Bugs in ECI for QM and QM/MM (point charge embedding only).

[gmondragn]

I'm trying to use ECI module on an 60 atoms two molecule system. To test ECI module beyond the example, I submitted a QM calculation with only my 60 atoms. Atom types are C, H, N, O, and Cl.

I created BD and BA corresponding directories according to SHARC manual, and my ECI template is not different from the example in SHARC directory. ECI iterates correctly over the point charges and it calculates correctly all my requested states. After that, it displays some warnings, then displays an error and suddenly ECI stops:

/projects/donglab/software/SHARC4/bin/SHARC_ECI.py:944: ComplexWarning: Casting complex values to real discards the imaginary part
mu[z][(s1,s2)] = child.QMout['dm'][:,i1,i2].astype(float)
/projects/donglab/software/SHARC4/lib/ECI.py:419: ComplexWarning: Casting complex values to real discards the imaginary part
energies = [ f.H[s.Z][f.states[s.Z].index(s), f.states[s.Z].index(s)].astype(float) for s in states ]
Traceback (most recent call last):
File "/projects/donglab/software/SHARC4/bin/SHARC_ECI.py", line 1047, in
SHARC_ECI().main()
~~~~~~~~~~~~~~~~^^
File "/projects/donglab/software/SHARC4/lib/SHARC_INTERFACE.py", line 292, in main
self.run()
~~~~~~~~^^
File "/projects/donglab/software/SHARC4/bin/SHARC_ECI.py", line 979, in run
self.ECIjobs[Z].run()
~~~~~~~~~~~~~~~~~~~^^
File "/projects/donglab/software/SHARC4/lib/ECI.py", line 321, in run
self.calculate_Hmat() # Construct entire ECI-CT Hamiltonian for all multiplicities
~~~~~~~~~~~~~~~~~~~^^
File "/projects/donglab/software/SHARC4/lib/ECI.py", line 987, in calculate_Hmat
bas2 = self.extract_subbasis_from_global_indices( self.log, f2.mol.basis, [ f2.mol.atom_symbol(i) for i in range(f2.mol.natm)], shells_2 )
File "/projects/donglab/software/SHARC4/lib/ECI.py", line 748, in extract_subbasis_from_global_indices
shells = mole_basis[atom]
~~~~~~~~~~^^^^^^
KeyError: 'Cl7'

My QM.log file is truncated after the error occurs.

Now, a different error arises when I include an external point charge embedding by adding "point_charges" keyword in QM.in file. The calculation starts, somehow it's not parsing my file with the point charge coordinates and charges. It displays the following error:

Traceback (most recent call last):
File "/projects/donglab/software/SHARC4/SHARC402/bin/SHARC_ECI.py", line 1053, in
SHARC_ECI().main()
~~~~~~~~~~~~~~~~^^
File "/projects/donglab/software/SHARC4/SHARC402/lib/SHARC_INTERFACE.py", line 292, in main
self.set_coords(QMinfilename)
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^
File "/projects/donglab/software/SHARC4/SHARC402/bin/SHARC_ECI.py", line 714, in set_coords
if QMin.coords['pccoords'] != None:
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

I would appreciate any help on these issues!

[tpitesa]

Hi,

Regarding the first error - can you send us your QM directory (ideally in .tar format)? I can't reproduce this KeyError and can't logically see how it can occur, so I have to try with your input.

Regarding the external-PCs error - this is a bug in SHARC_ECI.py and will be fixed in the next release. Thanks for reporting!
As a quick fix, try to replace QMin.coords['pccoords'] != None: with if QMin.coords['pccoords'] is not None: throughout SHARC_ECI.py (it appears on multiple places).
Not sure whether it'll work (can't try it at the moment), but it actually might.

Best,
Tomislav

[tpitesa]

The first error comes from inconsistencies between basis-set dict and atomic_symbols() function of PySCF with lower/upper-case letters in atomic symbols (one gives Cl and the other one expects CL).
We never tried the final code on a system with a two-letter element, so we didn't catch this :)

As a quick fix:

• find def extract_subbasis_from_global_indices(... in $SHARC/lib/ECI.py,
• add workatoms = [ atom.upper() for atom in atoms ] as the first line of the function,
• a few lines below, replace for atom in atoms with for atoms in workatoms.

We'll include the fix in the next release.

Tomislav