Technical Ceramics: Alumina, Zirconia, Silicon Nitride & Silicon Carbide — The 33 µm Flaw That Caps a 350 MPa Alumina, the 597 K Thermal-Shock Parameter That Makes Si₃N₄ a Turbine Material, and the 18% Sintering Shrinkage That Sets Every Mold Dimension

Metals are designed against yield; ceramics are designed against the largest flaw in the part. That single difference reorganises everything — the 99.5% alumina datasheet says 350–379 MPa flexural strength, but the Griffith equation says a 33 µm defect is enough to break it at 350 MPa, so the number in the catalogue is a population parameter, not a material constant. A 4× volume increase at Weibull m = 10 drops 800 MPa to 696 MPa. Thermal shock separates the grades faster than strength does: R = σ(1−ν)/(Eα) gives 92 K for alumina, 325 K for 3Y-TZP, 597 K for Si₃N₄ and 191 K for SSiC, and the measured R1 of gas-pressure-sintered Si₃N₄ is 590–620 K — which is why silicon nitride survives a turbine start and alumina does not. Densification is where the money is: 55% green density to 99.5% fired is 17.9% linear shrinkage, so every mold and every CAD model is scaled by 1.218 before the first part exists. Diamond grinding removes ceramic at 100–500 J/mm³, 20–100× the specific energy of steel, and the Bifano limit — d_c = 0.15(E/H)(K_IC/H)² — is 0.26 µm for alumina but 0.063 µm for SiC, which is the difference between a ground surface and a crumbled one. This guide works the full stack: grade-by-grade datasheets, Weibull design allowables and proof testing, transformation toughening (3–5 vol% expansion worth 4.6 MPa√m), R-curves and slow crack growth, pressing/CIP/tape casting/CIM/AM, sintering and HIP schedules, ductile-regime grinding, twenty design rules, AlN power-module substrates at 0.037 K/W, Si₃N₄ hybrid bearings at 496 N versus 1,209 N of ball centrifugal load at 30,000 rpm, inspection, fractography, failure modes, and an Indian cost model at ₹270 per part against ₹200 for machined 316.

·

Technical Ceramics: Alumina, Zirconia, Silicon Nitride & Silicon Carbide

Every guide in this series has sized a part by a strength number and then subtracted a safety factor. This one cannot start that way, because a ceramic does not have a strength number. It has a strength distribution, and the distribution is set by the largest processing defect that survived inspection.

That is not a philosophical distinction. It changes the drawing, the process route, the inspection plan and the price. A 99.5% alumina datasheet quotes 350–379 MPa flexural strength (ASTM F417); the same body will fail at 210 MPa if the grinding direction leaves a 60 µm edge chip in the tensile surface, and it will hold 500 MPa if the part is small enough that the stressed volume contains no large flaw at all. The catalogue number is the 63rd percentile of a Weibull population, not a property of Al₂O₃.

Five numbers frame the whole subject.

The first is 33 µm. For 99.5% alumina at 350 MPa with K_{IC} = 4 MPa√m and a surface crack Y = 1.12, the Griffith relation gives a critical depth of 33 µm. A part that is "350 MPa material" is a part whose largest surface flaw is smaller than a third of a hair. Everything in ceramic manufacturing — powder size, spray-dried granule defects, die-wall friction cracks, sintering pores, diamond-grinding scratches — is an attempt to keep that number true.

The second is 696 MPa. That is what 800 MPa becomes when the part volume is four times the test bar, at a Weibull modulus of m = 10. Size scaling is not a correction factor bolted onto the end of the calculation; it is the reason a 4 mm test bar and a 400 mm kiln roller cannot share a datasheet.

The third is 597 K. That is R = \sigma_f(1-\nu)/(E\alpha) for gas-pressure-sintered silicon nitride, and the measured thermal-shock parameter R_1 for the same grade is 590–620 K. Alumina's R is 92 K. This one index explains why silicon nitride turbocharger rotors and glow plugs exist, why alumina is a liner and not a turbine blade, and why 3Y-TZP — the strongest ceramic in the family at 900–1,200 MPa — is a poor choice for thermal cycling despite its toughness.

The fourth is 17.9%. A green body pressed to 55% of theoretical density and sintered to 99.5% shrinks linearly by 1-(0.55/0.995)^{1/3} = 0.179. Every mold cavity, every tape-cast outline, every "print then sinter" STL is scaled by 1/(1-0.179) = 1.218 before anything is made. Miss the number and the part is 22% oversize; get it right to ±0.2% and you can hold ±0.5% as-fired.

The fifth is 4.16 GPa. That is the peak Hertzian contact pressure of a 10 mm silicon nitride ball pressed against steel at 500 N. Ceramic rolling elements do not fail by yield; they fail by subsurface shear at roughly 0.31p_0, and the whole hybrid-bearing business — 496 N of centrifugal load for a ceramic ball versus 1,209 N for the same ball in steel at 30,000 rpm — is an exercise in keeping that pressure below the material's fatigue limit.

This guide runs in thirteen sections:

1. The Physics: Why a Ceramic Has No Yield Strength

1.1 The bond is the problem

Metals are metallic-bonded: delocalised electrons, close-packed planes, and dislocations that glide at a Peierls stress of order 10^{-4} to 10^{-3} of the shear modulus. That is why a 316 stainless part yields at 205 MPa, work-hardens, and gives you a warning before it breaks. The plastic zone at a notch root redistributes stress; the part bends before it fractures; the design allowable can be a fraction of yield.

Ceramics are ionic or covalent. Al₂O₃ is a corundum lattice of Al³⁺ in octahedral sites and O²⁻ in hexagonal close packing, with a band gap of 8.8 eV; Si₃N₄ is a covalently bonded network of SiN₄ tetrahedra sharing corners; SiC is 88% covalent in a zinc-blende/wurtzite stacking. Dislocation motion requires breaking and re-forming directional bonds, so the Peierls stress is a substantial fraction of the shear modulus and slip systems are few. At room temperature, a polycrystalline ceramic is effectively elastic to failure:

There is no notch blunting, no load redistribution, no audible warning. A ceramic part stores elastic energy right up to the moment it releases it, which is why ceramic fracture is explosive and why the fragment velocity is high enough to be a hazard in its own right.

1.2 Griffith, and the flaw that sets the strength

Inglis (1913) showed that an elliptical hole of length 2a and tip radius \rho concentrates stress by 1 + 2\sqrt{a/\rho}. Griffith (1921) turned that into an energy balance: a crack grows when the elastic energy released exceeds the surface energy created. Irwin's stress-intensity formulation gives the working form used in every ceramic design office:

The square root is the whole story. Halve the flaw and you gain only 41% strength. Double the toughness and you gain 100%. Increase hardness without increasing toughness and you gain nothing at all in strength — which is why SiC, at 2,500 HV, is not stronger than alumina at 1,600 HV.

Worked for 99.5% alumina, K_{IC} = 4 MPa√m, Y = 1.12 for a surface semi-elliptical crack:

Now run the same equation for 3Y-TZP at 1,000 MPa with K_{IC} = 9 MPa√m: a_c = 32.1 µm. The zirconia is nearly three times stronger, but it tolerates exactly the same flaw size, because its toughness rose by the same factor as its strength. This is the most useful sanity check in ceramic design: strength and toughness usually move together, so the critical flaw size does not change much between grades. What changes is the stress you can carry at that flaw size.

1.3 The four flaw populations

A ceramic part fails from whichever of these is largest in the tensile surface, and each one belongs to a different part of the process:

Flaw population · Size range · Where it comes from · How to remove it

Powder / granule defects · 20–200 µm · Spray-dryer hollow granules, agglomerates, foreign inclusions (Fe, Si) · Powder selection, granule classification, magnetic separation, isostatic pressing

Forming defects · 50–500 µm · Die-wall shear cracks, binder-rich seams, weld lines in CIM, tape-cast pinholes · Pressing cycle, mold design, feedstock rheology

Sintering defects · 5–100 µm · Residual porosity, large grain clusters, exaggerated grains, HIP-inaccessible surface pores · Sintering schedule, dopants, HIP

Machining defects · 5–100 µm · Grinding scratches, edge chips, subsurface median cracks, grit pullout · Ductile-regime grinding, edge chamfers, polishing, proof testing

The machining population is the one engineers control last and underestimate most. A ground alumina surface contains a residual stress layer 2–5× the grit depth, with median cracks extending normal to the surface. The strength of a ground 99.5% alumina is typically 250–320 MPa; the same body polished is 350–400 MPa. The grinding operation can cost 30% of the strength before the part ever sees a load.

1.4 What this does to the design method

Design question · Metal answer · Ceramic answer

Failure criterion · Yield (\sigma_y), then fatigue · Fracture from the largest flaw (K_{IC}, \sigma_f)

Material property used · R_{p0.2}, S_{ut} · Weibull \sigma_0 and m at a stated volume and surface condition

Safety factor · 1.5–3 on yield · 2–4 on the design allowable, plus proof testing

Stress concentration · Plastic redistribution, K_t partially relieved · Not relieved at all; K_t applies in full

Failure warning · Yielding, deformation, cracking · None

Inspection · Dimensions, NDT of welds · Density, penetrant, proof test, CT, fractography

Most important drawing note · Material spec and heat treatment · Material grade, process route, surface finish, and proof load

The practical consequence: a ceramic drawing is a process specification, not a material specification. "Al₂O₃ 99.5%" tells you almost nothing. "Al₂O₃ 99.5%, isostatically pressed, sintered 1,650 °C 2 h in air, HIP 1,500 °C/150 MPa, diamond ground D46, edges chamfered 0.3×45°, proof tested to 1.5× service load, 100% fluorescent penetrant" tells you what the part will actually do.

