Bias-based free energy estimates averaged over 32 runs of 8 ns
for four different metadynamics choices designed to give different
levels of resolution of the high-energy regions. Clockwise from top
left: WTMetaD, TTMetaD, 15 kJ/mol transition-referenced MBMetaD, 42.5
kJ/mol minimum-referenced MBMetaD. All methods use parameters as defined
in the text. Contours are placed every 5 kJ/mol. The estimates are
all but identical in low-energy regions despite stark differences
in the high-energy regions. WTMetaD is least controllable and gives
the most resolution of high energy regions, TTMetaD is more controllable
but would give more and more resolution of high energy regions given
additional convergence time, and both forms of MBMetaD are fully controllable,
converging without resolving the high energy regions.