Patreon for glass fusing creators — 2026 edition

Silicate glass chemistry COE mechanism network formers and modifiers, devitrification cristobalite nucleation crash-cool prevention and borax surface spray, color chemistry transition metal oxides copper cobalt iron manganese selenium-cadmium and gold colloidal ruby, dichroic glass thin film interference vacuum deposition and documentation, frit particle size and fusing behavior, firing schedule mass effect and segment programming, and the Apple Tax.

Glass fusing Patreons retain when they deliver the materials science and glass chemistry layer that kiln-opening reveals and “use COE 90 only” warnings structurally compress away. Here is that layer: why silicate glass has a coefficient of thermal expansion in the first place (SiO₂ network former held together by bridging oxygen bonds, network-modifying oxides Na₂O and K₂O breaking those bonds and increasing thermal expansion, stabilizing oxides CaO and BaO regulating viscosity), what COE actually predicts about cracking and where the prediction fails (how push-rod dilatometry measures expansion, why glasses with the same nominal COE from different manufacturers can still stress-crack in thick assemblies), devitrification at the crystallographic level (cristobalite crystal nucleation on silica-rich surfaces in the 595–704°C range, what crash cooling actually does to the nucleation kinetics, why borax surface spray dissolves the crystalline surface layer before it can grow), color chemistry in glass as d-orbital electronic transitions (copper Cu²⁺ blue-green in oxidation vs Cu⁺ red in reduction and why most studio kilns cannot reach true reduction, cobalt CoO blue consistent in any atmosphere, iron Fe³⁺ amber in oxidation vs Fe²⁺ blue-green in reduction, selenium-cadmium sulfide orange-red pigment and its thermal stability ceiling), dichroic glass manufacturing by physical vapor deposition (how film thickness determines reflected color through thin-film interference, what the documented film composition and base glass type allow patrons to reproduce), frit particle size effects on flow and surface texture in firing, how total glass mass shifts firing schedule requirements independently of piece size, and exactly how much the Apple Tax costs a glass fusing creator earning $200–$500 per month from a 55–80% iOS audience.

1. Silicate glass chemistry — why glass has a COE at all

Glass is an amorphous solid: a supercooled liquid that has cooled too quickly to form the ordered crystal lattice that characterizes minerals like quartz or feldspar. The structural backbone of soda-lime silicate glass — the family used in Bullseye (COE 90) and Spectrum/Oceanside (COE 96) fusing glass — is a three-dimensional network of SiO₄ tetrahedra linked by bridging oxygen atoms. Each silicon atom bonds to four oxygen atoms; each oxygen atom bridges two silicon atoms. This continuous covalent network gives glass high viscosity near its melting point (preventing rapid crystallization) and high bond strength at room temperature (explaining hardness and brittleness).

Network-forming oxides are those that participate in this tetrahedral structure: SiO₂ (primary, ~70% by weight in soda-lime glass), B₂O₃ (present in borosilicate glass and some Bullseye formulations, forms BO₃ triangles or BO₄ tetrahedra depending on composition), Al₂O₃ (forms AlO₄ tetrahedra when charge-balanced by alkali oxides, increasing durability). Network-modifying oxides break bridging oxygen bonds: Na₂O and K₂O donate their oxygen atoms to the network, converting bridging oxygens to non-bridging oxygens (each Na⁺ or K⁺ ion sits in an interstitial site and is charge-balanced by one non-bridging oxygen). Non-bridging oxygens are more mobile than bridging oxygens, which is why adding soda (Na₂O) lowers glass viscosity at a given temperature — it disrupts the rigid network. It also raises the coefficient of thermal expansion: the non-bridging oxygens allow greater vibrational amplitude of the network atoms with increasing temperature, meaning the glass expands more per degree than pure silica glass (which is almost entirely bridging oxygen and has a very low COE of approximately 0.55 × 10⁻⁶/°C compared to soda-lime’s ~9 × 10⁻⁶/°C).

Stabilizing oxides CaO, MgO, and BaO occupy modifier positions but introduce larger, higher-field-strength cations that stiffen the network around them, raising the glass transition temperature (T_g) and improving chemical durability without as much expansion increase as pure alkali modifiers. Bullseye glass uses a barium-containing formulation that contributes to its higher T_g (~520°C annealing point) compared to soda-lime window glass (~510°C). The specific oxide recipe determines exactly where the annealing point, strain point, softening point, and working point fall on the temperature scale — and these four temperatures define the kiln schedule. The coefficient of thermal expansion for any glass composition is the integral of the expansion rate across the temperature range from room temperature to the strain point, measured by push-rod dilatometry: a glass sample rod is heated at a controlled rate while a push-rod records dimensional change in real time, and the slope of the expansion-vs-temperature curve at each temperature gives the instantaneous COE. The reported single-number COE (e.g., 90 × 10⁻⁷/°C, often written as COE 90) is the average over the room-temperature-to-strain-point range.

