How a brushed glaze becomes glass
The question this note answers is how a brushed layer becomes a hard colored glass, and why studio effects appear instead of a flat coat. The route is one high-temperature history read in five steps. The coat that goes on the ware is an aqueous mineral suspension, not a solution. Heat drives off water, decomposes carbonates and organics, and opens a silicate eutectic so the particles become a liquid that must wet the ware and stay there. On cooling, viscosity rises until that liquid freezes as an amorphous network — a glass, which is a frozen unequilibrated melt. Color is then a species problem: the same metal is a different optical object dissolved in the network, locked in a crystal, or clustered as a colloid. The textures studios name — crystals, oil-spot, copper red, crawling, rutile streaks — are the same melt physics stopped before the layer becomes a uniform glass.
1. What you brush on
A studio glaze is ground minerals in water: feldspar, silica (flint or quartz), clay, often a frit, and a colorant. The water is a vehicle. It does not dissolve the batch the way a solvent dissolves a dye. The solids remain particles, kept from settling by the clay and by whatever gum or deflocculant the recipe uses, and they are carried onto the ware as a wet film.1,2
That is the same operational distinction the pigment post drew for paint. A dye dissolves; a pigment stays a particulate phase in a binder. A glaze slip is on the pigment side of that line: you are brushing a suspension.3,4 The analogy stops there. In paint the binder dries or polymerizes around particles that keep their own crystal structure and their own refractive index. In a glaze the particles are raw material for a melt. Feldspar supplies alkali and alumina; silica is the network former; clay both suspends the slip and contributes alumina and silica; a frit is a glass already melted and ground, added so some of the flux arrives pre-reacted and less soluble. The colorant — an oxide, a carbonate, a spinel stain — is just another particle until the kiln dissolves it, or fails to.
2. What heat does
The first thing the kiln does is remove the vehicle. Free water leaves, then chemically bound water from clays and hydrates. Carbonates (whiting, dolomite, some colorant salts) and any leftover organics decompose, releasing gas that must escape while the layer is still permeable. Only after that does the chemistry that makes glass begin: a eutectic in the silicate batch melts, and the remaining solids dissolve into that liquid.1,5
Pure silica is not a studio material in the melt sense. Its liquidus sits far above what a pottery kiln can hold as a working liquid. The alkali and alkaline-earth fluxes — sodium and potassium from feldspar, calcium and magnesium from limestone and dolomite — open the silica network by creating non-bridging oxygens. Historical lead oxide is a separate heavy-metal modifier that does the same non-bridging-oxygen job; it is not an alkali or an alkaline earth. Boron, added as borax or a frit, is a different cut: \(\mathrm{B}_2\mathrm{O}_3\) is itself a low-melting network former, entering as borate units rather than by that modifier mechanism, which is how §3 classifies B. Either route drops the mixture into the kiln range, and a film a fraction of a millimeter thick can become a liquid without the ware itself melting away.5,6
That liquid has a job that a bulk glass melt does not. It must wet the ware, flow enough to heal pinholes and brush marks, and remain on vertical surfaces instead of draining to the foot. Wetting is a surface-energy problem. The melt spreads when the bare-body surface energy is at least the sum of the glaze–body interfacial energy and the glaze free-surface energy (\(\gamma_{\text{body-air}} \ge \gamma_{\text{body-glaze}} + \gamma_{\text{glaze-air}}\)). A glaze–body interface that is merely cheaper than those two free surfaces taken separately is the weaker adhesion condition, and it does not by itself guarantee a sheet. Flow is a viscosity problem. Both are set by composition and temperature, and both have to stay inside a narrow window — fluid enough to level, stiff enough to stay put.1,5
3. Hard glass from a frozen melt
On cooling, the melt does not freeze the way water freezes. There is no single temperature at which a crystal lattice appears. Viscosity rises continuously. Atomic rearrangements that would build a periodic silicate slow and then stop, and the liquid is left as an extended, non-periodic network of \(\mathrm{SiO}_4\) tetrahedra sharing corners — and, where the batch supplies them, \(\mathrm{AlO}_4\) tetrahedra and borate units that are a mixture of trigonal \(\mathrm{BO}_3\) and tetrahedral \(\mathrm{BO}_4\). That is Zachariasen’s picture of oxide glass, and it is still the useful first description of a glaze coat: the atoms are linked by forces of the same kind as in a crystal, but without periodicity or symmetry.6,7