2. The Grades: What Each Ceramic Is Actually For

2.1 Alumina (Al₂O₃)

Alumina is the workhorse: cheap, hard, electrically insulating, chemically inert, and available in purities from 85% to 99.9%. The purity controls the amount of glassy silicate phase at the grain boundaries, and the glassy phase controls everything that matters at high temperature.

Property · 96% Al₂O₃ · 99.5% Al₂O₃ · 99.9% Al₂O₃

Density (g/cm³) · 3.70–3.75 · 3.89–3.92 · 3.95

Flexural strength (MPa) · 200–300 · 350–379 · 350–400

Young's modulus (GPa) · 300–320 · 370–372 · 380

Fracture toughness (MPa√m) · 3.5–4.5 · 4.0–5.0 · 3.5–4.5

Vickers hardness (HV) · 1,100–1,300 · 1,600–1,650 · 1,600–1,700

Thermal conductivity (W/m·K) · 20–25 · 30–32 · 35

CTE, 25–1,000 °C (10⁻⁶/K) · 7.4–8.0 · 8.2 · 8.0

Dielectric strength (kV/mm) · 9 · 8.7–17* · 17

Dielectric constant (1 MHz) · 9.0 · 9.8 · 9.9

Max service temperature (°C) · 1,400–1,500 · 1,700 · 1,750

Relative cost (×96%) · 1.0 · 2.5–4 · 5–8

\* 8.7 kV/mm is the ASTM D116 value on a 6.25 mm specimen; thin sections (0.5–1 mm) measure 15–20 kV/mm because the breakdown path is shorter and the electrode area smaller. Always quote thickness with dielectric strength.

The 96% grade is the wear-tile and insulator grade: Al₂O₃ grains bonded by a calcium-magnesium-alumino-silicate glass. The glass gives it 200–300 MPa and limits it to about 1,450 °C, but it also makes the material cheap (₹400–900/kg in tile form). The 99.5% grade uses 0.05–0.2% MgO as a grain-growth inhibitor — the classic sintering aid that keeps the final grain size at 2–5 µm instead of 20 µm — and delivers 350–379 MPa with a dielectric loss of 2\times10^{-4}. That is the grade for feedthroughs, vacuum insulators, wear plates and thick-film substrates. The 99.9% grade is for semiconductor and optical work where Na and Fe at the 100 ppm level are contamination.

2.2 Zirconia (ZrO₂)

Zirconia is the tough one. Pure ZrO₂ undergoes a martensitic tetragonal→monoclinic transformation on cooling with a 3–5% volume expansion and 7–8% shear, which shatters a large-grained body. Alloying with 3 mol% Y₂O₃ (3Y-TZP) retains the tetragonal phase metastably at room temperature; a propagating crack then triggers the transformation in a zone ahead of its tip, and the expansion compresses the crack shut. That is transformation toughening, and it is worth 4–6 MPa√m of apparent toughness.

Property · 3Y-TZP · Mg-PSZ · ZTA (10–20% ZrO₂ in Al₂O₃)

Density (g/cm³) · 6.00–6.05 · 5.7–5.9 · 4.1–4.3

Flexural strength (MPa) · 900–1,200 · 500–700 · 600–800

Young's modulus (GPa) · 200–210 · 200–210 · 340–360

Fracture toughness (MPa√m) · 8–10 · 8–12 · 6–8

Vickers hardness (HV) · 1,200–1,300 · 1,000–1,100 · 1,400–1,600

Thermal conductivity (W/m·K) · 2.4–2.5 · 2.5 · 20–25

CTE (10⁻⁶/K) · 10.5 · 10.0 · 8.5

Max service temperature (°C) · 1,000 · 1,000 · 1,200

Thermal-shock R (K) · 325 · 300 · 157

The weakness is not mechanical, it is hydrothermal. At 200–300 °C in the presence of water or steam, the tetragonal phase transforms spontaneously — low-temperature degradation (LTD) — grain by grain, from the surface inward. The transformation is accompanied by microcracking and grain pullout; strength falls, surface roughness rises, and the process is autocatalytic. ISO 13356 codifies an accelerated test: 5 h at 134 °C and 2 bar in an autoclave, which is taken as roughly equivalent to 15 years at body temperature, with acceptance criteria of ≤25 mol% monoclinic phase and ≤20% strength loss. A 3Y-TZP that passes the 800 MPa / Weibull ≥8 entry requirement can still fail the aging test if the grain size exceeds 0.5 µm or the yttria distribution is inhomogeneous.

Consequences for design: never use 3Y-TZP in steam above 150 °C, in hot water, or in a condensing humid environment; specify a 0.5–1 µm grain size with a certified aging test; do not grind it with a coarse wheel (grinding triggers transformation and raises the monoclinic fraction before the part is even in service); and for medical or dental work, require ISO 13356 or ISO 6474-2 compliance with a batch certificate.

2.3 Silicon nitride (Si₃N₄)

Silicon nitride is the structural ceramic: the best combination of strength, toughness and thermal shock in the family. It is not a single material. Gas-pressure-sintered (GPS) grades run 700–900 MPa with K_{IC} 6–8; hot-isostatically-pressed (HIP) grades reach 900–1,200 MPa with K_{IC} 8–9; sintered reaction-bonded grades are cheaper and weaker.

Property · GPS Si₃N₄ · HIP Si₃N₄ · High-conductivity Si₃N₄

Density (g/cm³) · 3.20 · 3.18–3.26 · 3.2

Flexural strength (MPa) · 900–1,200 · 760–830 (4-pt) · >800 (ball-on-ring)

Young's modulus (GPa) · 300–310 · 300–310 · 336

Fracture toughness (MPa√m) · 8.0–9.0 · 6.2–6.5 · —

Vickers hardness (HV) · 1,500–1,600 · 1,530–1,560 · —

Thermal conductivity (W/m·K) · 20–24 · 22 · 80 ± 10

CTE (10⁻⁶/K) · 3.0–3.8 · 3.2 · 3.0–3.5

Thermal-shock R (K) · ~597 · ~600 · —

Max service temperature, air (°C) · 1,200 · 1,200 · 800

Dielectric strength (kV/mm) · 19 · 19 · >15

The toughening mechanism is microstructure: β-Si₃N₄ grows as elongated, interlocking grains in a glassy grain-boundary phase, and crack propagation requires either breaking those grains or pulling them out. The result is a rising R-curve and a toughness that is 2× alumina's with a lower elastic modulus. It is also why Si₃N₄ fails "gracefully" for a ceramic — spalls rather than shatters — and why it is the only ceramic used in rolling bearings at scale.

The limitation is oxidation above about 1,000 °C in air: a silica-rich glass forms on the surface, and at 1,200 °C the grain-boundary glass softens and strength falls. Silicon nitride is a 1,200 °C material, not a 1,600 °C material, and its high-temperature capability depends on the sintering additives (Y₂O₃ + Al₂O₃ + MgO systems) and on whether the glass devitrifies.

2.4 Silicon carbide (SiC)

Silicon carbide is the hard, thermally conductive, chemically resistant grade. Two variants dominate:

Property · SSiC (pressureless sintered) · RBSC (reaction bonded / SiSiC)

Composition · ≥98% SiC, B/C aids · 85–92% SiC + 8–15% free Si

Density (g/cm³) · 3.10–3.16 · 3.00–3.05

Flexural strength (MPa) · 350–500 · 250–300

Young's modulus (GPa) · 410–430 · 330

Fracture toughness (MPa√m) · 3.5–4.5 · 3.5–4.0

Vickers hardness (HV) · 2,400–2,800 · 1,800–2,200

Thermal conductivity (W/m·K) · 120–170 (up to 270 semiconductor grade) · 40–120

CTE (10⁻⁶/K) · 4.0–4.5 · 4.2–4.5

Max service temperature, air (°C) · 1,500–1,650 · 1,350–1,400

Thermal-shock R (K) · 191 · ~200

SSiC sinters at 2,050–2,200 °C in argon with 0.5–1.0% boron and 0.5–1.5% carbon as sintering aids, reaching ≥98% theoretical density. RBSC is made by infiltrating a porous SiC–carbon preform with molten silicon at about 1,450 °C; the silicon reacts with carbon to form secondary SiC and the excess fills the remaining porosity, so the part is near-net-shape with essentially zero shrinkage — the reason large, complex RBSC kiln furniture and heat-exchanger tubes are affordable. The free silicon limits service temperature to about 1,350 °C and reduces corrosion resistance in strong alkalis.

Silicon carbide's combination of 120+ W/m·K and 4.2×10⁻⁶/K gives it the highest thermal-shock resistance of the structural grades after Si₃N₄, and its hardness (2,500 HV) makes it the standard for mechanical seal faces, slurry nozzles, wear plates and semiconductor process components. It is also the standard susceptor and wafer-carrier material in CVD and epitaxy, where high purity and thermal conductivity matter more than strength.

2.5 Aluminium nitride (AlN)

AlN exists for one reason: 170–230 W/m·K of thermal conductivity in an electrical insulator, with a CTE of 4.5×10⁻⁶/K that matches silicon and GaN. Beryllium oxide (BeO) is slightly better thermally (200–250 W/m·K) and is effectively obsolete because beryllia dust is a chronic lung toxin. AlN is the replacement.

