<HTML><BODY style="word-wrap: break-word; -khtml-nbsp-mode: space; -khtml-line-break: after-white-space; ">Dear Chun-Min,<DIV><BR class="khtml-block-placeholder"></DIV><DIV>cp2k is a mixed gaussian-plane wave code. Though in principle calculations, with PW based codes, are translational invariant, in</DIV><DIV>practice they are not because of the resolution of the GRID and the corresponding representation of the density on the grid (have a look at the paper on QS (Comp. Phys. Comm.) cited at the end of the run of cp2k if you want more technical details). The difference is generally small (as you've seen 0.5 mH) and in your case you're also using a pretty small cutoff. If you use a larger cutoff this difference should become smaller.</DIV><DIV><BR class="khtml-block-placeholder"></DIV><DIV>As I said this behavior is not peculiar of cp2k but is common to all PW based codes.</DIV><DIV><BR class="khtml-block-placeholder"></DIV><DIV>Teo</DIV><DIV> <BR><DIV><DIV>On 5 Oct 2007, at 22:34, chun wrote:</DIV><BR class="Apple-interchange-newline"><BLOCKQUOTE type="cite"><P style="margin: 0.0px 0.0px 0.0px 0.0px"><FONT face="Helvetica" size="3" style="font: 12.0px Helvetica"><SPAN class="Apple-converted-space"> </SPAN>&MGRID</FONT></P> <P style="margin: 0.0px 0.0px 0.0px 0.0px"><FONT face="Helvetica" size="3" style="font: 12.0px Helvetica"><SPAN class="Apple-converted-space"> </SPAN>CUTOFF 158</FONT></P> <P style="margin: 0.0px 0.0px 0.0px 0.0px"><FONT face="Helvetica" size="3" style="font: 12.0px Helvetica"><SPAN class="Apple-converted-space"> </SPAN>&END MGRID</FONT></P> </BLOCKQUOTE></DIV><BR></DIV></BODY></HTML>