<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<meta name="Generator" content="Microsoft Exchange Server">
<!-- converted from text --><style><!-- .EmailQuote { margin-left: 1pt; padding-left: 4pt; border-left: #800000 2px solid; } --></style>
</head>
<body>
<meta content="text/html; charset=UTF-8">
<style type="text/css" style="">
<!--
p
        {margin-top:0;
        margin-bottom:0}
-->
</style>
<div dir="ltr">
<div id="x_divtagdefaultwrapper" dir="ltr" style="font-size:12pt; color:#000000; font-family:Calibri,Helvetica,sans-serif">
<p><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">Dear Malte,</span></p>
<p><br>
</p>
<p><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">This is not an easy question. Let me express my opinion.
</span><br>
</p>
<p><br>
</p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">>
</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">I read many times in papers that I should not compare total energies of
</span><br>
<span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">> </span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">DFT+U calculations with different U values.
</span><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">Can you give some references please?<br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">Well the problem is that in literature often people still use
</span><b><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">empirical</span></b><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif"> values of U.</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">
 In this case, I agree that the comparison of energies will depend strongly on U which one chooses based on some arguments.</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif"> But if one computes U from first principles, then U
</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">is a response property of each system (and the response is different in each system, hence different U</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">)</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">
 then it makes sense to me to compare energies (but I am not aware of any theoretical proof that it is allowed to do so</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">)</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">.
</span><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">For example, in LiCoO2 the computed value of U for Co-3d is
</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">6.91 eV, while U for Co</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">-3d in CoO is 4.55 eV (both using 'atomic' orbitals and some pseudos on top of PBEsol
 - see PRB 101, 064305 (2020)</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">). So in both cases we consider Co-3d states, but the value of U is very different. This is so because Co-3d states require different corrections when they
 are in different chemical environments (these Co-3d states hybridize differently with ligands in different compounds). If you use an average value of
</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">5.73 eV for both systems, then this value is not a response property of any of these systems (I do not know what it is in this case</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">).
</span><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">One can also see DFT+U as DFT with a Hubbard
<b>functional</b>: the U correction is different for different systems (why should it be the same?</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">).</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif"> It's
 like hybrids: people tune the value of alpha in PBE0 for each system (alpha is related to the inverse of the dielectric constant</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">), and alpha is different for different systems (because
 the dielectric constant is different). So here is the same, U is different for different systems: different systems need different amount of the corrections to restore the piece-wise linearity
</span><font size="2"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">of the energy</span></font><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">
</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">(PRB </span>
<span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">71, 035105 (2005)</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">)</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif"> and alleviate self-interactions.</span><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">The comparison of
</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">energies with different
<b>ab initio</b> U values was done </span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">in these papers: PRB 99, 094102 (2019); PRR 2, 023313 (2020); PRM 3, 033801 (2019).</span></span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">>
</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">What should I do in case of complex hull calculations of a phase diagram?
</span><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">I would
<b>compute</b> U for each phase and compare energies.<br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">> With QE it is now possible to calculate the U values for each composition by the hp.x code. Can I
</span><br>
<span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">compare the resulting total energies with each other safely?</span></span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">Well there is no general consensus on this issue, in part because still many people use empirical U. I think that if U is computed ab initio
 (i.e. it is really a response property of each material</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">) then it makes sense to compare energies with different U values (U computed for each system</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">).
 But people are still investigating this issue.</span></span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">I will be happy to hear other opinions, and to hear corrections/remarks
</span><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">if something what I said is misleading.
