Docs update, forcefield elastic convenience maker, forcefield enum hydration (materialsproject#1072)

* update EOS docs

* udpate docs with implementation details

* add convenience constructor method for forcefield ElasticMaker

* allow MLFF enum to treat str(MLFF.) as valid member

* precommit

* add small MLFF test

Author: esoteric-ephemera

docs/user/codes/vasp.md

@@ -270,9 +270,39 @@ Afterwards, equation of state fits are performed with phonopy.
 
 ### Equation of State Workflow
 
-An equation of state workflow is implemented. First, a tight relaxation is performed. Subsequently, several optimizations at different constant
+An equation of state (EOS) workflow is implemented. First, a tight relaxation is performed. Subsequently, several optimizations at different constant
 volumes are performed. Additional static calculations might be performed afterwards to arrive at more
-accurate energies. Then, an equation of state fit is performed with pymatgen.
+accurate energies. Then, an EOS fit is performed with pymatgen.
+
+The output of the workflow is, by default, a dictionary containing the energy and volume data generated with DFT, in addition to fitted equation of state parameters for all models currently available in pymatgen (Murnaghan, Birch-Murnaghan, Poirier-Tarantola, and Vinet/UBER).
+
+#### Materials Project-compliant workflows
+
+If the user wishes to reproduce the EOS data currently in the Materials Project, they should use the atomate 1-compatible `MPLegacy`-prefixed flows (and jobs and input sets). For performing updated PBE-GGA EOS flows with Materials Project-compliant parameters, the user should use the `MPGGA`-prefixed classes. Lastly, the `MPMetaGGA`-prefixed classes allow the user to perform Materials Project-compliant r<sup>2</sup>SCAN EOS workflows.
+
+**Summary:** For Materials Project-compliant equation of state (EOS) workflows, the user should use:
+* `MPGGAEosMaker` for faster, lower-accuracy calculation with the PBE-GGA
+* `MPMetaGGAEosMaker` for higher-accuracy but slower calculations with the r<sup>2</sup>SCAN meta-GGA
+* `MPLegacyEosMaker` for consistency with the PBE-GGA data currently distributed by the Materials Project
+
+#### Implementation details
+
+The Materials Project-compliant EOS flows, jobs, and sets currently use three prefixes to indicate their usage.
+* `MPGGA`: MP-compatible PBE-GGA (current)
+* `MPMetaGGA`: MP-compatible r<sup>2</sup>SCAN meta-GGA (current)
+* `MPLegacy`: a reproduction of the atomate 1 implementation, described in
+  K. Latimer, S. Dwaraknath, K. Mathew, D. Winston, and K.A. Persson, npj Comput. Materials **vol. 4**, p. 40 (2018), DOI: 10.1038/s41524-018-0091-x
+
+  For reference, the original atomate workflows can be found here:
+    * [`atomate.vasp.workflows.base.wf_bulk_modulus`](https://github.com/hackingmaterials/atomate/blob/main/atomate/vasp/workflows/presets/core.py#L564)
+    * [`atomate.vasp.workflows.base.bulk_modulus.get_wf_bulk_modulus`](https://github.com/hackingmaterials/atomate/blob/main/atomate/vasp/workflows/base/bulk_modulus.py#L21)
+
+In the original atomate 1 workflow and the atomate2 `MPLegacyEosMaker`, the k-point density is **extremely** high. This is despite the convergence tests in the supplementary information
+of Latimer *et al.* not showing strong sensitivity when the "number of ***k***-points per reciprocal atom" (KPPRA) is at least 3,000.
+
+To make the `MPGGAEosMaker` and `MPMetaGGAEosMaker` more tractable for high-throughput jobs, their input sets (`MPGGAEos{Relax,Static}SetGenerator` and `MPMetaGGAEos{Relax,Static}SetGenerator` respectively) still use the highest ***k***-point density in standard Materials Project jobs, `KSPACING = 0.22` Å<sup>-1</sup>, which is comparable to KPPRA = 3,000.
+
+This choice is justified by Fig. S12 of the supplemantary information of Latimer *et al.*, which shows that all fitted EOS parameters (equilibrium energy $E_0$, equilibrium volume $V_0$, bulk modulus $B_0$, and bulk modulus pressure derivative $B_1$) do not deviate by more than 1.5%, and typically by less than 0.1%, from well-converged values when KPPRA = 3,000.
 
