Alumina purity grade 85% (Al2O3) delivers flexural strength around 296 MPa and dielectric strength of 9.4 kV/mm, while 99.5% grade pushes to 379 MPa flexural, 30.0 W/(m·K) thermal conductivity, and roughly 15 kV/mm dielectric strength [S1][S4].
The grade-to-property gap is large enough that one factory call-out "99% alumina" can change wear life, dielectric margin, and cost by more than a factor of two. This guide compares 85%, 96%, 99%, and 99.5% alumina against flexural strength, thermal conductivity, dielectric strength, max use temperature, and unit cost, and maps each to the duty class where it pays back.
Why Purity Grade, Not Brand, Drives Part Performance
Alumina purity is the mass fraction of Al2O3 in a sintered body; the remainder is sintering aids (SiO2, MgO, CaO) and secondary oxides that control grain growth and densification [S1][S4]. Those additives are not "dirt" in a negative sense, but they do govern glassy-phase content, which in turn governs dielectric loss, chemical inertness, and high-temperature creep [S1][S4]. Moving from 85% to 99.5% is not a linear "more pure equals better" shift; it is a microstructure step that changes strength, insulation, and machinability differently at each level [S4].
Two supplier datasets in the same August 2026 guide illustrate why single-source numbers must be treated as a starting point: the CoorsTek-style table lists 94% alumina at 22.4 W/(m·K) thermal conductivity and 8.3 kV/mm dielectric, while Accuratus reports 18 W/(m·K) and 16.7 kV/mm for the same nominal grade [S1]. Dielectric strength in particular does not rise monotonically with purity; sample thickness, electrode geometry, and test method shift the absolute number, which is why most alumina datasheets specify the test coupon (commonly 0.25 in. thick) alongside the kV/mm figure [S1].
Property Matrix: 85% vs 96% vs 99% vs 99.5% Alumina
Across 85–99.5% Al2O3, density climbs from 3.42 g/cc to 3.90 g/cc, flexural strength from 296 MPa to 379 MPa, thermal conductivity from 16.0 W/(m·K) to 30.0 W/(m·K), and maximum no-load service temperature from 1400°C to 1750°C [S1]. Dielectric strength does not track purity linearly: in one dataset it rises from 9.4 kV/mm at 85% to a peak near 8.7 kV/mm at 99.5%, while a second dataset puts 99.5% at 16.9 kV/mm and 96% at 14.6 kV/mm [S1]. The takeaway is that dielectric must be read off a supplier's datasheet for the actual part thickness, not assumed from purity alone.
Decision criteria, lined up against the four grades that cover most industrial B2B buying: 85% alumina wins on raw cost (lowest additive control, looser sintering, coarser grain) and is the right call for grinding media, chute liners, and wear tiles where dielectric is irrelevant [S1][S3]. 96% alumina is the workhorse, balancing 358 MPa flexural, 24.7 W/(m·K) thermal, and 1700°C ceiling for medical equipment, wear-resistant nozzles, and connector housings [S1][S3]. 99% alumina (often sold as "99.5% nominal" with binder accounting) hits the high-300s MPa flexural, low-30s W/(m·K) thermal, and is specified for pump seals, textile guides, and mechanical wear parts where dielectric loss matters [S1][S4]. 99.5% alumina is the electronic-substrate grade, delivering 379 MPa flexural, 30 W/(m·K) thermal, and roughly 15 kV/mm dielectric for EV power modules, semiconductor insulators, and high-voltage feedthroughs [S1][S4].
Where 85% Alumina Earns Its Slot (and Where It Fails)
At 85% Al2O3, density is 3.42 g/cc, flexural strength 296 MPa, thermal conductivity 16.0 W/(m·K), max no-load use 1400°C, and dielectric strength 9.4 kV/mm [S1]. The lower additive cost and coarser grain make 85% the cheapest alumina ceramic per kg and the easiest to sinter to size, so it dominates grinding balls, wear-resistant accessories, and structural tiles in mining and aggregate service [S1][S3].
Failure modes the engineer has to respect at 85%: limited corrosion resistance, poor thermal stability above ~1400°C no-load, and the lowest dielectric margin in the alumina family, so it is the wrong call for any insulating or vacuum application [S7]. If the part sees a brazed metallized interface, the coarser grain actually helps the braze wet the surface, which is one reason 85–95% grades are common in metallized ceramic-to-metal assemblies [S4]. Alumina ceramic grades span from 85% to 99.9%+ Al₂O₃, with each purity level offering different combinations of flexural strength, thermal conductivity, dielectric strength, and cost for specific applications.
