Docs / What we have found

Two rows, one name, one compartment — and it means one of two opposite things

Part of What we have found, the class where one name in one compartment holds two rows — which is either a substance published twice or two substances sharing a name.

One substance twice: carbon tetrachloride

EF 3.1 publishes carbon tetrachloride twice in each of five air compartments. One row is CFC-10, carrying CAS 56-23-5, EC 200-262-8, a general comment and some fifty synonyms. The other is Carbon tetrachloride, carrying no registry number, no EC number, no synonyms and no comment at all.

Both rows hold the same characterisation under the same nine method identifiers. Six of the nine agree exactly. The rest differ only in how many digits were kept — human toxicity, cancer is 0.000040473 on one row and 0.0000405 on the other: the same number to five figures and to three.

Which row survives is not a detail. Sorting by "most factors, then by identifier" ties on the factor count and then picks the bare row — the one with no identifier of any kind, and the one no other list maps onto. ecoinvent's correspondence lands on the identified row, reaching it from 3.8 as Methane, Tetrachloro-, R-10 and from 3.12 as Carbon Tetrachloride. Keeping the anonymous row would have stranded every one of those mappings, so this pair — and 46 others — carries an explicit written ruling naming the survivor rather than being settled by a sort (#526, #530).

Two rows of the same substance are often a chemical name beside an industry designation — Methyl chloroform beside HCFC-140, perfluorohexane beside PFC-51-14, Carbon tetrachloride beside CFC-10 — which is the codes-as-names problem showing up as duplication. The rest are plainer: the same name written twice in one compartment, 28 substances' worth, with kresoxim methyl and one oxazolidinyl carbamate duplicated in all thirteen compartments each, methylene chloride in nine and diethyl ether in five.

The two kinds, and how many of each

Carbon tetrachloride is the first of the two kinds, and the two need opposite treatment.

A flow list is supposed to have one row per substance per compartment. When it has two, the two rows are either the same substance written down twice, or two different substances that happen to share a name. From the outside they look identical, and they need opposite treatment: the first pair has to be merged, and the second must never be.

Measured over EF 3.1's 94,062 flows, and excluding pairs that this project's own renames created, 171 (name, compartment) pairs hold two rows. They split almost evenly, and the split is on whether the registry numbers agree:

Two substances once: nine names doing double duty

The other 104 groups are the dangerous ones, because merging them would be wrong. Nine names in EF 3.1 each cover two different chemicals in the same compartment, and in every case the pair is a substance and a close relative of it:

Fumarate and maleate are the trans and cis forms of the same acid ester and behave differently; CaHPO₄ and Ca(H₂PO₄)₂ carry different amounts of phosphorus per kilogram; L-lactide and the racemate polymerise to different plastics. Nothing in the flow name distinguishes any of them.

The decision. Identity is keyed on the registry number, so these pairs stay apart on their own: each row keeps its number, its structure and its factors, and they are published as two substances rather than fused.

What is still wrong. Keeping them apart is not the same as telling them apart on sight. Six of the nine pairs are published under one and the same preferred label today — both 1,3-benzenediamine records are called 1,3-benzenediamine, both esters Bis(2-ethylhexyl) But-2-enedioate, both lactates Methyl 2-hydroxypropanoate. The records are distinct and correct; the names on them do not say so. Three of the nine were separated by a better name arriving from the registries (Calcium hydrogen phosphate, Lactide, Tridemorph), which is what the other six still need. See Known limitations for how this reads in the published output.

What it cost, by source list:

ecoinvent has none: one flow per substance per compartment, throughout. BAFU has none either, though it looks otherwise at first — its repeated names are regional variants of one flow (Water, FR, Water, DE, …) and pairs measured in becquerels and kilobecquerels, which is a units finding rather than a duplication.