Material selection
Which technical ceramic: alumina, SiC, Si3N4, AlN or zirconia?
Among the five ceramics, Y-TZP zirconia shows the highest bending strength (1,000 to 1,240 MPa), AlN and bulk SiC move heat best, Si3N4 expands least, and alumina has the highest stored maximum use temperature (1,750 °C, no load). Every figure belongs to a named grade, source and test basis, so quote the basis with the number.
- Pick zirconia when the part must take bending or impact load and stay dry: Y-TZP shows a fracture toughness of 10 to 13 MPa·m½ in two datasheets, the highest in our data. Blades and pump and valve parts are stored uses, but its hardness (12.7 GPa Knoop) is below alumina and SiC, so wear is a use-list entry, not a hardness advantage. Pick Si3N4 instead for bearings, or for repeated heating and cooling on expansion grounds alone: Si3N4 has the lowest expansion of the five, and the data hold no measured thermal shock figure for it (see the thermal shock section). Its SN-GP grade has 6.2 MPa·m½ toughness.
- Pick AlN or SiC when heat has to leave the part. Our data puts AlN at 140 to 180 W/(m·K) (a typical range, one supplier) and bulk SiC at 140 to 150 W/(m·K). AlN insulation rests on dielectric strength (16 to 20 kV/mm in two datasheets); its resistivity is in the sources but not entered. Our data stores no resistivity for SiC either, and sintered SiC is often not an insulating ceramic (general engineering knowledge, not in our data), so check before swapping SiC in.
- Unsure: start with alumina if the part is an electrical insulator or wear part that carries little bending load. Then ask every supplier for flexural strength with the test method and temperature, and for the basis of any maximum temperature, before you compare grades.
The five ceramics at a glance: ranges across grades, sources and test temperatures. The row labelled YSZ is 3Y-TZP zirconia only, and SiC's 3.165 g/cm³ density shows rounded to 3.17. Si3N4 density and modulus include porous reaction-bonded NC-350; dense Si3N4 is 3.16 to 3.37 g/cm³ and 290 to 320 GPa. Heat and expansion are in the sections below.
| Material | density (g/cm³) | flexural strength (MPa) | elastic modulus (GPa) |
|---|---|---|---|
| Alumina (Al2O3) | 3.72 to 3.9 | 379 | 370 |
| Silicon carbide (SiC) | 3 to 3.17 | 346 to 480 | 370 to 421 |
| Silicon nitride (Si3N4) | 2.53 to 3.37 | 670 to 900 | 180 to 320 |
| Aluminium nitride (AlN) | 3.25 to 3.3 | 300 to 375 | 276 to 330 |
| YSZ | 6.02 to 6.07 | 1,000 to 1,240 | 195 to 210 |
- 1Will the part run hotter than 1,500 °C?YesAlumina, if unloadedAlumina has the highest stored rating, 1,750 °C, but with no load; Y-TZP tops out at 1,500 °C and Si3N4 at 1,400 °C. Our data holds no creep or hot-strength figure, so a loaded part needs hot-strength data from the supplier. SiC and AlN have no comparable rating (SiC's 1,160 °C is a reaction-bonded limit from a generic page), so our data cannot rank them above 1,500 °C.
- 2Does the part take bending or impact load at up to 800 °C, the lower Y-TZP rating?YesY-TZP zirconiaY-TZP shows the highest bend strength and toughness here. Check two things first. Hot water or steam can age it, and no service limit is in our data; as general engineering knowledge, not in our data, ageing is a low-temperature effect, so 800 °C is a dry rating and a humid part well below it may be affected. For sudden temperature changes go to question 4.
- 3Must heat leave the part quickly?YesAlN or SiCChoose AlN if the part must also insulate electrically (16 to 20 kV/mm) and stay dry. Choose SiC if it need not, or if hardness and stiffness matter more; check its resistivity, since our data stores none and sintered SiC is often not an insulator (general engineering knowledge, not in our data).
- 4Does the part see repeated heating and cooling, or rolling contact?YesSi3N4Si3N4 has the lowest expansion of the five, the only basis our data gives for thermal cycling (no measured thermal shock figure), and the bearing tests. Name the route, because the numbers move with it.
- OtherwiseAluminaIf none applies, alumina is the baseline hard electrical insulator. Keep tensile stress low and avoid fast thermal cycling.
Which technical ceramic is strongest in bending?
