Updated 28 September 2026

Data schema

Twenty-six fields, four enumerations, one record per material grade. The machine-readable version of this page ships as schema-1.6.0.json in every release and is what cinderq validate checks against.

Fields

Required means the record is rejected without it. Everything else may be empty.
Field	Type	Unit	Req.	Notes
id	string	-	yes	Pattern `CC-\d{4}`. Assigned once, never reused.
name	string	-	yes	Generic grade description in lower case, not a trade name.
family	enum	-	yes	One of eleven values, listed below.
aliases	list of string	-	no	Other names for the same class. Empty list, never null.
al2o3_pct	number	% by mass	no	As reported. Sources rarely agree past one decimal.
sio2_pct	number	% by mass	no	Includes free silica; no distinction is made.
mgo_pct	number	% by mass	no	Reported as MgO even for dolomitic grades.
cr2o3_pct	number	% by mass	no	Present only in the chrome-bearing family.
c_free_pct	number	% by mass	no	Free carbon after coking, not total carbon in the mix.
max_service_c	integer	°C	yes	Continuous limit, rounded to the nearest 10 °C.
max_service_atmosphere	enum	-	yes	The atmosphere `max_service_c` applies to.
air_limit_c	integer	°C	no	Separate limit in air for grades whose main limit is for a reducing or inert atmosphere.
k_1000_w_mk	number	W/(m·K)	no	Thermal conductivity. Read together with the next two fields.
k_ref_c	integer	°C	no	Temperature the conductivity figure belongs to. Usually 1000; 400 for grades not rated that high.
k_method	enum	-	no	Mandatory since v1.6.0 whenever `k_1000_w_mk` is present.
bulk_density_g_cm3	number	g/cm³	no	Dry. For castables, after the drying schedule named in the source.
open_porosity_pct	number	% by volume	no	Apparent porosity by water or kerosene displacement.
ccs_mpa	number	MPa	no	Cold crushing strength at room temperature.
hmor_mpa	number	MPa	no	Hot modulus of rupture. The test temperature lives in the source, not in this field - a known weakness.
rul_t05_c	integer	°C	no	Refractoriness under load, temperature at 0.5% subsidence.
thermal_shock_cycles	integer	cycles	no	Cycles to failure. Meaningless without the next field.
shock_protocol	enum	-	no	Which cycling test produced the count.
slag_index	integer	1–5	no	Project-assigned ordinal. See the caveat below.
source_ids	list of string	-	yes	At least one id from `sources.csv`.
confidence	enum	-	yes	`high`, `medium` or `low`, assigned by the rules on the methodology page.
updated	date	ISO 8601	yes	Date of the last accepted change to this record.

An empty cell is not a zero

This is the single most common way to misread the dataset, so it gets its own rule set:

An empty value - null in JSON, nothing between the commas in CSV - means no source we hold reports that property for that grade. It says nothing about the material.
A zero means a source reported zero. In practice this only happens for c_free_pct and cr2o3_pct, where a source states explicitly that the grade is carbon-free or chrome-free.
An empty list in aliases or source_ids is written as [], never as null. For source_ids an empty list is invalid and the validator rejects it.
Pass --include-nulls to the reader if you want the empty cells kept in CSV output; by default rows are dropped from a filter when the filtered field is empty, which is a choice you should be aware of.

Enumerations

family

fireclay silica high-alumina corundum silicon-carbide magnesia magnesia-carbon chrome-bearing zirconia-bearing carbon-graphite insulating

Family is a filing decision, not chemistry. A dense 90% alumina castable sits in high-alumina next to a brick, because people look for them together. insulating cuts across chemistry entirely and holds anything whose job is to not conduct: firebrick below 1.2 g/cm³, boards, blankets, microporous panels.

max_service_atmosphere

oxidising: Air or combustion products with free oxygen. The default, and what 318 of the 412 records are quoted for.
reducing: Carbon monoxide rich or otherwise oxygen starved. Used for most carbon-bearing grades.
inert: Argon, nitrogen or similar. Mostly graphite records.
vacuum: Six records only, all from one vacuum-degasser paper.

k_method

hot-wire: Parallel or cross-array hot wire, the usual method for bricks and castables.
panel: Calorimeter or guarded hot plate on a built panel. Reads lower than hot wire on porous grades.
calculated: Derived by the source from density and composition rather than measured. 46 records are like this and all carry confidence of at most medium.

shock_protocol

water-quench-1000: Heat to 1 000 °C, quench in water, repeat until failure. Harsh, and the most commonly published.
air-quench-1000: Same heating, forced air cooling. Cycle counts run several times higher than the water figure for the same grade, which is exactly why the two must not be mixed.
prism-950: Prism test at 950 °C with cooling in still air, mostly from older European literature.

One record, in full

{
  "id": "CC-0137",
  "name": "corundum brick, 99% Al2O3",
  "family": "corundum",
  "aliases": ["sintered alumina brick", "AL99"],
  "al2o3_pct": 99.1,
  "sio2_pct": 0.3,
  "mgo_pct": null,
  "cr2o3_pct": 0.0,
  "c_free_pct": 0.0,
  "max_service_c": 1750,
  "max_service_atmosphere": "oxidising",
  "air_limit_c": null,
  "k_1000_w_mk": 3.0,
  "k_ref_c": 1000,
  "k_method": "hot-wire",
  "bulk_density_g_cm3": 3.10,
  "open_porosity_pct": 17.5,
  "ccs_mpa": 85.0,
  "hmor_mpa": 9.5,
  "rul_t05_c": 1700,
  "thermal_shock_cycles": 12,
  "shock_protocol": "water-quench-1000",
  "slag_index": 4,
  "source_ids": ["S-004", "S-019"],
  "confidence": "high",
  "updated": "2026-06-28"
}

The CSV carries the same fields in the same order, with lists joined by a semicolon:

id,name,family,aliases,al2o3_pct,sio2_pct,mgo_pct,cr2o3_pct,c_free_pct,max_service_c,
max_service_atmosphere,air_limit_c,k_1000_w_mk,k_ref_c,k_method,bulk_density_g_cm3,
open_porosity_pct,ccs_mpa,hmor_mpa,rul_t05_c,thermal_shock_cycles,shock_protocol,
slag_index,source_ids,confidence,updated

CC-0137,"corundum brick, 99% Al2O3",corundum,"sintered alumina brick;AL99",99.1,0.3,,0.0,
0.0,1750,oxidising,,3.0,1000,hot-wire,3.10,17.5,85.0,9.5,1700,12,water-quench-1000,4,
"S-004;S-019",high,2026-06-28

The header line is one line in the real file; it is wrapped here to fit the page. Quoting follows the usual CSV convention, UTF-8 without a byte order mark, LF line endings.

On the slag index. It is the only field in the dataset we invented rather than transcribed, and it compresses a judgement into one digit. A 5 for magnesia-carbon describes resistance to basic steelmaking slag; a 5 for fused cast AZS describes resistance to molten glass. Those two numbers are not on the same scale and should never be compared, averaged or plotted together. We keep the field because readers kept asking for a first sort, and we would rather define it in the open than let people infer it from density.

Field stability

Within a major version, a field is never removed and its meaning never changes; new fields may appear. Three fields are marked provisional in the JSON schema and may still change shape before v2.0.0: air_limit_c, hmor_mpa and slag_index. If you build something on the dataset, pin the release and read the changelog before you move.