The convergency criterion of dimer

[lgxyz]

Hi, everyone, I was puzzled by the DIMCAR of the example given in the http://theory.cm.utexas.edu/vtsttools/dimer/.
Here is the copy of DIMCAR in the page:
Step Force Torque Energy Curvature Angle
1 3.26055 1.42551 -38.08043 -3.67928 7.32459
1 3.26055 0.47200 -38.08043 -3.89665 0.48560
2 0.65134 0.32662 -38.17792 -2.88463 1.05846
3 0.18678 0.22588 -38.19421 -2.88144 0.23646
4 0.20597 0.11406 -38.19315 -2.82977 0.14029
5 0.02071 0.58961 -38.19131 -3.08022 0.61217
6 0.01329 0.06599 -38.19131 -3.05717 0.14982
7 0.00990 0.11718 -38.19130 -3.07347 0.26844
8 0.00208 0.11183 -38.19130 -3.04074 0.14479
In the OUTCAR of the example, the step of iteration is "Iteration 35( 8)", and here the step in DIMCAR is only 8. What's the difference between the two?
Also, in the OUTCAR the FORCES(max atom) is 0.000795, and here the FORCE is 0.00208. The EDIFFG in the INCAR is set to -0.001.
So what does 0.00208 mean?

I have recaculated the example using the the same setup files.
Next is the DIMCAR that I recaculated, which is not equal with the above, pls give some suggestions.
Step Force Torque Energy Curvature Angle
1 3.26060 1.44172 -38.08043 -3.68256 7.36457
1 3.26060 0.56890 -38.08043 -3.93920 0.60458
2 0.65132 0.34607 -38.17788 -2.88912 0.76967
3 0.18381 0.28476 -38.19402 -2.88573 0.34103
4 0.19504 0.18946 -38.19310 -2.87075 0.24326
5 0.02674 0.58299 -38.19131 -3.12494 0.59608
6 0.00776 0.04358 -38.19131 -3.07804 0.06099
7 0.00557 0.03777 -38.19130 -3.06899 0.11713
8 0.00066 --- -38.19130 --- ---

Any suggestion I will appreciate very much! Many thanks!

[graeme]

Convergence as defined by EDIFFG is based upon the max atom force (the highest force on any atom), where in your case 0.026 < 0.030 determined convergence. The force that is output in the DIMCAR file is the total force on the dimer. Note that the dimer force can be slightly different from what is reported in the OUTCAR because the dimer force (0.08665) has a projected component and is for the center of the dimer, whereas the OUTCAR contains the force (0.065420) for whatever geometry is being using for dimer convergence. This would all be less confusing if we used the total force for convergence, but we have stuck with the vasp default.