auto-calculate ADDED_MOS based on basissets

Author: dev-zero

aiida_cp2k/calculations/__init__.py

@@ -22,6 +22,7 @@
     validate_pseudos_namespace,
     write_basissets,
     write_pseudos,
+    estimate_added_mos,
 )
 from ..utils import Cp2kInput
 
@@ -139,7 +140,7 @@ def prepare_for_submission(self, folder):
         :return: `aiida.common.datastructures.CalcInfo` instance
         """
 
-        # pylint: disable=too-many-statements,too-many-branches
+        # pylint: disable=too-many-statements,too-many-branches,too-many-locals
 
         # Create cp2k input file.
         inp = Cp2kInput(self.inputs.parameters.get_dict())
@@ -167,6 +168,21 @@ def prepare_for_submission(self, folder):
                                self.inputs.structure if 'structure' in self.inputs else None)
             write_basissets(inp, self.inputs.basissets, folder)
 
+        # if we have both basissets and structure we can start helping the user :)
+        if 'basissets' in self.inputs and 'structure' in self.inputs:
+            try:
+                scf_section = inp.get_section_dict('FORCE_EVAL/DFT/SCF')
+
+                if 'SMEAR' in scf_section and 'ADDED_MOS' not in scf_section:
+                    # now is our time to shine!
+                    added_mos = estimate_added_mos(self.inputs.basissets, self.inputs.structure)
+                    inp.add_keyword('FORCE_EVAL/DFT/SCF/ADDED_MOS', added_mos)
+                    self.logger.info(f'The FORCE_EVAL/DFT/SCF/ADDED_MOS was added with an automatically estimated value'
+                                     f' of {added_mos}')
+
+            except (KeyError, TypeError):  # no SCF, no smearing, or multiple FORCE_EVAL, nothing to do (yet)
+                pass
+
         if 'pseudos' in self.inputs:
             validate_pseudos(inp, self.inputs.pseudos, self.inputs.structure if 'structure' in self.inputs else None)
             write_pseudos(inp, self.inputs.pseudos, folder)

aiida_cp2k/utils/datatype_helpers.py

@@ -160,6 +160,31 @@ def validate_basissets(inp, basissets, structure):
             kind_sec["ELEMENT"] = bset.element
 
 
+def estimate_added_mos(basissets, structure, fraction=0.3):
+    """Calculate an estimate for ADDED_MOS based on used basis sets"""
+
+    symbols = [structure.get_kind(s.kind_name).get_symbols_string() for s in structure.sites]
+    n_mos = 0
+
+    # We are currently overcounting in the following cases:
+    #   * if we get a mix of ORB basissets for the same chemical symbol but different sites
+    #   * if we get multiple basissets for one element (merged within CP2K)
+
+    for label, bset in _unpack(basissets):
+        try:
+            _, bstype = label.split("_", maxsplit=1)
+        except ValueError:
+            bstype = "ORB"
+
+        if bstype != "ORB":  # ignore non-ORB basissets
+            continue
+
+        n_mos += symbols.count(bset.element) * bset.n_orbital_functions
+
+    # at least one additional MO per site, otherwise a fraction of the total number of orbital functions
+    return max(len(symbols), int(fraction * n_mos))
+
+
 def write_basissets(inp, basissets, folder):
     """Writes the unified BASIS_SETS file with the used basissets"""
     _write_gdt(inp, basissets, folder, "BASIS_SET_FILE_NAME", "BASIS_SETS")

aiida_cp2k/utils/input_generator.py

@@ -56,6 +56,69 @@ def add_keyword(self, kwpath, value, override=True, conflicting_keys=None):
 
         Cp2kInput._add_keyword(kwpath, value, self._params, ovrd=override, cfct=conflicting_keys)
 
+    @staticmethod
+    def _stringify_path(kwpath):
+        """Stringify a kwpath argument"""
+        if isinstance(kwpath, str):
+            return kwpath
+
+        assert isinstance(kwpath, Sequence), "path is neither Sequence nor String"
+        return "/".join(kwpath)
+
+    def get_section_dict(self, kwpath=""):
+        """Get a copy of a section from the current input structure
+
+        Args:
+
+            kwpath: Can be a single keyword, a path with `/` as divider for sections & key,
+                    or a sequence with sections and key.
+        """
+
+        section = self._get_section_or_kw(kwpath)
+
+        if not isinstance(section, Mapping):
+            raise TypeError(f"Section '{self._stringify_path(kwpath)}' requested, but keyword found")
+
+        return deepcopy(section)
+
+    def get_keyword_value(self, kwpath):
+        """Get the value of a keyword from the current input structure
+
+        Args:
+
+            kwpath: Can be a single keyword, a path with `/` as divider for sections & key,
+                    or a sequence with sections and key.
+        """
+
+        keyword = self._get_section_or_kw(kwpath)
+
+        if isinstance(keyword, Mapping):
+            raise TypeError(f"Keyword '{self._stringify_path(kwpath)}' requested, but section found")
+
+        return keyword
+
+    def _get_section_or_kw(self, kwpath):
+        """Retrieve either a section or a keyword given a path"""
+
+        if isinstance(kwpath, str):
+            kwpath = kwpath.split("/")  # convert to list of sections if string
+
+        # get a copy of the path in a mutable sequence
+        # accept any case, but internally we use uppercase
+        # strip empty strings to accept leading "/", "//", etc.
+        path = [k.upper() for k in kwpath if k]
+
+        # start with a reference to the root of the parameters
+        current = self._params
+
+        try:
+            while path:
+                current = current[path.pop(0)]
+        except KeyError:
+            raise KeyError(f"Section '{self._stringify_path(kwpath)}' not found in parameters")
+
+        return current
+
     def render(self):
         output = [self.DISCLAIMER]
         self._render_section(output, deepcopy(self._params))

test/test_gaussian_datatypes.py

@@ -778,3 +778,86 @@ def test_without_kinds(cp2k_code, cp2k_basissets, cp2k_pseudos, clear_database):
 