Property · AlN (PCAN1000 class)

Thermal conductivity (W/m·K) · 170 (200 and 230 grades available)

CTE (10⁻⁶/K) · 4.5

Young's modulus (GPa) · 310–330

Flexural strength (MPa) · 300–450

Fracture toughness (MPa√m) · 3.0–3.5

Dielectric constant (1 MHz) · 8.8–9.0

Dielectric strength (kV/mm) · 15–20

Density (g/cm³) · 3.26

Max service temperature, air (°C) · 800–1,000 (oxidation-limited)

The oxygen content of the powder and the sintering additive (Y₂O₃, CaO) control the conductivity: oxygen dissolved in the AlN lattice scatters phonons, so a 1 wt% oxygen pickup can drop 170 W/m·K to 90. That is why AlN substrate prices track powder purity and why a cheap AlN is not the same material as a 170 W/m·K substrate.

2.6 The specialists

Material · Key numbers · Use

Macor (machinable glass-ceramic) · 94 MPa flexural, 1.46 W/m·K, CTE 9.3×10⁻⁶/K, no shrinkage, machinable to ±0.013 mm with carbide · Prototype insulators, vacuum fixtures, feedthrough bodies

Mg-PSZ · 500–700 MPa, K_{IC} 8–12, 5.8 g/cm³ · Wear parts needing maximum toughness, not strength

ZTA · 600–800 MPa, K_{IC} 6–8, 4.2 g/cm³ · Cutting tools, wear parts, cost between alumina and TZP

SiC/SiC CMC · 2.7–3.0 g/cm³, 250–400 MPa, k 10–30 W/m·K, to 1,300 °C · Turbine shrouds, combustor liners, re-entry thermal structures

B₄C · 2,500–3,000 HV, 2.52 g/cm³ · Nozzles, armor, neutron absorbers

PZT (functional) · d_{33} 300–600 pC/N · Actuators, sensors, ultrasonic transducers

BaTiO₃ (functional) · \varepsilon_r 2,000–5,000 · MLCC dielectrics

ZnO (functional) · Nonlinear I–V, \alpha 30–50 · Varistors, surge arresters

The functional ceramics are a different business — electronic-grade powders, co-firing with base-metal electrodes, and tolerance measured in ppm — but they share the same powder-processing and sintering physics as the structural grades.

3. Statistics as Design: Weibull, Volume and Proof Testing

3.1 The distribution

Ceramic strength follows the two-parameter Weibull distribution, which is the weakest-link model applied to a volume containing a random population of flaws:

where P_f is the probability of failure at stress \sigma, V is the stressed volume (or effective volume), V_0 is a reference volume, \sigma_0 is the characteristic strength (63.2% failure probability at V = V_0), and m is the Weibull modulus — the shape parameter that measures scatter.

The modulus is the design-relevant number, not the strength. m = 5 is a poor, variable process; m = 10 is a good commercial grade; m = 15–20 is an excellent, tightly controlled grade. A silicon nitride with m = 15 has a strength distribution narrow enough that the 1% quantile is roughly 60% of the median; at m = 5 the same 1% quantile falls to about 30%.

3.2 Size scaling

Because failure is governed by the largest flaw in the stressed volume, a larger part is weaker:

Worked: a 3 × 4 × 20 mm four-point bend specimen has a uniform-moment gauge volume of 240 mm³ and measures 800 MPa at m = 10. A production part with four times that stressed volume — a 40 mm long roller, say — carries

Lose 13% of the strength by making the part bigger, with no change whatsoever to the material. This is why a supplier's test-bar data cannot be applied directly to a 300 mm kiln roller, and why ISO 20501 / ASTM C1239 require the effective volume and the test method to be stated with the strength.

3.3 Three-point versus four-point bending

The two tests do not measure the same number. In three-point bending the moment peaks on a single line under the centre roller, so the highly stressed volume is small. In four-point bending the moment is constant between the inner rollers, so a much larger volume sees the peak stress — and therefore more flaws see it. For the same material, four-point strengths are consistently lower than three-point strengths, typically by 10–25% at m = 10. Never compare a 4-point datasheet number with a 3-point drawing requirement.

ISO 14704 and ASTM C1161 specify the specimen (3 × 4 × 45 mm is standard), the spans (20/40 mm for 4-point), and the loading rate (0.5 mm/min). A test programme needs at least 30 specimens for a usable m estimate — ASTM C1239 says so explicitly, and the confidence interval on m from 10 specimens is so wide that the number is not worth quoting.

3.4 The design allowable

Invert the Weibull equation for a target failure probability:

For m = 10 and P_f = 10^{-3}, the first bracket is 0.501; for P_f = 10^{-4} it is 0.398. So a design allowable of half the characteristic strength buys one failure in a thousand, and 40% of characteristic strength buys one in ten thousand — before any safety factor for load uncertainty, surface damage in service, or subcritical crack growth.

The industry convention is a design factor of 2–3 on top of the statistical allowable for non-critical parts, and 4 for pressure-retaining or human-rated ceramic components. That sounds conservative until you remember there is no yield: the first overload is the last overload.

3.5 Proof testing: buy certainty with one load cycle

Proof testing loads every part to a stress above the service stress, then screens out the ones that fail. The survivors have a bounded maximum flaw size, and their failure probability at service stress drops by the ratio:

With m = 10 and a proof stress twice the service stress, the exponent gives 2^{10} = 1{,}024 — a part population that would have failed at 1 in 10 in service now fails at 1 in 10,000. Proof testing at 1.5× service (the usual specification) gives 1.5^{10} = 57.7.

Two caveats. First, proof testing damages the part: the proof load can extend a subcritical crack, so a proof-tested part is not the same part it was before. Second, it only works if the proof stress state matches the service stress state — a proof test in uniaxial compression tells you nothing about a part that fails in bending. For critical ceramic components, proof test at temperature, in the same fixturing, in the same stress mode.

4. Thermal Shock: The Index That Separates the Grades

4.1 The parameter

A sudden temperature change on the surface of a ceramic generates a biaxial stress proportional to the constrained thermal strain:

Failure occurs when \sigma_{th} reaches \sigma_f, which defines the critical temperature difference and the material index:

R has units of kelvin and is a pure material index. A second index adds conductivity, because a material that conducts heat well never develops the full \Delta T across its section:

Material · \sigma_f (MPa) · E (GPa) · \nu · \alpha (10⁻⁶/K) · k (W/m·K) · R (K) · R'' (W/m)

96% Al₂O₃ · 250 · 310 · 0.22 · 7.6 · 22 · 79 · 1,740

99.5% Al₂O₃ · 350 · 370 · 0.22 · 8.0 · 30 · 92 · 2,760

ZTA · 600 · 350 · 0.22 · 8.5 · 22 · 157 · 3,450

3Y-TZP · 1,000 · 205 · 0.30 · 10.5 · 2.5 · 325 · 813

SSiC · 400 · 410 · 0.16 · 4.3 · 120 · 191 · 22,900

Si₃N₄ · 800 · 310 · 0.26 · 3.2 · 30 · 597 · 17,900

AlN · 350 · 320 · 0.24 · 4.6 · 170 · 181 · 30,800

Read the table in two directions. On R alone, zirconia looks three times better than alumina and half as good as silicon nitride. On R'' — the index that accounts for heat flow — zirconia is the worst structural ceramic in the list, because its thermal conductivity is 2.5 W/m·K. A zirconia part quenched in water develops the full \Delta T across a thin surface layer, and the surface goes into tension before the bulk has any chance to follow. That is why 3Y-TZP survives 1,000 MPa of steady bending and cracks when a hot one is dropped into a bucket.

4.2 Checking the arithmetic against the catalogue

For 99.5% alumina at \Delta T = 200 K, fully constrained:

That is twice the 350 MPa strength, so a fully constrained 200 K quench must crack — and the measured water-quench \Delta T_c for CoorsTek AD-995 is exactly 200 °C, from a test in which the constraint is partial and the surface stress is biaxial. The index and the test agree within the uncertainty of the heat-transfer coefficient, which is the only reason the index is useful.

For gas-pressure-sintered Si₃N₄ the datasheet gives R_1 = 590–620 K; the formula gives 597 K. At \Delta T = 600 K the constrained stress is about 800 MPa, equal to the strength. The material is at its limit at 600 K and comfortable at 300 K. Alumina is at its limit at 200 K. That is the entire difference between a ceramic that can be a turbocharger rotor and one that cannot.

4.3 The Biot number, and why measured \Delta T_c is size-dependent

The R parameter assumes the surface reaches the bath temperature instantly — an infinite Biot number. Real parts have a finite h and a finite thickness, so the surface stress is reduced by a factor that depends on

Water quenching (h \approx 10{,}000–30{,}000 W/m²K) on a 5 mm alumina section gives Bi = 5–15, close to the infinite-constraint assumption, which is why water quench is the standard test. Air cooling (h \approx 50–200 W/m²K) gives Bi = 0.05–0.2, and the same part survives a much larger \Delta T. This is why a ceramic that survives an air-cooling cycle can fail in a water quench at the same nominal temperature difference, and why thermal-shock specifications must state the coolant, the flow rate and the section thickness.