 ### LOBSTER
 

src/atomate2/forcefields/__init__.py

@@ -3,6 +3,10 @@
 from __future__ import annotations
 
 from enum import Enum
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from typing import Any
 
 
 class MLFF(Enum):  # TODO inherit from StrEnum when 3.11+
@@ -17,6 +21,16 @@ class MLFF(Enum):  # TODO inherit from StrEnum when 3.11+
     Nequip = "Nequip"
     SevenNet = "SevenNet"
 
+    @classmethod
+    def _missing_(cls, value: Any) -> Any:
+        """Allow input of str(MLFF) as valid enum."""
+        if isinstance(value, str):
+            value = value.split("MLFF.")[-1]
+        for member in cls:
+            if member.value == value:
+                return member
+        return None
+
 
 def _get_formatted_ff_name(force_field_name: str | MLFF) -> str:
     """

src/atomate2/forcefields/flows/elastic.py

@@ -3,11 +3,24 @@
 from __future__ import annotations
 
 from dataclasses import dataclass, field
+from typing import TYPE_CHECKING
 
 from atomate2 import SETTINGS
 from atomate2.common.flows.elastic import BaseElasticMaker
+from atomate2.forcefields import MLFF, _get_formatted_ff_name
 from atomate2.forcefields.jobs import ForceFieldRelaxMaker
 
+if TYPE_CHECKING:
+    from typing import Any
+
+    from typing_extensions import Self
+
+# default options for the forcefield makers in ElasticMaker
+_DEFAULT_RELAX_KWARGS: dict[str, Any] = {
+    "force_field_name": "CHGNet",
+    "relax_kwargs": {"fmax": 0.00001},
+}
+
 
 @dataclass
 class ElasticMaker(BaseElasticMaker):
@@ -62,16 +75,12 @@ class ElasticMaker(BaseElasticMaker):
     symprec: float = SETTINGS.SYMPREC
     bulk_relax_maker: ForceFieldRelaxMaker | None = field(
         default_factory=lambda: ForceFieldRelaxMaker(
-            force_field_name="CHGNet",
-            relax_cell=True,
-            relax_kwargs={"fmax": 0.00001},
+            relax_cell=True, **_DEFAULT_RELAX_KWARGS
         )
     )
     elastic_relax_maker: ForceFieldRelaxMaker | None = field(
         default_factory=lambda: ForceFieldRelaxMaker(
-            force_field_name="CHGNet",
-            relax_cell=False,
-            relax_kwargs={"fmax": 0.00001},
+            relax_cell=False, **_DEFAULT_RELAX_KWARGS
         )
     )  # constant volume relaxation
     max_failed_deformations: int | float | None = None
@@ -89,3 +98,44 @@ def prev_calc_dir_argname(self) -> str | None:
         Note: this is only applicable if a relax_maker is specified; i.e., two
         calculations are performed for each ordering (relax -> static)
         """
+
+    @classmethod
+    def from_force_field_name(
+        cls,
+        force_field_name: str | MLFF,
+        mlff_kwargs: dict | None = None,
+        **kwargs,
+    ) -> Self:
+        """
+        Create an elastic flow from a forcefield name.
+
+        Parameters
+        ----------
+        force_field_name : str or .MLFF
+            The name of the force field.
+        mlff_kwargs : dict or None (default)
+            kwargs to pass to `ForceFieldRelaxMaker`.
+        **kwargs
+            Additional kwargs to pass to ElasticMaker.
+
+        Returns
+        -------
+        ElasticMaker
+        """
+        default_kwargs: dict[str, Any] = {
+            **_DEFAULT_RELAX_KWARGS,
+            **(mlff_kwargs or {}),
+            "force_field_name": _get_formatted_ff_name(force_field_name),
+        }
+        return cls(
+            name=f"{str(force_field_name).split('MLFF.')[-1]} elastic",
+            bulk_relax_maker=ForceFieldRelaxMaker(
+                relax_cell=True,
+                **default_kwargs,
+            ),
+            elastic_relax_maker=ForceFieldRelaxMaker(
+                relax_cell=False,
+                **default_kwargs,
+            ),
+            **kwargs,
+        )

