Beginner perfumer · 102
794 records give a flash point of 0 °C, and every one of them is something that cannot burn
· 12 min read
The flash point field has 33.9% coverage, and its most common value appears 794 times: '32.00 °F. TCC (0.00 °C.) (est)'. A third of those 794 have no carbon in the formula — sodium hydroxide, sulfuric acid, titanium dioxide, calcium carbonate, chitosan. They have no flash point because they do not burn. And 0 °C is the most hazardous reading a regulation recognises: it means this ignites at freezing. The default says the opposite of the truth. Worse, not one of those 794 records carries a GHS hazard statement — the only field that could correct it is empty exactly where the lie is. And 56.6% of all flash point values are marked (est), which turns out to be almost entirely a property of materials nobody sells: est records have a supplier median of 1, non-est records 5.
About a 4 minute read.
Last time I sorted out three ways to fill a field with nothing, and said the dangerous one is the default template — because it looks like a plausible number.
This is the most extreme case of that kind.
The flash point field's most common value
The field has 14,316 values (33.9%). Its most common one appears 794 times:
32.00 °F. TCC ( 0.00 °C. ) (est)
Exactly 0 °C, and marked estimated.
The best-supplied of those 794:
| Suppliers | Material |
|---|---|
| 53 | Sodium hydroxide |
| 38 | Sulfuric acid |
| 35 | Potassium hydroxide |
| 31 | Phosphoric acid |
| 30 | Tannic acid |
| 29 | Titanium dioxide |
| 29 | Chitosan |
| 28 | Nitric acid |
| 26 | Calcium carbonate |
These things have no flash point.
A flash point is the lowest temperature at which the vapour above a liquid is concentrated enough to ignite. Sodium hydroxide does not burn. Titanium dioxide does not burn. Calcium carbonate is limestone.
Putting a number on it
I used "does the formula contain carbon" as an objective proxy:
| With a formula | No carbon | |
|---|---|---|
| The 0.00 °C group | 797 | 268 = 33.6% |
| All other flash point records | 12,723 | 131 = 1.0% |
34×.
(The other two thirds are carbon-containing things that also do not readily burn — tannic acid, chitosan, quaternary ammonium salts and similar polymers and salts.)
And 0 °C is the most hazardous class there is
This is the point.
Most regulations put liquids flashing below 23 °C in the highest flammability category, and below 38 °C in "flammable liquid." A flash point of 0 °C means this ignites at freezing temperature.
A default that stands for "has no flash point" is being rendered as "extremely flammable."
A placeholder says "no data." This default says something specific and wrong — and wrong in the most dangerous direction.
And the one field that could correct it is empty
I checked how many of those 794 records carry a GHS hazard statement.
None.
A material that genuinely ignited at 0 °C would carry H224 or H225. Not one of these does — because they are not flammable, so they never got the label.
Two fields combine to invert the situation: one supplies a false hazard figure, and the other, which should say "actually, no," is blank.
I measured in article 100 that GHS is the emptiest field in the database (0.9%) — this is where that emptiness costs the most.
And half of all flash points are estimates
Measured along the way: 8,109 of the 14,316 flash point values carry "(est)" — 56.6%.
And its relationship with supplier count is identical to the assay default's:
| Records | Median suppliers | Share with 10+ | |
|---|---|---|---|
| With (est) | 8,109 | 1 | 4.2% |
| Without (est) | 6,207 | 5 | 32.0% |
Measured values belong to materials people sell. Estimates belong to materials nobody does.
That is the same pattern for the third time (assay, flash point, and the coverage jumps for appearance and uses): this database's record quality for "products" and its record quality for "entries" are two different things.
So what actually deserves caution
Removing both the estimates and that default, and looking only at measured flash points for materials with 15+ suppliers — the median is 92.2 °C, and only 92 records (6.6%) fall below 38 °C.
The lowest of them:
| Flash point | Suppliers | Material |
|---|---|---|
| 2.2 °C | 21 | Ethyl crotonate |
| 2.2 °C | 20 | Trimethylamine |
| 3.3 °C | 27 | Methyl isobutyrate |
| 4.4 °C | 37 | 2-methyl butyraldehyde |
| 7.2 °C | 66 | Diacetyl |
| 9.4 °C | 41 | Ethanol |
Diacetyl and ethanol are things many people already have on the bench. Everybody knows ethanol's 9.4 °C; diacetyl at 7.2 °C is lower still — it is a flammable liquid at room temperature.
In practice
- A flash point of "0.00 °C (est)" is not data, it is a default. Ask first whether the thing burns at all.
- Discount anything marked (est). 56.6% of flash points are estimated, and estimates concentrate on materials nobody sells.
- What you should check is GHS — but at 0.9% coverage, usually you cannot.
- To rank flammability by measured values, filter out the estimates and those 794 first.
What this doesn't establish
- "No carbon" is a crude proxy for "does not burn." There are carbon-containing things that do not (carbonates) and carbon-free things that do (hydrogen, phosphorus).
- I did not examine the database's estimation algorithm, so "0 °C is what it outputs for substances with no flash point" is inferred from the pattern.
- "Most regulations divide at 23 and 38 °C" is common knowledge, cited from no statute, and thresholds differ by jurisdiction.
- I looked at only the best-supplied dozen or so of the 794, not every record.
- The zero GHS statements may simply reflect those records being sparse overall, rather than "not flammable, so no label." I cannot separate those two explanations.
- This is a September 2026 snapshot.