One number, measured four ways
A polyurethane cure has one quantity worth tracking: how much of the isocyanate has reacted. Call it α. Everything a formulator cares about – whether the part can be demoulded, whether the modulus has stopped climbing, whether free isocyanate remains – is a question about α or about the network it has built so far. The awkwardness is that no instrument measures it directly. Each technique measures something else and infers α from a calibration:
- DSC reheats a sample and integrates the exotherm that is left. If an uncured specimen gives ΔH0 and a partly cured one gives ΔH, then α = 1 − ΔH/ΔH0. It is the most direct route and the one that needs the least thought, because the heat comes from the reaction itself.
- FTIR follows the isocyanate stretch at roughly 2270 cm−1, which sits in an otherwise empty window and disappears as the reaction runs. The band area has to be ratioed against something that does not change – the aliphatic C–H stretches near 2900 cm−1 are the usual choice – because path length and contact pressure vary from spot to spot and would otherwise be read as conversion.
- DMA or TMA measure the glass transition, which climbs as the network tightens. The DiBenedetto equation turns Tg back into α. This is the least direct route and the most sensitive to calibration, but it is the only one that works on a finished part.
(Tg − Tg0) ÷ (Tg∞ − Tg0) = λα ÷ (1 − (1 − λ)α)
Three routes to one number is not redundancy, it is the test. They rest on different physics and fail differently, so when they agree the number is probably right, and when they diverge the pattern of divergence says which one to distrust. That is what this page is for: enter what the instruments reported, and see all three side by side along with what the stoichiometry says is even possible.
Stoichiometry sets the ceiling
Before any measurement means anything, the formulation decides how far α can go. Isocyanate conversion is capped by whatever it has to react with. Run the isocyanate in excess and the hydroxyl runs out first, which caps α at 1/r where r is the ratio of isocyanate to hydroxyl equivalents. At index 105 that ceiling is 0.952, so a specification demanding 95 % NCO conversion is asking the formulation to go essentially to its limit, and a specification demanding 98 % is asking for something arithmetically impossible. That is worth checking before anyone blames the oven.
Water belongs in that arithmetic, and it counts twice. Water consumes one isocyanate to give a carbamic acid, which loses CO2 to leave an amine, and that amine consumes a second isocyanate to give a urea. So each mole of water removes two equivalents of NCO, and it does it while generating gas. A polyol at 200 ppm water is not dry: in the example below those 200 ppm shift the effective index by more than two points, which is larger than most people's weighing error. This page adds 2 × mol H2O to the hydroxyl equivalents before computing anything.
The gel point follows from the same equivalents. For a mixture of functionalities it is αgel = 1/√(r (fw−1)(gw−1)), with the functionalities averaged over groups rather than molecules, because a triol offers three chances to enter the network and a diol two. Water enters this as a difunctional hydroxyl component, which is one more reason wet polyol changes the answer.
The formulation
Only used to turn moles of elastic strand into a concentration.
The conversion the cure has to reach to pass.
Calibration and measurements
Each route needs its own two-point calibration: the value at zero conversion and, for the Tg route, the value at full cure. Run an uncured specimen and a fully cured one and these come straight off those two traces.
Roughly ΔCp∞/ΔCp0. Fit it if you have a Tg at a conversion you trust.
Kinetics and time to target
Urethane formation is second order, one isocyanate and one hydroxyl per bond. Integrating that for an arbitrary ratio gives a straight line through the origin:
kapp is k[OH]0 lumped together so it comes out per hour, which is the form worth quoting because it is what sets the cure schedule. Fitting several temperatures gives the activation energy from ln k = ln A − Ea/RT, and with Ea in hand the cure time at any other temperature follows.
The network at that conversion
Conversion on its own does not tell you whether a part is usable; the network does. Past the gel point, the same equivalents and functionalities that gave αgel also give the concentration of elastically effective strands by the Miller–Macosko recursive method, counting a junction only when at least three of its arms reach the infinite network. Two attached arms makes a chain extender and fewer makes a dangling end, and neither carries stress. The rubbery modulus follows as E′ = 3νeRT.
