The work vs. CV functions for each of 30 SMD replicates are given in color in Panel A as is their Jarzynski average (solid black line). For clarity, only the Jarzynski average, the bias corrected Jarzynski average (solid red line)

the square root of the mean square error (MSE

as defined in Gore,

*et al.* [24], dashed red lines), and the definition of the CV are shown in Panel B. Note that CV steering is from 8.92 Å to 3.52 Å, i.e., right to left for the formation of the carbinolamine. The large scatter in “barrier height” seen for the individual work functions and the scatter in the work values at the end of the simulation (left axis) are normal and expected for these kinds of methods [39]. Such barrier heights apply only to the pathway followed in that trajectory, and the individual work functions are not good estimates of the pathway

G function for the reaction. The purpose of the exponential averaging in the Jarzynski algorithm is to properly weight the individual work functions so that functions with large peaks contribute only to a degree appropriate to their probability of occurrence. The estimate of

G along the reaction pathway should use the Jarzynski average or some other comparable function such as its bias corrected version (the better choice) [24].

## PubMed Commons