedral B₁₂ clusters rather than a simple network, because with three valence electrons and four available orbitals it is electron-deficient and has to share bonding electrons between three atoms at a time. Boron chemistry needed its own bonding theory, and the boranes that came out of it earned William Lipscomb the 1976 Nobel Prize.

Doping is why the gap matters commercially

A pure semiconductor is a mediocre conductor and not much use. What makes the band gap valuable is that a small gap can be manipulated. Adding phosphorus to silicon introduces atoms with one more valence electron than the lattice needs, and that surplus electron sits in a level just below the empty band, where almost no energy is required to free it — an n-type material. Adding boron introduces atoms one electron short, creating vacancies just above the filled band that neighbouring electrons can hop into — p-type. Put the two in contact and current passes one way and not the other, and every diode, transistor and solar cell descends from that asymmetry.

The concentrations involved are the reason purity dominates the industry. Useful doping levels are around one part in ten million, so an unintended impurity a hundred times rarer than that still matters.

The word used to mean the opposite

"Metalloid" has been repurposed. In the early nineteenth century Berzelius and his contemporaries used it for what would now be called a nonmetal — literally "metal-like", applied to the newly isolated sodium and potassium, which were shiny and conducting but so light and soft that many chemists refused to accept them as true metals. The modern sense, meaning an element intermediate between metal and nonmetal, settled only in the twentieth century, and older sources using the word in its original sense are still in circulation. A category with no agreed membership has also had no stable meaning.

Polonium is not really one of these

Every family has a member that fits badly, and here it is the one added purely by position. Polonium conducts like a genuine metal — its resistivity rises with temperature in the ordinary metallic way — and its structure is unique in the periodic table: it is the only element known to adopt a simple cubic lattice under ambient conditions, an arrangement so inefficient at filling space that nothing else uses it.

Polonium is on metalloid lists because it sits on the staircase and because its neighbours are metalloids, not because it behaves like one. Astatine is even weaker as a case: too little of it has ever existed at once for its bulk electrical properties to be measured, and recent calculations suggest condensed astatine would be metallic. Both are examples of the same error the category invites — reading a position as a property.

Rectifiers that worked for thirty years before anyone could say why

The commercial history of this category opens roughly sixty years before the physics that explains it, and silicon is in it from the beginning.

In 1874 Ferdinand Braun touched a fine metal point to a galena crystal and found that current passed one way and not the other. Galena is lead sulfide and no kind of metalloid, but the effect he had stumbled into became this category's first industry. The crystal detector — the front end of every early radio set, a mineral chip and a springy wire — settled on two materials, and both were built on silicon. Henry Dunwoody patented a detector made of silicon carbide in March 1906, using the abrasive sold as carborundum, and Greenleaf Whittier Pickard patented one made of the element itself later that year, having worked through thousands of mineral samples to find it rather than reasoning his way to it. Dunwoody's filing had priority; both types stayed in service for years.

Nobody involved could account for the asymmetry. The band picture that explains it arrived in the 1930s, and the barrier model that describes a metal touching a semiconductor came at the end of that decade — by which time the devices had been sold commercially for a generation. That is the sharpest argument for why this awkward, ill-defined category earned a name: its members were doing something that neither the metals nor the nonmetals could do, and doing it before anyone had the vocabulary to say what.

No two of them come from the same industry

Purification rather than extraction is where the cost sits for anything electronic, and silicon works through that sequence at length; the question of why germanium got the first transistor and then lost everything belongs with silicon against germanium.

The more telling fact sits upstream of all that. These six share no supply chain whatsoever. Silicon is reduced from the commonest oxide on the planet, and its raw material has never once been the constraint. Boron is lifted from evaporite beds in a handful of arid basins. Antimony has an ore of its own and a producer list short enough to be a policy problem rather than a market one. Tellurium, arsenic and germanium are not produced on purpose at all — each is skimmed from the refining of a base metal somebody wanted for other reasons.

So "the metalloids" is not a market. A shortage in one implies nothing about the others. Three of the six cannot be made to appear faster however high the price climbs, because nobody anywhere is deciding how much of them to make. Of the three that can, silicon is limited by essentially nothing, and the other two are limited by where the deposits happen to be rather than by how much is in the ground.

Their smaller uses have as little in common. Antimony trioxide works as a flame-retardant synergist, and antimony hardens the lead in a battery grid. Boron reaches the world as borax and as borosilicate glass, while boron-10 — separated isotopically, not chemically — absorbs neutrons well enough to serve in control rods and shielding. Tellurium goes into cadmium telluride thin-film photovoltaics. Arsenic's largest legitimate use is gallium arsenide, which switches faster than silicon and emits light, which silicon essentially cannot.

Where the category misleads

The commonest error is treating metalloid and semiconductor as the same word. They are not. Gallium arsenide, indium phosphide, silicon carbide and cadmium telluride are all important semiconductors and none of them is a metalloid; grey tin is a semimetal and is not on anybody's list; and boron is a metalloid whose gap is wide enough that it makes a poor semiconductor.

The second is expecting a fixed count. Tables that state "there are seven metalloids" are reporting a convention, and a different textbook will state six or eight with equal confidence and equal justification.

The third is assuming the staircase means something physical about the elements immediately either side of it. Aluminum sits below the line and is a textbook metal, but its oxide is amphoteric like a metalloid's; selenium sits above the line and is a nonmetal, but its grey allotrope is a photoconductor that ran the entire xerographic printing industry for decades. The line is a drawing aid on a chart, and the elements have not read it.

The most useful way to hold all this is to treat metalloid as a description of behaviour in a particular situation rather than as a permanent property of an atom. Antimony was a metal to a typefounder and is a metalloid to a chemist; tin is a metal at room temperature and a semiconductor well below it; carbon is a transparent insulator as diamond and a conductor as graphite. The question worth asking of an element near the staircase is not which side it belongs on, but which of the five tests above it passes, and under what conditions it passes them.