Figure 11. Normalization effects in the analysis of differently sized simulation boxes.

(A) Schematic illustration of the approximate normalization of water diffusion coefficients in smaller boxes against the largest box. Note that no actual rebuilding of simulation boxes was performed, but only the water diffusion coefficients in the smaller boxes were proportionally combined with the bulk water diffusion coefficients. (B) Water diffusion in simulation boxes of varying size. A small effect on the diffusion of bulk water (in the pure water simulations) is observed due to the periodic boundary effects, as described by Yeh and Hummer, 2004 (visualized as triangles). For the Yeh and Hummer correction, a shear water viscosity of $3.08\times {10}^{-4}$ kg m⁻¹ s⁻¹ was used. A much stronger effect on the water diffusion constant is observed when a large protein, in this case hemoglobin, is added to the simulation box (square symbols). The apparent strong box size effect manifests due to the fact that only little bulk water (capable of diffusing with the bulk-like diffusion constant) is present in the smaller boxes. To account for the difference in protein to water ratio the smaller boxes (9 nm and 12 nm) were renormalized to the level of 15 nm box by employing the value of bulk water diffusion weighted by the difference in the number of water molecules between the smaller boxes and 15 nm box (circle symbols). The dashed line markes the bulk water diffusion value for water simulation in 9 nm box. (B) Water radial distribution function (RDF) for differently sized hemoglobin boxes, normalized to a sphere of 8.5 nm diameter.