ifilot / dftcxx Goto Github PK

Running the program for certain molecules (e.g. CO) results in slight oscillations in the computed energies in the SCF procedure. As convergence is not achieved, the program keeps looping infinitely, so I suggest imposing a maximum number of iterations to prevent this behaviour.

Hi,

I'm trying to compile the project on windows and I noticed something:

In double SH::legendre (int n, double x) you have at line 58:

double v[n];

That would allow using indices for v from 0 to n-1. Despite that, v[n] is written and returned.

The simple solution would be: double v[n + 1]; (as you already have in SH::legendre_p).

Another issue is that the code as it is creates problems when compiling with VisualC++ (error because n is not a constant), The solution would be:

double *v = new double[n+1];

and just before return:

    const double val = v[n];

    delete[] v;

    return val;

and similarily in SH::legendre_p

PS Nice work. How long did it take you to implement it? I'm thinking of adding DFT to my Hartree-Fock program for molecules, I did implement some DFT code before, but with plane-waves basis: https://compphys.go.ro/dft-for-a-quantum-dot/

Adrian

Replace all instances of the EigenSolver with SelfAdjointEigenSolver. The latter ensures that the coefficient matrix is unitary, the diagonalization algorithm is faster and all eigenvalues and eigenvectors are automatically sorted.