Nine Nines: Inside a Modern Gold Refinery

The bar arriving on the loading dock does not look like money. It looks like a dirty brick: pitted, grey-gold, stamped with a mine code and a weight in a hand-punched font. This is doré, the semi-refined alloy poured at the mine site, and depending on the geology behind it, somewhere between seventy and ninety per cent of its mass is gold. The rest is silver, copper, occasionally a trace of platinum, and whatever else the ore was carrying.

Turning that brick into a bar the wholesale market will accept without argument is an industrial process about a century and a half old in its essentials. It is also, increasingly, a documentation process. Refiners today spend as much effort proving where a bar's contents came from as they do removing the copper.

Weigh, sample, argue

Nothing enters the plant unweighed. The doré is logged to the hundredth of a gram on scales calibrated daily against certified masses. Then it is melted, homogenised and sampled — because a bar's average composition tells you nothing if the gold has segregated during cooling.

The classic sampling method is the dip sample: a rod plunged into the melt, withdrawn fast, quenched. More common now is drilling the solidified bar at prescribed points, top and bottom, corner and centre, and combining the swarf. The resulting sample is split three ways — one for the refiner, one for the supplier, one sealed as an umpire sample in case the first two disagree. Disagreements over a tenth of a per cent on a four-hundred-ounce bar are worth arguing about.

Fire assay: the referee

The determination that settles disputes is fire assay, a technique that predates the periodic table by two thousand years and has never been bettered for accuracy on gold. The sample is wrapped in lead foil, fused in a crucible with litharge and flux, and poured. The lead collects the precious metals and sinks; the base metals dissolve into a glassy slag.

The lead button is then transferred to a cupel — a small porous dish of bone ash or magnesia — and heated in a stream of air. Lead oxidises and is absorbed into the cupel wall, wicking away like water into blotting paper. What remains is a bead of gold and silver, a couple of millimetres across, sitting in the dish like a drop of solder. Parting the bead in nitric acid dissolves the silver. The gold that survives is weighed on a microbalance.

Stage one: the Miller process

Francis Bowyer Miller patented his method in 1867, and it is still how most of the world's gold takes its first serious step towards purity. Molten doré is held at around 1,100°C and chlorine gas is bubbled through it via a ceramic lance.

The chemistry is a matter of preference. Chlorine bonds with base metals before it bonds with silver, and with silver before gold. Zinc, iron, copper and lead convert to chlorides that either volatilise off as fume or float to the surface as a molten salt layer that can be skimmed. Silver becomes silver chloride and joins the skim. Gold, being the least reactive thing in the crucible, is left behind.

The operator's signal that the reaction is complete is visual: the fume above the melt shifts colour as chlorine begins attacking gold itself, producing a reddish gold chloride vapour that means the run has gone one minute too far. Modern plants automate the endpoint with spectroscopy, but the older hands still watch the smoke.

• Speed: a Miller run takes roughly ninety minutes to two hours per charge.
• Ceiling: purity plateaus around 99.5 per cent — good, and not good enough.
• Limitation: platinum-group metals do not chlorinate usefully and remain in the gold.
• Byproduct: silver chloride skim is reduced and recovered separately, and is often the more voluminous product.

Stage two: Wohlwill electrolysis

To cross from 99.5 to 99.99, the metal goes into a cell. Emil Wohlwill's 1874 process casts the Miller output as anodes, suspends them in an electrolyte of gold chloride in hydrochloric acid, and applies a current. Gold dissolves from the anode and plates onto a cathode of thin pure gold sheet or titanium.

Selectivity comes from electrochemistry. At the cell's operating potential, gold deposits and the platinum-group contaminants stay in solution. Silver, awkwardly, precipitates as insoluble chloride and falls to the bottom of the tank as part of the anode slime — which is recovered, because anode slimes from a large refinery are among the richest precious-metal feedstocks in existence.

The cost of the Wohlwill process is time and inventory. A cell cycle takes days, and during those days a substantial quantity of gold is locked up in solution and in partially dissolved anodes. Refiners describe this as 'metal in process', and financing it is a real line on the balance sheet. The alternative, the Fizzer cell and related aqua regia routes, trades throughput for a smaller locked inventory and is favoured by smaller operations.

The cast, and what is stamped on it

Finished gold is remelted and cast to specification. For the wholesale market the format is the Good Delivery bar: nominally 400 troy ounces, actual weight between roughly 350 and 430 ounces, minimum fineness 995.0, surfaces free of cavities, edges that allow safe handling. Each bar carries a serial number, the refiner's mark, the fineness and the year of manufacture.

That set of marks is what allows a bar to move between vaults in the loco London system without being re-assayed at every transfer. Accreditation is not permanent; refiners submit to periodic proficiency testing and can be suspended. The list of accredited refiners is, in effect, the market's trust registry.

Recycling, which is most of the business

It surprises people that a large share of refinery throughput never came out of the ground that year. Scrap jewellery, industrial sweeps, spent catalysts and electronic waste all arrive in the same building, and the economics can be better than mined feed because the concentration is higher. A tonne of high-grade ore might carry a few grams of gold. A tonne of sorted circuit-board connectors carries orders of magnitude more.

Handling scrap is messier work. Feedstock arrives as everything from wedding rings to shredded phone housings, and the first task is not chemistry but sorting, incineration and homogenisation — turning a heterogeneous heap into something that can be sampled honestly. Only then does the metal join the same route the doré takes.

What the numbers mean

Investors treat four nines as a synonym for 'pure', but the last decimal points are earned at steeply rising cost. Going from 99.99 to 99.999 requires a second electrolytic pass or zone refining, and the market for five-nines metal is narrow: semiconductor bonding wire, certain optical coatings, scientific reference materials.

For everything else — coins, bars, jewellery alloys, central bank reserves — four nines is the end of the road, and it is close enough to the theoretical ideal that the remaining 0.01 per cent is mostly a rounding error in trace silver. The refinery's real product, in the end, is not purity. It is the credible assertion of purity, backed by an assay, a mark and an accreditation that can be taken away.