ABINIT, DMFT input variables:

Value of double counting used for DMFT. Only value 1 is activated for the moment and is the FLL double counting.

Number of iterations for the DMFT inner loop.

Mixing parameter for the simple mixing of the self-energy.

Number of Matsubara frequencies (linear mesh)

Number of frequencies in the log mesh.

Flag to read/write Self-Energy. If put to one, self-energy is written and read at each DFT iteration.

Choice of solver for the Impurity model.

• 0=> No solver and U=0, J=0 (see upawu and jpawu).
• 1=> LDA+U self-energy is used (for testing purpose)
• 2=> Hubbard one solver. The Hubbard one solver is an approximation which gives a rough description of correlated Mott insulators. It should not be used for metals.
• 5=> Use the Continuous Time Quantum Monte Carlo (CTQMC) solver CT-Hyb of ABINIT in the density density representation, CTQMC calculations are much more time consuming that Hubbard I calculations. Nevertheless, the calculation is fully parallelised.
• 6=> Continuous Time Quantum Monte Carlo (CTQMC) solver CT-Hyb of TRIQS in the density density representation.
• 7=> Continuous Time Quantum Monte Carlo (CTQMC) solver CT-Hyb of TRIQS with the rotationally invariant formulation.

Can be set to 1 only if in cubic symmetry. It enables one to carry a DFT+DMFT calculations only on t2g orbitals.

The LDA occupation matrix for correlated electrons can be computed directly. It can be compared to the calculation of the same quantity using LDA Green's function, a sum over Matsubara frequencies and a projection over correlated orbitals. Because the Matsubara grid is finite, the two quantities differ. If this difference is larger than dmft_tolfreq, then the code stops and an error message is given.

Tolerance for the variation of Local Charge during iterations of the DMFT Loop.
The default value is good for fast calculations. However, to obtain good convergence of the DFT Loop, the DMFT Loop needs a better convergence criterion.

dmftbandf is the last band taken into account in the Projected Local Orbitals scheme of DFT+DMFT. With dmftbandi, they define the energy window used to define Wannier Functions. (see Amadon, B., Lechermann, F., Georges, A., Jollet, F., Wehling, T. O., and Lichtenstein, A. I. Phys. Rev. B 77(20), (2008).)

dmftbandi is the first band taken into account in the Projected Local Orbitals scheme of LDA+DMFT. With dmftbandf, they define the energy window used to define Wannier Functions. (see Amadon, B., Lechermann, F., Georges, A., Jollet, F., Wehling, T. O., and Lichtenstein, A. I. Phys. Rev. B 77(20), (2008).)

Only for developer purposes. (Introduced by B. Amadon, v6.1.0)

Check the fast calculations during the Monte Carlo simulation with very slow but robust methods. Should only be used for debugging.

• 0=> No check.
• 1=> Check the overlap calculations (Impurity operator).
• 2=> Check the update of M matrix calculation (Bath operator).
• 3=> Check both.

Flag to compute statistics about segments and anti-segments during the simulation. Slow down the simulation.

• 0=> Nothing done
• 1=> Calculations performed and written in "Correlation.dat" file

Default is no global moves. The value of this variable is the modulo used to try a global move. A value of 5000 means that a global move is tried every 5000 Monte Carlo sweep.

Compute the statistical noise for each time slice of each green function. This is a good approximation only if there is enough Monte Carlo sweeps per cpu.

• 0=> Nothing
• 1=> Do it and write the noise in the "Gtau.dat" file.

The modulo used to measure the interaction energy and the number of electrons. Example : 2 means the measure is perform every two sweeps.

Print a latex file per cpu displaying the full simulation. This option should only be use with very small number (<1000) of Monte Carlo sweeps since it requires a lot of I/O band width.

• 0=> Nothing
• 1=> Write the "Movie_id.dat" file where id is the MPI rank of each process

Measure the time evolution of the number of electrons for each orbital and perform a fourier transform. The result can be plotted using the "Markov_id.dat" file

• 0=> Nothing
• 1=> Do it and write the noise in the "Markov_id.dat" file where id is the rank of each MPI process.

Print a file containing the statistic distribution of the number of segments per orbital. The maximal order taken into account dmftctqmc_order : 50 means that we have the statistic distribution from 0 to 50 segments. The result is written in the "Perturbation.dat" file.

Specify the number of Legendre polynomials used for the calculation of Green's function in CTQMC code from the library TRIQS. Default is 30. The value of coefficients are given in file whose name ending is "Legendre_coefficient.dat". (see also Phys. Rev. B 84, 075145 (2010))

Number of time slices used to represent the time green function. This value should be carefully chosen according to Niquist frequency and the tsmear value.

Number of Monte Carlo sweeps. Should be at least 10^6.

Seed to initilize the random number generator.
Should not be relevant except for testing purpose.
NOTE : If the CT-QMC (dmft_solv=5) is used on many CPUs, each CPU initializes its random number generator with dmftqmc_seed+rank where rank is the rank of the cpu in the MPI communicator.

Number of Monte Carlo sweeps for the thermalization