Y-TZP (yttria-stabilised tetragonal zirconia) is strongest in bending among the sintered grades in our data. Zirconia shows 1,000 MPa for Morgan's 3Y-TZP grade (test not printed) and 1,240 MPa for CoorsTek's sintered YTZP, both at 20 °C, and 1,192 MPa from one paper's own samples. CoorsTek also lists a HIPed YTZP at 1,720 MPa, not entered here. The lowest stored Y-TZP figure is above every stored flexural figure for the other four.
Si3N4 follows with 900 MPa for CoorsTek's hot-isostatically pressed NBD-200 (20 °C) and 670 MPa for Qsil's SN-GP, whose test footnote is ambiguous. SiC shows 346 MPa for NIST's sintered alpha-SiC (four-point, 22 °C) and 480 MPa for CoorsTek's UltraSiC; a 410 MPa reaction-bonded figure comes from a generic page with no named source. Alumina AD-995 is 379 MPa (ASTM F417, 20 °C), and AlN is 300 to 375 MPa across three CoorsTek grades, with 320 MPa as Accuratus's typical value.
Bending strength is not tensile strength: AD-995 alumina lists 262 MPa tensile (ACMA test #4), below its 379 MPa flexural figure. Toughness, in MPa·m½: Y-TZP 10 and 13, dense Si3N4 5.22 to 6.35, alumina 4 to 5, SiC 2.78 to 4, AlN 2.6 to 3.25. Methods differ, so read these as a rough order.
Weibull modulus, a measure of scatter, is stored only for NIST sintered SiC (11.0) and Qsil's SN-GP (20), so consistency cannot be compared across all five. Ask suppliers for it.
Not like for like: grades, test methods and temperatures differ, as the text lists, and some print none. CoorsTek's HIPed YTZP is not entered. The glasses are left out. The bar labelled YSZ is 3Y-TZP zirconia.
Data behind this chart
| Material | Low (MPa) | High (MPa) | Values |
|---|---|---|---|
| YSZ | 1,000 | 1,240 | 3 |
| Silicon nitride (Si3N4) | 670 | 900 | 2 |
| Silicon carbide (SiC) | 346 | 480 | 3 |
| Alumina (Al2O3) | 379 | 379 | 1 |
| Aluminium nitride (AlN) | 300 | 375 | 4 |
Which technical ceramic is stiffest and hardest?
Alumina and SiC are the stiffest of the five. AD-995 alumina is 370 GPa (ASTM C848, 20 °C). SiC entries run from 370 GPa (generic reaction-bonded page) to 421 GPa (NIST polycrystalline). AlN is 276 to 330 GPa, and the dense Si3N4 grades are 290 to 320 GPa, while porous reaction-bonded NC-350 is 180 GPa. Y-TZP is the least stiff dense ceramic here, at 195 to 210 GPa. Fused silica glass is far lower, and borosilicate lists 64 in kN/mm2.
Hardness cannot share one chart, because the profiles use different tests and units. On Vickers or unspecified GPa figures, NIST's sintered SiC is 23.9 GPa (22 °C), Kyocera's alumina of at least 99% purity is 15.2 GPa (JIS R 1610) and Qsil's SN-GP is 14.5 GPa (ambiguous method footnote). CoorsTek lists UltraSiC at 26 GPa and zirconia at 12.7 GPa, both Knoop, which is not interchangeable with Vickers. AlN is 1,100 kg/mm2 (test type not stated) and Morgan's Y-TZP 1,300 kg/mm2 at HV0.3. We do not convert units. SiC is hardest on this evidence: 23.9 GPa Vickers and 26 GPa Knoop, against alumina at 15.2 and Si3N4 at 14.5 GPa Vickers. Zirconia's Knoop 12.7 is below SiC's Knoop 26, the one like-for-like pair; AlN and the other zirconia entry use other tests, so we do not rank them.
Ranges run across grades, sources and methods; many entries print none. Si3N4's 180 GPa is porous reaction-bonded NC-350. Borosilicate is left out because its modulus is stored in kN/mm2.
Data behind this chart
| Material | Low (GPa) | High (GPa) | Values |
|---|---|---|---|
| Silicon carbide (SiC) | 370 | 421 | 4 |
| Alumina (Al2O3) | 370 | 370 | 1 |
| Aluminium nitride (AlN) | 276 | 330 | 4 |
| Silicon nitride (Si3N4) | 180 | 320 | 5 |
| YSZ | 195 | 210 | 3 |
| Fused silica (SiO2 glass) | 72.7 | 72.7 | 1 |
Which technical ceramic conducts heat best?
