The common thread
ATRP, RAFT, and FRP all propagate through a free radical chain end (shown in red below), the same reactive, indiscriminate species in all three. What separates a controlled radical polymerization (ATRP, RAFT) from an uncontrolled one (FRP) is entirely about what happens to that radical between monomer additions: ATRP and RAFT both add a reversible step that keeps most chains "parked" in an unreactive, dormant form (shown in blue) most of the time, so they grow in small, evenly shared increments instead of a few chains racing to completion while others never even start. ROMP is the odd one out mechanistically: it never forms a free radical at all, and gets its living character from a metal (shown in green) that stays covalently bonded to the growing chain end throughout.
ATRP: Atom Transfer Radical Polymerization
A dormant alkyl halide chain end is reversibly activated by a transition metal/ligand complex, which abstracts the halogen to briefly generate a radical chain end.
This is "controlled" because the equilibrium sits heavily toward the dormant Pn–X side (kact ≪ kdeact), so the instantaneous concentration of radicals in solution is kept very low at all times. Since termination is second order in radical concentration but propagation is only first order, suppressing [Pn•] suppresses termination much more than it slows chain growth. Nearly every chain survives, grows a little bit at a time, and ends up close to the same length.
What holds the equilibrium in that useful place is the persistent radical effect, which explains both the control and the main failure mode. The two radicals in the system are not equivalent: the propagating radical is transient and can terminate, while the X–Cu(II)/L deactivator is persistent and cannot terminate with itself. Every termination event therefore consumes two propagating radicals and leaves behind an equivalent of deactivator, so Cu(II) irreversibly accumulates. Since the rate depends on the ratio of activator to deactivator, Rp = kpKATRP[M][Pn–X] × [CuI/L] / [X–CuII/L], that buildup pushes the equilibrium further toward dormant and self-limits further termination. The system is self-correcting.
That same ratio is why you cannot fix the cost and purification problem by simply using less copper. Classical ATRP runs at roughly 1000 to 10,000 ppm of catalyst versus monomer. Drop the total catalyst far below the alkyl halide concentration and the persistent radical effect converts nearly all of your activator into deactivator, at which point the reaction stalls outright. ARGET/ICAR/SARA/eATRP and photoATRP solve it properly: a reducing agent (tin(II), ascorbic acid), a radical initiator like AIBN, a zerovalent metal, an electric current, or light continuously converts the accumulated Cu(II) back to Cu(I). Preserving the ratio rather than the absolute amount takes practical copper loadings from roughly stoichiometric with initiator down to parts per million versus monomer, and it limits catalyst-induced side reactions with the propagating chains, which otherwise cap the achievable molecular weight. That regeneration also confers useful oxygen tolerance, since the regeneration chemistry consumes small amounts of oxidized catalyst along with the oxygen that caused it.
Two practical numbers follow from this. First, well-defined ATRP tops out around 150,000 to 200,000 g/mol; above that, normal bimolecular termination becomes significant (worst at high conversion) and control is lost. Second, for ATRP block copolymers, stop the first block at roughly 90 to 95% conversion rather than 100%: as the monomer runs out the rate falls, termination is no longer suppressed, and it eats the halide chain ends you need to grow the second block.
RAFT: Reversible Addition Fragmentation chain Transfer
A thiocarbonylthio compound (the RAFT agent) shuttles the radical "identity" rapidly between all growing chains via degenerate chain transfer, rather than suppressing the total radical population the way ATRP does.
Unlike ATRP, RAFT doesn't lower the total radical concentration below a normal FRP level. Termination still happens at roughly the same rate it would without the RAFT agent; it's just a small loss relative to the much larger population of chains cycling through the main equilibrium. Some RAFT systems, notably the dithiobenzoates, are an exception and do show real rate retardation: the intermediate adduct radical [Pn–S–C•(Z)–S–Pm] can couple with a propagating radical or fragment only slowly, which is also why clean stars and other complex architectures are hard to build by RAFT. Every active chain adds a monomer, then quickly gets "capped" back to a dormant thiocarbonylthio terminated chain while a different chain takes its turn as the radical, so on average, all chains grow in lockstep even though only one is actively propagating at any instant. The Z group tunes the RAFT agent's reactivity for a given monomer class; the R group becomes the initial leaving and reinitiating fragment and ends up as the other chain end. The radicals don't even have to come from a thermal azo initiator; the next section shows the photochemical route.
