Differences

This shows you the differences between two versions of the page.

--- events:2018_summer_school:scf_setup [2018/06/17 08:46] – [Traditional Diagonalisation (TD)] mwatkins
+++ events:2018_summer_school:scf_setup [2020/08/21 10:15] (current) – external edit 127.0.0.1
@@ Line 30: / Line 30: @@
       SCF_GUESS ATOMIC
     # SCF_GUESS        RESTART
-      EPS_SCF      1.0E-7
+      EPS_SCF      1.0E-5
       &MIXING
         ALPHA 0.4
@@ Line 155: / Line 155: @@
 &GLOBAL
   PROJECT H2O-32
-  RUN_TYPE ENERGY
+  RUN_TYPE MD
   PRINT_LEVEL MEDIUM
 &END GLOBAL
@@ Line 172: / Line 172: @@
 </note>
 ===== Orbital Transformation (OT) =====
+We can see the effect of changing to the OT method by simply changing the SCF section.
+<code>
+&SCF
+  SCF_GUESS ATOMIC
+  EPS_SCF 1.0E-06
+  MAX_SCF 20
+  &OT ON
+    MINIMIZER DIIS
+    PRECONDITIONER FULL_ALL
+    ENERGY_GAP 0.001
+  &END OT
+  &OUTER_SCF
+    MAX_SCF 2
+  &END OUTER_SCF
+&END SCF
+</code>
+<note>**TASK**
+See how OT compares with TD for the different basis sets you ran previously.
+</note>
+=====Other parameters=====
+There are other factors that influence the effectiveness of the setups.
+<note>**TASK**
+Exchange correlation functional.
+We were using LDA (the particular parameterization is called PADE in CP2K). Try changing the functional to `PBE` (in the `XC` section) and see how this changes convergence.
+</note>
+<note tip>
+Also change the pseudopotential you are using in the KIND sections so it matches the functional.
+</note>
+<note>**TASK**
+System size affects the efficiency too. You can find larger water boxes in the ''${main directory of my cp2k installation}/tests/QS/benchmarks'' directory, or {{https://sourceforge.net/p/cp2k/code/HEAD/tree/trunk/cp2k/tests/QS/benchmark/|online}}.
+Change the CELL parameters and the coordinates of the atoms and see how the methods scale.
+You could also explore how the methods scale with number of processors used.
+</note>
+<note>**TASK**
+And of course, probably the most important 'parameter' is the system itself. Water is quite well behaved - largely meaning it is closed shell and has a large band gap. Try similar tests on a system you are interested in!</note>