AlN and bulk SiC are the heat movers in our data. Accuratus gives AlN a typical 140 to 180 W/(m·K) at room temperature for an unnamed commercial grade. For bulk SiC, CoorsTek lists 150 W/(m·K) at 20 °C for UltraSiC and a generic reaction-bonded page lists 140. The CoorsTek AlN grades have no conductivity stored (its table prints the unit as W/m only), so the data cannot say which of AlN or SiC wins for a grade.
The 4H polytype of SiC, the semiconductor form, is quoted higher: 4.9 W/cmK in the Toshiba table and 3.7 W/cmK in a general overview, with no explanation of the gap. Use 140 to 150 for bulk ceramic parts; the chart leaves SiC out for this reason.
Alumina AD-995 is 30 W/(m·K) at 20 °C. Si3N4 is 24 (Qsil SN-GP) to 29 (CoorsTek NBD-200), both at 20 °C. Zirconia is a thermal insulator at 2.2 (CoorsTek) to 3 (Morgan), like fused silica (1.3 to 1.4) and borosilicate (1.2).
SiC is left off because its stored entries mix bulk ceramic values with 4H semiconductor-polytype values in W/cmK, which the chart would convert into a false peak far above the bulk value. AlN is one supplier's typical range for an unnamed grade.
Data behind this chart
| Material | Low (W/(m·K)) | High (W/(m·K)) | Values |
|---|---|---|---|
| Aluminium nitride (AlN) | 140 | 180 | 1 |
| Alumina (Al2O3) | 30 | 30 | 1 |
| Silicon nitride (Si3N4) | 24 | 29 | 2 |
| YSZ | 2.2 | 3 | 2 |
| Fused silica (SiO2 glass) | 1.3 | 1.4 | 2 |
| Borosilicate glass | 1.2 | 1.2 | 1 |
Which technical ceramic expands least?
Among the five, Si3N4 expands least on a like-for-like basis. From room temperature to 1,000 °C, Qsil's SN-GP shows 3.2 ppm/K, against 4.4 for SiC, 5.5 to 5.6 for the CoorsTek AlN grades, 8.2 for alumina and 10.3 for CoorsTek's Y-TZP over 25 to 1,000 °C. Do not mix in the other entries: SN-GP's 1.4 is near room temperature, NBD-200's 2.9 and Accuratus's AlN 4.5 print no range, and Morgan's Y-TZP 9 covers 25 to 400 °C.
Fused silica expands far less than every ceramic here, at 0.55 ppm/K (20 to 320 °C). BOROFLOAT 33 borosilicate, 3.25 ppm/K (20 to 300 °C), is close to Si3N4's 3.2 but over a different temperature range. Neither is structural: borosilicate's bending strength is 25 MPa, or 7 MPa for safety conditions (DIN 52 292), and fused silica has no flexural entry.
Not like for like. The AlN bar shows only Accuratus's 4.5 with no temperature printed; the three CoorsTek AlN grades, 5.5 to 5.6 over 25 to 1,000 °C, are not drawn. The Si3N4 bar starts at a near-room-temperature 1.4, and the Y-TZP bar includes Morgan's 9 (25 to 400 °C). The builder draws near-room-temperature or unlabelled values first; the text gives the comparable ones.
Data behind this chart
| Material | Low (ppm/K) | High (ppm/K) | Values |
|---|---|---|---|
| Fused silica (SiO2 glass) | 0.55 | 0.55 | 1 |
| Silicon nitride (Si3N4) | 1.4 | 3.2 | 3 |
| Borosilicate glass | 3.25 | 3.25 | 1 |
| Silicon carbide (SiC) | 4.4 | 4.4 | 1 |
| Aluminium nitride (AlN) | 4.5 | 4.5 | 1 |
| Alumina (Al2O3) | 8.2 | 8.2 | 1 |
| YSZ | 9 | 10.3 | 2 |
Which technical ceramic survives thermal shock?
Our data cannot rank the five for thermal shock: suppliers test it differently and two have no figure. Kyocera's alumina of at least 99% purity withstood a 200 °C sudden difference (JIS R 1648). For Y-TZP, Morgan lists 200 °C with no test printed and CoorsTek 350 °C from a water quench. Qsil gives SN-GP a calculated thermal shock parameter R1 of 534 K, not a measured difference. We hold no figure for SiC or AlN.
As general engineering knowledge, not in our data, low expansion and high conductivity both help a part survive sudden temperature change, but that is not a ranking.
How hot can each technical ceramic run?
