Respected Prof. Baroni<br> Sorry for that I didn't browsed in the mailing list's archive carefully!<br> I think you didn't understand me due to my english language problems.<br>1. I did a relax calculation for a slab with 10 layers, 20 C atoms. Than I got 20 relaxed positions.
<br>2. I did a scf calculation using the 20 positions. And pw.x found "No symmetry".<br>3. I inputed the structure into Material Studio, and it find 8 symmetry operations and 7 inequivalent atoms. The tolerance is
0.01A.<br>4. I used 8 symmetry operations to act on the 7 positions, then got 56 positions(some of them are superpositions).<br>5 I selected 20 positions in the primary unit cell(there were only 20 positions in one primary unit cell) and input them into the PWscf's run_scf script. Then
pw.x find 8 symmetry operations. <br> This method is equivalent to modify the atom positions manually in order to make the pw.x to find symmetry operations!<br> The manual say that :<br><br>>the number of significant figures in the atomic positions is not
large enough.
In file <br>><tt>PW/eqvect.f90</tt>, the variable <tt>accep</tt> is
used to decide whether a rotation is a symmetry operation.<br>>Its current value (<span class="MATH">10<sup>-5</sup></span>
) is quite strict: a rotated atom must
coincide with another atom to <br>>5 significant digits.
You may change the value of <tt>accep</tt> and recompile.<br><br> In the last email I said that I had changed the value of <tt>accep</tt> and recompiled. But it made no use to find any symmetry operations. So I have no method other than to modify the atom positions by symmetry manually.
<span style="font-family: monospace;"><br> Now I think the problem has been solved!<br> Many thanks again and forgive me to bring so many troubles!<br>Best wishes!<br>Yingli Niu<br><br><br></span><pre>On Aug 31, 2006, at 04:01:40 +0200, Prof. Baroni wrote:
</pre>>> Dear Pro. Baroni<br><pre>>><i> Thanks a lot for your help!<br></i>>><i><span style="font-style: italic;"></span> I have said that It was no use to set accep=3D1.0D-2<br></i>>><i><span style="font-style: italic;">
</span> and recompiled pw.x. The program still could not find<br></i>>><i> any symmetry operation.<br></i>>><i> Then I selected 7 inequivalent atoms, then used 8<br></i>>><i> symmetry operations:<br></i>
>><i> x,y,z<br></i>>><i> -x,-y,z<br></i>>><i> -x+1/2,y+1/2,-z<br></i>>><i> x+1/2,-y+1/2,-z<br></i>>><i> -x+1/2,-y+1/2,-z<br></i>>><i> x+1/2,y+1/2,-z<br></i>>><i> x,-y,z
<br></i>>><i> -x,y,z<br></i>>><i> to transform the 7 inequivalent atoms to 20 atoms in<br></i>>><i> the primary unit cell.<br></i>><br>>I can hardly understand what you did. If by applying some symmetry =20
<br>>operations to a bunch of positions ("atoms") you obtain a larger =20<br>>number of atoms, doesn't this mean that those operations are not =20<br>>"symmetrry operations"???<br>>><i> Then I ran scf. I use Norm-Conserving PP and chose
<br></i>>><i> ecutwfc=3D70,ecutrho=3D600.<span style="font-style: italic;"></span><br></i>><br>>there is ABSOLUTELY no point in using ecutrho>ecutwfc with norm =20<br>>conserving PP's<br>><br>>>
<i> There appeared message "warning:symmetry operation #<br></i>>><i> 3/4/5/6 not allowed fractional translation:" in the<br></i>>><i> out file as discussed before. Only 4 symmetry<br></i>>>
<i> operations without inversion could be found!<br></i>>><i> I set ecutrho =3D 500 and calculated it again. Then<br></i>>><i> all 8 symmetry operations was found!<br></i>>><i> The stress tensor did not change too much!
<br></i>>><i> stress tensor(kbar) from original atom positions:<br></i>>><i> 49.20 0.01 0.07<br></i>>><i> 0.01 -16.72 0.16<br></i>>><i> 0.07 0.16 2.61<br>
</i>>><i> stress tensor(kbar) from changed atom positions:<br></i>>><i> 49.98 0.00 0.00<br></i>>><i> 0.00 -16.13 0.00<br></i>>><i> 0.00 0.00 7.30<br></i>
>><i> All the nondiagonal elements changed to 0.00.<br></i>>><i> I think changed accpe from 1.0D-5 to 1.0D-2 will be<br></i>>><i> harmful for the calculations. So I reset it to 1.0D-5<br></i>>>
<i> and recompiled.<br></i>>><i> I have no other method to make the program find out<br></i>>><i> all 8 symmetry operations than this method.<br></i><br>>Have you browsed in the mailing list's archive, as I suggested? =20
<br>>Please, let me NOT rediscuss this point that I and others discussed =20<br>>the last time a few days ago<br>><br>>><i><br></i>>><i> Another question:<br></i>>><i> I use Yahoo mail box. How can I post the reply
<br></i>>><i> message next to the message which is replied in the<br></i>>><i> mailing list. I alway started a new discussion line<br></i>>><i> with reply message! Sorry for that!<br></i>>><br>>Please, ask Terry Semel. He certainly knows better than me.
<br>><br>>SB<br>><br>>---<br>>Stefano Baroni - SISSA & DEMOCRITOS National Simulation Center - =20<br>>Trieste<br>>[+39] 040 3787 406 (tel) -528 (fax) / stefanobaroni (skype)</pre>