4.4 Design rules for thermal shock

  1. Minimise the constrained dimension. Thermal stress scales with section thickness; a 3 mm wall survives what a 10 mm wall cannot.
  2. Avoid tensile stress on the surface. Put the ceramic in compression (a shrink-fitted metal band, a supporting sleeve), and radius every transition.
  3. Use Si₃N₄ or SiC for cyclic thermal load. Alumina for steady-state insulation, zirconia for wear at moderate temperature.
  4. Heat and cool slowly. A 5 °C/min ramp is standard for alumina above 1,000 °C; 2 °C/min for thick sections.
  5. Do not clamp a ceramic between materials with a higher CTE. A steel bolt through an alumina flange will crack the flange on heat-up; use Belleville washers and a clearance fit.
  6. Test the actual assembly, not the material. The R index ranks materials; only a system-level quench test ranks designs.

5. Toughness: Where the 4.6 MPa√m Comes From

5.1 Transformation toughening

The stress field at a crack tip in 3Y-TZP exceeds the critical transformation stress over a zone of radius h (typically 1–5 µm). The tetragonal grains in that zone transform to monoclinic, expanding 3–5% in volume and 7–8% in shear, and the surrounding untransformed material pushes back. The crack faces are clamped shut by a compressive stress of order 300–800 MPa. The toughening increment is approximated by

With E = 205 GPa, \varepsilon_T = 0.05 and h = 2 µm:

Add the 3 MPa√m base toughness of the untransformed lattice and you get 7.6 MPa√m — the measured range is 8–10, and the difference is the contribution of microcrack toughening and the R-curve. The equation also shows why zirconia toughness collapses above 800 °C: the transformation is thermally activated in reverse, and the metastable tetragonal phase becomes stable, so \Delta K \to 0. A zirconia part that is tough at 20 °C is not tough at 900 °C, and a fire is enough to destroy the transformation zone permanently.

5.2 Bridging and deflection in Si₃N₄

Silicon nitride's toughness comes from microstructure rather than phase transformation. β-Si₃N₄ grows as elongated hexagonal prisms 1–5 µm long; the interlocking network forces cracks to either cut through grains (transgranular, expensive) or follow the glassy boundary around them (intergranular, tortuous). Both paths consume energy, and the elongated grains behind the crack tip bridge the faces and shield the tip. The result is a rising R-curve: K_{IC} measured from a long crack is 6–9 MPa√m, while the toughness at initiation from a sharp machined flaw is closer to 4–5. That difference is why a Si₃N₄ part is more tolerant of a 50 µm grinding scratch than its long-crack toughness suggests, and why the same part is less tolerant of a 500 µm casting pore than the R-curve suggests.

SiC has no comparable mechanism: equiaxed grains, no transformation, no bridging. Its toughness is 3.5–4.5 MPa√m and its strength is flaw-limited at 350–500 MPa. It is harder than Si₃N₄ by 60% and tougher by half as much, which is exactly the trade you make when you specify it for a seal face rather than a bearing.

5.3 Slow crack growth

Ceramics do not fatigue in the metal sense, but they do grow cracks subcritically under sustained load in the presence of water or stress corrosion. The crack velocity follows

with n between 20 and 40 for structural ceramics in water (alumina 25–35, Si₃N₄ 25–40, zirconia 20–30). Because n is large, the time to failure is extremely sensitive to stress: a 10% increase in stress can cut life by a factor of 3–10. The practical design rule is to keep the sustained stress below 50–60% of the short-term strength in humid service, and to use ASTM C1368 constant-stress-rate testing to get n and A for the actual environment.

Static fatigue is the reason proof testing must be done at the service temperature and humidity, and the reason a ceramic part that passes a room-temperature proof test can still fail after a year at 200 °C in steam.

6. From Powder to Part: Forming, Sintering, Shrinkage

6.1 The powder

A structural ceramic starts as a powder with a specification that matters more than the datasheet:

Parameter · Typical requirement · Why it matters

Median particle size D_{50} · 0.3–1.5 µm (alumina 0.5–0.8; zirconia 0.2–0.5; SiC 0.5) · Sintering rate and final grain size

Specific surface area (BET) · 5–15 m²/g · Reactivity; high BET means agglomeration risk

Purity · 99.5–99.9% (Na₂O < 0.05%, Fe₂O₃ < 0.03%, SiO₂ < 0.1%) · Glassy phase and dielectric loss

Agglomerate content · < 1% above 20 µm · Agglomerates become strength-limiting flaws

Sintering aids · MgO 0.05–0.2% (alumina); Y₂O₃ 3 mol% (zirconia); Y₂O₃/Al₂O₃/MgO (Si₃N₄); B + C (SiC) · Densification vs grain growth

Spray-dried granules (50–150 µm, spherical, free-flowing) are what actually fill a die. A hollow granule collapses unevenly during pressing and leaves a lens-shaped density gradient that becomes a delamination after sintering — the classic "press crack" that only appears after firing.

6.2 Forming routes

Route · Green density (% TD) · Shape capability · Tooling cost · Volume · Notes

Uniaxial die pressing · 50–60 · Simple, prismatic, ≤ 50 mm · ₹50k–3L · 1k–1M · Fastest; density gradients with L/D > 2

Cold isostatic pressing (CIP) · 55–65 · Large, complex, near-net · ₹1–5L (bags) · 100–100k · 150–300 MPa; uniform density; green machining possible

Tape casting · 45–55 · Flat, 20 µm–1 mm thick · ₹1–3L (doctor blade) · 10k+ · Slurry viscosity 2,000–4,000 cP; organics 15–25 wt%

Slip casting · 50–60 · Hollow, large, complex · ₹20k–1L (plaster molds) · 10–10k · 12–24 h per cast; sanitaryware, crucibles

Ceramic injection molding (CIM) · 55–62 (solids loading) · Small, complex, 3D · ₹3–15L · 10k–10M · Feedstock 50–62 vol% powder; 18–22% shrinkage

Extrusion · 50–60 · Constant cross-section · ₹50k–2L · 1k+ · Tubes, rods, honeycomb

Additive manufacturing · 40–60 · Freeform, internal channels · none · 1–1k · Binder jet, SLA slurry, material extrusion, DIW

CIM is the process that makes small, complex ceramic parts economic — ferrules, nozzles, dental abutments, watch cases. The feedstock is 55–62 vol% powder in a multi-component binder (paraffin wax + polyethylene/polypropylene backbone + stearic acid), injected at 150–185 °C and 50–150 MPa into a hardened steel mold, then debound in two stages: solvent (heptane at 40–60 °C, removing 70–80% of the binder) and thermal (to 500–600 °C, burning out the backbone). The brown part is 40–45% porous and fragile; the sintering step then closes that porosity and shrinks the part 18–22% linearly.

6.3 The shrinkage equation

Densification is volume conservation of the solid phase. If the green body has relative density \rho_g and the fired body \rho_s, isotropic linear shrinkage is

Worked: 55% green to 99.5% fired gives S = 1-(0.55/0.995)^{1/3} = 0.179, or 17.9%. The mold scale factor is

so a 20.000 mm finished dimension is molded at 24.360 mm. Every dimension on the mold, the tape-cast outline and the printed STL is multiplied by that factor, and the factor is material- and process-specific: 55% → 17.9%, 58% → 16.5%, 60% → 15.4%. A 1% error in green density is a 0.3–0.4% error in the fired dimension — which is why the as-fired tolerance for CIM is ±0.5–1.0% and why precision features are always diamond-ground after sintering, not molded.

Differential shrinkage is the other half of the problem. A wall-thickness change from 3 mm to 8 mm produces a green-density gradient, and the thin section densifies first; the thick section then pulls it into warpage. Design rules: keep wall thickness uniform within ±20%, radius every internal corner, and avoid isolated thick bosses. If the geometry cannot comply, the part must be HIPed or machined from a simple billet.

6.4 Sintering schedules

Material · Atmosphere · Temperature · Hold · Density target · Notes

96% Al₂O₃ · Air · 1,500–1,600 °C · 1–2 h · 97–98% TD · Glassy phase densification

99.5% Al₂O₃ · Air / H₂ · 1,600–1,700 °C · 2 h · 99% TD · MgO-doped; grain 2–5 µm

3Y-TZP · Air · 1,380–1,480 °C · 2 h · >99.5% TD · Grain must stay < 0.5 µm

GPS Si₃N₄ · N₂, 1–10 MPa · 1,750–1,850 °C · 2–4 h · >99% TD · Gas pressure suppresses decomposition

SSiC · Ar · 2,050–2,200 °C · 0.5–1 h · ≥98% TD · B + C aids

RBSC · Vacuum · 1,450–1,550 °C · — · 99% · Si infiltration, zero shrinkage

AlN · N₂ · 1,800–1,900 °C · 2–4 h · >99% TD · Y₂O₃/CaO aids; oxygen control

Three densification routes go beyond pressureless sintering:

The post-sintering microstructure target is a balance: fine enough grains to keep strength (alumina 2–5 µm, zirconia < 0.5 µm, Si₃N₄ 0.5–2 µm) and dense enough to eliminate the pores that act as flaws. Oversintering grows grains and drops strength; undersintering leaves porosity and drops strength. The window is typically 50–100 °C wide, which is why the furnace uniformity specification (±5 °C) matters as much as the setpoint.

7. Diamond Machining: The Ductile-Regime Limit

7.1 Ceramics are ground, not cut

Diamond is the only tool material harder than the structural ceramics (10,000 HV vs 2,500 for SiC), and grinding is the only economical way to hold a tolerance on a fired part. The physics is different from metal cutting: the specific grinding energy for alumina is 100–500 J/mm³, against 5–15 J/mm³ for steel — a factor of 20–100. Almost all of that energy becomes heat in a shallow surface layer, and the heat plus the mechanical load generates the subsurface damage that limits strength.

