Docs / What we have found

The name is a part number, and the part number is not an identifier

Part of What we have found, the class where a flow's whole name is a trade designation — HFC-134a, CFC-113 — which no registry issues and no database can resolve.

What the code does not tell you

Fifty of EF 3.1's part-numbered flows shipped with no identifier at all — no registry number, no EC number, no formula, no structure, nothing but the designation and a characterisation factor. (E)-HFC-1225ye is one: EF ships five flows and ten factors under it, and nothing downstream could place the substance. What resolves it is the code itself, used as evidence rather than as a name: the digits decode to C₃HF₅, PubChem's record for CID 6329539 lists (E)-HFC-1225ye among its synonyms and reports the same formula, and the two agreeing is what justifies attaching 5595-10-8 to the flow.

And the designations are not always right. Two of EF's are internally contradictory:

• HCFC-140 — described with the trichloroethane flows — puts an F for fluorine in the name of a molecule that has none, and drops the isomer letter that separates it from the chemical it keeps being confused with.
• HFC-1234yf is a typo for HFO-1234yf. The leading digit of a four-digit refrigerant number is its count of double bonds, so 1234yf names an olefin, and the HFC family is saturated by definition: the compound cannot be an HFC. EF itself uses the right prefix elsewhere — its polyhaloalkene flow, which is this same substance, carries hfo-1234yf in its own synonym list. PubChem records the wrong spelling as a synonym, which makes it findable, not correct.

The decision. Where a registry names the number, the substance is published under the chemistry — HFC-134a becomes 1,1,1,2-Tetrafluoroethane — and the designation is kept as an alternative label, so an inventory written against EF 3.1 still finds it. That is done for 70 of the 118. The remaining 48 keep the part number, and the reasons are written down rather than left as an oversight: for 26 there is no chemical name to be had, and for the other 22 the available name is worse than the code — it names a mixture as one isomer, or it is itself another part number, or four isomeric ethers share one registry number and renaming them would publish four substances under one name (#521, #202; the full accounting is in Known limitations).

A designation is not an identifier

Reading a code as evidence works because the codes are systematic. Using one as a name does not.

Refrigerants, blowing agents and fire suppressants are traded by designation: HFC-134a, CFC-113, Halon-1211. The designations are systematic — the digits encode the counts of carbon, hydrogen and fluorine, the trailing letters the isomer — but they are issued by no registry, they are not unique across the families that use them, and a database cannot resolve one. When a flow list uses a designation instead of a chemical name, the substance has no name anything can look up.

EF 3.1 names 118 substances this way. Only EF does: measured over the extracted inputs, 714 EF flows have a designation as their entire name, and in ecoinvent and BAFU that number is zero — those lists append the designation to a chemical name instead (Ethane, 1,1,2-trichloro-1,2,2-trifluoro-, CFC-113), which loses nothing.

When the designation is the only half that identifies

Appending the designation to a chemical name loses nothing if the reader knows to read the tail. Sometimes the tail is all there is.

ecoinvent 2 named a family of fluorinated ethers as structural prose with the designation on the end, and gave every member of a family one registry number. Stepwise 2006, a SimaPro method file of that era, ships those rows verbatim — and within a family the prose and the number are identical, so the designation is the only thing that differs:

The file contradicts itself in its own numbers. Three rows it says are one substance — one name, one number, one formula, C4H3F7O — are characterised 576 and 963 kg CO₂-eq per kilogram. EF 3.1 ships the same three substances under the same shared number and gives them GWP100 of 576, 963 and 980. What the designation encodes is the arrangement of the ether, which is exactly what those numbers are sensitive to: the shared registry number is not a claim that these are one chemical, it is a family's number written on each member, and the trailing code is the member.

Twelve rows are in this shape, across four families: 406-78-0 on three, 382-34-3 on four, 1885-48-9 on three and 84011-06-3 on two. Both halves of the identification fail together on all twelve — the number reaches every member of the family, and the prose reaches none of them, because it is nobody's published name. A row that states a name and a registry number, both correctly copied, identifies nothing.

The decision. Where a registry number has found several substances and the whole name has matched none of them, the last comma-segment of the name is read against those candidates' published names — and only against those, so it can settle a family the number already reached and can never reach outside it. The row records cas+designation, so a match made on that reading is distinguishable from one made on a name the vendor shipped. Asked of Stepwise 2006's 6,064 rows it moves exactly these twelve, each to its designation's own substance: 5,061 matched becomes 5,073 (#202). EF's four shared numbers are separately a contested-CAS question, and all four carry a ruling that the number cannot stand for the whole family.

What it cost, by source list: