<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<style type="text/css" style="display:none;"><!-- P {margin-top:0;margin-bottom:0;} --></style>
</head>
<body dir="ltr">
<div id="divtagdefaultwrapper" style="font-size:12pt;color:#000000;font-family:Calibri,Helvetica,sans-serif;" dir="ltr">
<div id="divtagdefaultwrapper" dir="ltr" style="font-size: 12pt; color: rgb(0, 0, 0); font-family: Calibri, Helvetica, sans-serif, EmojiFont, "Apple Color Emoji", "Segoe UI Emoji", NotoColorEmoji, "Segoe UI Symbol", "Android Emoji", EmojiSymbols;">
<p style="margin-top:0; margin-bottom:0"><span id="ms-rterangepaste-start"></span></p>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
Dear Dr. Delugas and Dr. Paulatto,</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
<br>
</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
Thanks for the replies!</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
<br>
</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
We did test the option "nosym = .true.". Theoretically, the calculation will only be performed on an imbalanced set of k-points in BZ, and charge symmetrization will also be completely shut off. The total energy might diverge, if the charge density is optimized
to another minimum.</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
<br>
</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
And the real test cases also support the guess, especially in high symmetry case. In primitive FCC Cu, using "nosym = .true." and a set of k-points in IBZ of the symmetry of Fm3m with correct weights, the converged total energy has a difference of ~46 mRy,
compared to normal MP converged value. (By "converged", I mean k-point density convergence.) In very low symmetry structure, though, the difference is not such obvious, below ~1 mRy.</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
<br>
</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
Best Regards,</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
Yunzhe Wang</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
Mueller Research Group</div>
<div style="font-family:Calibri,Helvetica,sans-serif; font-size:16px; margin-top:0px; margin-bottom:0px">
Johns Hopkins University</div>
<span id="ms-rterangepaste-end"></span><br>
<p></p>
</div>
<hr tabindex="-1" style="display:inline-block; width:98%">
<div id="divRplyFwdMsg" dir="ltr"><font face="Calibri, sans-serif" color="#000000" style="font-size:11pt"><b>From:</b> Pietro Delugas <pdelugas@sissa.it><br>
<b>Sent:</b> Friday, November 16, 2018 3:41:30 AM<br>
<b>To:</b> Yunzhe Wang<br>
<b>Subject:</b> Re: [QE-developers] Patch to modify the pw.x behavior at user-provided k-point list</font>
<div> </div>
</div>
<div style="background-color:#FFFFFF">
<p><font size="-1">Dear Yunzhe <br>
</font></p>
<p><font size="-1"><br>
</font></p>
<p><font size="-1">if your kpoint are provided as a list, the k-point set has the right symmetry and the proportions between the weight is the right one, maybe you can try using nosym = .true. and use_all_frac = .true.
<br>
</font></p>
<p><font size="-1"><br>
</font></p>
<p><font size="-1"> </font><br>
</p>
<div class="x_moz-cite-prefix">On 15/11/18 19:30, Yunzhe Wang wrote:<br>
</div>
<blockquote type="cite">
<div id="x_divtagdefaultwrapper" dir="ltr" style="font-size: 12pt; color: rgb(0, 0, 0); font-family: Calibri, Helvetica, sans-serif, EmojiFont, "Apple Color Emoji", "Segoe UI Emoji", NotoColorEmoji, "Segoe UI Symbol", "Android Emoji", EmojiSymbols;">
<p style="margin-top:0; margin-bottom:0">Dear Professor Giannozzi,</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">I agree that additional flags will unnecessarily increase the complexity to maintain. My original thought is to not disturb the original behavior of pw.x.</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="">The easier way to stop the expansion by lattice is to change the line 553 of setup.f90 (github) from</p>
<p style=""> CALL irreducible_BZ(nrot, s, nsym ...)</p>
<p style="">to </p>
<p style=""> CALL irreducible_BZ(nsym + nsym_na, s, nsym ....)</p>
<p style=""><br>
</p>
<p style="">The original code is to expand by the lattice point group, and the latter is to expand by the crystal point group, with the discarded glide plane / screw axis included. Our server doesn't have the prior knowledge of the FFT grid used in pw.x and
therefore returns the k-points in IBZ of the crystal. If pw.x decides to throw away some symmetry, the IBZ should be enlarged to the lower symmetry one, hence with the "+nsym_na" part.</p>
<p style=""><br>
</p>
<p style="margin-top:0; margin-bottom:0">An example would be the one that an actual user reported to us during the beta testing:</p>
<p style="margin-top:0; margin-bottom:0">The structure is SiC, which she downloaded from ICSD data base. It's hexagonal with 12 atoms in the unit cell, but due to the limited precision that ICSD has (to 10^-4 Angstrom), both pw.x and our server detected it
as an orthorhombic structure. Our server returned a mesh with 14 asymmetrical k-points, and QE expanded it, with the point group of a hexagonal lattice, to 84 distinct k-points. The total energies are -394.9476363 Ry and -394.9476383 Ry (a difference of 0.002
mRy) , with the expansion by pw.x shut off and switch on respectively. And the calculation time on 12 processors are 1.55 minutes (14 k-points) and 8.43 minutes (k-points), respectively. </p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">I know in actual serious calculation for publication, the atomic positions would be much precise, and symmetry will be inspected. But a similar case would still be present if a user is working on a slab or a supercell
which the crystal symmetry and the unit cell symmetry are very different. An expansion by the lattice point group, will increase the calculation time by a lot. Also, for low symmetry structure, the k-points will increase, averagely, by 1/5 - 1/3, in serveral
test cases.</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">I attached below the input file for the SiC for your reference (the pseudo-potential files are not attached):</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>&CONTROL</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> calculation = 'scf' , </div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> pseudo_dir = './' ,</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> disk_io = 'none' ,</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> verbosity = 'high',</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>/</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div><br>
</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>&SYSTEM</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> ibrav = 0 ,</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> ntyp = 2 , </div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> nat = 12 , </div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> occupations = 'smearing' ,</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> smearing = 'marzari-vanderbilt' ,</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> degauss = 0.02 ,</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> ecutwfc = 50 ,</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>/</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div><br>
</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>&ELECTRONS</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> startingwfc = 'random' ,</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>/</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div><br>
</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>&IONS</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>/</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div><br>
</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>&CELL</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>/</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div><br>
</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>CELL_PARAMETERS bohr</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> 5.82224621542067 0.00000000000000 0.00000000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> -2.91112310771033 5.04221312964210 0.00000000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> 0.00000000000000 0.00000000000000 28.58172981466878</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div><br>
</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>ATOMIC_SPECIES</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> Si 28.0855 Si.GGA-PBE-paw.UPF</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> C 12.0107 C.pbe-rrkjus.UPF</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div><br>
</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>ATOMIC_POSITIONS crystal</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> Si 1.00000000000000 1.00000000000000 0.50000000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> Si 0.00000000000000 0.00000000000000 0.00000000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> Si 0.66670000000000 0.33340000000000 0.66640000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> Si 0.33330000000000 0.66660000000000 0.16640000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> Si 0.33330000000000 0.66660000000000 0.83290000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> Si 0.66670000000000 0.33340000000000 0.33290000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> C 0.66670000000000 0.33340000000000 0.54120000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> C 0.33330000000000 0.66660000000000 0.04120000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> C 0.33330000000000 0.66660000000000 0.70800000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> C 0.66670000000000 0.33340000000000 0.20800000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> C 0.00000000000000 0.00000000000000 0.87460000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div> C 0.00000000000000 0.00000000000000 0.37460000000000</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div><br>
</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div># Server version 2018.08.08. ||BETA_MODE|| K-points for C1Si1 with MINDISTANCE=20.0 Angstroms and INCLUDEGAMMA=AUTO FORCEDIAGONAL=false has 112 total points and 14 distinct points. Effective minDistance 21.34579415247883 Angstroms.</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>K_POINTS crystal</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>14</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.07142857142857 0.90178571428571 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.21428571428571 0.70535714285714 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.35714285714286 0.50892857142857 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.50000000000000 0.31250000000000 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.64285714285714 0.11607142857143 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.92857142857143 0.72321428571429 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.07142857142857 0.52678571428571 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.21428571428571 0.33035714285714 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.50000000000000 0.93750000000000 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.64285714285714 0.74107142857143 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.07142857142857 0.15178571428571 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.21428571428571 0.95535714285714 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.64285714285714 0.36607142857143 0.25000000000000 8.0</div>
