DETErmine_bath n :

This flag starts n identical copies of a procedure aiming to determine a suitable temperature (or temperature range) for all four annealing modes. This we do as follows : the temperature is set to a sufficiently remote (very high) value²⁰of T₀ = 0.6250. The system is equilibrated at this temperature for $\Delta$t=10000 time steps and the mean value of the resulting R-factors (say, $\bar{R}_{0}^{}$) is saved. Then the temperature is set to T₁ = T₀/2 = 0.31250 and the procedure repeated up to and including T₁₃ = 0.0000763 which will give us $\bar{R}_{{13}}^{}$. For every pair of average R-factors $\bar{R}_{n}^{}$, $\bar{R}_{{n+1}}^{}$, we calculate $\Delta$$\bar{R}_{n}^{}$ = $\bar{R}_{n}^{}$ - $\bar{R}_{{n+1}}^{}$ and then we do a cubic spline interpolation on the resulting $\Delta$$\bar{R}_{n}^{}$'s and from that we determine the temperature T₀ for which $\Delta$$\bar{R}$ is maximum²¹. A graphical representation of the results from this calculation are written in the file temp.ps (a postscript file, only produced when the keyword POSTSCRIPT is on).

Footnotes

... value²⁰

The probability of accepting a move against the gradient is exp(- |$\Delta$R|/T). Now, |$\Delta$R| is the absolute value of the difference between two values of the conventional R factor and, so, takes values of the order of 10^-1. For, say, T = 10, the probability of accepting a move against the gradient is 0.990, ie practically all moves are accepted and so the system is thermally disordered.

... maximum²¹

The quantity $\Delta$$\bar{R}$/$\Delta$T is analogous to the specific heat from statistical mechanics. A large value of $\Delta$$\bar{R}$/$\Delta$T signifies that a ``phase'' transition is occurring at the corresponding temperature, and so the cooling rate should be as low as possible at this temperature range.