</span><br>
</span></font></p>
<p><font size="2"><span style="font-size:10pt"><br>
</span></font></p>
<p><font size="2"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">Greetings,</span></font></p>
<p><font size="2"><span style="font-size:12pt; font-family:Arial,Helvetica,sans-serif">Iurii</span></font><br>
</p>
<p><br>
</p>
<div id="x_Signature">
<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">
<div name="x_divtagdefaultwrapper" style="font-family:Calibri,Arial,Helvetica,sans-serif; font-size:; margin:0">
<div name="x_divtagdefaultwrapper" style="font-family:Calibri,Arial,Helvetica,sans-serif; font-size:; margin:0">
<font size="3" face="'Times New Roman', Times, serif" color="808080">--<br>
Dr. Iurii TIMROV<br>
Postdoctoral Researcher<br>
<font color="808080"><font face="'Times New Roman', Times, serif"></font></font></font></div>
<font color="808080"></font>
<div name="x_divtagdefaultwrapper" style="font-family:Calibri,Arial,Helvetica,sans-serif; font-size:; margin:0">
<font size="3" face="'Times New Roman', Times, serif" color="808080">STI - IMX <font color="808080">
<font face="'Times New Roman', Times, serif">- THEOS</font></font></font><font size="3" face="'Times New Roman', Times, serif" color="808080"> and NCCR - MARVEL<br>
</font></div>
<div name="x_divtagdefaultwrapper" style="font-family:Calibri,Arial,Helvetica,sans-serif; font-size:; margin:0">
<font size="3" face="'Times New Roman', Times, serif" color="808080"><font size="3" face="'Times New Roman', Times, serif" color="808080">Swiss Federal Institute of Technology Lausanne (EPFL<font color="808080"><font face="'Times New Roman', Times, serif">)</font></font></font><br>
</font></div>
<font color="808080"></font>
<div name="x_divtagdefaultwrapper" style="font-family:Calibri,Arial,Helvetica,sans-serif; font-size:; margin:0">
<font size="3" face="'Times New Roman', Times, serif" color="808080">CH-1015 Lausanne, Switzerland<br>
+41 21 69 34 881</font></div>
<div name="x_divtagdefaultwrapper" style="font-family:Calibri,Arial,Helvetica,sans-serif; font-size:; margin:0">
<a href="http://people.epfl.ch/265334" id="LPNoLP" tabindex="0">http://people.epfl.ch/265334</a><br>
</div>
</div>
</div>
</div>
</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> users <users-bounces@lists.quantum-espresso.org> on behalf of Giuseppe Mattioli <giuseppe.mattioli@ism.cnr.it><br>
<b>Sent:</b> Thursday, October 29, 2020 3:52:48 PM<br>
<b>To:</b> Quantum ESPRESSO users Forum<br>
<b>Subject:</b> Re: [QE-users] Thermodynamics with DFT+U</font>
<div> </div>
</div>
</div>
<font size="2"><span style="font-size:10pt;">
<div class="PlainText"><br>
Dear Malte<br>
I suppose that, as a trivial rule of thumb, if U values are not so  <br>
different then you can use some average value for all systems.<br>
HTH<br>
Giuseppe<br>
<br>
Quoting Malte Sachs <malte.sachs@chemie.uni-marburg.de>:<br>
<br>
> Dear all,<br>
><br>
> I read many times in papers that I should not compare total energies  <br>
> of DFT+U calculations with different U values. What should I do in  <br>
> case of complex hull calculations of a phase diagram? With QE it is  <br>
> now possible to calculate the U values for each composition by the  <br>
> hp.x code. Can I compare the resulting total energies with each  <br>
> other safely?<br>
><br>
> Best regards,<br>
> Malte<br>
><br>
> -- <br>
> Malte Sachs<br>
> Anorganische Chemie, Fluorchemie<br>
> Philipps-Universität Marburg<br>
> Hans-Meerwein-Straße 4<br>
> 35032 Marburg (Paketpost: 35043 Marburg)<br>
> Tel.: +49 (0)6421 28 - 25 68 0<br>
> <a href="http://www.uni-marburg.de/fb15/ag-kraus/">http://www.uni-marburg.de/fb15/ag-kraus/</a><br>
<br>
<br>
<br>
GIUSEPPE MATTIOLI<br>
CNR - ISTITUTO DI STRUTTURA DELLA MATERIA<br>
Via Salaria Km 29,300 - C.P. 10<br>
I-00015 - Monterotondo Scalo (RM)<br>
Mob (*preferred*) +39 373 7305625<br>
Tel + 39 06 90672342 - Fax +39 06 90672316<br>
E-mail: <giuseppe.mattioli@ism.cnr.it><br>
<br>
_______________________________________________<br>
Quantum ESPRESSO is supported by MaX (<a href="http://www.max-centre.eu">www.max-centre.eu</a>)<br>
users mailing list users@lists.quantum-espresso.org<br>
<a href="https://lists.quantum-espresso.org/mailman/listinfo/users">https://lists.quantum-espresso.org/mailman/listinfo/users</a></div>
</span></font>
</body>
</html>