PET-RAFT: Photoinduced Electron Transfer RAFT
The RAFT equilibrium doesn't care where its radicals come from. In PET-RAFT, a visible light photocatalyst (a metal polypyridyl complex, a porphyrin, or an organic dye such as eosin Y at parts per million versus monomer) replaces the thermal initiator entirely: the photoexcited catalyst transfers an electron to the thiocarbonylthio compound, which fragments to release the propagating radical directly from the dormant chain end.
That has practical consequences. Activation only happens while the light is on, so chain growth can be paused and resumed at will; that on/off temporal control is unique to the photochemical methods. The whole polymerization runs at room temperature. And because there is no azo initiator, there are essentially no initiator derived chains: every chain starts and ends on the CTA, which is why PET-RAFT gives such high end group fidelity for block extension. With an added tertiary amine such as triethylamine, eosin Y systems even tolerate residual oxygen through a reductive quenching cycle, allowing polymerization without rigorous degassing. Catalyst choice sets the working wavelength: eosin Y and fluorescein respond to green or blue light, Ru(bpy)₃²⁺ and fac-Ir(ppy)₃ to blue, and porphyrins such as ZnTPP reach into the red. The calculator's RAFT tab has a PET-RAFT mode that doses the photocatalyst in ppm for you.
Key reference: Xu, Shanmugam, Duong, and Boyer, Polym. Chem. 2015, 6, 5615 to 5624 (organo-photocatalysts including eosin Y and fluorescein at 10 to 100 ppm for PET-RAFT of methacrylates, including oxygen tolerant operation with triethylamine).
ROMP: Ring Opening Metathesis Polymerization
No radicals here at all. ROMP runs on chemistry (the Chauvin mechanism): a metal alkylidene reacts with a strained cyclic olefin through a [2+2]/retro[2+2] cycloaddition sequence, opening the ring and handing the reactive carbene off to the new chain end.
Simplified metallacyclobutane intermediate: the four membered ring that forms and immediately opens on every cycle. When it opens, one metal–carbon bond and the carbon–carbon bond across from it break; the metal keeps its other carbon as the regenerated carbene (M=C) and the remaining two carbons become the new backbone alkene (C=C), so the metal never leaves the chain.
Because the metal never leaves the chain end between cycles, ROMP is "living" in the same practical sense as ATRP/RAFT (predictable Mₙ, low dispersity, chain extension possible) but for a completely different mechanistic reason. There's no dormant/active equilibrium to manage, just a metal that keeps finding the next ring to open. Leaving the reaction to run too long without quenching risks secondary metathesis (chain scission/backbiting) rather than radical termination. The driving force is the relief of ring strain in the monomer, which is why highly strained rings like norbornene polymerize so readily and so fast; very large, essentially unstrained macrocycles can still be polymerized, but there the driving force is entropic instead. The flip side is cyclohexene: a nearly strain-free six-membered ring, it gives only very-low-MW oligomers and is effectively unpolymerizable by ROMP, a concrete reminder that ring strain, not the metathesis chemistry itself, is the thermodynamic driver. The practical limit on ROMP is catalyst death, which is the main termination pathway. It bites hardest exactly where you would want living behavior most: as each block nears full conversion the reaction becomes monomer starved, the polymerization slows, and the longer the active catalyst sits there the more of it decomposes. That is why ROMP diblocks are routine while clean tri-, tetra-, and higher multiblocks are genuinely difficult, and it is a good argument for not chasing the last few percent of conversion on an early block. A closely related technique, ADMET (acyclic diene metathesis), uses the same metal carbene chemistry but builds chains from linear diene monomers with loss of ethylene instead of opening a ring. Unlike ROMP, though, ADMET is a step-growth polymerization (the same class as the polyurethanes further down this page), not a chain reaction, so high molecular weight only shows up at very high conversion. It isn't one of this calculator's tabs, but it's worth knowing it's a mechanistic cousin of ROMP.
Free Radical Polymerization (FRP)
The baseline case: a radical is generated once and propagates until it dies. There's no reversible step holding chains back, which is exactly why FRP isn't a controlled/living technique.
