Smelting Aluminium

A Metal Bound to Oxygen

Aluminium is the most abundant metal in the Earth’s crust, about 8 % by mass, yet there is no metallic aluminium in nature: it is always bound to oxygen, silicon and other elements in minerals. Smelters start from alumina, Al₂O₃, refined from bauxite.

Alumina is very stable. Forming a mole of it gives out about 1,676 kJ of heat, and it does not melt until about 2,050 °C. The usual chemical ways of taking oxygen away hardly work on it.

By heat of formation at 25 °C, aluminium gives out about 559 kJ for each mole of oxygen atoms it binds, iron about 275 kJ (computed here): aluminium holds its oxygen far more tightly.

Heat of formation of Al₂O₃ about −1,676 kJ/mol · melts at about 2,050 °C · per mole of oxygen: aluminium about 559 kJ, iron about 275 kJ (computed)

Sold by the Ounce

In 1825 Ørsted made a tiny amount of aluminium; in 1827 Wöhler reduced it with potassium and got an impure powder. In the 1850s Deville in France switched to sodium and made it in batches, yet in 1869 the whole world made only about 2 t.

As late as the early 1880s aluminium was sold by the troy ounce, at a retail price reported to be higher than silver’s. In the late 1850s a pound cost about $17, close to silver; a pound of gold then cost about $300 (from $20.67 an ounce). On the recorded prices, aluminium was never dearer than gold.

In 1884 the Washington Monument was finished with a solid aluminium tip of 100 ounces and 8.9 inches, the largest piece of aluminium cast to that time; before it went up it was shown at Tiffany’s in New York. Aluminium cost about $1.10 an ounce that year. At the front left, a replica, not to the pot’s scale.

Late 1850s: about $17 a pound, close to silver · gold about $300 a pound (converted) · 1884: about $1.10 an ounce

Two Inventors, One Year

In 1886 two young men, both born in 1863, hit on the same idea independently: instead of melting alumina, dissolve it in molten cryolite and pass a current. In the United States Hall made his first globules of aluminium on 23 February; in France Héroult filed his patent on 23 April.

Hall’s application of 9 July describes dissolving alumina in a fused bath of the fluorides of aluminium and of a metal more electropositive than aluminium, then passing an electric current through it. It also notes that with a carbon anode the released oxygen combines with the carbon, which is gradually consumed and must be renewed.

In 1888 Hall and his partners started a works in Pittsburgh, later the Aluminum Company of America (Alcoa). They called the cells pots and the electrolyte bath, words still used today. By 1891 a pound cost less than $1. Both men died in 1914.

Hall: 23 February 1886 · Héroult: French patent filed 23 April 1886 · US patent 400,766 (filed 1886, granted 1889)

Dissolved, Not Melted

The bath is mostly cryolite, Na₃AlF₆, 75⁠–⁠80 %, with excess aluminium fluoride, 9⁠–⁠12 %, calcium fluoride, 4⁠–⁠7 %, and 2⁠–⁠4 % of dissolved alumina. The additives lower its melting point.

Pure cryolite melts at about 1,011 °C, and the pot works at about 960 °C, more than a thousand degrees below alumina’s melting point: the alumina need not melt; dissolved in the salt, it can be split.

The alumina comes from the Bayer process: bauxite is digested in hot caustic soda, the red mud filtered off, aluminium hydroxide precipitated and calcined at about 1,000 °C. A tonne of aluminium takes about 1.93 t of alumina in practice, 1.89 t by the formula.

Cryolite 75⁠–⁠80 % · AlF₃ 9⁠–⁠12 % · CaF₂ 4⁠–⁠7 % · Al₂O₃ 2⁠–⁠4 % · about 960 °C

The Metal Lies Beneath

Liquid aluminium is a little denser than the bath, and the two don’t mix: the metal sinks onto the carbon cathode in a layer 10⁠–⁠20 cm deep, under 15⁠–⁠20 cm of bath. Electrochemically, the cathode is really the top of the metal.

The anodes’ undersides are only 4⁠–⁠5 cm above the metal. The current runs from the anodes through that layer of bath into the metal, then through the steel collector bars at the bottom of the carbon blocks and out through both sides of the shell.

Molten bath attacks the side walls; even carbon would not last long. The side insulation is set so the walls lose just enough heat to freeze a layer of bath on their inner faces; this ledge protects them. The metal covers the bottom, so a pot is never tapped dry.

Metal 10⁠–⁠20 cm · bath 15⁠–⁠20 cm · anode–cathode distance 4⁠–⁠5 cm · a frozen ledge on the side walls

Oxygen Burns the Carbon

Overall, 2Al₂O₃ + 3C → 4Al + 3CO₂. Put simply: at the cathode, aluminium ions take electrons and become aluminium; at the anode, oxide ions give up electrons and join the carbon as carbon dioxide at once. In fact the dissolved aluminium and oxygen sit in complex ions, not as bare Al³⁺ and O²⁻.

The carbon anode is not used up by the current: it is a reactant, burnt by the oxygen. Taking part in the reaction, the carbon lowers the voltage needed by about 1 V.

At 960 °C the reaction needs at least about 1.2 V; without carbon, giving off oxygen gas, it would need about 2.2 V (computed here from thermodynamic data).

2Al₂O₃ + 3C → 4Al + 3CO₂ · at least about 1.2 V (960 °C) · about 2.2 V without carbon (computed)

A Kilogram a Minute

The bath has to hold 2⁠–⁠4 % alumina. A hopper on the pot carries it, and a few point feeders sit on the centre line: a crust breaker punches a hole in the crust, and the feeder drops about 1 kg into it every minute or so.

