[QE-users] Transferred hyperfine coupling constants in TM-oxides
kris.harris at mcmaster.ca
Wed Apr 18 20:22:45 CEST 2018
Thanks for the reply and the PP.
I had hoped to get close here given Mali et al. reported such excellent
agreement for this exact system using your code in 2010. Though thanks for
the hint tip that I may not find continued agreement with other systems at
this level of theory.
I tried the alternate PP you sent and get somewhat different results,
though the two calculations are closer to each other than either is to the
experiment. Perhaps this suggests I am doing something wrong?
The Capsoni paper I mentioned has:
Li1: A = 4.32 MHz
Li2/Li3: A = 2.13 MHz/ 2.07 MHz (crystallographically very similar)
Reading chemical shifts from Fig. 4 of Mali et al's Quantum Espresso
calculations and using their Eq. 1 (approximate as magnetic susceptibility
not known) to convert A values yields:
Li1: delta = 1425 ppm --> A = 4.14 MHz
Li2/Li3: delta = 800/760 ppm --> A = 2.33/2.21 MHz (for the 2 sites)
The results I get with this new PP are (Li1 here is the site well-known to
have the larger A value):
----- Fermi contact in MHz -----
bare GIPAW core-relax total
Li 1 -9.209256 8.650557 0.000000 -0.558699
Li 2 -12.302726 4.367716 0.000000 -7.935010
Li 3 -12.374583 4.261266 0.000000 -8.113317
They aren't too sensitive to ecutwfc. Interestingly, the "GIPAW" term is
nearly exactly double the correct answer. I sort-of get the idea of your
core-relaxation theory, but I haven't managed to read anything to
understand the breakdown into the "bare" vs "GIPAW" components.
thanks for any further insight,
On Wed, Apr 18, 2018 at 7:19 AM, Davide Ceresoli <davide.ceresoli at cnr.it>
> Dear Kris,
> the core-relaxation feature of GIPAW is extremely experimental.
> It's based on a perturbation approach which might not be adequate
> in all cases.
> Could you try with the attached pseudopotential? this one has 3 valence
> electrons for Li and a cutoff between 120 and 150 Ry should be sufficient.
> This way, the polarization of the 1s electrons should be described at a
> better level.
> However, we are also working on similar Li-battery materials, comparing to
> experiments and Crystal14+B3LYP calculations, and in some cases we are
> not able to reproduce the Li hyperfine cc.
> Good luck for your calculations.
> *Density Functional Theory (DFT) Calculation of Materials Properties*
> A *special issue* of *Materials* (ISSN 1996-1944). Guest Editors: D.
> Ceresoli and A. Ponti
> If you are willing to contribute to this special issue, please click here
> The deadline for manuscript submissions is *15 July 2018*
> On 04/17/2018 06:45 PM, Kris Harris wrote:
> I'm starting out with QE and am attempting to learn by repeating some
> literature results for hyperfine coupling constants for Li in Li2MnO3.
> Capsoni et. al in 1997 Phys. Rev. B. list the experimental value for A as
> 4.31 MHz for one site (there are 2 other sites at essentially half that A)
> Mali. et al. were able to reproduce that value very well in Chem. Comm. in
> 2010 (DOI:10.1039/c003065a) using QE with the GIPAW module. I'm attempting
> to reproduce that result, following their method as close as possible.
> I used dist.x to attempt to verify the structure, and did some convergence
> tests with respect to ecutwfc and k-points to make sure the calculations
> was reasonable. I also verified that the binary I have reproduced the H2O+
> test case included in the distribution.
> The result I get is quite a long ways off (0.06 MHz and -7 MHz instead of
> ~4.3 and ~2.15 MHz).
> Any help on what I've managed to do wrong here would be greatly
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the users