next up previous
Next: More Efficient Methods: Conjugate Up: The Car-Parrinello Method Previous: Car-Parrinello Lagrangian

Orthonormality

The orthonormal constraints in the equations of motion are vital to give the correct electronic states in the molecular dynamics method. If the equations are solved in the absence of the orthonormal constraints the time evolution of the one electron wavefunctions are found to be either oscillatory or convergent to a single degenerate ground state[31], depending on initial conditions. The initial wavefunctions will only converge to different eigenstates if orthogonality is imposed.

Equation (3.46) ensures that the tex2html_wrap_inline5478 remain orthogonal at all instants in time. However, to ensure this, the Lagrange multipliers must vary continuously with time and so the implementation of these equations require the evaluation of tex2html_wrap_inline5774 at infinitely small time separations. To make the calculation possible, in practice, tex2html_wrap_inline5774 are held constant throughout each timestep of integration. This leads to non-orthogonality of the wavefunctions at the end of each timestep which then requires a separate orthogonalisation step. This is done by Gram-Schmit orthogonalisation

  eqnarray776

and normalisation

  eqnarray782

where the orthonormalised set tex2html_wrap_inline5778 is generated from the linearly independent set tex2html_wrap_inline5780 obtained from integration of the equations of motion.

As a result of this, the constraints of orthogonality are imposed each time the electronic equations of motion are integrated. tex2html_wrap_inline5774 can the be approximated by the expectation values of the eigenstates, tex2html_wrap_inline5784 giving the equation of motion of the form

  eqnarray790

The wavefunctions are then an exact eigenstate when their accelerations tex2html_wrap_inline5786 are zero which are found self-consistently.



Stewart Clark
Thu Oct 31 19:32:00 GMT 1996