Hidden Clay Minerals Steal Magnesium Yield Before the Pidgeon Process Even Starts

Magnesium is the lightest structural metal in widespread industrial use, prized for its low density of roughly 1.74 grams per cubic centimeter and its favorable mechanical and chemical properties. It underpins die casting, aluminum alloying, steel desulfurization, and a growing range of applications in biomedicine, new energy vehicles, and aerospace. Global primary production approaches one million tonnes per year, and China dominates the supply, accounting for approximately 85 percent of world output. Nearly all of that metal is made by the Pidgeon process, a vacuum silicothermic reduction route in which magnesium oxide from calcined dolomite is reduced by ferrosilicon at temperatures above 1100 degrees Celsius, with small additions of calcium fluoride acting as a mineralizer to speed the reaction.

For decades, engineers and researchers have optimized the Pidgeon process by studying heat transfer, reaction kinetics, furnace configuration, and thermal field distribution. Yet nearly all of this work shares a hidden assumption: that calcined dolomite behaves as an ideal binary mixture of magnesium oxide and calcium oxide, produced by the complete decomposition of the mineral dolomite, CaMg(CO3)2. Under that ideal framework, the silicothermic reaction requires a one-to-one molar ratio between the two oxides, and the maximum magnesium producible is set by whichever oxide is the limiting reagent. The calcium silicate larnite, Ca2SiO4, is understood as the thermodynamic sink that forms during reduction itself, driving the equilibrium toward metallic magnesium release.

A new study published in Results in Engineering by Aldemir de Melo Sotero and colleagues at the Universidade Federal do Pará in Brazil challenges that assumption in a way that could reshape how magnesium producers evaluate their raw materials. The team investigated two natural dolostones from Brazil with sharply contrasting clay contents: the Alcântara Dolostone, drawn from a Cretaceous coastal formation in the state of Maranhão, which contains the hydrated magnesium-aluminum clay mineral palygorskite at levels of 2.7 and 11.1 weight percent in two variants, and a Silurian dolostone more than 400 million years old that is essentially pure dolomite at 99.8 weight percent. The pure sample served as a mineralogical reference, while the clay-bearing rocks exposed a yield-destroying chemistry that bulk analysis cannot see.

The problem begins with what palygorskite does when heated. This phyllosilicate loses adsorbed and channel water below about 300 degrees Celsius, dehydroxylates between roughly 400 and 500 degrees Celsius, and then undergoes structural collapse above approximately 800 degrees Celsius, releasing poorly ordered, chemically reactive silica. That temperature range overlaps almost exactly with the calcination stage of the Pidgeon process, in which dolomite decomposes into lime and periclase. The freshly released reactive silica does not wait for the reduction stage. It reacts immediately with free calcium oxide, and to a lesser extent magnesium oxide, forming calcium and magnesium silicate phases before any reducing agent is introduced. The same silicate chemistry that is productive and mechanistically necessary during reduction becomes a yield-limiting constraint when it occurs during calcination, because it permanently removes oxides from the pool available for reduction.

To quantify this sequestration, the researchers designed a full 2³ factorial experiment varying charge mass from 1 to 10 grams, calcination temperature from 900 to 1100 degrees Celsius, and residence time from 1 to 8 hours, with center-point replicates, for a total of 12 runs per sample. They characterized the raw materials and calcined products using wavelength-dispersive X-ray fluorescence, simultaneous differential scanning calorimetry and thermogravimetry, and powder X-ray diffraction coupled with Rietveld quantitative phase analysis. The Rietveld method, a full-pattern fitting approach that iteratively refines a calculated diffraction profile against the measured pattern, delivered weight fractions for every crystalline phase with refinement agreement indices showing weighted profile R-factors between 5.4 and 9.0 percent and goodness-of-fit values close to unity.

The phase assemblages revealed a far more complex silicate chemistry than previously reported for clay-bearing dolostones. In the moderately clay-rich sample, larnite formed under every experimental condition, reaching weight fractions up to 43.8 percent, while diopside, MgCaSi2O6, gehlenite, Ca2Al2SiO7, and a magnesium silicate with Mg3Si2O9 stoichiometry appeared at higher temperatures and longer residence times. The gehlenite confirmed that aluminum released from palygorskite structural collapse also partitions into silicate phases. The Mg3Si2O9 phase, which is thermodynamically unstable at calcination temperatures, was interpreted as a retrogressive product crystallizing during cooling, when freshly calcined periclase reacts with residual amorphous silica. In the pure Silurian reference sample, by contrast, the calcined products consisted exclusively of periclase and lime, with no silicate phases detected in any run, establishing a clean baseline.