The COE compatibility problem: if two glasses with different COEs are fused together, they behave as a single piece during the viscous fusing stage but become bonded at the strain point on cooling. Below the strain point, the glass is essentially rigid — it cannot flow to relieve stress. As the piece cools further, the higher-COE glass contracts more than the lower-COE glass per degree, and because the two are bonded, each exerts a tensile force on the other. If the stress exceeds the glass’s tensile strength (typically 25–65 MPa for soda-lime glass, much lower than its compressive strength of 1,000–2,000 MPa), the piece cracks. The crack may appear immediately on kiln opening, or hours later as residual stress slowly finds a fracture path — a phenomenon called “delayed cracking” or “spontaneous breakage.” The tolerance between glasses within a certified-compatible family: Bullseye certifies glasses to within approximately ±1 unit of COE 90 for standard compatibility, measured against a reference test. Cross-manufacturer COE 90 glasses may differ by 1–3 units, which is usually fine for thin two-layer assemblies but can produce stress in assemblies of 6mm or more total thickness because the absolute stress scales with the thickness of the assembly.

2. Devitrification — cristobalite nucleation and how crash cooling prevents it

Devitrification is the partial crystallization of glass during firing: a normally amorphous glass surface develops a matte white or iridescent haze caused by crystals of cristobalite (a polymorph of SiO₂) and other silicate minerals nucleating and growing on the glass surface. It is the most common quality problem in glass fusing, most visible on pieces that have been fired multiple times, and the mechanism explains both why it happens and exactly what prevents it.

Crystal nucleation requires two conditions: thermodynamic driving force (the crystalline phase must be lower in free energy than the amorphous phase at that temperature) and kinetic mobility (atoms must be able to rearrange from amorphous to crystalline order). In a normal cooling curve, glass passes through the nucleation window — the temperature range where both conditions are met — too quickly for significant crystal growth. The critical temperature range for cristobalite nucleation in soda-lime silicate glass is approximately 595–704°C: above 704°C the glass is too fluid for crystal lattices to stabilize; below 595°C the atomic mobility is too low for crystal growth to proceed at meaningful rates. The glass surface is particularly susceptible because the surface concentration of SiO₂ is higher than the bulk (sodium and potassium volatilize from the surface at fusing temperatures, leaving a silica-enriched layer), and cristobalite nucleates preferentially on silica-rich surfaces.

Crash cooling works by minimizing the time spent in the 595–704°C nucleation window. A standard firing schedule holds the piece at the fusing peak temperature, then drops temperature as rapidly as the kiln elements allow (opening the lid or kiln door slightly, or simply cutting all power and letting the kiln cool freely). The goal is to pass through 704°C to 595°C in the minimum practical time, then hold at the annealing soak temperature (516°C for Bullseye) where atomic mobility is too low for crystal growth but thermal stress relief can still occur. Document the crash-cool protocol: whether the kiln lid was partially opened, at what temperature lid opening occurred, how long the drop from peak to 516°C took. The same kiln model in a cold versus warm studio takes different crash-cool times, and that difference matters for devitrification risk.

Devitrification surface spray (Bullseye Glass Spray A or Overglaze, ZYP Coatings Boron Nitride, or diluted borax solution) works through a different mechanism. The spray deposits a thin flux layer on the glass surface before firing. During firing, this flux dissolves the nascent cristobalite crystals as they nucleate: borax (Na₂B₄O₇) melts at 743°C and acts as a glass-forming flux that incorporates surface crystallites into a thin vitreous layer rather than allowing them to grow. The surface spray does not prevent nucleation — it chemically dissolves crystals as they form. Document spray concentration (Bullseye Glass Spray at label dilution vs 2× dilution), application method (fine mist vs thin coat), whether the piece was fired immediately after spraying or allowed to dry, and whether devitrification was observed despite the spray. Spray applied too heavily produces a visible bloom or changes the surface texture; too lightly is ineffective on pieces with high surface-to-volume ratio or fired multiple times.

