Beginner perfumer #121: the 208 materials with two boiling points at one pressure

About a 4 minute read.

The short version

• 208 materials carry two or more boiling points at the same pressure
• Median gap 5 °C, quartiles 2 to 18; 48 exceed 20 °C
• 187 (90%) have one value marked (est) and one not. Keep the unmarked one
• Only 20 have neither marked, and only 6 of those differ by more than 20 °C
• The other 14 differ by 0 to 5 degrees and are duplicate entries. Four have identical values
• The two largest: 1,8-cineole (77 suppliers) at 87 °C apart, and dimethyl sulfide (81 suppliers), a sign error

The cheapest check yet

I have built up a run of internal consistency checks over these rounds: the boiling point field equalling the melting point, a reduced-pressure boiling point exceeding the atmospheric one, round endpoints in specification ranges, and lining a material up against its family.

This one is simpler still: a material can only have one boiling point at a given pressure. If the bp field holds two numbers both marked @ 760.00 mm Hg, one of them is spare.

208 records, and most are fine

	Records
Two or more boiling points at one pressure	208
Gap >1 °C	170 (82%)
Gap >5 °C	98 (47%)
Gap >20 °C	48 (23%)
Gap >50 °C	26 (12%)

The median gap is 5 °C. Most are two literature values for the same material, and a few degrees apart is ordinary.

And 90% of the time the check answers itself

This part I did not expect. Look at whether the two values carry (est):

	Records	Share
One marked (est), one not	187	90%
Neither marked	20	10%
Both marked (est)	1	0.5%

Nine times in ten it is an estimate sitting beside a measurement.

And the bigger the gap, the stronger the pattern: of the 48 records more than 20 °C apart, 43 (90%) are an estimate against a measurement.

So the rule is simple: when the two values differ, keep the one without (est).

For example, para-anisyl formate (21 suppliers) reads:

220.00 °C. @ 760.00 mm Hg | 254.00 to 255.00 °C. @ 760.00 mm Hg (est)

Keep 220. That 35 °C gap is the distance between an estimator and a measurement, not a data error.

Of the remaining 20, only 6 are a real problem

The 20 records with neither value marked are the ones where two measurements disagree:

Gap	Suppliers	Material	The two values
87.0 °C	77	1,8-cineole	176–177 / 90–91
77.0 °C	4	5-acetyl-2,3-dihydro-1,4-thiazine	259–319 / 242
74.8 °C	81	dimethyl sulfide	37.30 / −37.50 to −37.30
66.0 °C	2	fresh decan-2-one	323–325 / 259
62.0 °C	40	2-methyl undecanal	171 / 233
20.0 °C	67	maple furanone	102–103 / 83–86 @0.5 mmHg

These six, and no more. The other 14 differ by 0 to 5 degrees:

• laevo-linalool (25 suppliers): 198 and 197–199. The second contains the first: one value written two ways
• dipentene (37 suppliers): 176 and 178
• Four have identical values: cyclohexyl butyrate 212 and 212, 4-butyl pyridine 209 and 209. Straightforward duplicate entries

Of 208 records, 6 need attention: 0.014% of the database.

The dimethyl sulfide one is a sign error

37.30 °C. @ 760.00 mm Hg | -37.50 to -37.30 °C. @ 760.00 mm Hg

The same digits with a minus sign.

Dimethyl sulfide boils at 37 °C; it is something that leaves the bottle at room temperature, with a vapour pressure of 647 mmHg. Its melting point field reads −98 °C and its flash point −36.67 °C. So 37.30 is right and that −37.30 has a spare minus sign.

This is the third sign error I have found in this database. The first was lime oil's optical rotation reading +35 to −41, and the second was my own regex reading acetaldehyde's −53 as 53.

Minus signs fall off easily, and they turn up easily too.

What is 1,8-cineole's 90 °C

This one has 77 suppliers and is thoroughly mainstream, being the main constituent of eucalyptus.

176.00 to 177.00 °C. @ 760.00 mm Hg | 90.00 to 91.00 °C. @ 760.00 mm Hg

The 176–177 is right: eucalyptol boils at 176 °C, and its vapour pressure of 1.900 mmHg fits that bracket.

Where does 90–91 come from? I checked the record's other fields: melting point 1–2 °C, flash point 50 °C. None of them is 90.

I do not know where that number came from, and that is this check's limit. It can tell you a spare number is present; it cannot tell you where the number belongs.

The rule you can run yourself

Open any material's boiling point field:

1. Are there two or more values? Look at their pressure annotations
2. Are the two pressures the same? If so, one is spare
3. Does one carry (est)? Then keep the other. That resolves nine cases in ten
4. Neither marked? Go to other fields. Vapour pressure, melting point and flash point all corroborate

Step 3 is the point. That little (est) marker is one of the most useful pieces of metadata in this database, and it also sits on 96% of vapour pressures and 96% of appearance fields. A great deal of my own reasoning has been built on estimates.

What this doesn't establish

• I looked up no correct boiling points. This is entirely a comparison within fields.
• "Keep the one without (est)" is a rule of thumb. It assumes measurements beat estimates, which usually holds but is not a theorem; measurements get miscopied too.
• The figure of 6 depends on my 20 °C threshold. At 10 °C a few more join.
• I did not find the source of 1,8-cineole's 90 °C.
• I swept only the bp field. mp, flash_point and specific_gravity could carry the same problem; I have not checked.
• Automated curation of chemical datasets has its own literature (PMID 27885862, PMID 23237381). What I ran here is four steps you can check by eye, not work at that level.