Stored ratings: alumina 1,750 °C (AD-995) and 1,700 °C (96% grade), both with no load; Y-TZP 1,500 °C for CoorsTek and 800 °C for Morgan, whose basis is not printed, so the range is a supplier difference; Si3N4 1,200 °C in air and 1,400 °C inert for SN-GP, and 1,400 °C for NBD-200; SiC 1,160 °C, a mechanical maximum from a generic page with no named source; AlN none. Fused silica lists 1,080 °C on a generic page.
Biscay and co-authors report that SiC's silica scale is porous in dry oxygen between 500 and 800 °C and seals the surface between 1,200 and 1,400 °C. Accuratus says AlN oxidises at the surface above 700 °C in air and is protected by an oxide layer up to 1,370 °C; those are page statements, not use limits.
A maximum use temperature is not a strength at temperature. Every flexural figure stored for the five is at room temperature or has no temperature printed.
Not like for like: the bases differ, as the text lists. SiC and AlN are not drawn because their entries use a different property name or do not exist.
Data behind this chart
| Material | Low (°C) | High (°C) | Values |
|---|---|---|---|
| Alumina (Al2O3) | 1,750 | 1,750 | 1 |
| YSZ | 800 | 1,500 | 2 |
| Silicon nitride (Si3N4) | 1,200 | 1,400 | 3 |
| Fused silica (SiO2 glass) | 1,080 | 1,080 | 1 |
Where is each technical ceramic used, and what goes wrong?
Uses in the profiles: alumina in semiconductor process chambers, wafer handling, laser devices and power tubes, and in heat, corrosion and wear parts. SiC in wear parts, in chemical-vapour-deposited mirrors (NASA Langley LIDAR, delivered 1990) and in power diodes. Si3N4 in bearings (a NASA evaluation of rolling elements) and propulsion parts. AlN in substrates, IC packages, heat sinks, heater plates and chucks. Y-TZP in blades, cutting edges, pump and valve parts. Density: Y-TZP 6.02 to 6.07 g/cm³, alumina 3.72 to 3.9, AlN 3.25 to 3.3, SiC 3 to 3.165 and dense Si3N4 3.16 to 3.37.
Each has its own trap. Alumina is brittle, and its profile has no outgassing result. AlN hydrolyzes slowly in water; PubChem says water decomposes it to ammonia and aluminium hydroxide. Y-TZP loses strength in hot water or steam, but the evidence is thin: a cited study reports about 12% flexural loss, quoted in a dental paper whose autoclave test (134 °C for 5 h) stands for about 30 years in the mouth. The profile has no service-temperature limit for humid exposure. PubChem lists H350i (may cause cancer by inhalation) for silicon carbide, though 70.3% of reports say it does not meet GHS criteria.
Space evidence is thin and mostly ground tests under narrow conditions. NASA tested Si3N4 balls in 57-mm turbopump bearings, and the report describes spalling with the cause unknown. In one test reactor, alumina of unstated purity was held at about 1,200 K to 2.3 to 2.6 × 10²¹ neutrons/cm²; its electrical resistance rose 3 to 5 times and grain boundaries weakened, while earlier tests found a five to tenfold rise or decreases. Synthetic fused silica showed no change in transmission in one solar-cover-glass test at 1 × 10¹⁵ 1 MeV electrons/cm².
Terms used
- Flexural strength
- the stress at which a ceramic bar breaks when it is bent, also called modulus of rupture. Three-point, four-point and biaxial set-ups give different numbers (general engineering knowledge, not in our data).
- Elastic modulus
- stiffness: how much stress it takes to stretch or bend a material by a given amount. Higher means stiffer.
- Fracture toughness
- resistance to crack growth. Low values mean a small flaw can break the part suddenly.
- Weibull modulus
- a measure of how consistent ceramic strength is from part to part. A higher number means less scatter.
- Sintering
- pressing fine powder into shape and heating it until the grains fuse. The route (reaction-bonded, hot-pressed, gas-pressure-sintered) changes the numbers.
- Vickers and Knoop hardness
- two indentation tests with different diamond shapes. Their numbers are not interchangeable.
- Thermal shock
- cracking caused by a sudden temperature change. Suppliers test and report it in different ways.
- Hydrolysis
- a slow chemical breakdown caused by water.
What to check before you order
- Name the grade and route, not the material: AD-995 or AD-96 alumina, SN-GP or NBD-200 Si3N4, AlN 180, AlN-HP or AlN-HTC, 3Y-TZP, and direct-sintered, reaction-bonded or chemical-vapour-deposited SiC. Ask for the density of that grade.
- Ask for flexural strength with the test method (three-point, four-point or biaxial), specimen size, test temperature, and whether the value is typical or a guaranteed minimum. None of the values in our data is labelled a guaranteed minimum.
