Last updated: 2026-11-05
Bottom line. Look up thermal shock for borosilicate 3.3 and you will meet three different figures — 280 °C on our Glassware page, 205 °C from a journal paper, and about 166 °C in an encyclopaedia entry. They do not contradict each other; the conditions differ — different formulation, different specimen, different test method. They are therefore not directly comparable, and none of them is an operating permit for a particular vessel.
[!] Every figure here comes from published literature or our own existing pages — not from RichTech measurements. Sources are listed at the foot of the page.
| Source | Value | Conditions |
|---|---|---|
Our Glassware page (page states data based on ISO 3585) |
280 °C | Material-level nominal value; test method not stated — refer to product documentation |
| Bulletin of the Chinese Ceramic Society, 2025, 44(7): 2647-2655 (DOI: 10.16552/j.cnki.issn1001-1625.2024.0974) | 205 °C | That paper's specific formulation: SiO₂ 82.7 / B₂O₃ 12.2 / Na₂O 4.0 / CaO 1.1 (mol%), with α = 3.33×10⁻⁶ °C⁻¹ for that formulation |
| English encyclopaedia entry, "Borosilicate glass" | about 166 °C (330 °F) | General narrative, no test conditions given; background only |
To be precise: 280 is a material-level nominal figure (our Glassware page, stated as based on ISO 3585, with no test method given); the paper's 205 applies to that paper's specific formulation; the encyclopaedia's 166 is general narrative. The conditions differ, so they cannot be compared directly.
The paper's formulation has α = 3.33×10⁻⁶ °C⁻¹ — very close to the nominal 3.3 but not identical. A small difference in expansion shifts the calculated ΔT appreciably.
What a piece of glass can actually survive depends on the internal temperature gradient. A thick-walled item has a larger gradient across its wall and cracks more readily for the same material; a thin-walled item performs better thermally but has lower mechanical strength and is more fragile. These are two different failure modes.
Three methods are common, and their results cannot be converted into one another:
Calculation uses theoretical strength. Real components carry scratches and microcracks, so their strength is usually below the theoretical value — which makes a calculated ΔT either optimistic or conservative depending on the strength value chosen.
A new item and a scratched old one differ greatly in thermal-shock tolerance — surface microcracks are where cracks start.
It is within the usual range for this material and is a material-level nominal value. Treating it as "any 3.3 vessel will take 280 °C" goes beyond what it means — our Glassware page states it as based on ISO 3585, with no test method given.
Because that paper measured its own formulation and reported that formulation's expansion coefficient. The figure holds for that formulation; it cannot be transferred directly to another manufacturer's product.
Ask the supplier for product-specific test evidence, not material-level general data. This is also why we do not publish a single site-wide "continuous service temperature".
They cannot be ranked that way. A larger or smaller number does not equal more or less conservative — even the test methods may differ. The conservative approach is to leave margin in your process, not to pick one number as a limit.
Figures here come from published literature and our own existing pages, not from RichTech measurements.