rsspolymlp 0.4.5__tar.gz → 0.4.6__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (81) hide show
  1. {rsspolymlp-0.4.5/src/rsspolymlp.egg-info → rsspolymlp-0.4.6}/PKG-INFO +1 -1
  2. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/pyproject.toml +1 -1
  3. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/rss_summarize.py +5 -5
  4. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/gen_struct_rep.py +28 -22
  5. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/identify_uniqstr.py +13 -7
  6. rsspolymlp-0.4.6/src/rsspolymlp/analysis/struct_matcher/niggli_reduce.py +203 -0
  7. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/primitive_transform.py +27 -42
  8. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/redrep_position.py +104 -22
  9. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/struct_match.py +5 -5
  10. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/cli_rsspolymlp_utils.py +5 -12
  11. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/rsspolymlp.py +2 -2
  12. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/rsspolymlp_utils.py +12 -12
  13. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/common/interface_vasp.py +39 -0
  14. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/rss/load_logfile.py +5 -0
  15. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/rss/optimization_mlp.py +71 -97
  16. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/rss/random_struct.py +7 -2
  17. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6/src/rsspolymlp.egg-info}/PKG-INFO +1 -1
  18. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp.egg-info/SOURCES.txt +1 -2
  19. rsspolymlp-0.4.5/src/rsspolymlp/analysis/struct_matcher/redrep_lattice.py +0 -143
  20. rsspolymlp-0.4.5/src/rsspolymlp/utils/vasp_util/write_poscar.py +0 -49
  21. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/LICENSE +0 -0
  22. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/README.md +0 -0
  23. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/setup.cfg +0 -0
  24. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/__init__.py +0 -0
  25. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/__init__.py +0 -0
  26. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/eos.py +0 -0
  27. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/ghost_minima.py +0 -0
  28. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/phase_analysis.py +0 -0
  29. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/__init__.py +0 -0
  30. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/assignment_algo.py +0 -0
  31. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/chiral_spg.py +0 -0
  32. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/struct_matcher/test/gen_lattice.py +0 -0
  33. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/analysis/utils/convert_data_format.py +0 -0
  34. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/__init__.py +0 -0
  35. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/cli_rsspolymlp.py +0 -0
  36. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/cli_rsspolymlp_devkit.py +0 -0
  37. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/cli_rsspolymlp_plot.py +0 -0
  38. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/rsspolymlp_devkit.py +0 -0
  39. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/api/rsspolymlp_plot.py +0 -0
  40. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/common/__init__.py +0 -0
  41. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/common/atomic_energy.py +0 -0
  42. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/common/composition.py +0 -0
  43. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/common/interface_pypolymlp.py +0 -0
  44. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/common/interface_spglib.py +0 -0
  45. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/common/property.py +0 -0
  46. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/dataset/compress_vasprun.py +0 -0
  47. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/dataset/gen_mlp_dataset.py +0 -0
  48. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/dataset/split_dataset.py +0 -0
  49. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/estimate_cost.py +0 -0
  50. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/model_selection/binary/gen_single_model.py +0 -0
  51. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/model_selection/elemental/gen_hybrid_model.py +0 -0
  52. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/model_selection/elemental/gen_single_model.py +0 -0
  53. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/model_selection/n_feature.py +0 -0
  54. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/model_selection/pypolymlp_gridsearch.py +0 -0
  55. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/model_selection/reduce_grid.py +0 -0
  56. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/pareto_opt_mlp.py +0 -0
  57. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/mlp_dev/polymlp_dev.py +0 -0
  58. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/rss/__init__.py +0 -0
  59. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/rss/eliminate_duplicates.py +0 -0
  60. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/__init__.py +0 -0
  61. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/lammps_utils.py +0 -0
  62. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/matplot_util/__init__.py +0 -0
  63. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/matplot_util/custom_plt.py +0 -0
  64. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/matplot_util/examples/__init__.py +0 -0
  65. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/matplot_util/examples/example.py +0 -0
  66. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/matplot_util/examples/template.py +0 -0
  67. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/matplot_util/make_plot.py +0 -0
  68. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/opt_sscha/opt_geometry_sscha.py +0 -0
  69. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/opt_sscha/opt_position_sscha.py +0 -0
  70. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/opt_sscha/run_opt_sscha.py +0 -0
  71. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/opt_sscha/sscha_property.py +0 -0
  72. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/pymatgen_utils.py +0 -0
  73. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/vasp_util/__init__.py +0 -0
  74. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/vasp_util/api.py +0 -0
  75. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/vasp_util/gen_incar.py +0 -0
  76. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp/utils/vasp_util/gen_script.py +0 -0
  77. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp.egg-info/dependency_links.txt +0 -0
  78. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp.egg-info/entry_points.txt +0 -0
  79. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp.egg-info/requires.txt +0 -0
  80. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/src/rsspolymlp.egg-info/top_level.txt +0 -0
  81. {rsspolymlp-0.4.5 → rsspolymlp-0.4.6}/tests/test_cli_help.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: rsspolymlp
3
- Version: 0.4.5
3
+ Version: 0.4.6
4
4
  Summary: A framework for random structure search using polynomial MLPs
5
5
  Author-email: Hayato Wakai <wakai@cms.mtl.kyoto-u.ac.jp>
6
6
  Maintainer-email: Hayato Wakai <wakai@cms.mtl.kyoto-u.ac.jp>
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
4
4
 