7.2 The Bifano limit

Bifano, Dow and Scattergood (1991) showed that a brittle material can be removed by plastic flow — "ductile-regime" grinding — if the uncut chip thickness stays below a critical depth:

Material · E (GPa) · H (GPa) · K_{IC} (MPa√m) · d_c (µm)

99.5% Al₂O₃ · 370 · 15.0 · 4.0 · 0.263

3Y-TZP · 205 · 12.7 · 8.0 · 0.961

Si₃N₄ · 310 · 15.5 · 6.0 · 0.450

SSiC · 410 · 25.0 · 4.0 · 0.063

AlN · 320 · 11.0 · 3.2 · 0.369

Silicon carbide's 63 nm limit is the reason it is the most difficult structural ceramic to grind: a D46 wheel has 46 µm mean grit, and even at a 5 µm infeed each active grit cuts several tenths of a micron, well above the ductile limit. The surface therefore fractures at the microscale, and the strength of a ground SiC part is 250–320 MPa against 400–500 MPa polished. Zirconia, with a 0.96 µm limit, grinds almost like a metal and holds strength well — provided the wheel is kept sharp and the part is not overheated, because grinding also triggers the t→m transformation that starts LTD.

7.3 Wheel and parameter specification

A working specification for precision alumina:

Parameter · Value · Reason

Abrasive · Diamond, resin bond · Only harder material; resin absorbs shock

Grit · D46 (≈ 320 mesh) for roughing, D15 for finishing · Grit depth sets the flaw size

Concentration · 75–100 · Enough grits per unit area to share the load

Wheel speed · 25–35 m/s · Higher speed lowers chip thickness at fixed feed

Depth of cut · 5–20 µm rough, 1–5 µm finish · Below the ductile limit for finish passes

Table feed · 100–2,000 mm/min · Sets the undeformed chip thickness

Coolant · Water-based, high flow, filtered · Removes heat, flushes swarf

Specific removal rate for a 10 mm wide grind, 10 µm depth and 10 m/min table speed:

At 200 J/mm³ that is 3.3 kW at the wheel — which is why ceramic grinding spindles are 5–15 kW and why coolant flow, not wheel speed, is usually the limiting factor.

7.4 Subsurface damage and edge quality

Grinding leaves three zones: an outer layer of fractured and pulverised grains (1–3× grit depth), a median-crack zone extending 2–5× grit depth normal to the surface, and a residual-stress layer. For D46 roughing, expect 50–150 µm of damage; a D15 finish pass removes the roughing damage but introduces 10–30 µm of its own. The rule is to remove at least 2× the previous step's damage depth in each subsequent pass, which is why a precision alumina part may see D46 → D25 → D15 → lap.

Edge chipping is the other strength killer. A 90° edge on a ceramic has no plastic zone to blunt the contact stress, so a chip 100 µm deep is a 100 µm flaw in the highest-stress region. Specification: chamfer all edges 0.2–0.5 mm × 45°, or radius 0.3 mm minimum, and inspect at 10× before penetrant testing.

7.5 Alternatives to diamond grinding

8. Design Rules: Twenty Numbers

  1. Design to the tensile surface. Ceramics are 5–10× stronger in compression (2,500 MPa for alumina) than in tension. Put the tensile stress where the volume is smallest and the surface is best.
  2. Radius every internal corner. Minimum 0.5 mm; 1 mm preferred. A sharp corner is a K_t of 3–5 with no plastic redistribution.
  3. Keep walls uniform within ±20%. Differential shrinkage warps anything more.
  4. Minimum wall 1 mm, preferred 2–3 mm. Below 1 mm, handling and pressing defects dominate.
  5. Avoid isolated thick sections. A 10 mm boss on a 3 mm wall will crack at the junction during sintering.
  6. Chamfer all edges 0.2–0.5 × 45°. An unchamfered edge is a chip waiting to happen.
  7. Holes: depth ≤ 5× diameter, diameter ≥ wall thickness. Diamond drilling is slow and the exit edge spalls; specify a sacrificial backing plate.
  8. No molded threads below M6. Use metal inserts, or ground threads with a 1.5× diameter relief.
  9. Metal inserts: shrink-fit with CTE compensation. A steel insert in alumina at 200 °C sees \Delta\alpha \approx 4\times10^{-6}/K and a hoop stress of 100–200 MPa; use a compliant sleeve or a brazed Mo-Mn metallisation.
  10. Specify surface finish by function. As-fired R_a 0.8–3 µm; ground 0.2–0.8 µm; lapped 0.02–0.1 µm. A 0.4 µm finish is worth 20–30% strength over 1.6 µm.
  11. As-fired tolerance ±0.5% (CIM ±0.5–1.0%), ground ±0.01–0.05 mm, lapped ±0.001–0.005 mm. Do not put a ±0.02 mm tolerance on an as-fired dimension; it doubles the price.
  12. Flatness 1–3 µm/mm as-fired, 0.5 µm/mm ground, 0.1 µm/mm lapped. Seal faces need the last number.
  13. Perpendicularity and concentricity: 0.1% of the dimension as-fired, 0.02 mm ground.
  14. Avoid tensile stress from clamping. A bolted joint needs a Belleville washer and a clearance hole 0.5 mm oversize; direct bolt preload will crack the part.
  15. Design for proof testing. Include a defined proof-load direction and a fixture concept in the drawing; a part that cannot be proof tested cannot be certified.
  16. No impact loads, no point loads, no line contacts. Use a compliant pad or a spherical seat.
  17. For thermal cycling, specify Si₃N₄ or SiC, not zirconia or alumina. Check R'', not just R.
  18. For wear, hardness wins; for strength, toughness wins. Alumina for sliding wear, zirconia for impact wear, SiC for abrasive slurry, Si₃N₄ for rolling contact.
  19. For electrical insulation, specify volume resistivity at temperature, not just at 20 °C. Alumina drops from 10¹⁵ Ω·cm at 20 °C to 10⁶ Ω·cm at 1,000 °C.
  20. Put the process route on the drawing. Grade + forming + sintering + HIP + grinding + proof test + inspection. A ceramic drawing without a process route is a request for a quote, not a specification.

8.1 Tolerances and cost

Feature · As-fired · Ground · Lapped · Cost driver

Linear dimension · ±0.5% · ±0.02 mm · ±0.002 mm · Number of grinding setups

Flatness · 2 µm/mm · 0.5 µm/mm · 0.1 µm/mm · Part size, wheel wear

Parallelism · 0.5% · 0.02 mm · 0.005 mm · Fixturing

Surface finish R_a · 0.8–3 µm · 0.2–0.8 µm · 0.02–0.1 µm · Time per cm²

Hole diameter · ±1% · ±0.02 mm · — · Depth/diameter ratio

Edge chip · ≤ 0.3 mm · ≤ 0.05 mm · ≤ 0.01 mm · Handling, chamfer

The cost rule of thumb: as-fired is 1×, ground is 2–4×, lapped is 5–8×. Every tolerance tighter than ±0.05 mm on a ceramic part is a grinding operation, and grinding is charged by the hour (₹400–1,200/h in India), not by the part.

9. Applications, With the Numbers Attached

9.1 Wear: alumina liners and the Archard ratio

Sliding wear follows the Archard relation:

where V is wear volume, W the normal load, L the sliding distance, H the hardness, and k a dimensionless wear coefficient that depends on the tribosystem. For alumina against steel in abrasion, k \approx 10^{-4} and H = 16 GPa; for 1045 steel on steel, k \approx 10^{-3} and H = 2.5 GPa. The ratio k/H is 6.25×10⁻¹⁵ against 4×10⁻¹³ — a factor of 64 in favour of alumina, before any geometry change.

In slurry erosion the measured advantage is smaller, typically 10–30×, because erosion is a combination of cutting, deformation and brittle chipping rather than pure sliding. A 25 mm alumina tile in a cyclone feed chute outlasts a 10 mm mild-steel liner by 2–4 years against 3–6 months — which is why 92–96% alumina tiles, bonded into a steel shell with epoxy, are the default for mining and cement plants in India. The tile itself costs ₹400–900/kg; the downtime it avoids costs ₹50,000–5,00,000 per day.

9.2 Zirconia: plungers, dies and dental

3Y-TZP's combination of 1,000 MPa and 8–10 MPa√m makes it the material for:

The caution is temperature: do not use zirconia above 150 °C in steam, and do not autoclave it repeatedly.

9.3 Silicon carbide: seal faces and heat exchangers

SiC is the standard mechanical-seal face material for pumps handling abrasive or corrosive media. Its 120–170 W/m·K conductivity is the reason: a seal face generates heat at the interface, and the face temperature rise is

For a heat flux of 1 MW/m² through a 5 mm face, alumina (k = 30) gives 167 K, SiC (k = 120) gives 42 K. The lower face temperature keeps the lubricating film from vaporising, which is why SiC seals run at PV values of 10–20 MPa·m/s where alumina seals are limited to 3–5. The pairing is usually SiC against carbon-graphite, not SiC against SiC, because like-on-like contact has poor emergency-running behaviour.

RBSC heat-exchanger tubes exploit the same conductivity: 120 W/m·K against 16 for stainless steel means a third of the area for the same duty, and the zero-shrinkage infiltration route makes 2 m long tubes economical.

