Dear Chen, The electric field is treated in Rydberg atomic units in pw.x in this way:
1)The polarization P(\Psi) is calculated as sqrt(2.)*<\Psi|r|\Psi>/Omega where \Psi is the slater determinant of the KS wavefunctions, Omega is the valume of the simulation cell, and the expectation value of the position operator r is calculated through the Berry's phase formulation. This assures that for the same wavefunctions the "numerical value" of the polarization is sqrt(2.) times larger then that calculated for the case of Hartree atomic units. The KS Hamiltonian in the presence of the electric field E contains a term proportional to Omega*E*\partial P(\Psi) /\partial \Psi = E* sqrt(2) *\partial *<\Psi|r|\Psi> /\partial \Psi This is the factor sqrt(2.) appearing in the routine h_epsi_her_set.f90 If you want to calculate the ratio P/E you must use both values in Rydberg atomic units: the electric dipole as reported by the code (in Rydberg atomic units) and the input numerical value of the electric field which is also given in Rydberg atomic units (see Example 31) Note that to obtain the corresponding numerical value of the electric field in Hartree atomic units you must divide by a factor of sqrt(2.) Best regards, Paolo Umari, DEMOCRITOS > Dear PWSCF users, > I have a question about the polarization unit which has confused me for > a long time. > The atomic unit of electric field is: e/a0^2 where 'e' is the electron > charge and a0 is the Bohr radius. The unit of polarization (i.e. dipole per > unit volume) should have the same unit as electric field. However from the > source code c_phase_field.f90(version4.0.2), the output of dipole per cell > has a factor sqrt(2) which is the electron in the Ry units. So 'e' becomes > sqrt(2). > Then if I want to calculate P/E which is dimensionless, shall I convert > the output of P into P/sqrt(2). Note when we input the electric field E, E > is in the atomic unit, i.e. 'e' is not converted into sqrt(2). > I am just confused that since you use atomic unit, why you explicitly > convert 'e' into sqrt(2) rather than implicitly retain it in the unit? > > Hanghui Chen > Department of Physics, > Yale University ---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program.