Network formers (Si, B, P, and Al when it substitutes for Si) build that framework. Network modifiers — the alkali and alkaline-earth fluxes of §2, and historical lead — break bridging \(\mathrm{Si{-}O{-}Si}\) linkages and lower viscosity. A glaze is then rigid, not because it crystallized, but because the viscosity became so large that the melt could no longer flow on any laboratory timescale. The coat is a frozen unequilibrated liquid. Its hardness — resistance to scratch or indentation — is a property of the bonded oxide network, not of that long relaxation time alone.1,5,6
Crystallization is the competing path, not a later decoration. If the melt is held, or cooled slowly, in a composition that has a stable crystalline phase — willemite, anorthite, a silica polymorph, an iron oxide — nuclei form and grow. A “glassy” coat is the kinetic outcome in which that path lost. A crystalline coat is the outcome in which it won, locally or everywhere. A matte surface is a different claim — an optical appearance that can also come from liquid–liquid phase separation, underfiring, retained particles, or surface roughness. Nothing in the cooling step changes the chemical identity of the oxides; it changes whether they are trapped in a random network or allowed to order.5,6
4. The same metal, different optical species
Color in a glaze is not a property of “iron” or “copper” as an element. It is a property of the optical species that element becomes: the oxidation state, the ligands, the coordination geometry, and whether the metal is a dissolved ion, a constituent of a crystal, or a metal colloid. The pigment post treated the ligand field of a crystal — oxide octahedra around \(\mathrm{Cr}^{3+}\) in viridian, tetrahedral \(\mathrm{Co}^{2+}\) in cobalt blue. In a glaze the ligand is the melt itself. Oxide ions of the silicate network occupy the coordination sphere, and the ligand-field splitting \(\Delta\) of the \(d\) orbitals is set by that disordered oxygen environment rather than by a lattice site. Charge transfer, usually oxygen-to-metal, sits at higher energy and is much more intense when it falls into the visible.8–10
Iron is the clearest case of one metal, two colors. In a reducing kiln atmosphere \(\mathrm{Fe}^{2+}\) dominates; its \(d\)–\(d\) absorption, with a strong near-infrared tail into the red, leaves the familiar blue-green of celadon. In oxidation \(\mathrm{Fe}^{3+}\) dominates; ligand-field bands are weaker, but oxygen–iron charge transfer and the coordination mixture (tetrahedral and octahedral \(\mathrm{Fe}^{3+}\)) push the glaze toward amber, yellow, and brown. Controlled firings of an Arita celadon batch move the same glaze from gray-yellow in oxygen to blue-green as the atmosphere becomes more reducing and \(\mathrm{Fe}^{2+}\) replaces \(\mathrm{Fe}^{3+}\). The hue is a redox ratio in a glass, not a different recipe.9–11
Cobalt dissolves into the silicate network as \(\mathrm{Co}^{2+}\). In ordinary alkali and lime glazes it occupies a tetrahedral oxide site and absorbs in the orange–red, so the coat is blue — the same ion, and the same qualitative site symmetry, that makes cobalt-blue spinel a pigment, now without the spinel lattice. Copper as \(\mathrm{Cu}^{2+}\) is green or turquoise in the network, another ligand-field ion with oxide ligands. Under reduction the same copper leaves the network as metal. The red of a copper-red glaze is not \(\mathrm{Cu}^{2+}\) and is not cuprite as the main colorant: it is colloidal metallic copper, grown to a size that absorbs and scatters in the green, leaving transmitted and reflected red. Too much growth makes the red muddy and then black. The metal is the same; the optical object is not.8,9,12
Chromium as \(\mathrm{Cr}^{3+}\) is a green network ion, the glaze analogue of chromium oxide green, again a \(d\)–\(d\) chromophore in an oxide field. Hexavalent chromium, when it is present, is a different species and a different color. Rutile is not mainly a dissolved-ion colorant. It is a titanium ore that carries some iron; studios add it for broken color. The melt cannot keep all of that titanium in solution on cooling, and the resulting crystals and local composition gradients streak and mottle a coat that would otherwise have been flat.1,9,10
5. Effects that are the melt not finishing
The textures that look like a separate craft are the physics of §3 and §4 stopped partway.