9.4 Silicon nitride: hybrid bearings and cutting tools

A 10 mm Si₃N₄ ball weighs 1.68 g against 4.08 g in bearing steel. At 30,000 rpm (ω = 3,142 rad/s) and a 30 mm pitch radius, the centrifugal force on each ball is

against 1,209 N for the steel ball — 59% less. That is the entire speed argument for hybrid bearings: less centrifugal load on the outer race means less contact stress, less skidding, less heat, and a DN limit that moves from about 1.0 million to 1.5–2.0 million mm·rpm. Add the electrical insulation (10¹² Ω·cm against 10⁻⁵ for steel), which blocks VFD shaft currents that flute steel races, and the 3–10× price premium becomes defensible in machine-tool spindles, traction motors and dental turbines.

The contact stress is the check that matters. A 10 mm ball at 500 N on a flat steel plate gives a Hertzian contact radius of 240 µm and a peak pressure of 4.16 GPa:

with 1/E^* = (1-\nu_1^2)/E_1 + (1-\nu_2^2)/E_2. Peak subsurface shear is about 0.31p_0 = 1.29 GPa at a depth of 0.48a ≈ 115 µm — and that is where rolling-contact fatigue initiates. Bearing-grade Si₃N₄ with K_{IC} 6–7 and a HIPed, pore-free microstructure survives that stress for 10⁹ cycles; a sintered grade with 1% porosity does not. This is why bearing balls are specified to ASTM F2094 / ISO 26602 and not to a generic "Si₃N₄" datasheet.

Silicon nitride cutting tools use the same toughness: K_{IC} 6–9 with 1,550 HV allows speeds of 300–1,000 m/min in grey cast iron and nickel superalloys, where alumina-based ceramics are limited to 300–600 m/min and carbide to 150–250. The failure mode is notch wear at the depth of cut, not crater wear, so the tool geometry and edge preparation matter more than the grade.

9.5 AlN substrates: the thermal resistance calculation

A 0.63 mm AlN substrate with 170 W/m·K and 100 mm² of active area has a through-thickness thermal resistance of

A 96% alumina substrate of the same thickness and area gives 0.254 K/W — 6.8× worse. For a 170 W IGBT die, the difference is 6.3 K against 43 K of junction-to-base rise, which is the difference between a 125 °C junction and a 165 °C one.

The CTE match matters as much as the conductivity. AlN at 4.5×10⁻⁶/K against silicon at 2.6×10⁻⁶/K gives a mismatch strain of 1.9×10⁻⁶/K; at ΔT = 200 K the stress in a constrained bond is

below AlN's 350 MPa strength and marginal against silicon's 150–200 MPa fracture strength — which is why die-attach and substrate thickness are co-designed, and why AlN rather than alumina is used for large-area power modules. Direct-bonded copper (DBC) and active-metal-brazed (AMB) copper are the two metallisation routes; AMB gives better thermal cycling because the braze layer is more compliant than the eutectic bond.

9.6 Functional ceramics

9.7 Ceramic additive manufacturing

Ceramic AM is not one process; it is four, and they differ in where the shrinkage happens:

Process · Solids loading · Green density · Shrinkage · Best for

SLA / DLP slurry · 40–60 vol% · 40–50% · 15–25% linear · Small, intricate, high-precision

Binder jetting · 50–60 vol% · 50–60% · 15–20% · Large, complex, moderate tolerance

Material extrusion · 50–60 vol% · 45–55% · 15–20% · Prototypes, coarse features

DIW / robocasting · 40–50 vol% · 40–50% · 15–25% · Lattices, functionally graded

The shrinkage is the same physics as CIM, which is why the CAD model must be pre-scaled by 1.18–1.25 in all three axes. The processes are economic below about 1,000 parts, where tooling is not justified; above that, CIM or pressing wins on every cost line. The real advantage of ceramic AM is geometry that cannot be molded — internal cooling channels, lattices, and patient-specific implants — and the real limitation is that the layer lines and the debinding schedule control the flaw population, so the Weibull modulus of an AM ceramic is typically 5–8 against 10–15 for CIM.

10. Inspection: What to Measure and What It Tells You

Test · Standard · Measures · Acceptance · When

Density / apparent porosity · ASTM C20, ISO 18754 · Archimedes density · ≥ 99% TD structural; ≥ 99.5% bearing · Every batch

Flexural strength · ISO 14704, ASTM C1161 · 3- or 4-point strength · Per drawing; report with Weibull · Qualification, batch sampling

Weibull modulus · ISO 20501, ASTM C1239 · Scatter, m · ≥ 8 medical, ≥ 10 structural · Qualification

Fracture toughness · ASTM C1421, ISO 23146 (SEVNB) · K_{IC} · Per material spec · Qualification

Hardness · ASTM C1327, ISO 14705 · Vickers HV · Per material spec · Qualification

Elastic modulus · ISO 17561, ASTM C1198 · Sonic resonance · Per material spec · Qualification

Dielectric strength · ASTM D116 · Breakdown kV/mm at thickness · Per drawing · Batch or type test

Permittivity / loss · ASTM D150 · \varepsilon_r, tan δ · Per drawing · Type test

Thermal conductivity · ASTM C408 · W/m·K · Per drawing · Type test

CTE · ASTM C372, ASTM E228 · 10⁻⁶/K · Per drawing · Type test

Surface finish · ISO 21920 (profile) · R_a, R_z · Per drawing · Every ground surface

Fluorescent penetrant · ASTM E1209 / E1417 · Surface-breaking cracks · No linear indication · 100% on critical parts

Ultrasonic C-scan · ASTM E2580 (ceramics) · Internal voids, delamination · Per acceptance standard · Large or thick parts

X-ray CT · ASTM E1441 / E1570 · Internal porosity, inclusions · 5–50 µm voxel · First article, critical parts

Proof test · Customer spec · Screens weak parts · No failure at proof load · 100% on critical parts

Fractography · ASTM C1678 · Origin, mirror size · Failure analysis · On every failure

10.1 Fractography: read the mirror

A ceramic fracture surface records its own stress. The smooth mirror region around the origin is bounded by mist and hackle, and the mirror radius r_m is related to the failure stress by

where A is the fracture-mirror constant (2.0–2.5 MPa·m^0.5 for alumina, 2.5–3.5 for zirconia, 3–4 for Si₃N₄). Measure the mirror radius in µm, and you recover the stress at failure to ±20% without a strain gauge:

For alumina at 300 MPa with A = 2.3: r_m = 58.8 µm. A part that failed at 150 MPa has a mirror four times larger, 235 µm — visible at 20× under a stereo microscope. This is how a supplier's "it broke in service" claim is tested against the actual load history, and why ASTM C1678 exists.

The origin itself is diagnostic: a 30–50 µm pore at the origin means sintering, a 100 µm half-penny crack at the edge means handling, a fan-shaped mirror at a grinding scratch means machining, and an origin at a grain-boundary glass pocket means powder contamination. Fractography is the only inspection method that tells you which process step to fix.

11. Failure Modes: Eight Ways a Ceramic Part Dies

Mode · Signature · Root cause · Fix

Fast fracture from a processing flaw · Mirror at a pore or agglomerate; strength 30–50% below datasheet · Powder, granule or sintering defect · Powder spec, CIP, HIP, better sintering

Machining crack · Origin at a grinding scratch, subsurface median crack · Too coarse a grit, too deep a cut, no finish pass · D46 → D15 sequence, ductile-regime parameters

Edge chipping · Origin at a 90° edge, 50–300 µm chip · No chamfer, handling, fixturing · 0.3 mm chamfer, padded fixtures

Thermal shock · Multiple cracks from a surface, no single origin · \Delta T > \Delta T_c, constrained section · Change material (R''), reduce thickness, slow ramp

Static fatigue (SCG) · Failure after months at 40–60% of short-term strength · Water + sustained tensile stress · Keep stress < 50% \sigma_f, proof test, protective coating

Contact / Hertzian · Cone crack or subsurface spall at a contact · Point load, ball impact, wrong radius · Spherical seat, compliant pad, larger contact area

Low-temperature degradation (zirconia) · Surface roughening, grain pullout, strength loss in steam · 200–300 °C + water, grain > 0.5 µm · ISO 13356 grade, avoid steam, Y-TZP with alumina doping

Electrical breakdown / tracking · Puncture at an electrode edge, carbonised path · Field enhancement, contamination, moisture · Radius electrodes, clean, specify resistivity at temperature

Two more that deserve their own line: oxidation of non-oxides (Si₃N₄ above 1,000 °C, SiC above 1,500 °C, AlN above 800 °C) and galvanic/crevice corrosion of the metal hardware that clamps the ceramic, which is usually diagnosed as a ceramic failure because the ceramic is what broke.

12. Cost and Sourcing

12.1 What drives the price

A ceramic part's price is set by four things, in this order:

  1. Grinding time. Diamond grinding at ₹400–1,200/h dominates the cost of any part with a tolerance tighter than ±0.05 mm.
  2. Yield. Sintering shrinkage variation, warpage and edge chips typical yield 70–90% for mature parts and 40–60% for new ones.
  3. Material grade. 96% alumina is ₹400–900/kg in tile form; 99.5% is ₹1,500–3,000/kg; 3Y-TZP ₹4,000–8,000/kg; Si₃N₄ ₹15,000–30,000/kg; AlN substrate-grade powder higher still.
  4. Tooling amortisation. A CIM mold at ₹10–15 lakh needs 50,000+ parts to amortise to ₹20–30/part; a pressing die at ₹2–3 lakh amortises over 20,000 parts.

