<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix">The best theoretical&nbsp; approach is the
      one mentioned by Laurence in his last message. <br>
      <br>
      About your last question ... It is clearly a matter of taste
      because whatever we do, an approach consisting to choose the Ueff
      value by ajusting it for one property is "semi-empirical". I never
      did such an asjustment of Ueff ... just based on the cell volume
      or structural properties and then use the Ueff for electronic
      properties. <br>
      However, it could make sense, because the structural properties
      are the consequence of the electronic structure (orbital
      interactions) and thus a good description of the structural
      properties should imply a good description of the electronic
      structure ... However, you can also artificially stabilize the
      experimental volume for bad reasons!!! It is why I prefer to have
      one more property (optical, magnetic, NMR, ...) to compare with my
      calculation in order to validate the Ueff value. <br>
      <br>
      I should mentioned that depending on what you are doing the Ueff
      value could be different! Indeed fixing the Ueff based on magnetic
      properties or optical properties will not lead to the same Ueff
      value ... This is mainly related to the fact that when we adjust
      the Ueff value based on an experimental value we correct not only
      the deficiency of DFT to properly describe the localized d- or
      f-states, but more ... In that sense adjusting the Ueff value on
      an experimental value is not correct but it remains a pragmatic
      approach (with clear limitations). <br>
      <br>
      In the other hand, the Ueff value deduded from the constrained
      approach will lead to results that will not perfectly reproduce
      the experiments, but these results will correspond to the best you
      can do using DFT with a "correct" Ueff correction.&nbsp; It is why I
      consider that it remains a matter of taste and why I feel that you
      need other experimental data to insure that your Ueff parameter is
      "relevant in a certain context". <br>
      <br>
      Best Regards<br>
      <br>
      Xavier<br>
      <br>
      <br>
      Le 3/1/2013 9:12 PM, Zsolt Rak a &eacute;crit&nbsp;:<br>
    </div>
    <blockquote
cite="mid:CACQkguoaX9m+9+yMZfdHVpno6z-SkqP7SXZB7ANSPSuupZp3PA@mail.gmail.com"
      type="cite">Dear Xavier, you said that you choose the Ueff value
      to reproduce one property and then you use this Ueff value to
      predict the others. My question is can I choose Ueff to reproduce
      a structural property (such as the lattice constant or cell
      volume) and than use that value of Ueff to predict electronic
      properties? <br>
      <br>
      <br>
      <br>
      <div class="gmail_quote">On Fri, Mar 1, 2013 at 3:04 PM, Laurence
        Marks <span dir="ltr">&lt;<a moz-do-not-send="true"
            href="mailto:L-marks@northwestern.edu" target="_blank">L-marks@northwestern.edu</a>&gt;</span>
        wrote:<br>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          If all you have are the atomic positions (and lattice
          constant) then the only unbiased approach would be to
          calculate the U, see <a moz-do-not-send="true"
            href="http://www.wien2k.at/reg_user/textbooks/Constraint_U.pdf"
            target="_blank">http://www.wien2k.at/reg_user/textbooks/Constraint_U.pdf</a>.