There's no dormant state and no equilibrium bringing a chain back once it's a radical. Each chain simply propagates at rate kp[M] until it terminates (or transfers), which typically happens in well under a second. Because initiator decomposition (kd) is slow and continuous, new chains keep starting throughout the whole reaction while earlier chains have already terminated, so the population is a statistical mix of chain lengths (kinetic chain length set by kp, kt, and the steady state radical concentration) rather than the narrow distribution set by the feed ratio that you get from ATRP/RAFT/ROMP. This is exactly why the calculator's FRP tab treats Mₙ as a simplified estimate rather than a precise design target.
Polyurethane Formation
This one has nothing in common with the four mechanisms above. There's no radical and no metal carbene, just a step-growth polyaddition: a nucleophile (an alcohol oxygen lone pair) attacks the electrophilic carbon of an isocyanate, and every new bond simply stitches two existing molecules together. It's addition chemistry, but the growing "chain" is built entirely from these individual addition steps between separate molecules rather than a chain end adding one monomer at a time. No small molecule is released, so strictly this is a polyaddition, yet polyurethanes are conventionally grouped with the condensation/step polymers because the urethane linkage (–NH–CO–O–) is structurally kin to the ester (–CO–O–) and amide (–NH–CO–) linkages.
The calculator's Step 1 caps a polyol (soft segment) with excess diisocyanate; Step 2 chain extends the leftover isocyanate ends, building the hard segment shown here between two diisocyanate units.
There's no chain end propagating through hundreds of monomer additions the way there is above. Instead, molecular weight builds through step growth: any two molecules with a free OH and a free NCO can react, at any point in the reaction, so early on the mixture is mostly short oligomers, and high molecular weight only shows up once conversion is very high. This is exactly why the calculator's polyurethane tab is built around a two step recipe instead of a target Mn and DP: cap a polyol with excess diisocyanate first (Step 1), so the free NCO ends are known and controlled, then react that prepolymer against a measured amount of chain extender (Step 2) to build the final hard segment length deliberately, rather than leaving stoichiometry to chance.
Emulsion Polymerization
This isn't a different chain-growth chemistry from FRP above. Once a chain is growing inside a particle, propagation and termination happen exactly the same way they do in bulk. What's different is where and how each chain gets started: the initiator is water soluble and generates radicals in the aqueous phase, not inside the monomer itself, and those radicals have to find and enter a surfactant micelle before a particle even exists.
The numbers trace the sequence. Surfactant (head toward the water, tail toward the oil) assembles into a monomer-swollen micelle (1); a water-phase radical enters and nucleates it (2); the particle then grows (3) as more monomer keeps diffusing in from the large droplet reservoir. Only a small fraction of micelles ever capture a radical, so far fewer particles form than there were micelles (only about 0.1% of them, typically 1016 to 1018 particles per liter); the rest give up their surfactant to stabilize the growing particles.
That snapshot is a single instant. Run the same reaction from 0 to 100% conversion and it passes through three distinct regimes, the Smith-Ewart intervals, and the particle count set during the first, short one is what locks in everything after it.
Interval I is nucleation: radicals are still finding micelles, so the particle number climbs and the rate climbs with it. It is the shortest interval (typically 2–15% conversion) and ends once the surfactant has all been absorbed onto particles and the micelles run out. Interval II starts once that particle number is locked in: monomer droplets still exist as a reservoir, feeding the particles through the water, so the rate holds roughly steady (or creeps up if the gel effect kicks in). Interval III begins the moment the droplets are exhausted; the particle number stays the same, but with no reservoir left the monomer concentration inside each particle falls, and the rate falls with it.
That's why the particle count from Interval I is the whole game: everything measured afterward (rate, final molecular weight, particle size) is downstream of a number that was fixed early and never revisited.
Because nucleation is essentially over once the micelles run out, the particle count set during that short window determines the final particle size for the rest of the reaction: the same total polymer volume divided among however many particles got started. That's the whole basis for the calculator's emulsion tab, which estimates that particle count from the surfactant and initiator charge (Smith-Ewart theory) and then reports the final diameter from a simple mass balance. Everything downstream of nucleation (propagation, transfer, termination) is ordinary free radical chemistry; the physics of getting a chain started is the only genuinely new mechanism here. This micellar picture holds when surfactant is well above its CMC; for styrene and MMA it accounts for over 99% of the particles. Below the CMC, or for more water-soluble monomers, a second route takes over: homogeneous nucleation, where oligomers growing in the water phase eventually become insoluble, precipitate, and pick up surfactant to become particles in their own right.