Where each method goes wrong
- DSC drifts low when the reaction is nearly over. At 95 % conversion the residual exotherm is 5 % of a figure that was never large, and the baseline under a broad shallow peak is a judgement call worth several percent of α. Worse, the residual scan itself cures the sample as it heats, so a slow scan on a nearly cured specimen reports less heat than a fast one. If DSC reads systematically below the other routes above about α = 0.9, suspect the baseline before you suspect the chemistry.
- FTIR is undone by what else absorbs near 2270 cm−1. The window is usually clean, but a carbodiimide from an overheated cure absorbs at 2130, a nitrile filler absorbs at 2240, and a CO2 bubble from the water side absorbs at 2340 with a band shape that leaks into a wide integration. Narrow the integration limits and state them. The internal reference has its own trap: aliphatic C–H near 2900 is the right choice only if nothing in the formulation changes there, which rules it out when an aliphatic plasticiser is migrating.
- The Tg route is only as good as λ. DiBenedetto is an interpolation with one shape parameter, and λ is usually taken as ΔCp∞/ΔCp0 from the same two calibration runs. If Tg∞ was measured on a specimen that was not actually fully cured, every α from this route is biased high, and nothing in the fit will tell you. The route is also insensitive where it matters least and most sensitive where the curve is steep: near full cure, a 1 °C error in Tg is worth far less conversion than the same error at half cure.
- Vitrification stops the reaction without stopping the clock. When Tg climbs past the cure temperature the material goes glassy and the rate falls off a cliff, because the reaction becomes diffusion controlled rather than chemically controlled. The second-order fit on this page knows nothing about that, so it will keep predicting conversion the sample cannot reach at that temperature. If the measured points fall progressively below the fitted curve at long times, vitrification is the first thing to suspect, and the fix is a post-cure above the final Tg, not more time.
- One rate constant is doing several jobs. The fit lumps the hydroxyl and water reactions into a single k, and they do not have the same rate or the same activation energy. With a dry polyol that is a small error; at a few hundred ppm it starts to matter, and a catalyst that favours one over the other makes it worse. Primary and secondary hydroxyls differ too, which is why a polyol blend rarely gives as clean a line as a single polyol.
A worked example
The values loaded above are a polyether triol of 1000 g/eq at 200 ppm water, cured with a difunctional isocyanate of 125 g/eq. Those 200 ppm are 1.11 × 10−4 mol of water against 0.01000 eq of hydroxyl, and because each mole takes two isocyanates they add 2.22 × 10−4 to the effective hydroxyl equivalents – enough to move the index from 100.0 down to 97.8. The gel point lands at α = 0.719 and, because the isocyanate is now the deficient group, αmax is 1.
The DSC, FTIR and Tg columns agree to within a few parts per thousand, which is what synthetic data looks like and what real data does not; see the spread column for what your own numbers do. The rate constant comes out near 0.30 h−1 at 60 °C and 0.54 h−1 at 70 °C, for an activation energy of about 56 kJ/mol, which is the usual range for a tin-catalysed urethane. Those two together say the cure reaches 95 % in about 53 hours at 60 °C and about 29 hours at 70 °C – the familiar result that ten degrees roughly halves a cure.
Now change the isocyanate mass from 1.250 to 1.342 g, leaving every measurement exactly as it was. The index goes to 105.0 and αmax falls to 0.9521, so the 0.95 target now sits a thousandth below the ceiling. More tellingly, the fit degrades: R2 at 60 °C drops from 1.0000 to 0.9154, and the activation energy falls from 55.9 to 42.8 kJ/mol. Nothing about the measurements changed – only the stoichiometry they are being interpreted against.
That degradation is the useful part. The second-order form only straightens out when r is right, so a cure plot that curves is often telling you the index is not what you think it is, and a wet polyol or a mis-weighed isocyanate is a likelier explanation than exotic kinetics. Check the stoichiometry before blaming the oven.
Related tools
- Polyurethane calculator. The other half of the problem: equivalent weights, prepolymer %NCO and chain extension, for designing the formulation this page then watches cure.
- Network formation. The same Miller–Macosko recursion with sol fraction and a gel-point sweep, for any mixture of functionalities rather than one cure.
- Crosslink density. Mc measured from swelling or the rubbery plateau, to check the network this page predicts against the one you actually made.
- Thermal analysis. What a DSC trace is telling you, including how to place a baseline under a broad cure exotherm.
- Step-growth polymerisation. Carothers, the gel point from a single formula, and why stoichiometry dominates everything in a condensation.