Industries / Mining and materials / Steel and metals / Precious metals refining
Chemical and electrolytic refining
Chemical and electrolytic refining is the work of separating gold, silver and the platinum-group metals from each other and from base metals, with salt, chlorine, acids, solvents or electric current, until each metal is pure enough to sell.
6
Milestones achieved, from 6th century BCE to 1997.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
6th century BCE
Salt parts gold from silver
At Sardis, refiners hammered electrum into thin sheets, packed them in salt inside coarse pots and heated them to about 800 °C for many hours, perhaps days. Salt vapor turned the silver into silver chloride, which soaked into the clay and left the gold "basically pure"; the silver was won back by smelting the used clay and bricks with lead and then heating off the lead. The Sardis Expedition dates the work to the first half of the sixth century BCE and calls the finds "the earliest known evidence for the cementation process anywhere."
LessonLater analysis found that gold from the nearby Pactolus stream "contains virtually no silver," so the workshop may have been parting old electrum coins and objects rather than river gold.
Credit
Lydian refiners; excavated in 1968 and 1969 under Andrew Ramage of Cornell University, published with Paul Craddock of the British Museum in 2000
1805
Platinum made workable
Platinum melts at 1,772 °C, beyond the furnaces of the time. William Hyde Wollaston dissolved platinum ore in aqua regia, a mix of nitric and hydrochloric acids, precipitated the platinum with ammonium chloride, then pressed, heated and hammered the powder into solid metal, a method now called powder metallurgy. He began selling malleable platinum in 1805, and along the way found palladium, which he put on sale before April 1803, and rhodium.
LessonWollaston kept the method secret until a month before his death in 1828; Percival Norton Johnson then used it, and Johnson Matthey grew into the main supplier of these metals.
Credit
William Hyde Wollaston, in partnership with Smithson Tennant, who found osmium and iridium in the leftover residue
1867
Chlorine through molten gold
Francis Bowyer Miller, an assayer at the Sydney branch of the Royal Mint, lodged a London patent in June 1867 for refining gold by passing chlorine gas through it while molten. Gold chloride does not survive above about 400 °C, but silver and copper form a slag of chlorides, and lead and iron leave as fumes. Within two years full-scale plants ran at the Sydney Mint and in Auckland. The method gives gold of 995 to 996 parts per thousand, fine enough for the London Good Delivery bar, and in 1995 the German refiner Degussa estimated that two-thirds of all gold ever refined had passed through it, by LBMA's account.
LessonChlorine cannot separate platinum from gold and does not reach the highest purity, so after many refiners had swapped their electrolytic tanks for Miller furnaces, demand for 999.9 gold, especially in East Asia, brought electrolysis back in the second half of the 20th century.
Credit
Francis Bowyer Miller
1878
Electric current purifies gold
In the Wohlwill process, impure gold hangs as the anode in a bath of hydrochloric acid and gold chloride. An electric current dissolves it, pure gold builds up on the cathode, and impurities stay in solution or fall to the bottom as slime, giving gold of about 99.99%, by Britannica's description. Britannica dates its introduction in Hamburg to 1878; LBMA gives 1874. Its inventor, Emil Wohlwill, was chief chemical engineer of the Norddeutsche Affinerie, today Aurubis, where by the company's account he had already built the first fully continuous copper electrolysis system in 1876.
Credit
Emil Wohlwill, Norddeutsche Affinerie
1971
Solvents sort the metals
Solvent extraction uses an oily liquid that does not mix with water to pull one metal at a time out of an acid solution. Inco's refinery at Acton in the UK had long separated gold, platinum, palladium and the rarer platinum-group metals by dissolving, precipitating and fusing them over and over, which gave a poor first-time yield and took up to six months for rhodium. Inco introduced solvent extraction for gold at Acton in 1971 and then applied it to the other metals; by 2002 several refineries dissolved everything at once and took the metals out one after another with solvents, SGS's reviewers wrote.
LessonIn this trade slow steps are expensive, because the metal held inside the refinery while it waits is worth so much.
Credit
Inco (International Nickel Company), Acton refinery; reviewed by Steve Cole and C. Joe Ferron of SGS Minerals Services
1997
Solvent refining at the mine
Mintek's Minataur process (Mintek Alternative Technology for Au Refining) dissolves the gold, pulls it into a solvent to leave impurities behind, and precipitates gold powder of 99.99% or 99.999%. The first full-scale plant, built for 24 metric tons of gold a year on one shift a day, was commissioned in May 1997 at Harmony Gold Mine in Virginia, South Africa. It refined sludge from the mine's own gold-recovery cells, which had previously been smelted into impure bars called doré and sent to a toll refiner, a refiner paid a fee to process someone else's metal.
