[QE-users] Help: Could you please teach me how to improve the accuracy of LDA+U calculation when reading maximally localized Wannier function files?

[yfang330]

Dear experts,

This is Yuan Fang. My research focuses on first principles simulation of infrared and Raman spectra of adsorbates at liquid/solid electrochemical interfaces. Recently, I plan to accurately calculate the adsorption energies of the Pt(111)(2×2)-COL structure (shown in Fig.1) by LDA+U calculation with reading maximally localized Wannier function of  specified bands (pmw.x tool and "U_projection_type='file'" tag). Since the method can’t automatically obtain the atomic forces in Quantum Espresso, I need manually calculate the atomic forces by the finite difference method to decide the next atomic movement and to optimize the structure finally.


Fig.1 the strcture of Pt(111)(2×2)-COL adsorption.


However, I find that the energy accuracy of LDA+U calculations with reading MLWF of specified bands is so low that the differential value (atomic force in red line of Fig.2) deviates the scf calculation without LDA+U (the black line) and fails to calculate atomic force. Although I tried increasing the kpoint density (from 4*4*1 to 6*6*1), the scf convergence threshold (from 1E-7 Ry to 1E-8 even 1E-10 Ry), cutoff energy (from 40 Ry to 120 Ry), the spreading of Wannier function (from 0.02 to 0.002) or replacing Methfessel-Paxton first-order spreading with Marzari-Vanderbilt cold smearing, all of them can't improve the energy accuracy obviously in Fig.2. Could you please teach me how to improve the the energy accuracy of LDA+U calculations with reading MLWF of specified bands?  Many thanks!
Fig.2 the force of carbon atom belonging to adsorbed CO molecule along with z axis as the fuction of differential displacement.


Best Regards,


Yuan Fang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20210802/13e6b970/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Graph137.png
Type: image/png
Size: 126086 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20210802/13e6b970/attachment.png>