High-Purity Alumina: Sodium Decides the Sale

High purity alumina is aluminum oxide refined to at least 99.99%, which the trade calls 4N. That number is the least informative column on the specification sheet.

A 4N grade carries a total impurity limit of roughly 100 parts per million; 5N carries 10 and 6N carries 1. Ordinary smelter-grade alumina, the Bayer process material used for refractories and spark plugs, runs at 99.6% to 99.9%. Material above that gap is sold as high purity alumina, or HPA, and the phrase covers products that behave nothing alike.

The practical consequence is narrow but consequential. Two suppliers can both quote 4N and be selling materials that cannot substitute for each other, because the specification that decides the purchase sits outside the headline purity number: crystal form, particle size, bulk density, and the sodium limit. The rest of this guide is about those columns.

What high purity alumina means in practice

The definition is a floor, not a standard. Altech’s 2020 market white paper states it plainly: HPA is aluminum oxide with a minimum purity of 99.99%, defined as 4N, equivalent to 100 ppm of total impurities. The grades are defined by impurity budgets, and the document gives all three:

Producing it means escaping the Bayer process rather than improving it. Bayer alumina is made by digesting bauxite in sodium hydroxide, and sodium is precisely the element that ruins the applications HPA exists for. The industrial routes listed by Sumitomo Chemical in its own R&D paper include thermal decomposition of ammonium alum, thermal decomposition of ammonium aluminum carbonate hydroxide, underwater spark discharge, vapour-phase oxidation, and hydrolysis of aluminum alkoxide. CRU’s 2018 commentary lists the alkoxide route, a modified Bayer route using sodium hydroxide, and choline hydrolysis.

The alkoxide route dominates commercial supply, and its logic explains the cost structure. High purity aluminum metal is dissolved in alcohol to form an alkoxide, the alkoxide is purified by distillation, hydrolysed to a hydrated alumina, and calcined into the final crystal structure. Sumitomo has run this process since 1981, starting with a 250 tonne per year plant and reaching 1,500 tonnes per year by 2004.

That sequence is worth pausing on, because it is circular in an expensive way: the plant first purifies aluminum, then converts the purified metal back into an oxide. The duplication is part of why alkoxide HPA is costly.

4N is a floor, not a specification

The clearest evidence that grade is a floor comes from a manufacturer’s own grade table rather than from a market report.

Sumitomo’s 2007 paper has two grade tables. Table 1 lists seven high purity alumina grades; six are shown here, five of them above 99.99% purity. They are not interchangeable:

Read the purity column: it barely separates anything. Read the others and you have six different materials. Surface area spans a factor of roughly forty. Crystal form runs across three phases. Sodium differs by more than threefold between the standard and the fine grades.

The crystal form is not a labelling choice. α-alumina is the stable high-temperature phase; hydrated alumina such as boehmite converts to it through γ, δ and θ intermediates, and that transition requires temperatures above 1,200 °C because the oxygen packing has to rearrange from cubic closest-packed to hexagonal closest-packed. Nucleation of the α phase is the rate-determining step, and because nuclei form sparsely, grain growth after nucleation is abrupt, producing dendritic α particles of micron size unless the process is controlled.

Control here means seeding with α crystals, managing water vapour pressure in the calcination atmosphere, and limiting impurities, which affect the transition temperature. It also means deagglomeration afterwards, since the calcined product arrives as agglomerates that must be milled before it has a usable particle size distribution.

Sodium, phase and particle size are the three numbers that appear on a purchase order and not on a certificate of purity.

Why sodium decides more than the headline number

If 4N is a floor, the interesting question is why anyone would mislabel it. The answer is that the cheap route is unlikely to reach it, and the label is hard to audit.

Three routes can be compared on one variable, which is where the sodium ends up:

A modified Bayer process keeps the inexpensive bauxite-derived feedstock, but it retains elevated sodium in the alumina crystal structure. Sodium is exactly the impurity that damages sapphire growth and separator performance. Altech’s white paper puts the commercial consequence bluntly: a product from that route is unlikely to be genuine 4N quality.

