Hello, I am a doctoral student at CNRS-strasbourg (IPCMS)

I was interested in abinit because I would like to analyze certain quantities such as 1-local spectral functions for 5ds orbitals of Mn (MnPc-molecule) 2-Matsubara self-energies at different T (K) of the five orbitals 3-the imaginary part of the hybridization function

I first did the DFT + U calculation with VASP or the value of u = 4eV and j = 0.9ev and I read in the tutorial that for the LDA + DMFT calculation these values will be U = u-4 / 3d and J = 7 / 6d or for LDA + DMFT on abinit I have U = 2.8eV and J = 1.05eV please could you confirm that I understood correctly?

and second in the tutorial there in the analyze of the spectral function there is a sentence which is not understandable for me could enlighten me? because I don't get the same curve of the spectral function (and those even by multiplying by 27.2 in the optimal_spectral_function*.DAT file: the sentence is "Change the unit from Hartree to eV, and then, you have the spectral function : "here it is necessary to convert which value in Hartree? U? J? or the eigenvalues (which I also multiply by 27.2 and not always obtained the same curve)

I prepared my input file to successively do an LDA and LDA + DMFT calculation the LDA calculation (Nstep1 = 90) ends in 1 hour but the LDA + DMFT calculation (just an iteration takes me about 10 hours) if it could you please inspect my input file to be able to help me modify or adjust it for optimal calculation and analysis? at the end of the calculation I got this warning:

--- !WARNING

src_file: m_global.F90

src_line: 229

message: |

WARNING in QMC : Ctqmc_printQMC : at least one negative sign occured. There might be a bug in the CT-QMC

...

--- !WARNING

src_file: m_global.F90

src_line: 229

message: |

WARNING in QMC : Ctqmc_printQMC : bad statistic according to Noise. Increase sweeps

for the analysis I used (for the tutorial) the code OMEGAMAXENT but did not see how to have the different properties mentioned above (your guide would be welcome) thank you in advance.

## LDA+DMFT(Spectral function for 5d +matsubara self-energies)

**Moderators:** jbeuken, Jordan, pouillon

**Forum rules**

Please have a look at ~abinit/doc/config/build-config.ac in the source package for detailed and up-to-date information about the configuration of Abinit 8 builds.

For a video explanation on how to build Abinit 7.x for Linux, please go to: http://www.youtube.com/watch?v=DppLQ-KQA68.

**IMPORTANT: when an answer solves your problem, please check the little green V-like button on its upper-right corner to accept it.**

### Re: LDA+DMFT(Spectral function for 5d +matsubara self-energies)

Dear user,

Concerning U and J, you can use the same value in DFT+U and in DFT+DMFT. Conventions are the same. So U=4 eV and J=0.9 eV.

Concerning spectral functions: the file contains two columns, the first one is the energy in Ha and the second is the spectral functions in Ha^{-1}. So you need to multiply the first one by 27.211 to have it in eV.

Concerning your input. I suggest to use dmftctqmc_gmove=0 unless you need it.

You are selecting all bands in the calculation of Wannier function, is it necessary ? You could select only the bands on which a non negligeable weight of d orbitals is present. To find them, use pawfatbnd in your LDA calculation.

DFT+DMFT is more time consuming than DFT so an increase of computer time is expected.

Concerning the precision of CTQMC, you can have a look to the Green's function to check if it is noisy or not.

Best regards

Bernard

Concerning U and J, you can use the same value in DFT+U and in DFT+DMFT. Conventions are the same. So U=4 eV and J=0.9 eV.

Concerning spectral functions: the file contains two columns, the first one is the energy in Ha and the second is the spectral functions in Ha^{-1}. So you need to multiply the first one by 27.211 to have it in eV.

Concerning your input. I suggest to use dmftctqmc_gmove=0 unless you need it.

You are selecting all bands in the calculation of Wannier function, is it necessary ? You could select only the bands on which a non negligeable weight of d orbitals is present. To find them, use pawfatbnd in your LDA calculation.

DFT+DMFT is more time consuming than DFT so an increase of computer time is expected.

Concerning the precision of CTQMC, you can have a look to the Green's function to check if it is noisy or not.

Best regards

Bernard

Bernard Amadon

CEA

France

CEA

France

### Re: LDA+DMFT(Spectral function for 5d +matsubara self-energies)

Dear Bernard,

Thank you for your reply. how to have this spectral function (DMFT-DOS) project on each orbital (dz2; dxy, dxz, dyz, dx2-y2). I only have one k-points,

Franck O.

Thank you for your reply. how to have this spectral function (DMFT-DOS) project on each orbital (dz2; dxy, dxz, dyz, dx2-y2). I only have one k-points,

Franck O.