When the alumina runs short, an anode effect starts: a film of gas covers the undersides of the anodes, the voltage jumps from just over 4 V to tens of volts, and the cryolite itself is broken down, giving off CF₄ and C₂F₆. Over a hundred years CF₄ warms about 7,400 times as much as the same mass of CO₂, and it stays in the air for about 50,000 years.

Modern point-fed pots can run a week or even months without an anode effect. Alumina also covers the crust and the anode tops to keep the heat in, and in the fume treatment it takes up hydrogen fluoride before going back into the pots.

Alumina 2⁠–⁠4 % · fed about 1 kg a minute · anode effect: tens of volts · CF₄ warms about 7,400 times as much as CO₂ (100 years)

Hundreds of Kiloamperes, Four Volts

Modern potlines carry about 300 to 600 kA, yet each pot takes only 4.0⁠–⁠4.5 V. The current runs from the anode beam down the aluminium rods and steel stubs into the anodes, through the bath and the metal, and out through the steel collector bars and the flexibles into the cathode busbars.

One pot’s cathode busbars connect to the next pot’s risers. A hundred to four hundred pots in series make a potline, at well over 1,500 V; a potroom can be more than a kilometre long.

So much current makes a strong magnetic field that pushes the metal round, at about 5 cm a second in places. How the busbars run affects the metal’s waves and the energy used.

300⁠–⁠600 kA · 4.0⁠–⁠4.5 V a pot · 100⁠–⁠400 pots in series · a potline can exceed 1,500 V

One Pot, One Day

Each aluminium atom takes three electrons. By Faraday’s law, a kilogram of aluminium takes at least 2,980 Ah of charge; a kiloampere-hour gives at most about 0.336 kg.

A 400 kA pot at 95 % current efficiency makes about 3.1 t of aluminium a day (computed here). In a 2014 test, Rio Tinto’s AP60 pots ran at 570.7 kA and 95.9 %, and each made about 4,407 kg a day, almost exactly what Faraday’s law gives.

The metal is siphoned out into vacuum crucibles from time to time, about 99.7 % pure. The current efficiency falls short of 100 % mostly because aluminium dissolved in the bath is oxidised again by the carbon dioxide from the anodes.

1 kg of aluminium = 2,980 Ah (computed) · 400 kA at 95 %: about 3.1 t a day (computed) · AP60: 570.7 kA, about 4.4 t a day

Where the Four Volts Go

Of a pot’s four volts and more, only about 1.2 V actually splits alumina. One calculation for a 180 kA cell: reversible voltage 1.22 V, anode overvoltage 0.59 V, bubbles 0.26 V, the bath’s resistance 1.34 V, the cathode 0.03 V and the external conductors 0.74 V, 4.18 V in all.

The extra voltage turns into heat, and that heat is what keeps the bath at 960 °C, so the energy efficiency is only 45⁠–⁠50 %. A modern pot uses about 13⁠–⁠15 MWh of direct current per tonne, against a theoretical minimum of about 6 MWh; the first industrial cells needed more than 40 MWh.

Electricity is about a third of the cost of aluminium, so new smelters are built only where power is cheap. In 2019 about 55 % of the world’s smelter power came from coal and 26 % from hydro. Recycled aluminium takes about 5 % of the energy of primary metal.

Reversible voltage about 1.2 V · bath resistance about 1.3 V · 13⁠–⁠15 MWh of DC a tonne · theoretical minimum about 6 MWh

Four Weeks an Anode

By the formula, each tonne of aluminium burns 0.33 t of carbon and gives off 1.22 t of carbon dioxide. In practice it takes 0.40⁠–⁠0.45 t of anode carbon and the anodes give off about 1.5 t of CO₂: air burns some carbon too, and a layer has to be left to protect the steel stubs.

The anodes are lowered as they burn, and after 25⁠–⁠30 days, at about a quarter of their size, they are changed. A pot has from a dozen to forty, so on average one is changed every day. A new anode weighs about a tonne, pressed from petroleum coke and pitch and baked at about 1,150⁠–⁠1,200 °C; the spent butts are cleaned and go back to the anode plant.

With coal-fired power, the plant that makes the electricity gives off far more carbon dioxide than the anodes. In 2024 the world made about 72.8 million tonnes of primary aluminium, about six-tenths of it in China.

Carbon: 0.33 t per tonne by the formula, 0.40⁠–⁠0.45 t in practice · CO₂ from the anodes about 1.5 t per tonne · an anode lasts 25⁠–⁠30 days

Thermite: Aluminium Takes Iron’s Oxygen

Aluminium’s grip on oxygen works the other way too: lit, a mixture of aluminium powder and iron oxide reacts as aluminium takes the iron’s oxygen, Fe₂O₃ + 2Al → Al₂O₃ + 2Fe. Each mole of iron oxide gives out about 850 kJ (computed here), the temperature passes 2,000 °C, and the iron comes out liquid.

In 1895 the German chemist Hans Goldschmidt made it an industrial process (German patent 96317), and it was soon welding the tram tracks of Essen.

To weld a rail, liquid iron runs from a crucible into a sand mould clamped round the gap between two rails; in about 3 minutes it sets and joins them, and the head is ground smooth. The rail at the front has such a weld in its middle. Welded into long lengths, a track loses the clack of its joints.

Fe₂O₃ + 2Al → Al₂O₃ + 2Fe · about −850 kJ/mol (computed) · over 2,000 °C · 1895, Goldschmidt

A Pot Taken Apart

First a hood panel comes off; then an anode assembly (the carbon block, the stubs and the aluminium rod), a crust breaker, a feeder and the superstructure lift out, and last the cut cathode block slides out with its two steel collector bars; the shell, the lining, the bath and the metal stay.

The part under the pointer lifts out alone and is named; left alone, the parts lift out in turn.

Hood panel · anode assembly · crust breaker · feeder · superstructure · cathode block · the pot