          <div>
            <br>
          </div>
          <div>N.B., the DFT+U method is variational for a fixed U, but
            I don't think it is variational as a function(al) of U.<br>
            <br>
            <div class="gmail_quote">
              <div class="im">On Fri, Mar 1, 2013 at 1:58 PM, Zsolt Rak
                <span dir="ltr">&lt;<a moz-do-not-send="true"
                    href="mailto:zsolt.rak@gmail.com" target="_blank">zsolt.rak@gmail.com</a>&gt;</span>
                wrote:<br>
              </div>
              <blockquote class="gmail_quote" style="margin:0 0 0
                .8ex;border-left:1px #ccc solid;padding-left:1ex">
                <div>
                  <div class="im">
                    Let's suppose that the atomic positions (and lattice
                    constants) are the only information that I have. Is
                    it physically justified to use LDA+U/GGA+U to
                    optimize the volume? or to tune the U value to
                    reproduce the experimental lattice constants? Also,
                    is the DFT+U method based on the variational
                    principle?<br>
                    <br>
                  </div>
                  <div>
                    <div class="h5">
                      <div class="gmail_quote">On Fri, Mar 1, 2013 at
                        2:43 PM, Laurence Marks <span dir="ltr">
                          &lt;<a moz-do-not-send="true"
                            href="mailto:L-marks@northwestern.edu"
                            target="_blank">L-marks@northwestern.edu</a>&gt;</span>
                        wrote:<br>
                        <blockquote class="gmail_quote" style="margin:0
                          0 0 .8ex;border-left:1px #ccc
                          solid;padding-left:1ex">
                          You need more a-priori information than this,
                          for instance the bulk energy of related
                          compounds for which a U is relevant. Tuning
                          the U to reproduce known data is not by itself
                          spectacular science.<br>
                          <br>
                          <div class="gmail_quote">
                            <div>On Fri, Mar 1, 2013 at 1:24 PM, Zsolt
                              Rak <span dir="ltr">&lt;<a
                                  moz-do-not-send="true"
                                  href="mailto:zsolt.rak@gmail.com"
                                  target="_blank">zsolt.rak@gmail.com</a>&gt;</span>
                              wrote:<br>
                            </div>
                            <blockquote class="gmail_quote"
                              style="margin:0 0 0 .8ex;border-left:1px
                              #ccc solid;padding-left:1ex">
                              <div>
                                <div>I want to calculate the most
                                  accurate bulk energy and the a-priori
                                  information I have are the atomic
                                  positions.<br>
                                  <br>
                                  <br>
                                  <br>
                                </div>
                                <div class="gmail_quote">
                                  <div>On Fri, Mar 1, 2013 at 1:59 PM,
                                    Laurence Marks <span dir="ltr">&lt;<a
                                        moz-do-not-send="true"
                                        href="mailto:L-marks@northwestern.edu"
                                        target="_blank">L-marks@northwestern.edu</a>&gt;</span>
                                    wrote:<br>
                                  </div>
                                  <blockquote class="gmail_quote"
                                    style="margin:0 0 0
                                    .8ex;border-left:1px #ccc
                                    solid;padding-left:1ex">
                                    <div>
                                      <div>My two cents. Both LDA+U and
                                        GGA+U are wrong. That said, for
                                        f-/d- systems they are often
                                        better than LDA/GGA for some
                                        properties. The question you
                                        should ask yourself is what
                                        property are you trying to
                                        measure/predict, and what
                                        a-priori information (reference
                                        state) do you have that can be
                                        used?
                                        <div><br>
                                        </div>
                                        <div>For instance, if I want to
                                          calculate a surface energy
                                          then I would tune the U to
                                          give the most accurate bulk
                                          energy treating this as my
                                          a-priori information;
                                          similarly if I wanted to
                                          calculate the elastic behavior
                                          of a defect I would tune to
                                          the bulk elastic constants. In
                                          my opinion this is the only
                                          justifiable approach.
                                          <div>
                                            <div><br>
                                              <br>
                                              <div class="gmail_quote">On
                                                Fri, Mar 1, 2013 at
                                                12:47 PM, Zsolt Rak <span
                                                  dir="ltr">
                                                  &lt;<a
                                                    moz-do-not-send="true"
href="mailto:zsolt.rak@gmail.com" target="_blank">zsolt.rak@gmail.com</a>&gt;</span>
                                                wrote:<br>
                                                <blockquote
                                                  class="gmail_quote"
                                                  style="margin:0 0 0
                                                  .8ex;border-left:1px
                                                  #ccc
                                                  solid;padding-left:1ex">
                                                  <div>Dear wien2k
                                                    users, <br>
                                                    <br>
                                                    I am calculating the
                                                    properties of
                                                    several f-electron
                                                    compounds. I would
                                                    like to ask the
                                                    users' opinion about
                                                    the volume
                                                    optimization in an
                                                    f- or d-electron
                                                    system: which way is
                                                    better (or
                                                    physically
                                                    justified), with
                                                    LDA/GGA or with
                                                    LDA+U/GGA+U? In my
                                                    opinion, the
                                                    LDA+U/GGA+U
                                                    techniques were
                                                    developed to correct
                                                    band energies of
                                                    localized states, so
                                                    there is no
                                                    fundamental physical
                                                    reason to use
                                                    LDA+U/GGA+U methods
                                                    for volume
                                                    optimization.
                                                    However, we observe
                                                    a change in the
                                                    lattice parameters
                                                    when we go from
                                                    LDA/GGA to
                                                    LDA+U/GGA+U. Also,
                                                    from a brief search
                                                    of the literature we
                                                    found that, in many
                                                    cases, people tune
                                                    the Hubbard-U
                                                    parameter to
                                                    reproduce the
                                                    experimental lattice
                                                    constants.