5
5
  [project]
6
6
  name = "rsspolymlp"
7
- version = "0.4.5"
7
+ version = "0.4.6"
8
8
  description = "A framework for random structure search using polynomial MLPs"
9
9
  license = "MIT"
10
10
  authors = [
@@ -38,7 +38,7 @@ class RSSResultSummarizer:
38
38
  num_process: int = -1,
39
39
  backend: str = "loky",
40
40
  assign_method: Literal["redrep", "greedy", "HK", "pymatgen"] = "redrep",
41
- primitive_symprecs: list[float] = [1e-2],
41
+ tol_sm: list[float] = [1e-2],
42
42
  output_poscar: bool = False,
43
43
  thresholds: list[float] = None,
44
44
  parse_vasp: bool = False,
@@ -52,7 +52,7 @@ class RSSResultSummarizer:
52
52
  self.num_process = num_process
53
53
  self.backend = backend
54
54
  self.assign_method = assign_method
55
- self.primitive_symprecs = primitive_symprecs
55
+ self.tol_sm = tol_sm
56
56
  self.output_poscar = output_poscar
57
57
  self.thresholds = thresholds
58
58
  self.parse_vasp = parse_vasp
@@ -124,7 +124,7 @@ class RSSResultSummarizer:
124
124
  axis_tol=axis_tol,
125
125
  pos_tol=pos_tol,
126
126
  assign_method=self.assign_method,
127
- primitive_symprecs=self.primitive_symprecs,
127
+ tol_sm=self.tol_sm,
128
128
  num_process=self.num_process,
129
129
  pre_analyzer=self.analyzer,
130
130
  )
@@ -220,7 +220,7 @@ class RSSResultSummarizer:
220
220
  axis_tol=axis_tol,
221
221
  pos_tol=pos_tol,
222
222
  assign_method=self.assign_method,
223
- primitive_symprecs=self.primitive_symprecs,
223
+ tol_sm=self.tol_sm,
224
224
  num_process=self.num_process,
225
225
  pre_analyzer=self.analyzer,
226
226
  )
@@ -281,7 +281,7 @@ class RSSResultSummarizer:
281
281
 
282
282
  print(" - Converting reduced crystal structure representation...")
283
283
  generator = StructRepGenerator(
284
- primitive_symprecs=self.primitive_symprecs,
284
+ tol_sm=self.tol_sm,
285
285
  )
286
286
  struct_reps_list = generator.generate_struct_reps_parallel(
287
287
  rss_results,
@@ -7,6 +7,7 @@ from typing import Optional
7
7
  import joblib
8
8
  import numpy as np
9
9
 