Alkoxide processing avoids the problem structurally. Distillation purifies the alkoxide before hydrolysis, so the separation happens before the alumina exists. The extra cost does not buy a better number on the certificate. It buys a different impurity profile.

Verifying the claim is the weak link. As the same white paper notes, 4N HPA is difficult to assay: detection tolerances are low and test methods vary across the industry. A buyer comparing two certificates may be comparing two measurement systems. That is a technical fact about the material, and it explains how mislabelled product can persist in a market without anyone intending fraud.

Sapphire substrate and separator coating want opposite powders

The two largest applications, sapphire growth and separator coating, consume the same chemistry in incompatible physical forms.

Sapphire growth takes beads or pellets. A sapphire boule is pulled from a melt above 2,000 °C, and the feedstock has constraints that have nothing to do with purity. Absorbed water oxidises the molybdenum crucible at that temperature. In continuous feeding, particles must not fuse to each other or they clog the apparatus. And bulk density determines how much material fits into the furnace, which sets the productivity of the whole run. Sumitomo’s sapphire grade is a spherical α-alumina of roughly 2 mm, and the company later developed a high-density version at 2.0 g/cm³ by improving particle density and packing.

That density is a commercial specification in its own right. Two alumina powders at identical purity, one at 1.0 and one at 2.0 g/cm³, produce different output from the same furnace.

Then the boule is cut into LED sapphire, the wafer substrate for GaN-based LEDs, where alumina’s value is a close lattice match to the gallium nitride grown on top of it and heat resistance at growth temperature. The older Verneuil process gave poor crystallinity; edge-defined film-fed growth was developed to get both crystallinity and industrial throughput.

Battery separator coating takes powder, and wants the opposite properties. Here high-purity alumina powder is deposited as a 3 to 5 µm layer on a polyolefin separator, at 40 to 120 grams per kilowatt-hour of cell capacity, with a mean particle size below 1 µm so the coating stays even and the separator stays permeable. The material at this point is an aluminum oxide powder, and its particle size distribution is critical to the coating line. The point is thermal: a plain polyolefin separator transitions at 135 to 165 °C, while a ceramic-coated separator withstands above 220 °C, which is past the roughly 200 °C at which cells typically run away.

Sapphire accounted for about 78% of 4N demand in 2018, with separator coating at 2,030 tonnes. Both are growing from the same base. An alumina plant tooled for 2 mm beads would need a different finishing line to serve the coating customer. The other product families in this chemistry, from flame-retardant hydrate to ceramic grades, are set out in what alumina is used for.

What high purity alumina costs to make, and why the cheap route is not free

Four cost figures circulate for high purity alumina, and they describe four different things. They should be read separately.

Altech’s white paper estimates cash cost at about US$15 per kilogram for the alkoxide route. CRU reports that Nippon Light Metal, working from alumina hydrate rather than purified metal, ran at under US$10,000 per tonne in 2018. The same CRU commentary cites a 2013 presentation by a major Chinese producer giving a total cost of US$12 per kilogram, with energy at 35%, raw materials at 26% and labour at 22%. Altech, which is developing a kaolin-leach plant, states its own projected cash cost at US$8.6 per kilogram.

That last number belongs to a company selling the route, and should be read that way. The structural argument behind it is sound, though: leaching aluminous clay and crystallising the product avoids purifying aluminum only to oxidise it again.

Prices sit above these costs. Altech’s review of CRU pricing puts the 4N range at US$15 to US$100 per kilogram, with a 2018 average for 4N and above of about US$24. Sumitomo material averaged around US$28, with a higher band near US$35 likely to apply to 4N5 and 5N specifications. Product claiming 4N but assaying at 3N tended to trade below US$15.

The substitution story matters more than the price spread. Separator makers have already moved some volume to boehmite and 3N alumina, which carry roughly 5,000 and 1,000 ppm of impurities respectively, an order of magnitude or two above 4N. Those coatings are used in lower-density cells. For high-density automotive cells the impurity load is the thing being avoided, which is why the 4N floor survives there despite the cost.

China’s share is a tonnage fact, not a trust fact

Two statements about China circulate and appear to conflict.

