Updates

I. The Main App — 19 commits

63 files changed, +15,953 lines. Source repository: stull-atlas

17 New Z-Axis Options (all UMF oxides)

Before this sprint, 8 oxides were available as Z-axis choices. Now there are 23 individual oxides, organized by role in the glaze melt.

Fluxes — R₂O (Alkali)

OxideRole in the Glaze
Li₂O newMost powerful flux per mole. Lowers viscosity dramatically. Small amounts shift color response, reduce thermal expansion. From spodumene, petalite, lithium carbonate.
Na₂OPrimary alkali flux. High expansion (crazing risk). Develops color in cobalt/copper. From soda feldspar, nepheline syenite, soda ash.
K₂OSlightly less active than Na₂O. Lower expansion. Slightly different color development. From potash feldspar.

Fluxes — RO (Alkaline Earth + Divalent)

OxideRole in the Glaze
CaOWorkhorse flux. Glossy at high calcium, dry/matte at high alumina. Promotes color clarity. From whiting, wollastonite, dolomite.
MgOButtery mattes. Smooths surfaces. Promotes opacity. From dolomite, talc, magnesium carbonate.
SrO newSimilar to CaO but wider firing range, different crystal effects. Lead-free alternative for bright colors. From strontium carbonate.
BaOSatin surfaces, enhances blues. Toxic in soluble form. From barium carbonate.
ZnOPromotes crystallization (zinc crystals at high %), increases fluidity, reduces thermal expansion. From zinc oxide.
PbO newHistorical low-fire flux. Brilliant color response, wide firing range. Toxic — flagged in safety metrics. From lead bisilicate (fritted).

Glass Formers all new

OxideRole in the Glaze
TiO₂Promotes crystallization, opacity (rutile). Yellow/cream colors. Breaks up glaze surfaces. From rutile, titanium dioxide.
ZrO₂Opacifier. Stiffens melt, promotes whiteness. From zircopax, ultrox, zirconium silicate.
SnO₂Classic opacifier (tin glazes). White, opaque. More expensive than zirconium. From tin oxide.

Colorants & Misc all new

OxideRole in the Glaze
MnO / MnO₂Browns, purples, plums. Speckle when coarse. Manganese dioxide stronger than monoxide.
NiOGrays, browns, blue-greens depending on base glaze. Mutes other colorants.
CuO / Cu₂OGreens in oxidation, reds in reduction (copper reds). Cu₂O is the reduced form.
CoOStrongest colorant — intense blues at 0.5–2%. Works in nearly any base.
Cr₂O₃Greens (forest to lime depending on base). Chrome-tin pinks with SnO₂.
P₂O₅Opalescence, milky blues (chun/jun glazes). Promotes phase separation. From bone ash.

Computed Values new

AxisWhat It Shows
Total Flux MolesSum of all flux oxide moles — a single “how much flux?” number.
Thermal Expansion (COE)Appen coefficient of expansion approximation — predicts crazing/shivering. Uses published coefficients for 11 oxides.

12 Computed Z-Axis Metrics

These go beyond individual oxides to compute derived properties from the full UMF, using published ceramic science.

MetricFormula BasisWhat a Potter Learns
NBO/TMysen & Richet modelThe fundamental measure of glass structure. NBO/T = 0 means fully polymerized (hard, durable). NBO/T = 4 means completely depolymerized (fluid, run-prone). Most glazes: 0.5–2.5.
Optical BasicityDuffy & Ingram, 21 oxidesPredicts color response. High basicity shifts Fe toward amber/brown, Cu toward blue. Low basicity gives greens. Explains why the same 2% iron looks different in two glazes.
Flux EntropyShannon information entropyMeasures flux diversity. Entropy = 0 means one flux dominates. High entropy = many fluxes in balance — more complex, interesting surfaces.
CaO:MgODirect ratioCaO-dominant (glossy, color-bright) vs MgO-dominant (buttery matte, opacity).
Combined AlkaliLi₂O + Na₂O + K₂OTotal soluble flux load. High = glossy, fluid, high expansion. Low = more stable.
Na₂O:K₂ODirect ratioTwo glazes with same total alkali behave differently if one is sodium-dominant vs potassium-dominant. Different color, expansion, eutectic behavior.
Viscosity IndexFlux-weighted sum / glass formersQuick estimate of melt viscosity. High = runny (dripping risk). Low = stiff (crawl risk).
Surface TensionDietzel field-strength modelPredicts crawling and beading. High surface tension melts pull back from edges.
DurabilitySiO₂ / (Na₂O + K₂O + Li₂O)Chemical resistance. Below ~3.0, the glaze dissolves in dishwater. Above ~5.0, highly durable.
Total ColorantSum of transition metalsTotal color load regardless of which oxide. A different perspective than individual oxides.
Fe:Ti RatioFe₂O₃ / TiO₂High = cleaner iron colors. Low = broken, variegated surfaces (rutile blues, aventurine).

