FREQUENTLY ASKED QUESTIONS
Q:
When integrating an atomic basin TOPXD gives the following error message:
  PATHE2: OSCILLATION OF PATHS
  PATHE2: THE ATTRACTOR OF THIS PATH WAS PROBABLY NOT INCLUDED IN THE CLUSTER
A:
     Did you check the list of atoms reached into the feeler rays determination step?
     Do you think that some neighbouring atoms were missed?  If so, you have to increase NVI parameter in order to include the missing atoms into the list of possible attractors of trajectories of the charge density.
     The  OSCILLATION OF PATHS  message may also appear in some cases where the integration will be anyhow successful. In many instances it represents just a warning. Especially, if you noticed that the list of neighboring atoms (after the feeler ray step) corresponds to your expectations.
     Check the list of atoms reached in the feeler ray step. Increase NVI parameter, if needed, and then start again. Once you have used a very large NVI value, leave your calculation to try to end its task (even if the message appear many times).
 
Q: What grid should be used for integration of atomic basins and how does it affect the computing time ?
A:
    In order to obtain satisfactory results you should use something like:
        64*48*120 (phi*theta*radial) for second-row atoms
        32*24*96   (phi*theta*radial) for hydrogen atoms (if not involved in H-bond)
        48*32*96   (phi*theta*radial)for hydrogen atoms (if involved in H-bond)
     Computing time is roughly proportional to n(phi)*n(theta) points. The number of radial points is very important for the precision, but hardly affects the total integration time, as it is operative only in the integration step and NOT in the ZFS determination (which takes about 95% of the total time).
 
Q:  Integration of an atomic basin takes a very long time. What options do we have to speed up the calculation ?
A:
     Unfortunately the integration step is very very long (especially the ZFS determination which takes about 95% of this time). You can try with the other proposed method (which is much faster), but it often fails.
          Using the indirect method (the one you are using) you can save some time by decreasing the accuracy of the surface determination. It is 0.001 (see parameter ACCUR in subroutine TOPON)
          One could try to increase it up to 0.003 (no more than 0.005, I would say). You loose somewhat in precision, but you certainly increase in speed.
          You could compare the results of these two computations on one of the atoms you have already integrated, , etc. using:
                a) first test      : 64*48*120  ACCUR=0.001
                b) second test : 64*48*120  ACCUR=0.003
         Then you can decide if it is worth varying such a parameter and how much you can vary it.
 
Q:
How do I check the accuracy of the integration ?
A:
     Check the value of the integrated Lagrangian. For "exact" integration it should vanish (for the divergence theorem).
      In practice :
          a) It should be less than 5*10-5 for H atoms, possibly around 1*10-5. A value of 1*10-4 could be perhaps acceptable, but not too precise.
          b) For second row atoms (C,N,O, etc,) it should not exceed 1*10-3. Possibly 1*10-4
 
Q: You've mentioned that the computing time increases by a factor of  planes, but how does the NVI parameter affect the elapsed time?
A:
     It will affect it, but in a very limited way, especially after the feeler ray step. Indeed the atoms reached during the feeler ray step are put at the top of the list of the NVI reachable atoms. So that the DO loop in PATHEN and PATHEN2 (these DO run on the 3*NVI*STAR_MULTIPLICITY coordinates of the possible  attractors) are in most cases (>99%) terminated much before the end of the loop.
     In practice you shouldn't notice a CPU time increase with NVI increase. Rather you could notice a decrease, if you have added an attractor that had to be enclosed. In this case the path oscillation is avoided and CPU time considerably saved.
 
Q:
When trying to increase the number of points to something like 128*96, TOPXD gives the following error message:
ERROR **** YIELD**** GAUSS QUADRATURE NOT AVAILABLE - NUMBER OF POINTS= 128
A:
    Actually the current programs works with these number of points for the quadrature:
         1, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 36, 40, 48, 64, 96, 120
     So 128 it is not a valid option. 120*96 should work.
 
Q: Sometimes I have problems with integrated Lagrangian, which stays above 1*10-3 despite the fact that I use ACCUR=0.001 and grid as large as 96*64. I remember that you've mentioned that decreasing the number of points might help, but when I reduce these numbers to 48*32 or 64*48 it still doesn't help. These problems usually occur with carbon and nitrogen atoms, never with oxygens or hydrogens. What do I do ?
A:
    What about electroneutrality? Are you very far from it ?
     The fact that one may get problems with carbon or nitrogens and never with oxygens or hydrogen atoms seems to indicate that the former have more complicate ZFS's than the latter. You could try to solve such a problem, by increasing the radius of the beta sphere for such atoms, thus reducing the size of the remaining part of the atomic basin.
     You could use for the beta sphere something like the distance of the closest bond critical point multiplied by 1.15 (the program then reduces this number by 20%). Furthermore, the increase (inside the code) of the number of theta and phi points in the inner beta sphere might help. Please contact us and we will send you instructions on how to do it…