</blockquote>
<blockquote style="margin:0 0 0 40px; border:none; padding:0px">
<div>0.21428571428571 0.58035714285714 0.25000000000000 8.0</div>
</blockquote>
</blockquote>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
</div>
<hr tabindex="-1" style="display:inline-block; width:98%">
<div id="x_divRplyFwdMsg" dir="ltr"><font face="Calibri, sans-serif" color="#000000" style="font-size:11pt"><b>From:</b> Paolo Giannozzi
<a class="x_moz-txt-link-rfc2396E OWAAutoLink" href="mailto:p.giannozzi@gmail.com" id="LPlnk767164" previewremoved="true">
<p.giannozzi@gmail.com></a><br>
<b>Sent:</b> Thursday, November 15, 2018 9:04:03 AM<br>
<b>To:</b> General discussion list for Quantum ESPRESSO developers<br>
<b>Cc:</b> Yunzhe Wang<br>
<b>Subject:</b> Re: [QE-developers] Patch to modify the pw.x behavior at user-provided k-point list</font>
<div> </div>
</div>
<div>
<div dir="ltr">
<div>Dear Yunzhe</div>
<div><br>
</div>
<div>as a rule I am against adding new input variables - there are already hundreds of them - unless there are no alternatives. From what I understand, your procedure produces a list of k-points that are already in the Brillouin Zone reduced using the symmetry
of the crystal. If so, the best solution is to ensure that the code does not spoil it when computing the actual k-point grid from the input k-point grid. It would be useful to have a few examples showing what happens. </div>
<div><br>
</div>
<div>Paolo<br>
</div>
<br>
<div>
<div class="x_x_gmail_quote">
<div dir="ltr">On Wed, Nov 14, 2018 at 8:26 AM Yunzhe Wang <<a href="mailto:ywang393@jhu.edu" target="_blank" id="LPlnk187655" class="OWAAutoLink" previewremoved="true">ywang393@jhu.edu</a>> wrote:<br>
</div>
<blockquote class="x_x_gmail_quote" style="margin:0 0 0
.8ex; border-left:1px #ccc solid; padding-left:1ex">
<div dir="ltr">
<div id="x_x_m_-1999253779495694799m_660957958707217870divtagdefaultwrapper" dir="ltr" style="">
<p style="margin-top:0; margin-bottom:0">Dear Professors and Developers,</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">Sorry about the bothering long mail!</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">Our group has built a server to generate the least-number k-point mesh for a user-given length of the corresponding supercell size. (By "corresponding", I mean a calculation with a dense k-point mesh for a primitive
cell, is equivalent to, a calculation with a large supercell with 1 k-point). Currently, the server is able to read the user input file and append the returned K_POINTS card at the end.</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">However, during the beta testing, we found that QE had the subroutine irreducible_BZ() to force the mesh being expand by the lattice point symmetries first, and then being reduced by the true crystal symmetry. Our server
already returns a mesh complying with the crystal symmetry, but will be expanded quite a lot (even more in the low symmetry case) by irreducible_BZ(). </p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">So my question is, is there a specific reason to let the mesh being expanded by the lattice symmetry first? Our beta test results show the two cases, expanded and not-expanded (initial K_POINTS card in both cases provided
by our server), has nearly the same total energy. And the converged energy by the not-expanded one is also very close to that of the MP generated k-point mesh.</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">We have wrote a small patch to include an additional flag "kpuser" to switch on and off the expansion, and when it's off, we leave the behavior of QE untouched. Do you think it's ok that we release the patch and let users
of our server to patch their local QE distribution? Or we open an issue on QE and let the experts of QE to scrutinize this feature, if you think it's valuable to be incorporated in?</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0">Best Regards,</p>
<p style="margin-top:0; margin-bottom:0">Yunzhe(Phil) Wang, <span style="font-size:12pt">Graduate Student</span></p>
<p style="margin-top:0; margin-bottom:0">Mueller Research Group,</p>
<p style="margin-top:0; margin-bottom:0">Johns Hopkins University</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
<p style="margin-top:0; margin-bottom:0"><br>
</p>
</div>
</div>
_______________________________________________<br>
developers mailing list<br>
<a href="mailto:developers@lists.quantum-espresso.org" target="_blank" id="LPlnk356199" class="OWAAutoLink" previewremoved="true">developers@lists.quantum-espresso.org</a><br>
<a href="https://lists.quantum-espresso.org/mailman/listinfo/developers" rel="noreferrer" target="_blank" id="LPlnk110363" class="OWAAutoLink" previewremoved="true">https://lists.quantum-espresso.org/mailman/listinfo/developers</a><br>
</blockquote>
</div>
<br clear="all">
<br>
-- <br>
<div dir="ltr" class="x_x_m_-1999253779495694799gmail_signature">
<div dir="ltr">
<div>
<div dir="ltr">
<div>Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,<br>
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy<br>
Phone +39-0432-558216, fax +39-0432-558222<br>
<br>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<br>
<fieldset class="x_mimeAttachmentHeader"></fieldset>
<pre class="x_moz-quote-pre">_______________________________________________
developers mailing list
<a class="x_moz-txt-link-abbreviated OWAAutoLink" href="mailto:developers@lists.quantum-espresso.org" id="LPlnk370565" previewremoved="true">developers@lists.quantum-espresso.org</a>
<a class="x_moz-txt-link-freetext OWAAutoLink" href="https://lists.quantum-espresso.org/mailman/listinfo/developers" id="LPlnk503960" previewremoved="true">https://lists.quantum-espresso.org/mailman/listinfo/developers</a>
</pre>
</blockquote>
</div>
</div>
</body>
</html>