Flux Triaxial Explorer

An interactive ternary diagram where potters visualize the three flux families (R₂O / RO / B₂O₃) and see how their glazes distribute across that space. Clicking a point shows what recipes live in that flux ratio neighborhood.

Explorer Bridge — Shift+Click to Blend

While exploring the 2D or 3D Stull chart, shift+click selects glazes (orange borders). Selected glazes appear in a new “Blend” sidebar tab. Click “Line Blend” / “Triaxial” / “Quadaxial” and it launches the blend calculator pre-filled with those recipes. This bridges the gap between discovering glazes and blending them.

Data Accuracy Overhaul

Three commits fixing edge cases in data representation: z-clamp, fluxRatio, cone range, null-cone handling. 2D colorscale now matches 3D. Robust surface interpolation. Cone labels use full ceramic notation (06 through 13) with a formatCone() utility.

Glaze Dataset Upgrade

Replaced entire dataset: 9,712 recipes from a full SQL export, up from 3,214. 3× the data. Every visualization, search, and blend calculation now operates on a significantly richer dataset.

Database Architecture — 40 Tables

Three commits built the full Supabase schema: 40 tables covering glazes, recipes, users, profiles, explorations, materials, kilns, firing schedules, and more. This is the foundation for user profiles, personal recipe databases, and social features.

II. Digitalfire Archive — 8 deploy commits

ChangeImpact
Database deliverySwitched from 8 parallel chunks (122 MB total) to single gzipped file (26 MB). 79% smaller download.
Search fixedSwitched to FTS5-enabled sql.js WASM build (1.2 MB). Full-text search across entire Digitalfire corpus now works in-browser, no server required.
Rich contentInternal links between articles, markdown rendering, dead-link cleanup.
PolishHidden pricing (everything free), hidden stats display, cache-busted all assets, fixed logo linking to /digitalfire/.

III. Community Atlas — Brand New Section

852Nodes
1,821Connections
619Individual Pages
3Explorer Views

Built from the ceramics-community knowledge graph. Every connected person, school, organization, place, and technique gets their own page showing all connections.

DirectoryPagesContents
/community/people/297Individual ceramicists
/community/schools/193Schools, universities, residencies
/community/orgs/54Organizations, publications, companies
/community/places/44Studios, galleries, venues
/community/techniques/30Firing methods, glaze techniques

Three explorer views: force-directed graph, geographic map, and 3D globe.

IV. The 8 Phases to 120 Z-Axis Options

Currently live: 40 axes (23 oxides + 17 computed). The roadmap takes this to ~124 axes in 8 phases. Each phase adds new lenses to the 3D Stull chart.

Phase A High-Impact Structural Metrics — 6 axes

Straightforward published formulas. Highest practical value.

AxisWhat It Predicts
ASIAlumina Saturation Index. Al₂O₃ / (Na₂O + K₂O + Li₂O + CaO). ASI < 1 = peralkaline (glossy, fluid). ASI > 1 = peraluminous (matte, crystallization). Maps directly to the Stull chart’s left-right axis.
Refractive IndexAppen model, 7 oxide coefficients. Observable optical property — brilliance, luster, depth. Why some glazes look “alive” and others look flat.
Glass DensityAppen molar volume model. Physical weight of the glass. Affects application thickness and glaze-body fit.
Crawl RiskSurface tension × flux / glass formers. Flags glazes likely to bead up and crawl.
Craze RiskCOEglaze − COEbody. The single most practical axis for functional ware. Positive = crazing. Negative = compression.
FluidityFlux-effectiveness-weighted sum. Weights each flux by actual melting power (Li₂O × 3.0, Na₂O × 2.0, MgO × 0.6).

