Constant temperature run

In this mode you define a temperature T which is kept constant for the whole length of the calculation. The maximum move sizes (how much the parameters that we will test in the next move may deviate from the current parameters) are also kept constant for the whole run and equal to : max($\Delta_{{\rm t}}^{}$) = 2d_min/max(a, b, c), max($\Delta_{{\kappa}}^{}$) = 2d_min, where d_min is the minimum Bragg spacing of the input data, a, b, c are the unit cell translations of the target structure (in Å), max($\Delta_{{\rm t}}^{}$) is the maximum possible offset for any of the molecular translations (in fractional units) and max($\Delta_{{\kappa}}^{}$) is the maximum possible offset (in degrees) for the $\kappa$ angle of any of the molecular orientations¹⁸.

The constant temperature protocol can be quite efficient if you know how to choose a suitable temperature (so that the system is neither thermally disordered, nor as cold as to get stuck inside a shallow local minimum). I have recently started trusting the automatic temperature determination performed by QS, and so, an (automatically determined) constant temperature run is well-worth trying if the default runs fails to find a convincing solution.

Footnotes

... orientations¹⁸

QS stores the molecular orientations in terms of the polar angles $\omega$,$\phi$,$\kappa$. A new molecular orientation is obtained from the previous one by rotating the molecule by $\kappa$ degrees about an axis defined by the $\omega$ and $\phi$ angles. The orientation of the new rotation axis is always chosen randomly in the interval 0 $\leq$ $\omega$ $\leq$ $\pi$/2 and 0 $\leq$ $\phi$ < 2$\pi$, ie the full half sphere. We obtain the new orientation by rotating about this axis by $\pm$$\xi$$\Delta_{{\kappa}}^{}$ degrees, where $\xi$, 0 $\leq$ $\xi$ $\leq$ 1 is a random number.