Three different questions
"End group analysis" covers three questions that need different instruments, and picking the wrong one is the usual reason an answer disappoints.
- What is on the end? An identity question. Mass spectrometry answers it outright; infrared and NMR answer it if you already suspect what you are looking for.
- How many chains carry it? A fidelity question, and the one that matters before a chain extension or a block. Needs a method that counts end groups against chains, not against mass.
- How long are the chains? A counting question. If every chain carries exactly one end group, counting end groups against repeat units gives Mn directly, with no calibration and no standards. That is the great virtue of the method and the source of its limits.
Pendant groups raise a fourth question. What fraction of the repeat units carry the modification? That is the same integration arithmetic pointed sideways instead of at the chain ends.
What the words refer to
Count them. Two end groups per chain however long it gets, and one pendant group for every repeat unit. That asymmetry is the whole method and its limit: the end-group signal is a constant while the backbone signal grows with the chain.
Which technique for which question
Eight methods, and the useful question is not which is best but which answers the question you actually have. The pill on each card is what it costs you to get a molecular weight out of it.
The catch. The end-group signal shrinks as the chain grows; see the ceiling below. Needs a relaxation delay long enough to be quantitative: a routine delay set for speed under-integrates the slowest-relaxing protons, which are often exactly the end group.
The catch. Only if fluorine is there. When it is, the method is far better than ¹H: fluorine is 100% abundant, the chemical shift range is enormous so nothing overlaps, and there is no background from solvent or backbone.
The catch. The sample must be derivatised with a phosphitylating reagent first, so every group you want to see has to react quantitatively. Worth it when "how much OH" is not a single question.
The catch. Band area depends on path length and concentration as much as on chemistry, so a number needs a calibration curve and an internal reference band. The exception is isocyanate: a strong, isolated band near 2270 cm−1 with nothing else nearby, which is why polyurethane cure is tracked this way as a matter of routine.
The catch. Returns an equivalent weight, not a molecular weight. Multiplying by the wrong functionality is the classic error. Needs a lot more sample than NMR and sees nothing about structure.
The catch. The definitive answer to "what is on the end", because you see the mass. But ionisation and detection favour some chain lengths over others, so averages from a broad sample are biased; treat it as an identity method first.
The catch. Needs ε for that group in that solvent, and it assumes the chromophore is intact. A thiocarbonylthio end group has a strong π→π* band near 300 nm and a weak visible band, which is why RAFT polymers are coloured and why fading is itself a report on end-group loss.
The catch. Sulfur from a RAFT agent or bromine from an ATRP initiator gives end-group content by mass. Blunt, and easily wrecked by residual catalyst, salt or solvent, because it cannot tell a chain end from a contaminant containing the same element.
Mₙ from ¹H NMR end-group integration
Count repeat units against chains. If a backbone signal comes from Hb protons per repeat unit and an end-group signal from He protons per chain, the ratio of the normalised integrals is the degree of polymerization:
For PMMA the OCH3 singlet is 3H.
Per chain, not per group: a telechelic with the same group at both ends contributes twice.
Both ends together: for ATRP, the initiator fragment plus the halogen.
Baseline drift and noise cost you a roughly fixed slice of the total integral. 0.1–0.5% is a realistic range on a clean spectrum.
The ceiling, and why it is where it is
One window of a PMMA spectrum at three chain lengths, each normalised to its backbone OCH3 singlet. The ester OCH2 quartet from the initiator fragment is drawn at forty times vertical expansion in all three panels, so the collapse you are watching is real and not a change of scale. Peak heights are integrated areas at equal linewidth; this is a schematic of one window, not a simulated spectrum.
That is the whole argument in one picture. The backbone peak does not care how long the chain is, because it grows in step with it. The end-group peak is the same two protons at every chain length, so as a fraction of the spectrum it falls as 1/DP: 2.6% of this window at DP 25, 0.66% at DP 100, 0.17% at DP 400. At DP 400 a 0.2% error in where you put the baseline is larger than the entire signal you are trying to measure, and the Mn that comes out is not wrong by a little.
Which peak you integrate is a decision, and it lands in He. The figure uses the ester OCH2 quartet, 2H per chain, because it sits clear of everything else. The gem-dimethyl of the same initiator fragment is 6H, three times the signal, and the table below assumes it. But it lands among the backbone α-methyl signals, which tacticity has already split into three. Three times the signal is worth having when you can integrate it cleanly, and He is an input on the calculator above for exactly this reason. The ceiling moves with the choice: about 20 kg/mol on the 6H signal, about 7 kg/mol on the 2H one.
End-group NMR is usually said to run out somewhere around 20 kg/mol. That is not a property of the spectrometer, it is arithmetic. The end-group signal is a fixed number of protons; the backbone grows with every unit added, so the end group's share of the spectrum falls as 1/DP. A baseline error that costs a fixed slice of the total integral therefore costs a growing fraction of the end-group peak, and that error lands directly on DP.
