grimme-lab/multicharge

Electronegativity equilibration model for atomic partial charges

multicharge's Issues

Read-in and Usage of System Charges

Multicharge does not currently take into account the total charge of the system via .CHRG or a command line argument.
This results in the partial charges being incorrect for charged systems.
The implementation should be straightforward. I will look into this.

Meson build file failing

The meson.build file in config/, especially line 106:

multicharge/config/meson.build

Line 106 in 4281891

if libs[0].startswith('-L')

seems to be fragile and breaks if no value had been provided for custom_libraries. I am not sure, probably, that variable should always have a defined value, when the code reaches that line, still a more informative error message would be helpful for the user. (Currently, it stops with [...]config\meson.build:106:13: ERROR: Index 0 out of bounds of array of size 0.)

Add note about git version in README.md

Problem
Meson build error “ERROR: Git command failed” with git version 1.8.3.1

Reason
Git version 1.8.3.1 has no parameter called "-C"

Solution
Update to a newer git version (e.g., 2.21.0) solves this.

cmake fails to find mstore

mstore is installed:

$ pkg info -l mstore
mstore-0.2.0:
	/usr/local/bin/mstore-fortranize
	/usr/local/bin/mstore-info
	/usr/local/include/mstore/GNU-10.3.0/mstore.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_amino20x4.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_but14diol.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_data_collection.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_data_record.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_data_store.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_heavy28.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_ice10.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_il16.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_mb16_43.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_upu23.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_version.mod
	/usr/local/include/mstore/GNU-10.3.0/mstore_x23.mod
	/usr/local/lib/cmake/mstore/Findmctc-lib.cmake
	/usr/local/lib/cmake/mstore/mstore-config-version.cmake
	/usr/local/lib/cmake/mstore/mstore-config.cmake
	/usr/local/lib/cmake/mstore/mstore-targets-release.cmake
	/usr/local/lib/cmake/mstore/mstore-targets.cmake
	/usr/local/lib/cmake/mstore/mstore-utils.cmake
	/usr/local/lib/libmstore.so
	/usr/local/lib/libmstore.so.0
	/usr/local/lib/libmstore.so.0.2.0
	/usr/local/libdata/pkgconfig/mstore.pc

but multicharge fails to find it:

-- Could NOT find mstore (missing: mstore_DIR)
-- Retrieving mstore from https://github.com/grimme-lab/mstore

Version: 0.1.2
OS: FreeBSD 13

Support for CMake build system

Explore the possibility to support CMake for this project as well beside meson and fpm.

Dependencies:

• mctc-lib (meson, cmake, and fpm enabled)
• mstore (meson and fpm enabled, see grimme-lab/mstore#4)

Segfault for large systems

If one does not increase the stack size for large systems (ulimit -s unlimited), a segmentation fault occurs.

The problem seems to stem from the parallel region in get_amat_0d.

gdb log

(gdb) bt
#0  0x0000000000412d81 in get_amat_0d::L_multicharge_model_mp_get_amat_0d__135__par_loop0_2_6 ()
    at ../src/multicharge/model.F90:135
#1  0x0000000002294eb3 in __kmp_invoke_microtask ()
#2  0x000000000225750c in __kmp_fork_call ()
#3  0x0000000002226ff8 in __kmpc_fork_call ()
#4  0x000000000041298a in multicharge_model::get_amat_0d (self=..., mol=..., amat=...)
    at ../src/multicharge/model.F90:135
#5  0x000000000041e3b7 in multicharge_model::solve (self=0x7fffffffd148, mol=..., cn=..., 
    dcndr=<Fehler beim Lesen der Variable: Location address is not set.>, 
    dcndl=<Fehler beim Lesen der Variable: Location address is not set.>, energy=..., 
    gradient=<Fehler beim Lesen der Variable: Location address is not set.>, 
    sigma=<Fehler beim Lesen der Variable: Location address is not set.>, qvec=..., 
    dqdr=<Fehler beim Lesen der Variable: Location address is not set.>, 
    dqdl=<Fehler beim Lesen der Variable: Location address is not set.>)
    at ../src/multicharge/model.F90:468
#6  0x00000000004089e6 in main () at ../app/main.f90:76
#7  0x00000000004056f2 in main ()

The error vanishes upon removing the OMP directives.

The problem was first raised in dftd4/dftd4#218.

Wrong energy gradient for charged systems

The gradient of the electrostatic energy does not match the numerical gradient for charged systems.

Examples are added in #25.