- Ask for the Weibull modulus, the number of test pieces and the toughness method.
- Ask what a maximum use temperature means: load, atmosphere and exposure time. Ask how thermal shock was tested.
- For zirconia in hot water or steam, ask for ageing data at your temperature. For AlN, ask for moisture limits. For SiC and Si3N4 dust, ask for the safety data sheet.
What do we not know yet?
- Fracture toughness and hardness are not charted: toughness is stored by different methods and unit spellings, and hardness in Vickers GPa, Knoop GPa and kg/mm2.
- No strength at temperature, creep or oxidation lifetime is stored.
- Among the five, Weibull modulus exists only for SiC and Si3N4 SN-GP, and a thermal shock figure only for alumina and zirconia (plus a calculated parameter for Si3N4). AlN has no use temperature, and its CoorsTek thermal conductivity is missing.
- The zirconia profile covers 3Y-TZP only. Eight-mole-percent YSZ, other stabilisers and alumina-toughened grades are not stored, and neither is a humid service-temperature limit.
- Space evidence is limited to NASA ground tests and one mirror report. No outgassing figure is stored for any of the five, and no cost or machining-difficulty data is stored.
Questions people ask
Is zirconia or alumina stronger?
Zirconia is stronger in bending. Y-TZP shows 1,000 to 1,240 MPa in two datasheets, against 379 MPa for AD-995 alumina, and its toughness is 10 to 13 MPa·m½ against 4 to 5. Alumina is rated to a higher temperature (1,750 °C against 800 to 1,500 °C) and conducts heat better (30 against 2.2 to 3 W/(m·K)).
Is silicon carbide or silicon nitride better for bearings?
Our data supports silicon nitride for bearings: NASA tested Si3N4 balls in turbopump bearings, and SiC has no bearing test here. On hardness, treat the comparison as indicative: NIST sintered SiC is 23.9 GPa Vickers, while Qsil's SN-GP is 14.5 GPa with an ambiguous method footnote.
Can aluminium nitride replace alumina?
Only after checks. Accuratus says metallisation lets AlN replace alumina and BeO in many electronic uses, and that is a supplier statement. AlN conducts 140 to 180 W/(m·K) against 30 for alumina, but its expansion is 5.5 to 5.6 ppm/K (25 to 1,000 °C) against 8.2, it hydrolyzes slowly in water, and metallisation is not in our data. Check all three.
Can zirconia be used in hot water or steam?
Not without ageing data. Y-TZP transforms from its tetragonal form in hot water or steam. The profile cites about 12% flexural strength loss, a figure a dental paper takes from another study, and that paper's autoclave test (134 °C for 5 h) is an accelerated one, equal to about 30 years in the mouth. The profile sets no humid service limit, so ask the supplier for ageing results for your grade.
Which has the lowest thermal expansion?
Fused silica glass, at 0.55 ppm/K from 20 to 320 °C, is lowest of everything compared here. Among the five ceramics it is silicon nitride, at 3.2 ppm/K for SN-GP from room temperature to 1,000 °C; the others run from 4.4 (SiC) to 10.3 (Y-TZP) over comparable ranges.
Are these ceramics hazardous to machine or handle?
The hazards sit with powder and dust, not finished parts. PubChem lists H350i for silicon carbide, though 70.3% of reports say it does not meet GHS criteria. AlN carries irritation codes and decomposes in water. Zirconia itself is Not Classified in PubChem (93.2% of reports); the yttria stabiliser powder carries H315, H319 and H335. Alumina is mostly unclassified, with minority H335 and H372 entries. Read the supplier safety data sheet.
Why do suppliers quote different numbers for the same ceramic?
Grade, purity, route and test method all change the figure. Alumina density is 3.72 for a 96% grade and 3.9 for 99.5%. Si3N4 density runs 2.53 to 3.37 across reaction-bonded to hot-pressed forms. Test methods, hardness loads and temperature ranges differ. Quote the grade, source and test basis together.
How this was made
Each material is a profile in the Space Mushroom database, drafted by AI assistants from cited pages; a script checks that each value appears in the source it cites, and each profile page lists its sources. Every number in this article is either resolved from those profiles when the page is built or checked against them, so it cannot drift from the data. Ranges run across the conditions each source reports (temper, form, heat treatment). A named human reviewer has not yet signed this article. Related: Alumina (Al2O3), Silicon carbide (SiC), Silicon nitride (Si3N4), Aluminium nitride (AlN), all collections, standards.
Last checked 25 October 2026. Next review when a profile changes.