Where polishing actually happens in flat glass
Most people outside the industry assume that window glass is polished. It is not — not any more. When Pilkington introduced the float process in 1959, it eliminated the grinding and polishing of plate glass faces altogether: both surfaces of float glass leave the tin bath fire-finished [1].
The polishing that remains in flat glass processing is concentrated exactly where the float process cannot help: edges, mirror and decorative bevels, crystal facets and the repair of surface scratches. These are the surfaces the end customer touches, inspects and judges. Producers themselves describe their rare-earth polishing ranges in these terms — plate glass, crystal glass, mirror production, and edge and facet polishing [2] — and the final step of bevelling has long been a felt-covered wheel running a slurry of water and optical polishing powder [3].
It is not a marginal activity. A peer-reviewed review published in 2018 estimated that of roughly 40,000 tonnes of rare-earth oxides consumed by the glass industry, about 16,000 tonnes were used for polishing [4]. Yet in most plants the polishing powder is bought as a commodity line item. This article explains why that is a costly simplification — and what to measure instead.
1. Why cerium oxide polishes glass: a chemical tooth, not just an abrasive
The reference explanation was published by Lee Cook in 1990 in the Journal of Non-Crystalline Solids [5]. Cook showed that glass polishing is not simple scratching at a smaller scale. The glass surface hydrates in water; the polishing particle bonds chemically with that hydrated layer; and material is removed as the particle moves on. Hoshino and co-workers later confirmed the mechanism on silica: CeO₂ particles form Si–O–Ce bonds with the surface, and silica is torn away in clusters rather than dissolved atom by atom [6].
Two practical consequences follow for any glass shop:
- Chemistry matters as much as grain size. Water quality, slurry pH, glass composition and the surface state of the particles all influence removal. Two powders with the same nominal particle size can polish at different speeds.
- “Cerium oxide” is not a specification. Commercial rare-earth polishing powders span an extremely wide composition range: the 2018 review cited above tabulates products from about 30 to 99.9 wt% CeO₂, with 0 to 35 wt% La₂O₃, and fluorine additions of about 3 to 9 mass % in some formulations [4]. Two bags with the same label can be very different products.
2. Preston’s law: why a powder change shows up as a machine problem
In 1927, F. W. Preston formalised what every polisher knows intuitively: removal rate is proportional to pressure multiplied by relative velocity, times a coefficient that captures everything else [7]. Written simply: removal rate = kₚ × pressure × velocity.
On a straight-line edger or a bevelling machine, pressure and velocity are set by the machine and the operator. The Preston coefficient kₚ is where the powder, the felt and the state of the slurry live. When a new powder lot behaves differently, or when a recirculated slurry ages, kₚ moves — and the line compensates with lower feed speed, more pressure or another pass. The symptom appears on the machine; the cause often sits in the slurry tank.
OPERATIONAL READING
When edge quality or feed speed drifts without a mechanical cause, check the slurry before the machine: concentration, pH, temperature, age of the recirculated batch and the powder lot in use. These are the variables inside kₚ.
3. The number that decides scratches is not the average particle size
Procurement specifications usually quote a median particle size (d50). The most rigorous public research on polishing defects — carried out at Lawrence Livermore National Laboratory (LLNL) on fused silica — shows that scratches and roughness are governed by the largest particles, not the average ones:
- When “rogue” particles were deliberately added to a 0.5 µm ceria slurry, the scratch density increased exponentially with rogue particle size at a fixed rogue concentration. Dried ceria agglomerates were among the rogue sources tested [8].
- In grinding, sub-surface damage depth began to increase at rogue particle fractions below 10⁻⁴ — fewer than one particle in ten thousand [8].
- In pad polishing, final roughness correlated quantitatively with the shape of the exponential tail of the particle size distribution; average-sized particles had negligible influence [9].
These studies concern precision optics, but the physics is identical on a mirror bevel or a polished edge: a visible scratch comes from a few oversized particles or agglomerates, not from the median. Two operational conclusions follow. First, ask for the tail of the distribution (d90, d99 or a maximum size), not only d50. Second, treat dried slurry crusts in pipes, tanks and splash guards as a scratch source — LLNL’s work identifies dried ceria agglomerates as exactly that [8].
