Live numbers
1. The catalytic cycle
Pd(0)L2 + Ar-X -> Ar-Pd(II)-X oxidative addition
Ar-Pd-X + R'-M (+ base) -> Ar-Pd-R' + M-X transmetalation (Heck: migratory insertion)
Ar-Pd-R' -> Ar-R' + Pd(0)L2 reductive elimination
The palladium is not used up: it ends each turn exactly as it began, so one atom can make thousands of new carbon-carbon bonds. These reactions earned the 2010 Nobel Prize in Chemistry (Heck, Negishi, Suzuki).
2. The rate equations
d[Pd0]/dt = -k1 [Pd0][ArX] + k3 [TM] - kd [Pd0]
d[OA]/dt = k1 [Pd0][ArX] - k2 [OA][R'M]
d[TM]/dt = k2 [OA][R'M] - k3 [TM]
d[Pd black]/dt = kd [Pd0] d[P]/dt = k3 [TM]
k(T) = k_ref exp[ -Ea/R (1/T - 1/353.15 K) ]
With substrate held fixed, the steps act like resistors in series, so the turnover frequency per palladium is TOF = 1 / (1/(k1 [ArX]) + 1/(k2 [R'M]) + 1/k3). The slowest step dominates, and that is the one the strip names.
3. Ligand bulk and temperature
Bulky, electron-rich phosphines speed up oxidative addition and reductive elimination, crowd the metal so transmetalation slows, and keep Pd(0) from clumping into inactive palladium black. In the model this is a smooth exponential factor in the bulk slider: an illustration of the trend, not a fit. Heating speeds every step (Arrhenius), but catalyst death has the largest activation energy, so too much heat wastes the catalyst.
Honest simplifications
Real cycles have more intermediates (ligand loss, base and solvent roles, side reactions such as homocoupling and protodeboronation). Rate constants, activation energies and ligand factors are illustrative values of a typical order of magnitude, chosen so a run takes a few hours; they do not describe one specific catalyst. One-electron and cluster pathways are not modelled.
Independent numerical checks