12.2 A worked ₹/part model

A 20 mm OD × 12 mm ID × 15 mm long alumina sleeve, 99.5%, 25 g, 5,000 pieces/year, ground bore and faces:

Cost element · Calculation · ₹/part

Powder · 25 g at ₹2,200/kg ÷ 0.85 yield · 65

Pressing · 300 t press, 15 s cycle, ₹800/h · 4

Sintering · 1,650 °C batch, 8 h, ₹9,000/batch ÷ 400 parts · 23

Diamond grinding (OD, ID, 2 faces) · 14 min at ₹900/h · 210

Edge chamfer · 2 min at ₹700/h · 23

Inspection (density, penetrant, dimensional) · 100% visual + sample · 18

Scrap allowance (12%) · 12% of the above · 41

Total · · 384

An equivalent 316L stainless sleeve, machined from bar, 12 min at ₹700/h plus ₹70 material, lands at ₹210–230. The ceramic is 1.7× the piece price and 5–20× the life in abrasive slurry — which is the only argument that matters in a life-cycle-cost calculation.

12.3 The Indian supply chain

India has a real technical-ceramics base, concentrated in a few clusters:

Lead times: 2–4 weeks for standard alumina wear parts, 6–10 weeks for CIM tooling plus first articles, 8–16 weeks for HIPed Si₃N₄ or AlN substrates. Quotes should always state the process route, the density, the strength test method and the surface finish — a quote that says only "alumina 99%" is not comparable.

13. Selection: The Decision Path

flowchart TD
    A[Application] --> B{Electrical insulation?}
    B -->|Yes, high voltage| C[Alumina 96-99.5%]
    B -->|Yes, thermal management| D[AlN 170-230 W/mK]
    B -->|No| E{Thermal cycling?}
    E -->|Severe, >300 K| F[Si3N4 or SSiC]
    E -->|Moderate| G{Wear mode?}
    G -->|Sliding abrasion| H[Alumina 92-99.5%]
    G -->|Impact / chipping| I[3Y-TZP or ZTA]
    G -->|Slurry erosion| J[SSiC or alumina]
    G -->|Rolling contact| K[Si3N4 bearing grade]
    E -->|None| L{Strength >600 MPa?}
    L -->|Yes| M[3Y-TZP or Si3N4]
    L -->|No| N[Alumina or SSiC]

    classDef hi fill:#1e3a5f,stroke:#38bdf8,color:#ffffff
    classDef mid fill:#f3f4f6,stroke:#111827,color:#111827
    class F,K,M hi
    class C,D,H,I,J,N mid

13.1 One-page summary

If you need… · Choose · Because

Cheapest hard, insulating, corrosion-resistant part · 96% alumina · ₹400–900/kg, 1,100–1,300 HV, 200–300 MPa

Precision insulator, feedthrough, vacuum part · 99.5% alumina · 350–379 MPa, 10⁻⁴ tan δ, 8.7 kV/mm

Maximum strength and toughness at room temperature · 3Y-TZP · 900–1,200 MPa, 8–10 MPa√m

Strength + thermal shock + rolling contact · Si₃N₄ (HIP/GPS) · 800–1,200 MPa, R 597 K, 3.2 g/cm³

Hardness + thermal conductivity + corrosion · SSiC · 2,500 HV, 120–170 W/m·K

Large complex shapes, zero shrinkage · RBSC · Si infiltration, 1,350 °C limit

Heat spreading in a power module · AlN · 170 W/m·K, CTE 4.5

Prototype machinable insulator · Macor · Carbide tools, ±0.013 mm, no shrinkage

Wear at 1,000 °C in air · Alumina or SiC · Si₃N₄ oxidises above 1,000 °C

Wear in steam at 200 °C · Alumina or SiC · Zirconia undergoes LTD

Lowest cost per part at 100k volume · CIM 3Y-TZP or alumina · 18–22% shrinkage, ±0.5% as-fired

Frequently Asked Questions

Why can't I use the datasheet strength directly?

Because a datasheet strength is a population statistic measured on a specific specimen geometry, surface finish and stressed volume — usually a polished 3 × 4 × 45 mm four-point bend bar. Your part has a different volume, a ground or as-fired surface, and a different stress distribution. Use \sigma_2 = \sigma_1(V_1/V_2)^{1/m} to scale the volume, derate 20–30% for a ground surface, and then apply a design factor of 2–3. The number that matters is the design allowable, not the catalogue number.

What Weibull modulus should I specify?

m \geq 8 for medical or human-rated parts (ISO 13356 requires ≥8 for Y-TZP), m \geq 10 for structural parts, and m \geq 15 for bearing-grade silicon nitride. Below 8, the strength distribution is so wide that a "safe" design stress determined from a small sample is not statistically defensible. Ask the supplier for the modulus, the number of specimens, and the test method — not just the mean strength.

Why does my alumina part break at 200 MPa when the datasheet says 350 MPa?

Three usual causes, in order: a grinding scratch or edge chip in the tensile surface (30–40% loss), a sintering pore or agglomerate at the fracture origin (20–40% loss), or a stressed volume much larger than the test bar (10–20% loss). Fractography answers the question in ten minutes — measure the mirror radius and look at the origin. If the origin is at a surface scratch, the fix is the grinding sequence, not the material.

Alumina or zirconia for a wear part?

Alumina for sliding abrasion, zirconia for impact and chipping. Alumina is harder (1,600 HV vs 1,250) and cheaper; zirconia is three times tougher and survives edge impacts that chip alumina. In slurry erosion, alumina usually wins on cost per hour; in a valve seat that sees grit impacts, zirconia wins on survival. Do not use zirconia above 150 °C in steam.

Why is silicon nitride used for bearings and not alumina?

Because it has the combination that rolling contact requires: K_{IC} 6–9 (against 4 for alumina), 3.2 g/cm³ density (against 3.9), 800–1,200 MPa strength, and a failure mode that spalls rather than shatters. A hybrid bearing with Si₃N₄ balls in steel rings runs at 1.5–2.0 million DN; an alumina ball would fracture at the first contamination particle. The specification is ASTM F2094 / ISO 26602, not a generic ceramic datasheet.

How much does sintering shrink, and how do I design for it?

For isotropic sintering, S = 1-(\rho_g/\rho_s)^{1/3}. A 55% green body fired to 99.5% shrinks 17.9% linearly; CIM parts typically shrink 18–22% and tape-cast parts 15–20%. Scale the mold by 1/(1-S) — 1.218 for 17.9% — and expect ±0.5% as-fired tolerance because green density varies. Grind anything that needs better than ±0.05 mm.

Can I machine a ceramic after sintering?

Only with diamond. Carbide tooling will not touch alumina, zirconia or SiC; diamond grinding, ultrasonic machining, laser or waterjet are the options. Green machining with carbide is possible before sintering, but the 18% shrinkage magnifies every error, so it is limited to features with ±0.3 mm tolerance. The economical route is near-net shaping plus diamond grinding of the functional surfaces only.

What is the strongest ceramic?

3Y-TZP, at 900–1,200 MPa flexural strength, is the strongest commercial structural ceramic. It is not the hardest (SiC and B₄C are harder), not the stiffest (SiC at 410 GPa), and not the most thermal-shock resistant (Si₃N₄). Strength, hardness, toughness and thermal shock are four different properties, and no single grade wins all four — which is why the selection table exists.

Why does zirconia crack when quenched, despite being the toughest ceramic?

Because toughness resists crack growth but does not prevent crack initiation from thermal stress, and zirconia's thermal conductivity is only 2.5 W/m·K. A quench develops the full \Delta T across a thin surface layer, and the resulting stress, E\alpha\Delta T/(1-\nu), reaches 700 MPa at \Delta T = 240 K — close to its strength. Its R is 325 K but its R'' is only 813 W/m, the worst of the structural grades. Use Si₃N₄ or SiC for thermal cycling.

How do I inspect a ceramic part for internal defects?

Density by Archimedes (ASTM C20) catches bulk porosity; ultrasonic C-scan at 10–25 MHz catches delamination and large voids (0.1–0.5 mm); X-ray CT at 5–50 µm voxel catches porosity and inclusions; fluorescent penetrant catches surface-breaking cracks; and proof testing catches the weak parts that all of the above miss. For critical parts, use all five, and add fractography on the first article to confirm the flaw population.

Is ceramic additive manufacturing cheaper than CIM?

Below about 1,000 parts per year, yes — there is no tooling. Above that, no: CIM's cycle time is seconds, the feedstock is cheaper per part, and the tolerance is tighter. Ceramic AM wins on geometry that cannot be molded (internal channels, lattices, patient-specific shapes) and on lead time for prototypes; it loses on unit cost, surface finish and Weibull modulus (5–8 against 10–15).

Previous guides in this series: Fatigue, Fracture Mechanics & Durability Design · Industrial Lubrication & Tribology · Abrasive Machining & Precision Grinding · Springs & Spring Design · O-Rings, Gaskets & Mechanical Seals · Industrial Furnaces, Kilns & Refractories · Metrology & Dimensional Inspection · Powder Metallurgy · Metal Additive Manufacturing · CNC Machining · Non-Destructive Testing · Corrosion Engineering.