3. Color chemistry in glass — transition metal oxides and firing atmosphere

Glass color arises from the selective absorption of visible light wavelengths by colorant ions dissolved in the silicate network. For transition metals (cobalt, copper, iron, manganese, chromium, nickel), color results from d-orbital electronic transitions: the colorant cation in the glass network is surrounded by oxygen ligands that split its d-orbital energy levels, and photons at specific wavelengths are absorbed to excite electrons between these levels. The color seen is the complement of the absorbed wavelength. The splitting energy — and therefore the absorbed wavelength and resulting glass color — depends on both the transition metal and the glass network chemistry surrounding it: the same metal produces different colors in different glass compositions. This is why colorant behavior in glass is not transferable from one glass system to another.

Copper (Cu). Cu²⁺ in an oxidizing glass network produces blue-green color (d-d transition absorption ~800–900 nm, red-orange absorbed, blue-green transmitted). Cu⁺ produces a different d-electron configuration that results in colorless glass in most formulations, but at high concentrations and in reducing conditions, cuprous copper can produce the deep red called “copper ruby” or “flambé” through a colloidal mechanism similar to gold ruby (submicroscopic copper metal particles scatter and absorb light preferentially). True copper reduction requires CO or H₂ atmosphere that most studio kilns cannot maintain. The turquoise-to-blue-green range of commercial copper-colored fusing glasses is produced by Cu²⁺ in an oxidizing melt. Re-firing copper-bearing glasses in a kiln with good ventilation is important because copper volatilizes as CuO at high temperatures, reducing color intensity in subsequent firings and contaminating adjacent glass surfaces.

Cobalt (Co). CoO in glass produces a consistent deep blue regardless of firing atmosphere. Co²⁺ is relatively stable under both oxidizing and reducing conditions, and its d-orbital transition energy in the silicate network corresponds to absorption in the orange-red range (~500–600 nm), transmitting blue. This is the most reliable of the transition metal colorants for consistent results across different kilns and firing schedules. Very small concentrations (0.01–0.1% CoO by weight) produce strong color: cobalt is approximately 10× more powerful per unit weight than iron and 3–4× more powerful than copper. Document the glass colorant identity (not just “cobalt blue” but the specific glass product name and lot) because cobalt-blue glasses with different flux compositions produce different tints of blue.

Iron (Fe). Iron is the most complicated glass colorant because Fe²⁺ and Fe³⁺ produce distinctly different colors and their ratio is highly sensitive to firing atmosphere. Fe³⁺ (ferric) absorbs strongly in the blue-ultraviolet range and weakly in the red, producing an amber or yellow-brown color. Fe²⁺ (ferrous) absorbs strongly in the red and yellow range, producing a blue-green color. Most commercial iron-containing glasses are fired in air (oxidizing conditions) and contain predominantly Fe³⁺, producing the amber-to-brown range. A glass that appears amber at room temperature in transmitted light will shift toward the blue-green side if reduced during firing. The shift is not reversible in a single additional firing unless reducing conditions are re-established. High-iron glasses (green bottle glass, which contains ~0.3–0.5% Fe₂O₃) can show visible color shifts after re-firing in a very tight kiln that accumulates reducing atmosphere from organic burnout of kiln wash.

Manganese (Mn). Mn³⁺ in silicate glass produces purple through absorption in the green (~490–530 nm) and in the yellow (~580–620 nm). Mn²⁺ is nearly colorless. Manganese in commercial glass colorants is typically added as MnO₂, which decomposes to MnO (Mn²⁺) at high temperatures in a reducing environment and can partially oxidize back to Mn³⁺ on cooling in air, giving residual purple. Manganese was historically used to decolorize bottle glass by oxidizing Fe²⁺ to Fe³⁺ (the two colors partially cancel), which is why antique window glass “solarizes” to pink-purple on long UV exposure: the UV light reduces some Mn back to Mn³⁺.

Selenium-cadmium sulfide (orange-red colorants). The orange, orange-red, and red-orange colors in commercial fusing glasses are typically produced not by a dissolved transition metal but by a colloidal pigment system: cadmium sulfide (CdS, yellow-orange, band gap ~2.4 eV) or cadmium sulfoselenide (CdS_xSe_1−x, orange to red as Se content increases, shifting band gap from 2.4 to 1.7 eV). These pigments are formed by precipitation of nanoscale crystals within the glass network during a “striking” heat treatment after initial forming. Commercial selenium-cadmium glass must be struck at an appropriate temperature to develop full color. The thermal stability ceiling for cadmium sulfoselenide is approximately 650–700°C: above this temperature the pigment particles begin to dissolve back into the glass melt, causing color loss. Repeated high-temperature firings degrade orange and red colors more than cobalt blue or copper green. Document any observed color fade in selenium-bearing reds and oranges across multiple firings.