The team then built a Rietveld-based stoichiometric model that converts crystalline phase weight fractions into molar quantities per 100 grams of calcined material, performs elemental balances to separate reducible from sequestered oxide fractions, and simulates the silicothermic reduction with commercial 75-grade ferrosilicon. The results were striking. In the pure dolostone, available calcium oxide and magnesium oxide appeared in nearly equimolar proportions, matching the ideal Pidgeon stoichiometry. In the low-clay sample, available calcium oxide ranged from 22.5 to 35.5 weight percent, and theoretical magnesium yield ranged from 0.401 to 0.626 moles per 100 grams. In the high-clay sample, available calcium oxide collapsed to as little as 3.0 weight percent under severe conditions, leaving more than 90 percent of the periclase without its calcium counterpart and driving theoretical magnesium yield down to 0.053 moles per 100 grams, nearly an order of magnitude below the pure reference. Bulk oxide chemistry would have overestimated recoverable magnesium by about 42 percent for the low-clay sample and by up to roughly 1200 percent for the clay-rich sample under the harshest conditions.

Statistical analysis of the factorial design added a second layer of insight. For the low-clay sample, residence time was the only significant factor for both magnesium yield and total silicate formation, with opposing effects that confirmed the mineralogical trade-off: longer heating simultaneously increases silicate formation and decreases magnesium availability. For the clay-rich sample, temperature became the dominant factor by a wide margin, with residence time second and a significant temperature-by-time interaction reflecting the self-limiting nature of silicate growth once available calcium oxide is exhausted. Composite desirability optimization identified low temperature, short residence time, and low charge mass as the optimal calcination window for both samples, but the operating window was far narrower for the clay-rich rock. Even at their respective optima, the clay-rich sample yielded about 12 percent less theoretical magnesium and generated about 24 percent more slag than the low-clay sample, a penalty attributable to raw material mineralogy that process adjustment can mitigate but never eliminate.

The practical implications reach directly into industrial operations. The authors argue that ferrosilicon dosage should be calculated on the basis of effectively available calcium oxide rather than total periclase content, since dosing against total magnesium oxide in a clay-bearing feedstock would produce a substantial and costly excess of reducing agent. They also recommend routine X-ray diffraction characterization of dolomite raw materials as an integral component of quality control, replacing reliance on bulk oxide chemistry alone, because two feedstocks with identical bulk compositions can differ dramatically in reducible magnesium depending on how much of their oxide inventory is already locked into silicates. The mechanistic framework extends beyond palygorskite to other phyllosilicate impurities such as kaolinite, smectites, chlorites, and illite, each of which releases reactive silica at calcination temperatures. The theoretical yields reported here represent a mineralogical ceiling on recoverable magnesium, and the authors note that experimental validation through representative silicothermic reduction trials is planned as a dedicated follow-up study. For an industry that produces nearly a million tonnes of magnesium a year under thin margins, the message is clear: the clay hidden in the rock can quietly consume the metal before the furnace ever gets its chance to make it.

Subject of Research: Quantitative assessment of clay-derived reactive silica sequestration of CaO and MgO during dolostone calcination and its impact on theoretical magnesium yield in the Pidgeon silicothermic reduction process

Article Title: Magnesium yield estimation based on Rietveld refinement of clay-bearing dolostone: implications for silicothermic reduction

Article References: Sotero, A. D. M., Oliveira, R. L. R., Brito, C. E. C., Albuquerque, A. R. L., Paz, S. P. A., & Angélica, R. S. (2026). Magnesium yield estimation based on Rietveld refinement of clay-bearing dolostone: implications for silicothermic reduction. Results in Engineering, 32, Article 113155. https://doi.org/10.1016/j.rineng.2026.113155

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113155

Keywords: magnesium, Pidgeon process, silicothermic reduction, dolostone, palygorskite, Rietveld refinement, calcination, larnite, reactive silica, X-ray diffraction, ferrosilicon, mineralogy