References

[1] W. D. Kingery, H. K. Bowen and D. R. Uhlmann, Introduction to Ceramics, 2nd ed., Wiley, 1976 — crystal chemistry, point defects, sintering theory and the Herring scaling laws. [2] M. W. Barsoum, Fundamentals of Ceramics, 2nd ed., Taylor & Francis, 2003 — bonding, dislocation behaviour in ceramics, and the Peierls stress argument for brittle behaviour. [3] D. W. Richerson, Modern Ceramic Engineering: Properties, Processing, and Use in Design, 4th ed., CRC Press, 2018 — the standard design reference for structural ceramics, including Weibull design allowables, proof testing and design rules. [4] B. Lawn, Fracture of Brittle Solids, 2nd ed., Cambridge University Press, 1993 — Griffith, Irwin, indentation fracture, and the mirror–mist–hackle fractography. [5] A. A. Griffith, "The phenomena of rupture and flow in solids," Philosophical Transactions of the Royal Society A, 221, 1921, pp. 163–198 — the energy criterion for crack growth. [6] G. R. Irwin, "Analysis of stresses and strains near the end of a crack traversing a plate," Journal of Applied Mechanics, 24, 1957, pp. 361–364 — the stress-intensity factor. [7] T. G. Bifano, T. A. Dow and R. O. Scattergood, "Ductile-regime grinding: a new technology for machining brittle materials," Journal of Engineering for Industry, 113(2), 1991, pp. 184–189 — the critical depth d_c = 0.15(E/H)(K_{IC}/H)^2 and the ductile-to-brittle transition. [8] S. M. Wiederhorn, "Subcritical crack growth in ceramics," in Fracture Mechanics of Ceramics, Vol. 2, Plenum, 1974 — the v = AK^n law and the stress-corrosion mechanism. [9] R. C. Garvie, R. H. Hannink and R. T. Pascoe, "Ceramic steel?", Nature, 258, 1975, pp. 703–704 — transformation toughening in zirconia. [10] R. H. J. Hannink, P. M. Kelly and B. C. Muddle, "Transformation toughening in zirconia-containing ceramics," Journal of the American Ceramic Society, 83(3), 2000, pp. 461–487 — the t→m transformation, the 3–5% volume change, and the toughening-zone analysis. [11] J. Chevalier, L. Gremillard, A. V. Virkar and D. R. Clarke, "The tetragonal–monoclinic transformation in zirconia: lessons learned and future trends," Journal of the American Ceramic Society, 92(9), 2009, pp. 1901–1920 — low-temperature degradation, grain-size dependence, and the ISO 13356 aging test. [12] P. F. Becher, "Microstructural design of toughened ceramics," Journal of the American Ceramic Society, 74(2), 1991, pp. 255–269 — crack bridging by elongated grains in Si₃N₄ and the R-curve. [13] G. Ziegler, J. Heinrich and G. Wötting, "Relationships between processing, microstructure and properties of dense and reaction-bonded silicon nitride," Journal of Materials Science, 22, 1987, pp. 3041–3086 — GPS/HIP Si₃N₄ grades, sintering additives, and the strength–toughness trade. [14] A. G. Evans and M. Y. He, "Strength and fracture of SiC and Si₃N₄ ceramics," Journal of the American Ceramic Society, 81(9), 1998 — high-temperature strength and oxidation limits of non-oxide ceramics. [15] ASTM C1161-18, Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature. [16] ASTM C1239-13(2019), Standard Practice for Reporting Uniaxial Strength Data and Estimating Weibull Distribution Parameters for Advanced Ceramics. [17] ISO 20501:2019, Fine ceramics (advanced ceramics, advanced technical ceramics) — Weibull statistics for strength data. [18] ISO 14704:2016, Fine ceramics — Test method for flexural strength of monolithic ceramics at room temperature; ISO 14705:2016, Test method for hardness of monolithic ceramics at room temperature; ISO 17561:2016, Test method for elastic moduli of monolithic ceramics at room temperature by sonic resonance. [19] ASTM C1421-18, Standard Test Methods for Determination of Fracture Toughness of Advanced Ceramics at Ambient Temperature; ISO 23146:2012, Fine ceramics — Test methods for fracture toughness of monolithic ceramics — Single-edge V-notch beam (SEVNB) method. [20] ISO 13356:2015, Implants for surgery — Ceramic materials based on yttria-stabilized tetragonal zirconia (Y-TZP) — the ≥500 MPa biaxial / ≥800 MPa four-point strength floor, Weibull ≥8, density ≥6.00 g/cm³, and the 134 °C/2 bar/5 h accelerated aging test with ≤25 mol% monoclinic phase. [21] ISO 6474-1:2019, Implants for surgery — Ceramic materials — Part 1: Ceramic materials based on high purity alumina; ISO 6474-2:2019, Part 2: Composite materials based on a high-purity alumina matrix with zirconia reinforcement. [22] ASTM F2094/F2094M-18, Standard Specification for Silicon Nitride Bearing Balls; ISO 26602:2017, Fine ceramics — Silicon nitride materials for rolling bearing balls. [23] ASTM C20-00(2015), Standard Test Methods for Apparent Porosity, Water Absorption, Apparent Specific Gravity, and Bulk Density of Burned Refractory Brick and Shapes by Boiling Water; ISO 18754:2020, Fine ceramics — Determination of density and apparent porosity. [24] ASTM C1327-15, Standard Test Method for Vickers Indentation Hardness of Advanced Ceramics. [25] ASTM C1678-10(2015), Standard Practice for Fractographic Analysis of Fracture Mirror Sizes in Ceramics and Glasses — the \sigma_f\sqrt{r_m} = A mirror relation and mirror constants. [26] ASTM C1368-18, Standard Test Method for Determination of Slow Crack Growth Parameters of Advanced Ceramics by Constant Stress-Rate Flexural Testing at Ambient Temperature. [27] ASTM D116-15, Standard Test Methods for Vitreous Ceramic Whitewares and Related Products (dielectric strength by ASTM D116 is cited in the AD-995 datasheet); ASTM D150-18, Standard Test Methods for AC Loss Characteristics and Permittivity (Dielectric Constant) of Solid Electrical Insulation. [28] CoorsTek, AD-995 Alumina (nom. 99.5% Al₂O₃) datasheet — 379 MPa flexural (ASTM F417), K_{IC} 4–5 MPa√m, 8.2×10⁻⁶/K CTE (25–1,000 °C), 8.7 kV/mm dielectric strength (ASTM D116), ΔT_c 200 °C; and Almath Crucibles, Alumina 99.5/99.6/99.8% properties table — 375–379 MPa, 350–370 GPa, 27.5–30 W/m·K, 3.8–3.9 g/cm³. [29] Precision Ceramics, CeramaSil-N PCSN3000 silicon nitride datasheet — 760–830 MPa, K_{IC} 6.2–6.5, 300–310 GPa, 22 W/m·K, 3.2×10⁻⁶/K, R_1 590–620 K, 1,200 °C in air; and CeramAlum AlN datasheet — 170/200/230 W/m·K grades, CTE matched to Si and GaN. [30] Fraunhofer IKTS, "Properties and Reliability of Silicon Nitride Substrates with AMB Copper Conductor," 44th International Symposium on Microelectronics, 2011 — GPSSN 900–1,200 MPa, K_{IC} 8–9, 20–24 W/m·K, 3.0–3.8×10⁻⁶/K, and the 80 W/m·K tape-cast high-conductivity grade. [31] SilMaterials, Sintered Silicon Carbide (RBSC/SSiC/re-crystallized) specification — SSiC 350–500 MPa, 3.5–4.5 MPa√m, 120–270 W/m·K, 2,500 HV; RBSC 300–400 MPa with 8–15% free silicon. [32] Great Ceramic, Advanced ceramic material properties table — 96%/99.5%/99.9% alumina, ZTA, 3Y-TZP, Mg-PSZ, GPSN/HPSN, SSiC, RBSC and Macor property ranges used in §2. [33] Microcertec, Materials properties data sheet — Macor 94 MPa, 1.46 W/m·K, and the zirconia/alumina/Si₃N₄ comparison set. [34] "Zirconia Injection Molding: Process, Challenges, and Design Guide for ZrO₂ CIM Parts," MIM Supplier technical note — 55–62 vol% solids loading, 18–22% linear shrinkage, 1,380 °C sintering, ±0.5–1.0% tolerance. [35] "Statistical analysis of process parameters influencing dimensional control in ceramic injection molding," Journal of Materials Processing Technology, 1998 — 19.3% shrinkage at 55 vol% solids loading versus 23.6% at 45 vol%, and the dominance of solids loading on dimensional variation. [36] D. Munz and T. Fett, Ceramics: Mechanical Properties, Failure Behaviour, Materials Selection, Springer, 1999 — thermal-shock parameters R, R', R'', Biot-number corrections, and the Weibull size-scaling relations. [37] Indian practice: CSIR-Central Glass & Ceramic Research Institute (CGCRI), Kolkata — advanced ceramics characterization and pilot tape casting; Carborundum Universal (CUMI) and Jyoti Ceramic Industries for alumina/zirconia wear parts; IS 13730 (ceramic insulators) and IEC 60672-3 for ceramic insulating materials classification.

Note on sources: the governing relations are Griffith/Irwin for strength, Weibull for statistics, R = \sigma(1-\nu)/(E\alpha) for thermal shock, Bifano for ductile-regime grinding, and S = 1-(\rho_g/\rho_s)^{1/3} for sintering shrinkage; the specifications are ASTM C1161/C1239/C1421/C1678, ISO 14704/20501/13356/26602 and ASTM F2094. Where the literature differs — 3Y-TZP toughness quoted anywhere from 6 to 10 MPa√m depending on the measurement method, Si₃N₄ strength from 700 to 1,200 MPa depending on sintering route, alumina dielectric strength from 8.7 to 20 kV/mm depending on thickness — this guide prints the grade and the test condition, because a ceramic property without its specimen geometry and process route is a number without a meaning.

More FabFlow blog posts