                                                    <br>
                                                    I would appreciate
                                                    further thoughts and
                                                    insights into this
                                                    issue.<br>
                                                    <br>
                                                    Thank you, <br>
                                                    Zs </div>
                                                </blockquote>
                                              </div>
                                              <br>
                                              <br clear="all">
                                              <div><br>
                                              </div>
                                            </div>
                                          </div>
                                          <span><font color="#888888">--
                                              <br>
                                              Professor Laurence Marks<br>
                                              Department of Materials
                                              Science and Engineering<br>
                                              Northwestern University<br>
                                              <a moz-do-not-send="true"
href="http://www.numis.northwestern.edu" target="_blank">www.numis.northwestern.edu</a>
                                              <a moz-do-not-send="true"
href="tel:1-847-491-3996" value="+18474913996" target="_blank">1-847-491-3996</a><br>
                                              "Research is to see what
                                              everybody else has seen,
                                              and to think what nobody
                                              else has thought"<br>
                                              Albert Szent-Gyorgi </font></span></div>
                                        <br>
                                      </div>
                                    </div>
                                    <div>_______________________________________________<br>
                                      Wien mailing list<br>
                                      <a moz-do-not-send="true"
                                        href="mailto:Wien@zeus.theochem.tuwien.ac.at"
                                        target="_blank">Wien@zeus.theochem.tuwien.ac.at</a><br>
                                      <a moz-do-not-send="true"
                                        href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien"
                                        target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a><br>
                                      <br>
                                    </div>
                                  </blockquote>
                                </div>
                                <br>
                              </div>
                            </blockquote>
                          </div>
                          <br>
                          <br clear="all">
                          <div>
                            <div>
                              <div><br>
                              </div>
                              -- <br>
                              Professor Laurence Marks<br>
                              Department of Materials Science and
                              Engineering<br>
                              Northwestern University<br>
                              <a moz-do-not-send="true"
                                href="http://www.numis.northwestern.edu"
                                target="_blank">www.numis.northwestern.edu</a>
                              <a moz-do-not-send="true"
                                href="tel:1-847-491-3996"
                                value="+18474913996" target="_blank">1-847-491-3996</a><br>
                              "Research is to see what everybody else
                              has seen, and to think what nobody else
                              has thought"<br>
                              Albert Szent-Gyorgi </div>
                          </div>
                          <br>
_______________________________________________<br>
                          Wien mailing list<br>
                          <a moz-do-not-send="true"
                            href="mailto:Wien@zeus.theochem.tuwien.ac.at"
                            target="_blank">Wien@zeus.theochem.tuwien.ac.at</a><br>
                          <a moz-do-not-send="true"
                            href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien"
                            target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a><br>
                          <br>
                        </blockquote>
                      </div>
                      <br>
                    </div>
                  </div>
                </div>
              </blockquote>
            </div>
            <div>
              <div class="h5"><br>
                <br clear="all">
                <div><br>
                </div>
                -- <br>
                Professor Laurence Marks<br>
                Department of Materials Science and Engineering<br>
                Northwestern University<br>
                <a moz-do-not-send="true"
                  href="http://www.numis.northwestern.edu"
                  target="_blank">www.numis.northwestern.edu</a> <a
                  moz-do-not-send="true" href="tel:1-847-491-3996"
                  value="+18474913996" target="_blank">1-847-491-3996</a><br>
                "Research is to see what everybody else has seen, and to
                think what nobody else has thought"<br>
                Albert Szent-Gyorgi
              </div>
            </div>
          </div>
          <br>
          _______________________________________________<br>
          Wien mailing list<br>
          <a moz-do-not-send="true"
            href="mailto:Wien@zeus.theochem.tuwien.ac.at">Wien@zeus.theochem.tuwien.ac.at</a><br>
          <a moz-do-not-send="true"
            href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien"
            target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a><br>
          <br>
        </blockquote>
      </div>
      <br>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
Wien mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Wien@zeus.theochem.tuwien.ac.at">Wien@zeus.theochem.tuwien.ac.at</a>
<a class="moz-txt-link-freetext" href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a>
</pre>
    </blockquote>
    <br>
  </body>
</html>