Figure 7.9 Li+-water interaction profile for the pure QM reference potential and the hybrid PAW/AMBER potential.
At close range, the steric component of the van der Waals potential begins to play a more dominant role. The position and depth of the potential minimum is governed by the delicate balance between the repulsive steric and the attractive electrostatic components. The PAW/AMBER potential compares surprisingly well to the reference potential, considering that the parameters used were not optimized for this purpose. Figure 7.9 shows that the position of the potential minimum at 1.9 Å is almost exactly reproduced by the hybrid potential. Unfortunately, the well depth is over estimated by 6 kcal/mol or approximately 20%. The good agreement between the PAW/AMBER potential and the reference potential suggests that by adjusting the van der Waals potential, we should be able to improve the hybrid potential.
Table 7.3 Comparison of the fitted charges from the PAW QM/MM calculation and a pure QM PAW calculation.
|
PAW QM/MM charges
|
|
pure QM PAW chargesa
|
Li-O distance (Å)
|
oxygen
|
hydrogen
|
|
oxygen
|
hydrogen
|
lithium
|
1.70
|
-1.27
|
0.63
|
|
-1.00
|
-0.51
|
0.98
|
1.90
|
-1.15
|
0.57
|
|
-0.97
|
0.50
|
0.97
|
2.50
|
-0.94
|
0.47
|
|
-0.90
|
0.46
|
0.98
|
3.00
|
-0.84
|
0.42
|
|
-0.84
|
0.43
|
0.99
|
a calculated with the same PAW settings as the PAW QM/MM simulation except that a 24.0 Å cubic simulation cell is used.
As the lithium ion is brought closer to the oxygen atom, there should be a polarization of the water molecule, such that charge density is drawn towards the oxygen atom. Displayed on the right-hand side of Table 7.2 are the fitted charges of the water molecule from the PAW QM/MM calculation. (The unperturbed charges from the PAW calculation of free water are -0.62e and 0.31e for the oxygen and hydrogen, respectively.) As the lithium ion is drawn closer to the oxygen, the charge on the oxygen atom steadily increases while the hydrogen charge diminishes, showing that there is indeed a polarization of the wave function in the PAW QM/MM model. The energetic cost of distorting the wave function from its unperturbed density can be easily calculated as the difference between (+) and the total interaction energy. The distortion energy, ∆Edist, as we have defined it is also presented in Table 7.2. In the long range, the distortion energy is zero and should increase for closer interactions. For example, at 12 Å the distortion is negligible, at 3 Å it is 2.1 kcal/mol or 15% of the interaction energy, while at the potential minimum the distortion energy amounts to 12.6 kcal/mol or 36% of the total interaction energy.
The polarization of the wave function in the QM/MM model can be examined by comparing the fitted charges from the PAW QM/MM calculation to the fitted charges from a pure QM PAW calculation. For this purpose, a series of pure QM PAW calculations on the lithium ion-water system have been performed in a 24.0 Å cubic simulation cell. The fitted charges are compared in Table 7.3. Even at a distance of 3.0 Å, the polarization is not trivial with the fitted charge on oxygen increases by about 0.2e compared to the free state. At this point, the QM/MM charges compare very well to those of the pure QM calculation differing by less than 0.01e. At the potential minimum where the Li-oxygen distance is 1.90 Å, the QM/MM fitted charge on oxygen has almost doubled from that in the free state to a value of -1.15 e. Compared to the pure QM PAW calculation which gives a fitted charge of -0.97e on oxygen the polarization is slightly too high. The over-polarization further increases as the distance is decreased. This discrepancy can be attributed to two factors. First, the pure QM calculation allows for a small charge transfer from the water molecule to the lithium ion which cannot occur in the QM/MM calculation. Secondly, the charge distribution of the lithium ion is more diffuse in the pure QM calculation compared to the point charge distribution in the QM/MM interaction. For these reasons, the polarization of the water molecule is slightly exaggerated in the short-range region in the combined QM/MM model.
Share with your friends: |