4. Slurry ages — and it ages silently
A polishing powder does not keep its performance in a recirculating system. The 2018 review notes that glass polishing powder “loses its polishing ability gradually during usage” because other compounds accumulate on it and because its particle size distribution changes [4]. In practice this means that a stable process depends as much on slurry management — density, pH, renewal or bleed-off, filtration — as on the powder bought.
Research laboratories treat this with discipline: in LLNL’s published convergent polishing protocol, the ceria slurry is prepared to a defined density, stabilised at pH 9.5, re-adjusted every 24 hours and filtered to remove large particles, which reduced scratching [10]. A flat glass line does not need laboratory tolerances, but it does need a written slurry routine.
5. The wheel is half of the system
The felt or synthetic wheel holds the slurry against the glass. Its structure, hardness and wear determine how much powder is present at the contact and how uniformly. Wheel makers publish very different service lives depending on construction — one supplier quotes roughly 40 run hours for basic spiral wheels up to more than 120 hours for advanced versions [11]. A powder must therefore be qualified with the wheel it will run on. A trial that changes both at once cannot tell you which one made the difference.
6. Supply chain: separating real risks from noise
Rare earths make headlines, and polishing buyers are often told that supply is at risk. The facts are more nuanced:
| Fact | What it means for a glass shop | Source |
| China mined about 270,000 t of the 390,000 t world rare-earth output in 2025 (≈ 69 %). | Upstream concentration is real. | USGS 2026 [12] |
| China accounts for about 91 % of global rare-earth separation and refining. | Most polishing oxide passes through Chinese refining, whoever the brand owner is. | IEA, Oct. 2025 [13] |
| April 2025 Chinese export controls cover samarium, gadolinium, terbium, dysprosium, lutetium, scandium and yttrium. | Cerium and lanthanum — the polishing elements — are not on the list. | USGS 2026 [12] |
| The broader October 2025 measures are suspended until 10 November 2026. | A date to watch, not a current restriction. | CIRS, Nov. 2025 [14] |
| Cerium oxide (99.5 %) averaged US$1.21/kg in 2024 and US$1.71/kg in 2025. | +41 % in one year on the raw oxide: price volatility, not availability, is the operational risk. | USGS 2026 [12] |
Note that the raw oxide price is not the price of a polishing powder: grading, calcination, particle engineering, packaging and documentation add value — and that is precisely where products differ. The defensible supply questions are therefore about price transparency, lot-to-lot consistency and documentation, not about whether cerium will be available.
7. How to evaluate a polishing compound — a practitioner’s checklist
| Ask for / measure | Why | Basis |
| Full particle size distribution, including d90 and d99 (or maximum size), for two recent lots | Scratches come from the tail, not the median | LLNL [8], [9] |
| Composition: CeO₂/total rare-earth oxide ratio, La₂O₃, fluorine | Products labelled “cerium oxide” range from ~30 to 99.9 % CeO₂ | Borra et al. [4] |
| Lot certificate of analysis and lot traceability | Lets you link a process drift to a specific lot | Good practice |
| Safety data sheet in the format and language of your country; transport classification | Legal obligation for the importer and user; avoids customs delays | Local regulation |
| Trial on your machine, your wheel, your glass — one variable changed at a time | Only a like-for-like trial isolates the powder’s effect (Preston’s kₚ) | Preston [7] |
| Record: feed speed achieved, visual quality under a fixed inspection standard, powder consumption, wheel hours, rework | These are the inputs of cost per good part | — |
8. The metric that matters: cost per good part
Price per kilogram appears once in the cost of a polished part. Process stability appears in every other term:
COST PER GOOD PART
(powder consumed × price) + (wheel cost ÷ wheel life × run time) + (machine and labour time × cycle time) + (rework rate × cost of one rework) — all divided by the number of good parts produced.
A cheaper powder that forces slower feed, shorter wheel life or more rework can easily cost more per good part than a more expensive one. Measure the formula above on your own line before and after any change of supplier. It is the only comparison that settles the argument.
Key takeaways
- Since the float process, flat glass polishing is about edges, bevels, crystal facets and repair — the surfaces customers judge [1][2].
- Cerium oxide polishes by a chemical-mechanical mechanism; chemistry and slurry state matter as much as grain size [5][6].
- Scratches are driven by the largest particles and agglomerates, not by d50 — specify the tail [8][9].