src/atomate2/forcefields/flows/eos.py

@@ -1,4 +1,4 @@
-"""Flows to generate EOS fits using CHGNet, M3GNet, or MACE."""
+"""Flows to generate EOS fits using machine learned interatomic potentials."""
 
 from __future__ import annotations
 
@@ -62,6 +62,7 @@ def from_force_field_name(
         cls,
         force_field_name: str | MLFF,
         relax_initial_structure: bool = True,
+        **kwargs,
     ) -> Self:
         """
         Create an EOS flow from a forcefield name.
@@ -72,6 +73,9 @@ def from_force_field_name(
             The name of the force field.
         relax_initial_structure: bool = True
             Whether to relax the initial structure before performing an EOS fit.
+        **kwargs
+            Additional kwargs to pass to ElasticMaker
+
 
         Returns
         -------
@@ -89,6 +93,7 @@ def from_force_field_name(
                 force_field_name=force_field_name, relax_cell=False
             ),
             static_maker=None,
+            **kwargs,
         )
 
 

tests/forcefields/flows/test_elastic.py

@@ -7,7 +7,10 @@
 from atomate2.forcefields.jobs import ForceFieldRelaxMaker
 
 
-def test_elastic_wf_with_mace(clean_dir, si_structure, test_dir):
+@pytest.mark.parametrize("convenience_constructor", [True, False])
+def test_elastic_wf_with_mace(
+    clean_dir, si_structure, test_dir, convenience_constructor: bool
+):
     si_prim = SpacegroupAnalyzer(si_structure).get_primitive_standard_structure()
     model_path = f"{test_dir}/forcefields/mace/MACE.model"
     common_kwds = {
@@ -16,10 +19,17 @@ def test_elastic_wf_with_mace(clean_dir, si_structure, test_dir):
         "relax_kwargs": {"fmax": 0.00001},
     }
 
-    flow = ElasticMaker(
-        bulk_relax_maker=ForceFieldRelaxMaker(**common_kwds, relax_cell=True),
-        elastic_relax_maker=ForceFieldRelaxMaker(**common_kwds, relax_cell=False),
-    ).make(si_prim)
+    if convenience_constructor:
+        common_kwds.pop("force_field_name")
+        flow = ElasticMaker.from_force_field_name(
+            force_field_name="MACE",
+            mlff_kwargs=common_kwds,
+        ).make(si_prim)
+    else:
+        flow = ElasticMaker(
+            bulk_relax_maker=ForceFieldRelaxMaker(**common_kwds, relax_cell=True),
+            elastic_relax_maker=ForceFieldRelaxMaker(**common_kwds, relax_cell=False),
+        ).make(si_prim)
 
     # run the flow or job and ensure that it finished running successfully
     responses = run_locally(flow, create_folders=True, ensure_success=True)

tests/forcefields/test_utils.py

@@ -4,6 +4,12 @@
 from atomate2.forcefields.utils import ase_calculator
 
 
+@pytest.mark.parametrize(("force_field"), [mlff.value for mlff in MLFF])
+def test_mlff(force_field: str):
+    mlff = MLFF(force_field)
+    assert mlff == MLFF(str(mlff)) == MLFF(str(mlff).split(".")[-1])
+
+
 @pytest.mark.parametrize(("force_field"), ["CHGNet", "MACE"])
 def test_ext_load(force_field: str):
     decode_dict = {