Phase B Classification & Prediction — 6 axes

Semi-empirical composites. Higher-level questions.

AxisWhat It Answers
Matteness Predictor“Will this glaze be matte?” Combines ASI with mattifying oxides minus glossy alkalis.
Boron RatioB₂O₃ / (B₂O₃ + SiO₂). Explains why two Stull-identical glazes look different: boron substituting for silica.
Compositional EntropyShannon entropy across ALL oxides. Simple 3-oxide vs complex 10-oxide glazes separate clearly.
Excess Alkali(R₂O) − Al₂O₃. The “free modifier.” Positive = glossy, fluid. Negative = peraluminous.
KNaO EquivalentUnified alkali “dose” normalized to Na₂O-equivalent based on ionic radius.
Leaching Risk(PbO + BaO + CuO) / (SiO₂ / R₂O). Food safety flag.

Phase C Glass Physics — 5 axes

Published additive models with full coefficient tables. Real material science.

AxisWhat It Measures
Young’s ModulusStiffness / scratch resistance. Makishima-Mackenzie 1973.
Vickers HardnessPractical abrasion resistance. Derived from Young’s modulus + Poisson’s ratio.
Poisson’s RatioLow = brittle, high = ductile. Predicts chip resistance.
Tg EstimateGlass transition temperature (°C). Where the glaze freezes on cooling. Annealing and thermal shock.
Thermal Shock ResistanceCombines fracture strength, modulus, Poisson, and expansion. The ovenware metric.

Phase D Advanced Glass Science — 5 axes

Topological constraint theory, Q-speciation, thermodynamics. PhD-level physics made into a slider.

AxisWhat It Reveals
Mean CoordinationAverage cation coordination number. 4 = tetrahedral network (strong). 6 = depolymerized (weak).
Constraints/AtomPhillips-Thorpe-Mauro model. nc = 3 means “isostatic” — the perfect glass (how Corning designed Gorilla Glass).
Q⁴ FractionFully cross-linked silica fraction. Correlates with durability and stiffness.
Q² FractionChain-like structures (metasilicates). High Q² = tendency to crystallize as wollastonite/enstatite.
Liquidus EstimateTemperature (°C) where last crystal dissolves. Above this = homogeneous melt. Below = crystals forming.

Phase E Exploratory Physical — 5 axes

Semi-quantitative but visually revealing.

AxisWhat It Shows
Devitrification IndexPredicts crystal mattes, aventurine, willemite. High = crystals likely on slow cooling.
Hydration EnergyRigorous durability — accounts for each oxide’s water reactivity.
Modifier FractionAbove the Greaves percolation threshold (~0.16–0.20), leaching jumps dramatically.
Fragility IndexStrong (m≈20, gradual viscosity change) vs fragile (m≈60+, sudden freeze). Predicts crystallization.
Fe Redox EstimateFe³⁺/Fe²⁺ ratio from optical basicity + temperature. Determines iron color: amber (ox) vs celadon green (red).

Phase F Temperature-Dependent & Atmosphere-Aware — 8 axes

Requires cone → Kelvin via Orton lookup, unlocking true thermodynamic models.

AxisWhat It Predicts
Firing TemperatureOrton cone → °C standard conversion. Foundation for all below.
GRD ViscosityGiordano-Russell-Dingwell 2008 model — actual viscosity in Pa·s at firing temperature. What glass factories use.
Liquidus OvershootTfiring − Tliquidus. Positive = fully melted. Negative = crystals survive. The single most predictive melt indicator.
Working RangeTemperature span where the glaze is workable. Wide = forgiving. Narrow = must nail the temperature.
Heatwork ExcessOver-fired or under-fired relative to chemistry? Over-fired = running. Under-fired = incomplete melt.
Volatilization IndexPredicts mass loss from volatile oxides. The glaze out of the kiln ≠ the glaze you batched.
Fe Color (Oxidation)Predicted iron color spectrum in oxidation: amber → brown → honey.
Fe Color (Reduction)Predicted iron color spectrum in reduction: celadon green → blue.