LessonRefining at the mine raised legal questions as well as chemical ones; the developers wrote that it might lead to changes in South Africa's law on marketing gold.
Credit
Angus Feather, Kathryn Sole and Leslie Bryson of Mintek; Harmony Gold Mine
Watch for
Gold without cyanide or mercury
A Flinders University team reported a two-step method that dissolves gold from ore and electronic waste with trichloroisocyanuric acid, a chemical sold to disinfect pools, activated by a halide catalyst, and then catches the gold from the solution on a sulfur-based polymer; heating or breaking down the polymer returns high-purity gold. They tested it on ore, electronic waste and other gold-bearing waste in the laboratory.
Achieved whena team outside Flinders, or a mine or recycler, runs it at pilot scale and publishes the gold recovered, its purity and the chemicals used per kilogram.
Claimed by
Maximilian Mann, Justin M. Chalker and colleagues, Flinders University, in Nature Sustainability, August 2025
Open
01
Separation data beyond one solvent
An open, cited dataset of how well each published extractant (a solvent chemical that picks up a metal) takes gold and each platinum-group metal (platinum, palladium, rhodium, iridium, ruthenium and osmium) out of the chloride solutions refiners use, with the conditions of every measurement, and a model trained on it that predicts extraction under untested conditions.
The briefCredit · unclaimed02
Refining routes side by side
A public table that compares, per kilogram of refined gold, the main routes (Miller chlorination, Wohlwill electrolysis, dissolving in aqua regia, solvent extraction such as Minataur, and the room-temperature chemistry now used on circuit boards) on energy, chemicals used, waste made, purity reached, days the metal spends inside the plant and the metal held there as stock, with every number linked to its source and every gap marked as a gap.
The briefCredit · unclaimed
Open 01 · the brief
Separation data beyond one solvent
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open, cited dataset of how well each published extractant (a solvent chemical that picks up a metal) takes gold and each platinum-group metal (platinum, palladium, rhodium, iridium, ruthenium and osmium) out of the chloride solutions refiners use, with the conditions of every measurement, and a model trained on it that predicts extraction under untested conditions.
Why it is still open
Refiners keep their recipes private, and open data are thin. In July 2026 Guillaume Zante posted on ChemRxiv a set of more than 18,000 data points for 72 elements extracted with a single solvent, tributyl phosphate, trained machine-learning models on it, and proposed a proof-of-concept pipeline of language models to pull such data from papers, including numbers shown only in figures. He names building reliable datasets from the literature as the main limit. Searches of OpenAlex, Zenodo and GitHub found no comparable set for the extractants used to separate the platinum-group metals.
Built with what exists
Open-access papers and patents; Zante's method as a template; a document-reading model such as Meta's Muse Ember to pull values and conditions (see the Note "The jobs agents repeat"); a second model that flags doubtful readings for a person to check (see the Note "Seven more flagged"); WebPlotDigitizer for values printed only in graphs; scikit-learn for the Extra Trees model that Zante found most accurate.
First step
Enter by hand the data from 20 papers on taking palladium and platinum out of hydrochloric acid, to fix the columns and time how long one paper takes.
Done looks like
At least 5,000 data points covering gold and all six platinum-group metals with at least ten extractants, each linked to its paper, page and figure; a hand check of a random 5% showing at least 95% correct; and the model's error on a held-out set of papers published.
Open 02 · the brief
Refining routes side by side
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table that compares, per kilogram of refined gold, the main routes (Miller chlorination, Wohlwill electrolysis, dissolving in aqua regia, solvent extraction such as Minataur, and the room-temperature chemistry now used on circuit boards) on energy, chemicals used, waste made, purity reached, days the metal spends inside the plant and the metal held there as stock, with every number linked to its source and every gap marked as a gap.
Why it is still open
Published figures cover only parts. Benjamin Fritz, Carin Aichele and Mario Schmidt of Pforzheim University of Applied Sciences collected plant data for the aqua regia route from German refineries in 2020, noting that common life-cycle databases held no data on recycling such gold; Fritz's 2026 dissertation at the Karlsruhe Institute of Technology compares mining, small-scale mining and recycling, not the refining steps against each other. Reviews describe the routes in words: SGS's 2002 review gives up to six months for rhodium in the classical route but no common table.
Built with what exists
Papers, patents and refiners' environmental and annual reports; openLCA, free life-cycle assessment software; the inventories Fritz and colleagues published; an agent that cites the page behind each number, built with a kit such as Cursor's Agent SDK (see the Note "Findings that cite notes").
First step
Fill the aqua regia row from Fritz and colleagues' published inventories, and list for each other route the three papers most likely to hold its numbers.
Done looks like
Five routes by eight measures, at least 70% of the 40 cells filled from primary sources and cited, and the table sent to at least three refiners for comment, with their replies published.