For PMMA made from ethyl 2-bromoisobutyrate, with a 3H backbone singlet and a 6H end-group signal:
| DP | Mn | End group's share of the spectrum | Mn error at 0.1% | at 0.2% | at 0.5% |
|---|---|---|---|---|---|
| 25 | 2,700 | 7.41% | 1% | 3% | 6% |
| 50 | 5,200 | 3.85% | 3% | 5% | 13% |
| 100 | 10,200 | 1.96% | 5% | 10% | 25% |
| 200 | 20,200 | 0.99% | 10% | 20% | 50% |
| 400 | 40,200 | 0.50% | 20% | 40% | 100% |
| 800 | 80,300 | 0.25% | 40% | 80% | 200% |
So the familiar ceiling is real and it has a number behind it: at Mn = 20 kg/mol a 0.2% integration error is already a 20% error in Mn, and at 40 kg/mol it is 40%. Two consequences worth acting on. Choose the end-group signal with the most protons per chain, not the most convenient one: six protons instead of two moves the whole table by a factor of three. And above about 20 kg/mol, stop treating an end-group Mn as a measurement and start treating it as a check on something else.
There is a second error the table does not show, and it runs one way. If some chains have lost their end group through termination, transfer or a failed initiation, the end-group integral is smaller than the chain count deserves, so DP and Mn come out too high. An end-group Mn above the target is therefore ambiguous in a way one below it is not: it can mean long chains, or it can mean dead ones. The distribution simulator shows what a dead-chain population does to the whole distribution, and GPC sees it directly as a low-molecular-weight tail.
Reading the ends off a mass spectrum
NMR counts end groups. Mass spectrometry reads them, which is a different and stronger thing: it answers the first of the three questions outright, and it does not need a calibration or a standard to do it. The whole logic is in the spacing and the offset.
A stretch of the MALDI spectrum of PMMA made from ethyl 2-bromoisobutyrate, sodiated. The comb envelope is the molecular weight distribution; the two series are two populations of the same polymer with different ends. Drawn from the same M0 and Mend the calculator above uses.
The spacing names the monomer. Consecutive peaks in one series differ by exactly one repeat unit, 100.12 for methyl methacrylate. That is a measurement you can make with a ruler, and it identifies what the chain is made of without reference to anything else.
The offset names everything that is not monomer. Take any peak, subtract the nearest whole number of repeat units, and what is left is the two end groups plus whatever cation carried the chain into the gas phase. Here 2420.7 − 22 × 100.12 leaves 218.0, which is the initiator fragment with its bromine (195.05) plus sodium (22.99). Get that arithmetic to come out and you have proved what is on the end, rather than assumed it.
A second series is a second population. The weaker comb sits 78.9 below the main one at every n, and 78.9 is a bromine traded for a hydrogen. Those chains are dead: they cannot be extended and they will not make a block. NMR would have folded them into the same averaged integral and reported a slightly high Mn; here they are a separate peak you can point at. This is why the grid above calls MALDI an identity method first.
What it will not give you is a trustworthy average. The relative heights of those two combs are set by how well each species ionises and how well the detector sees it at that mass, not only by how much of each is in the flask, and both effects vary across the distribution. Read the spectrum for what is there. Go back to NMR, or to the trace analyzer, for how much.
Degree of substitution on a pendant group
Post-polymerization modification asks the same question sideways: of all the repeat units, what fraction now carries the new group? Integrate a signal unique to the modified unit against one carried by every unit, modified or not.
Net change: the group gained minus whatever left.
Titration: equivalent weight, and the functionality trap
A titration counts reactive groups per gram of sample. That is an equivalent weight, and it becomes a molecular weight only when you multiply by how many of those groups each chain carries. Hydroxyl, acid and amine values are all reported as mg of KOH equivalent to one gram of sample, so they share one conversion; isocyanate is reported as a mass percent instead.
When the answers disagree
- End-group Mn above GPC Mn. Usually not a contradiction at all. A conventional GPC number is a polystyrene-equivalent molecular weight and can be off by tens of percent for any other chemistry; the end-group number is absolute. Convert the GPC value first on the calibration converter, then compare.
- End-group Mn above the target. Suspect lost end groups before celebrating long chains, because the arithmetic pushes Mn up when chains stop carrying the group being counted.
- Titration Mn below NMR Mn. Suspect the functionality. Any monofunctional impurity, or a chain that cyclised, changes groups per chain without changing the chemistry the titration sees.
- Everything disagrees on a branched or star polymer. Each arm carries an end group, so end-group counting returns arm length rather than molecular weight, while GPC returns something smaller than the true mass because a branched chain is compact. The chain dimensions page covers why.
Related tools
- Calculator. Sets the target Mn from the recipe, including the end-group term this page measures.
- GPC trace analyzer and calibration converter. The other route to Mn, and how to put the two on the same scale.
- Polyurethane calculator. Equivalent weights and index, where %NCO and hydroxyl value do their day job.
- Conversion monitoring. The same integration discipline applied to monomer consumption.