Crystalline glazes are a supersaturated melt allowed to nucleate. A zinc-rich, low-alumina composition is taken to a fluid melt and then held where willemite, \(\alpha\)-\(\mathrm{Zn}_2\mathrm{SiO}_4\), can grow. The glaze decomposes into zinc-rich and silica-rich regions; willemite spherulites grow from the former as plate-like sheaves constrained by the thickness of the coat. The crystals are not painted on. They are the equilibrium solid the glass of §3 declined to become, grown on purpose.6,13
Oil-spot and other iron-saturate coats are unfinished iron enrichment, not a printed decoration. In some lime–alumina–silica and alkali-bearing melts a miscibility gap is established: the homogeneous liquid splits into a silica-rich melt and a modifier- and iron-rich melt, and that second liquid can freeze as droplets — the route behind Jun opalescence — or feed later crystals.14,15 That phase separation is not required for every oil-spot. Iron-oxide decomposition can release oxygen bubbles that carry iron-rich material to the surface, where magnetite and hematite crystallize, without the homogeneous melt having entered a two-liquid field. The cited Jian and Tianmu coats combine those paths in specific batches; they do not make immiscibility the only mechanism.16
Reduction is the kiln atmosphere as a reagent. A reducing fire does not “add a special color”; it changes the oxygen activity of the melt so that \(\mathrm{Fe}^{3+}/\mathrm{Fe}^{2+}\) and \(\mathrm{Cu}^{2+}/\mathrm{Cu}^{0}\) shift, which is §4 read as a function of \(p_{\mathrm{O}_2}\) rather than of recipe. Copper red is the extreme case: the ion is reduced, then the metal atoms diffuse and coarsen into colloids. Celadon is the moderate case: the ion stays dissolved, but its oxidation state, and therefore its ligand-field spectrum, changes.9,11,12
Crawling, breaking, and rutile streaks are viscosity and surface tension written onto the ware. If the raw film cracks, or if the melt does not wet the body — dust, grease, a refractory undercoat, a melt that is too stiff — the liquid pulls into islands instead of remaining a sheet. Rhodes’s studio description is the same physics: the melt parts during heating and leaves bare clay, and the glazes that do this are the viscous, high-clay, matte ones, while fluid transparent melts tend to heal the same breaks.1,5 Breaking on a rim or a ridge is the melt flowing under gravity and surface tension until the coat is thin enough to show the body. Rutile streaks, as in §4, are local crystallization and composition gradients that the melt did not homogenize before it froze.
None of these is a second subject. A crystal growing from a supersaturated melt, a second liquid nucleating in a miscibility gap, a colloid coarsening after reduction, and a melt failing to wet are the glass-formation story of §3 and the species story of §4, encountered when the layer does not finish as a uniform frozen liquid.
6. Where the model stops
This note is a mechanism sketch for a chemist who has not thrown a pot. It does not choose a firing schedule, and it does not turn a batch into a Seger formula. Those are quantitative tools; they are not what makes the brushed layer glass. It also has no generated demonstration: the route is qualitative species and kinetic path, not a computed spectrum or a formula count from a published batch. A later figure could show the same objects in two representations; this draft does not invent one.
The redox story is a kiln-average. Oxygen activity is not uniform from firebox to stack, or from the surface of a thick coat to the body interface, so \(\mathrm{Fe}^{2+}/\mathrm{Fe}^{3+}\) is a local variable. The optical-species account also stops at the dissolved ion and the simple colloid. Some oil-spot and hare’s-fur color is scattering and interference from thin crystalline films, not only ligand-field absorption by leftover iron in the glass.16 A complete color model would have to carry both.
One mechanical limit belongs here because it is the same cooling history. If the glaze’s thermal expansion does not match the body’s, the frozen coat goes into tension or compression on the way down; tension crazes the glass. That mismatch is a thermoelastic fact, not a separate aesthetic, and it is not developed here.5
A reader who knows ceramic science will see what was left qualitative — liquidus surfaces, optical basicity, the details of spinodal versus binodal separation. That is the right place to push. The claim this note can defend is only the route: suspension, melt, frozen network, optical species, unfinished equilibration.
7. Summary
What is brushed on is an aqueous suspension of ground minerals. Heat removes the water and the gases, then a fluxed silicate eutectic melts and must wet the ware. Cooling raises the viscosity until that liquid is trapped as an amorphous oxide network — a glass, which is a melt that did not crystallize. Color is the optical species the metal becomes in that network, in a crystal, or as a colloid; celadon and amber are two iron ions, copper-green and copper-red are an ion and a metal particle. The studio effects are the same history stopped short: crystals from a supersaturated melt, a second liquid or an iron-rich surface film, a redox shift in the kiln atmosphere, and a melt whose viscosity and surface tension never produced a flat sheet.