Phase G Multi-Modal Data Axes — 27 axes

Beyond UMF chemistry — recipe, surface labels, statistics, geography, time.

CategoryAxes
Recipe / IngredientIngredient count, frit %, free silica fraction, clay content, LOI estimate, material diversity
Surface / LabelSurface mismatch (predicted vs labeled), surface probability P(matte), opacity predictor
StatisticalAnomaly score (Mahalanobis distance), feature density (k-NN), percentile rank, nearest surface boundary
Limit FormulasLimit distance (inside/outside UMF envelope), safety score (composite of all checks)
TaxonomyType centroid distance, nearest type boundary, type confidence
TemporalRecipe age (years), innovation score (distance from all prior recipes at time of creation)
Color / AppearanceColorant per flux (effective colorant concentration in the melt)
Interaction EffectsMixed alkali effect (peaks at 50:50 Na:K), boron N₄ fraction (trigonal → tetrahedral transition)
Fitness CompositesFood safety score, microwave score (low transition metals = safe)

Phase H Sensitivity & Robustness — 3 axes

The math of “fussy” vs “forgiving.”

AxisWhat It Reveals
Robustness ScoreMax property swing from ±0.01 change in any single oxide. Low = forgiving. High = tiny weighing error changes everything.
Critical OxideWhich single oxide causes the largest swing if mis-measured by ±5%? Tells you where to be most careful.
Error WidthGiven ±2% weighing error on all ingredients, what’s the range of possible outcomes? Monte Carlo propagation.

V. Full Version Roadmap

VersionNameStatusWhat It Adds
v3.4The CompassshippedWeighted similarity search + surface prediction
v3.5The GalleryshippedPhotos in exploration, visual browsing
v3.5.1The Instrumentsplanned120 Z-axis options in 8 phases (A–H above)
v3.6The WalkplannedRecipe interpolation — “how do I get from my glaze to that one?”
v3.6.3Wet BlendinghenryMix finished glazes by ratio, compute resulting UMF, sweep blend ratios
v3.6.4Flux Ratio ControlshenryPer-element flux breakdown as plottable axes, flux triangle
v3.7The ConstellationsplannedAuto-discover and name glaze families; fuzzy grouping henry
v3.8The Knowledge GraphvisionVisual graph navigation with photos — glazes as a network
v3.9The StudioplannedMaterial substitution, bidirectional recipe calculation
v3.10The Time MachineplannedHistorical glaze evolution, animated timeline
v3.11The Laboratoryplanned17 calculator/prediction engines: fit, viscosity, crawl, defect, color, batch, LOI, comparison, substitution
v3.12The KilnplannedFiring schedule builder, heatwork calculator, test tile notebook, QR tracker
v3.13The BenchplannedGlaze application physics, clay body calculator, recipe sheet generator
v3.14The MapplannedPhase diagram overlays, glaze layering predictor, historical glaze library
v3.15The VoiceplannedPlain-English glaze narrative generator
v3.16The What-If MachineplannedDual-mode cockpit: Studio Mode + Science Mode. One slider, everything moves.
v3.16.5Living Recipe CardplannedNCECA-ready printable handout with all data
v4.0The CommunityvisionShared knowledge, social features
v4.1.0Personal ProfileshenryEach potter gets their own page and recipe database
v4.1.1Free TierhenryNo paywall at launch — make it useful first
v4.2The ClassroomvisionWorkshop exercises, teaching tools, curriculum modules
v4.3The BridgevisionUniversal import: GlazeMaster, Matrix, Insight, OCR, Glazy sync
v3.4 Compass        → defines “similar” (distance function)
v3.5 Gallery        → shows what “similar” looks like (photos)
v3.5.1 Instruments  → 120 lenses on every glaze (physics, predictions, risk)
v3.6 Walk           → shows how to get there (recipe delta, wet blending)
v3.7 Constellations → groups become named places (fuzzy boundaries)
v3.8 Knowledge Graph→ the places become a navigable world
v3.11 Laboratory    → 17 prediction engines built on Instruments data
v3.16 What-If       → everything moves together when you change one thing
v4.0 Community      → other potters’ explorations enrich the landscape

© 2026 Stull Atlas - Thanks Henry