4. Dichroic glass — thin film interference, vacuum deposition, and what to document

Dichroic glass displays different transmitted and reflected colors simultaneously — it transmits one color and reflects the complementary color at the same angle of viewing. The effect is produced by physical vapor deposition (PVD): in a vacuum chamber at pressures around 10⁻⁵ to 10⁻⁷ torr, metallic oxides (titanium dioxide TiO₂, silicon dioxide SiO₂, zirconium dioxide ZrO₂, magnesium fluoride MgF₂, and various metallic elements) are vaporized by electron beam heating and deposited onto the glass surface one layer at a time. The alternating high- and low-refractive-index layers form a multilayer thin-film interference filter. Constructive interference of reflected light at specific wavelengths produces strong reflection at those wavelengths; destructive interference of other wavelengths allows them to transmit. The peak reflected wavelength is determined by: λ = 2 × n × d, where n is the refractive index of the layer and d is its physical thickness — a layer of TiO₂ (n ≈ 2.35) at 100 nm thickness produces peak reflection at 2 × 2.35 × 100 = 470 nm (blue). Changing the layer thickness by tens of nanometers shifts the reflected color through the entire visible spectrum.

Dichroic glass is available in Bullseye COE 90 and Oceanside COE 96 versions. The dichroic coating is applied to either a clear base glass or a black base glass: on clear base, both transmitted and reflected colors are visible; on black base, the transmitted light is absorbed by the black glass and only the reflected color is seen, which is why black-base dichroic appears mirror-like rather than transmissive. Document for each dichroic piece used: the COE family (90 or 96), the base glass color (clear or black), the film composition if known from the manufacturer (some manufacturers publish the film series, e.g., Wissmach, Coatings By Sandberg), the starting film color as seen in reflection under natural light, and the resulting color shift after firing — because heat does change dichroic appearance. Dichroic coatings are stable through standard fusing temperatures, but the optical appearance shifts as the glass surface texture changes (fire polishing creates a smoother surface that reflects more evenly) and as the coating melts partially into the glass matrix in very long or very hot firings. Fire-side vs non-fire-side placement documentation: the dichroic surface must face up (fire side) for most effects; placing it face-down against another glass traps the coating and produces a different, often brighter effect because light bounces between the coating and the glass surface above it.

5. Frit particle size — how coarseness determines flow and surface texture

Frit is crushed glass available in a range from powder (superfine, passing a 200-mesh sieve, particles <75 μm) through fine (70-mesh, ~200 μm), medium (20-mesh, ~850 μm), coarse (10-mesh, ~2mm), and coarse chunks (5–10mm irregular pieces). The particle size has a direct and predictable effect on firing behavior that is mechanistically distinct from how sheet glass fuses.

Powder and fine frit fuse at lower effective temperatures than coarse frit because the surface-area-to-volume ratio is higher: more glass surface area per unit volume is exposed to the kiln atmosphere, and the smaller particles have less internal thermal gradient, allowing them to soften and flow with less total heat input. Powder frit at the same peak temperature and soak time as sheet glass produces a smoother, more fully fused surface. Because powder frit is mixed as individual particles suspended in the base glass surface, it produces uniform color coverage and can be used to create gradients (less concentration toward the edge) or opaque coverage (higher concentration). Fine frit at 1–2 layers of coverage creates a slightly textured, sandy surface rather than a smooth film; document application density in grams per square centimeter for reproducibility.

Coarse frit and chunks retain their surface texture more than fine frit after a standard tack fuse because the thermal gradient within each large particle means the core heats more slowly than the surface, and the particle softens and slightly rounds without fully losing its identity. At full-fuse temperature, coarse frit eventually flows flat, but the transition between retained texture and fully flat requires documentation: a piece fired to 760°C with coarse frit on top shows the frit beginning to round and adhere but not flatten; the same piece fired to 793°C shows the frit significantly flattened. A color test tile fired at 5-degree increments through the 750–820°C range using coarse frit on clear base is one of the highest-value Patreon deliverables a fuser can produce: it shows the exact flow-texture relationship in that creator’s specific kiln, which cannot be derived from any general guide.

