In the present study, we try to determine if the dynamics of the B-DNA backbone phosphates (and especially their interconversions between their two distinct conformations B I and B II ) are fast enough to be sufficiently sampled in the course of molecular dynamics simulations in the nanosecond time range. For this purpose, we performed twelve 10-ns simulations of the Drew-Dickerson dodecamer d(CGCGAATTCGCG) 2 to investigate the dynamics of B I /B II interconversion. We forced the DNA backbone angles and with restraints to values that are characteristic for B I and B II , resulting in DNA double helices with all phosphates in the B I or B II substate. These restraints were removed after 10 ns, and unrestrained simulations at temperatures of 250, 275, 287.5, 300, and 325 K were performed for another 10 ns, which allowed us to analyze the dynamics of relaxation in detail. These simulations were compared to simulations of the undisturbed dodecamer at 250 and 300 K, as a reference for the equilibrated state. We found that the relaxation from the B II state is considerably fast, with high rate constants, and is dependent on temperature. From this temperature dependence of the rate constants, we calculated the activation energy necessary for the B II to B I transition to be 2.5 kcal/mol. Half-life times of the B II state derived from the relaxation process are in the range of 110-370 ps, which indicates that a simulation time of 10 ns is sufficiently long to investigate conformational transitions of the DNA backbone. The structures of the all-B I DNA are more similar to structures found for the DrewDickerson dodecamer by X-ray crystallography than the all-B II DNA. This fact is not astonishing, because the B I conformation has been observed to be privileged. Nevertheless, both structures are quite different from canonical A-or B-DNA. That observation is revealing, because we expected the all-B II DNA to be the transition state to canonical A-DNA or at least structurally very similar. Furthermore, we find that the relaxation of our rather-distorted starting structures is fast and, despite the large difference at the beginning, leads to a similar equilibrium, which, again, is similar to the undisturbed simulation.