The first: in 2016, global high purity alumina demand was 33,500 tonnes, supply was in surplus, and roughly 70% of it could be sourced domestically in China, according to CRU. Competition at the low-grade end was fierce and drove prices down. From 2017, environmental inspections raised Chinese production costs and some facilities were ordered to close.

The second: China barely exports HPA to Japan and South Korea, and overseas buyers do not trust Chinese supply. That statement reaches English readers through Altech’s white paper, which relays it from CRU. Altech is a prospective competitor to Chinese producers, so the claim arrives with an interest attached and should be handled as an attributed report rather than an established fact.

Both can be true at once, and the reason they can is the assay problem described earlier. A producer can hold a large share of tonnage while remaining absent from the export markets that pay for consistency, because those markets are buying a verified impurity profile rather than a purity percentage. Capacity is not the binding constraint. Verification is.

That also explains the shape of the capacity table. Altech’s 2020 projection put Hebei Pengda at 7,000 tonnes per year, of which 5,000 was reported as 4N and 2,000 as near-5N, against Sumitomo at 4,800 tonnes across Japan and South Korea, Sasol at 3,400, Baikowski at 2,100 and Nippon Light Metal at 1,000. Those are 2020 projections from a single document, not audited current capacities, and they should be read with that date attached.

What to specify before comparing two quotes

The useful output of all this is a short list of questions, each of which closes a gap that a purity number leaves open.

Name the application first. Sapphire feed and separator coating are different products with different acceptance criteria, and a quote that does not name the use cannot be evaluated.

Ask for the crystal form. α, θ and γ are not interchangeable, and the phase determines how the powder behaves in the customer’s process.

Ask for the sodium limit specifically, not just total impurities. Sodium is the element the modified Bayer route typically retains, and the one that damages both sapphire and cell performance.

Ask for particle size distribution or bulk density, whichever the application turns on. For sapphire, density governs furnace throughput; for coating, the sub-micron distribution governs layer uniformity.

Ask how purity was measured. Total impurity by ICP, a chosen element panel, and a supplier’s internal method are three different claims, and at 4N the difference between them is the difference between passing and failing.

Ask which route, and treat the answer as a statement about impurity profile rather than quality. Alkoxide, modified Bayer, hydrate and kaolin routes remove different things well; none of them is uniformly better.

The narrow conclusion is this. High purity alumina is a family of products sharing one floor, differentiated by crystal form, physical size and a single troublesome element. The grade tells you the floor. Everything that decides whether the material works sits above it.

Methodology

This guide uses one manufacturer’s R&D paper, one industry white paper whose author is a prospective producer, and one consultancy commentary. The Sumitomo paper is dated 2007 and its capacity figures are as of 2004. The Altech white paper is dated March 2020 and its market forecasts are not used here; its statements about Chinese supply are attributed and its commercial interest is stated in the text. CRU figures are 2016 to 2018 and are labelled as such. Cost figures from different routes and years are reported separately and are not combined into a range. No current market size, growth rate or current price is stated. A 2025 peer-reviewed study is listed for further reading and is not drawn on in the text.

Sources

1. Sumitomo Chemical, “Development of New High-Purity Alumina”, Sumitomo Kagaku vol. 2007-I. https://www.sumitomo-chem.co.jp/english/rd/report/files/docs/20070102_fth.pdf
2. Altech Advanced Materials, “The Global High Purity Alumina Market”, white paper, March 2020. https://www.altechadvancedmaterials.com/wp-content/uploads/2022/06/Doc-ATC-HPA-Market-White-Paper-Final-Mar-20.pdf
3. CRU Group, “High purity alumina, removing the separators between Chinese demand and overseas supply”, 2018. https://www.crugroup.com/en/communities/thought-leadership/2018/high-purity-alumina-removing-the-separators-between-chinese-demand-and-overseas-supply/
4. “Green Synthesis and Particle Size Control of High-Purity Alumina Based on Hydrolysis of Alkyl Aluminum”, Materials 18(9):2100, 2025 (listed for further reading; not drawn on in the text). https://pmc.ncbi.nlm.nih.gov/articles/PMC12073804/