- Slurry ages in recirculation; stability requires a written slurry routine [4][10].
- Cerium and lanthanum are not under Chinese export control; the real supply risks are price volatility, consistency and documentation [12][13][14].
- Compare suppliers on cost per good part, measured on your own line.
Working with Illumine Consultants
Illumine Consultants is an independent supplier of polishing compounds (CERITOP & LUCERIS ranges). We start with your process, not with a catalogue: your machine, wheel, glass and the defect you want to eliminate. We then propose a grade and a 1 kg evaluation sample, with a trial method designed to isolate the powder’s effect, followed by a technical debrief of your results.
About the author. Olivier M. is the founder of Illumine Consultants. Before founding the company, he worked in rare-earth polishing solutions at Solvay, supporting glass processors, mould shops and optical manufacturers. Illumine Consultants supplies cerium- and lanthanum-based polishing compounds with an application-first approach and a documented supply chain.
This article relies exclusively on public sources, listed below. Brand names are cited for descriptive purposes only; Illumine Consultants is not affiliated with the companies mentioned.
References
[1] Encyclopaedia Britannica, “Float-glass method”. britannica.com/technology/float-glass-method (accessed 28 Sept 2026).
[2] Treibacher Industrie AG, “Auerpol® – polishing agents”, product page. treibacher.com/en/products/auerpol-polishing-agents/ (accessed 28 Sept 2026). Brand cited for descriptive purposes only; no affiliation.
[3] “Beveled glass”, Wikipedia (description of traditional bevelling steps). en.wikipedia.org/wiki/Beveled_glass (accessed 28 Sept 2026).
[4] C. R. Borra, T. J. H. Vlugt, Y. Yang, S. E. Offerman, “Recovery of Cerium from Glass Polishing Waste: A Critical Review”, Metals 8(10), 801 (2018). doi:10.3390/met8100801.
[5] L. M. Cook, “Chemical processes in glass polishing”, Journal of Non-Crystalline Solids 120, 152–171 (1990). doi:10.1016/0022-3093(90)90200-6.
[6] T. Hoshino, Y. Kurata, Y. Terasaki, K. Susa, “Mechanism of polishing of SiO₂ films by CeO₂ particles”, Journal of Non-Crystalline Solids 283, 129–136 (2001). doi:10.1016/S0022-3093(01)00364-7.
[7] F. W. Preston, “The Theory and Design of Plate Glass Polishing Machines”, Journal of the Society of Glass Technology 11, 214–256 (1927).
[8] T. Suratwala, R. Steele, M. D. Feit, L. Wong, P. Miller, J. Menapace, P. Davis, “Effect of rogue particles on the sub-surface damage of fused silica during grinding/polishing”, Journal of Non-Crystalline Solids 354(18), 2023–2037 (2008). doi:10.1016/j.jnoncrysol.2007.11.015.
[9] T. Suratwala et al., “Microscopic Removal Function and the Relationship Between Slurry Particle Size Distribution and Workpiece Roughness During Pad Polishing”, Journal of the American Ceramic Society (2014). doi:10.1111/jace.12631.
[10] T. Suratwala et al., “Convergent Polishing: A Simple, Rapid, Full Aperture Polishing Process of High Quality Optical Flats & Spheres”, Journal of Visualized Experiments (2014). doi:10.3791/51965.
[11] Spartan Felt, “Glass Polishing – Solid and Spiral Mirror Ring for Glass Polishing”. spartanfelt.com/glass-polishing-wheels/ (accessed 28 Sept 2026). Manufacturer data.
[12] U.S. Geological Survey, Mineral Commodity Summaries 2026 — Rare Earths. pubs.usgs.gov/periodicals/mcs2026/mcs2026-rare-earths.pdf (accessed 28 Sept 2026).
[13] International Energy Agency, “With new export controls on critical minerals, supply concentration risks become reality”, commentary, 23 October 2025. iea.org.
[14] CIRS Group, “China Temporarily Suspends Export Controls on Key Raw Materials, Including Rare Earths, Lithium Batteries, and Diamonds”, November 2025. cirs-group.com (accessed 28 Sept 2026).
Working on flat glass edges, bevels or crystal?
See how we approach it — and request a 1 kg evaluation sample with a trial method.
See our cerium oxide polishing for flat glass → Request a sample