Where 99.5% Alumina Earns Its Slot (and Where It Fails)
At 99.5% Al2O3, density is 3.90 g/cc, flexural strength 379 MPa, thermal conductivity 30.0 W/(m·K), max no-load use 1750°C, and dielectric strength around 15 kV/mm (one supplier dataset) or 16.9 kV/mm (another) [S1][S4]. The glassy secondary phase is suppressed, the grain structure is finer, and chemical inertness approaches the limit of what polycrystalline alumina can deliver, which is why this grade is the workhorse for electronic substrates, semiconductor insulators, and high-voltage components [S1][S4].
Failure modes at 99.5%: the part is harder to machine (diamond grinding only), sintering shrinkage is tighter, and any deviation from the recommended sintering profile shows up as porosity that drops dielectric margin. Cost runs 2–4x the 85% grade for raw stock and rises again at finish-machining because of the wheel wear. For most industrial wear and structural parts, the 99.5% dielectric and thermal headroom is wasted; for EV power modules, semiconductor wafer handling, and high-vacuum insulators, that headroom is the difference between a 3-year and a 10-year service life [S4][S5]. For applications that sit between substrate and structural, the industrial ceramic reference covers broader selection logic.
Selection Rule: Match the Grade to the Failure Mode, Not the Spec Sheet Maximum
The cheapest grade that still meets the failure-driving property is the right grade: wear parts that fail by abrasion belong at 85–95% regardless of how clean the rest of the plant is, while insulators that fail by dielectric breakdown belong at 99% or higher regardless of how rough the mechanical environment is [S1][S4]. A common spec-trap is to call out "99% alumina minimum" on a drawing for a wear tile; the result is a part that costs 2–4x more, machines slower, and lasts no longer in service than an equivalent 95% tile [S1][S4].
A second trap is calling out "alumina" with no purity and no test method; with two supplier datasets putting 99.5% dielectric at 15 kV/mm and 16.9 kV/mm on the same nominal grade, the unannotated call-out cannot be enforced at incoming inspection [S1]. The drawing should pin purity in % Al2O3, density in g/cc, and dielectric strength in kV/mm with the test sample thickness, mirroring the format used in the supplier property table [S1][S2]. The broader zirconia ceramic comparison is useful when toughness, not hardness, is the failure driver; zirconia trades the temperature ceiling of alumina for fracture toughness alumina cannot match. For parts that pivot from ceramic to metal race (pump seals, bearings), the ceramic bearing page tracks the same purity-vs-cost trade in rolling-element form.
Limits, Verification, and What to Pin on the Drawing
Volume resistivity climbs roughly one order of magnitude across the same purity spread: 95% alumina is rated near 10^12 Ω·cm, 99% near 10^13 Ω·cm, and 99.7% near 10^14 Ω·cm at room temperature, which is the more reliable indicator of insulation performance than a single dielectric-strength number [S5]. At 95% the datasheet typically caps thermal stability at 1250°C for sustained service, 99% reaches 1500°C for short dwells, and 99.7% holds 1700°C, so any application with continuous exposure above 1500°C has effectively no grade lower than 99% as an option [S5].
Verification on the drawing: state purity as % Al2O3 minimum, density in g/cc (target 3.7–3.98 g/cc depending on grade), volume resistivity in Ω·cm at a stated temperature, dielectric strength in kV/mm at a stated sample thickness, and grain size in µm if the application is optical or electronic [S1][S5]. Density is the single fastest incoming-inspection check: a 99.5% coupon under 3.85 g/cc almost certainly missed the sintering window, and a 95% coupon over 3.65 g/cc is mislabeled stock [S1][S5].
Trackable next signals: supplier datasheet revisions that pin dielectric strength to a 0.25 in. or 1 mm test coupon (already standard at CoorsTek-style houses, still loose at smaller vendors) [S1]; and the gradual migration of EV power-module and semiconductor-handling demand from 99% to 99.5% nominal, which will tighten 99.5% capacity through 2026–2027 and likely pull lead times on the electronic-substrate grade [S4][S5].
Related analysis: 14-bit raw radiometric vs 8-bit AGC video: when radiometry wins over display-ready output.