     _, calc_node = run_get_node(CalculationFactory("cp2k"), **inputs)
     assert calc_node.exit_status == 0
+
+
+def test_added_mos(cp2k_code, cp2k_basissets, cp2k_pseudos, clear_database):  # pylint: disable=unused-argument
+    """Testing CP2K with the Basis Set stored in gaussian.basisset and a smearing section but no predefined ADDED_MOS"""
+
+    structure = StructureData(cell=[[4.00759, 0.0, 0.0], [-2.003795, 3.47067475, 0.0],
+                                    [3.06349683e-16, 5.30613216e-16, 5.00307]],
+                              pbc=True)
+    structure.append_atom(position=(-0.00002004, 2.31379473, 0.87543719), symbols="H")
+    structure.append_atom(position=(2.00381504, 1.15688001, 4.12763281), symbols="H")
+    structure.append_atom(position=(2.00381504, 1.15688001, 3.37697219), symbols="H")
+    structure.append_atom(position=(-0.00002004, 2.31379473, 1.62609781), symbols="H")
+
+    # parameters
+    parameters = Dict(
+        dict={
+            'GLOBAL': {
+                'RUN_TYPE': 'ENERGY',
+            },
+            'FORCE_EVAL': {
+                'METHOD': 'Quickstep',
+                'DFT': {
+                    "XC": {
+                        "XC_FUNCTIONAL": {
+                            "_": "PBE",
+                        },
+                    },
+                    "MGRID": {
+                        "CUTOFF": 100.0,
+                        "REL_CUTOFF": 10.0,
+                    },
+                    "QS": {
+                        "METHOD": "GPW",
+                        "EXTRAPOLATION": "USE_GUESS",
+                    },
+                    "SCF": {
+                        "EPS_SCF": 1e-05,
+                        "MAX_SCF": 3,
+                        "MIXING": {
+                            "METHOD": "BROYDEN_MIXING",
+                            "ALPHA": 0.4,
+                        },
+                        "SMEAR": {
+                            "METHOD": "FERMI_DIRAC",
+                            "ELECTRONIC_TEMPERATURE": 300.0,
+                        },
+                    },
+                    "KPOINTS": {
+                        "SCHEME": "MONKHORST-PACK 2 2 1",
+                        "FULL_GRID": False,
+                        "SYMMETRY": False,
+                        "PARALLEL_GROUP_SIZE": -1,
+                    },
+                },
+            },
+        })
+
+    options = {
+        "resources": {
+            "num_machines": 1,
+            "num_mpiprocs_per_machine": 1
+        },
+        "max_wallclock_seconds": 1 * 3 * 60,
+    }
+
+    inputs = {
+        "structure": structure,
+        "parameters": parameters,
+        "code": cp2k_code,
+        "metadata": {
+            "options": options,
+        },
+        "basissets": {label: b for label, b in cp2k_basissets.items() if label == "H"},
+        "pseudos": {label: p for label, p in cp2k_pseudos.items() if label == "H"},
+    }
+
+    _, calc_node = run_get_node(CalculationFactory("cp2k"), **inputs)
+
+    assert calc_node.exit_status == 0
+
+    # check that the ADDED_MOS keyword was added within the calculation
+    with calc_node.open("aiida.inp") as fhandle:
+        assert any("ADDED_MOS" in line for line in fhandle), "ADDED_MOS not found in the generated CP2K input file"

test/test_input_generator.py

@@ -185,3 +185,31 @@ def test_invalid_preprocessor():
     inp = Cp2kInput({"@SET": "bar"})
     with pytest.raises(ValueError):
         inp.render()
+
+
+def test_get_keyword_value():
+    """Test get_keyword_value()"""
+    inp = Cp2kInput({"FOO": "bar", "A": {"KW1": "val1"}})
+    assert inp.get_keyword_value("FOO") == "bar"
+    assert inp.get_keyword_value("/FOO") == "bar"
+    assert inp.get_keyword_value("A/KW1") == "val1"
+    assert inp.get_keyword_value("/A/KW1") == "val1"
+    assert inp.get_keyword_value(["A", "KW1"]) == "val1"
+    with pytest.raises(TypeError):
+        inp.get_keyword_value("A")
+
+
+def test_get_section_dict():
+    """Test get_section_dict()"""
+    orig_dict = {"FOO": "bar", "A": {"KW1": "val1"}}
+    inp = Cp2kInput(orig_dict)
+    assert inp.get_section_dict("/") == orig_dict
+    assert inp.get_section_dict("////") == orig_dict
+    assert inp.get_section_dict("") == orig_dict
+    assert inp.get_section_dict() == orig_dict
+    assert inp.get_section_dict("/") is not orig_dict  # make sure we get a distinct object
+    assert inp.get_section_dict("A") == orig_dict["A"]
+    assert inp.get_section_dict("/A") == orig_dict["A"]
+    assert inp.get_section_dict(["A"]) == orig_dict["A"]
+    with pytest.raises(TypeError):
+        inp.get_section_dict("FOO")