6. Firing schedule mass effect — why piece size and thickness require independent adjustments

Most published kiln schedules specify a peak temperature, ramp rate, and soak time for a given fuse level (tack, contour, full) in a particular glass system (COE 90 or COE 96). What they typically do not specify is the glass mass the schedule was calibrated for, because mass is variable in practice. Mass matters in two independent ways: as total stack thickness (which determines the thermal gradient from surface to bottom), and as total loaded weight (which affects how long the kiln’s elements must work to heat the thermal mass before the glass reaches the programmed temperature).

Stack thickness effect: a single 3mm sheet of glass heats through far more quickly than a 9mm stack (three 3mm layers). The 9mm stack has a larger thermal gradient between the glass touching the kiln shelf (which heats by conduction from the shelf) and the glass surface (which heats by radiation from the elements above). If the ramp rate is too fast for the stack thickness, the surface reaches peak temperature while the bottom layers are still significantly cooler, creating stress and potentially cracking the piece during the heating phase. For assemblies thicker than 6mm total, reduce initial ramp rates to 150–200°C/hour from the standard 333°C/hour; for assemblies over 12mm, reduce further to 100–150°C/hour. Document the total stack thickness in millimeters, not just the number of layers, because 3mm Bullseye and 4mm window glass produce different stack thicknesses for the same layer count.

Total mass effect: a kiln loaded with twenty 15cm square panels has roughly 20× the thermal mass of a kiln loaded with one panel. The kiln controller programs a ramp rate in degrees per hour, but the actual glass temperature lags the thermocouple reading when heating large mass. The thermocouple, located near the elements (not at the glass surface), reads the air temperature in the kiln — the glass temperature lags by a margin proportional to the mass. A kiln that reaches tack-fuse results at a peak temperature of 732°C for a single small panel may need to be programmed to 749°C or require a longer soak at 732°C to achieve the same tack level for a full shelf load. The solution is documentation: run a calibration firing for each distinct load type (single small panel, half-shelf load, full-shelf load) and note the programmed peak temperature and soak time that produced the target fuse level in each case. This is kiln-specific, load-specific calibration data that no manufacturer or general guide can provide — it is the highest-value technical deliverable in a glass fusing Patreon because it is simultaneously true, non-transferable, and genuinely useful to a patron trying to calibrate their own kiln using their own similar load.

7. Apple Tax — what glass fusing creators actually lose

Glass fusing creator iOS rates by platform. YouTube kiln-opening and glass fusing tutorial content: 55–68% iOS — above-average desktop share because kiln-timing tutorials are often viewed on a laptop or tablet next to the kiln controller, shifting some viewers to non-iOS platforms. Instagram kiln glass photography and finished piece reveals: 70–80% iOS. TikTok kiln-opening reveal and before-and-after content: 72–82% iOS. Pinterest, which drives significant traffic to glass fusing creators: mostly desktop (iOS share <40%), so Pinterest-primary creators have below-average iOS exposure. Glass fusing is a relatively small niche with highly engaged patrons who follow multiple creators and maintain long subscriptions because the technical documentation accumulates in value over time.

The Apple Tax at specific income levels. At $200/month with 62% iOS (YouTube-primary kiln tutorial creator): $200 × 0.62 × 0.30 = approximately $37.20/month ($446.40/year). At $300/month with 70% iOS (mixed YouTube and Instagram glass fusing creator): $300 × 0.70 × 0.30 = approximately $63/month ($756/year). At $350/month with 76% iOS (Instagram-primary kiln glass artist): $350 × 0.76 × 0.30 = approximately $79.80/month ($957.60/year). At $500/month with 78% iOS (large glass fusing Patreon with active Instagram following): $500 × 0.78 × 0.30 = approximately $117/month ($1,404/year).

Enable Patreon’s web-only billing toggle before October 31, 2026. Update all Instagram bio links and YouTube description Patreon links to the Patreon web URL (not the app deep link). Post a single “subscribe on the web” instruction with the direct URL to your tier list — one post to Instagram Stories pinned above the bio link. Test the subscription flow from Safari on iPhone before November 1. The web-only toggle is free to enable, requires no development work, and removes the 30% Apple fee entirely for any patron who subscribes via the browser rather than the iOS app. The enabling deadline is October 31; the mechanics work the moment the toggle is on.

See the glass fusing Patreon tier structure guide for COE compatibility documentation frameworks, firing schedule per-segment notation, slumping mold protocol documentation, and the two-tier structure that suits most kiln glass creators.