Seite - 97 - in Charge Transport in DNA - Insights from Simulations

7.3TestingwithChargeTransportCalculations Not onlywill this give a better understanding of the phase shift needed to obtain the correct correlations, but it is also a crucial parameter for theCTcharacteristics of the model. Therefore, IP differences between neighboring nucleobases were calculated fromMDsimulationsof20nsandanalyzed in termsof theirprobability distributions. Thephaseshiftof the IPwasparametrizedso that thewidthof these distributionsmatches thedata fromMDsimulations. Thefinalphaseshiftwasset to1.11perneighborforacosinefunctionwithaperiod of 2π. Figure7.4 shows thedistributions for thefirst twoneighboringnucleobases when this phase shift is applied. Note that the modeled standard deviation of the IP difference becomes inaccurate starting from the second-neighbor correla- tion. Thiswill have an effect in approaches relyingon the coherent transport, like the Landauer–Büttikermodel. However, the CTmodel shouldworkwith charge transfermethods,whichdescribenearest neighborhoppingpredominantly. 7.3 TestingwithChargeTransportCalculations With the EC and IP constructed, the time-dependent coarse-grainedHamiltonian can be set up, and the CT model can be tested with charge transport methods already. The easiest approach is to calculate the transmission function in every timestepandaverageover the simulation. As already noted, the CT model does not describe long-range correlations well. Thus, the calculation of a transmission functionwas done for a sequence of only three adenine nucleobases. Figure 7.5 shows the averaged transmission function in comparison to one obtained along an atomistic MD simulation. The overall formof the transmission function is reproduced accurately for thisDNA species, apparently. For longer sequences, the transmission functions assumes a triangular shape as a consequenceof awrongdistributionof IPdifferences starting from the A1-A4 neighbors. Clearly, this is anartifact of the simple correlationmodel. 97