10
+ from rsspolymlp.analysis.struct_matcher.niggli_reduce import niggli_reduce
10
11
  from rsspolymlp.analysis.struct_matcher.primitive_transform import (
11
12
  PrimitiveCell,
12
13
  get_primitive_cell,
@@ -32,8 +33,8 @@ class StructReps:
32
33
  original_positions: np.ndarray
33
34
  original_elements: np.ndarray
34
35
  spg_list: list[str]
35
- primitive_symprecs: list[float]
36
- redrep_symprecs: list[float]
36
+ tol_sm: list[float]
37
+ tol_redpos: list[float]
37
38
  normalize_axis: bool
38
39
  struct_path: Optional[str]
39
40
  struct_tag: Optional[int]
@@ -44,13 +45,13 @@ class StructRepGenerator:
44
45
 
45
46
  def __init__(
46
47
  self,
47
- primitive_symprecs: list[float] = [1e-2],
48
- redrep_symprecs: list[float] = [1e-4, 1e-2, 1e-1],
48
+ tol_sm: list[float] = [1e-2],
49
+ tol_redpos: list[float] = [1e-4, 1e-2, 1e-1],
49
50
  normalize_axis: bool = False,
50
51
  verbose: bool = False,
51
52
  ):
52
- self.primitive_symprecs = primitive_symprecs
53
- self.redrep_symprecs = redrep_symprecs
53
+ self.tol_sm = tol_sm
54
+ self.tol_redpos = tol_redpos
54
55
  self.normalize_axis = normalize_axis
55
56
  self.verbose = verbose
56
57
 
@@ -101,8 +102,8 @@ class StructRepGenerator:
101
102
 
102
103
  struct_rep_list = []
103
104
  for i, primitive_st in enumerate(primitive_st_set):
104
- redrep_symprecs = sorted(
105
- self.redrep_symprecs,
105
+ tol_redpos = sorted(
106
+ self.tol_redpos,
106
107
  key=lambda x: x if isinstance(x, (int, float)) else sum(x) / len(x),
107
108
  )
108
109
  struct_rep, log_redrep = self.generate_struct_rep(
@@ -129,8 +130,8 @@ class StructRepGenerator:
129
130
  original_positions=positions,
130
131
  original_elements=np.asarray(elements, dtype=str),
131
132
  spg_list=spg_list,
132
- primitive_symprecs=self.primitive_symprecs,
133
- redrep_symprecs=redrep_symprecs,
133
+ tol_sm=self.tol_sm,
134
+ tol_redpos=tol_redpos,
134
135
  normalize_axis=self.normalize_axis,
135
136
  struct_path=struct_path,
136
137
  struct_tag=struct_tag,
@@ -150,16 +151,16 @@ class StructRepGenerator:
150
151
  spg_number_set = []
151
152
  iter_refine = 0
152
153
  time_refine = 0
153
- for symprec in self.primitive_symprecs:
154
+ for tol in self.tol_sm:
154
155
  t_start = time.time()
155
156
  primitive_cell = get_primitive_cell(
156
- axis=axis, positions=positions, elements=elements, tol=symprec
157
+ axis=axis, positions=positions, elements=elements, tol=tol
157
158
  )
158
159
  symutil = SymCell(
159
160
  primitive_cell.axis,
160
161
  primitive_cell.positions,
161
162
  primitive_cell.elements,
162
- symprec=symprec,
163
+ symprec=tol,
163
164
  )
164
165
  spg_str = symutil.get_spacegroup()
165
166
  if spg_str is None:
@@ -167,8 +168,15 @@ class StructRepGenerator:
167
168
  else:
168
169
  spg_number = int(re.search(r"\((\d+)\)", spg_str).group(1))
169
170
 
171
+ axis_niggli, positions_niggli = niggli_reduce(
172
+ axis=primitive_cell.axis, positions=primitive_cell.positions, tol=tol
173
+ )
174
+ primitive_cell.axis = axis_niggli
175
+ primitive_cell.positions = positions_niggli
176
+
170
177
  primitive_st_set.append(primitive_cell)
171
178
  spg_number_set.append(spg_number)
179
+
172
180
  iter_refine += 1
173
181
  time_refine += time.time() - t_start
174
182
 
@@ -189,28 +197,26 @@ class StructRepGenerator:
189
197
  "metric_tensors": [],
190
198
  "axis_abc": [],
191
199
  "positions": [],
192
- "symprec": [],
200
+ "tol": [],
193
201
  }
194
202
  all_reps = {
195
203
  "metric_tensors": [],
196
204
  "axis_abc": [],
197
205
  "positions": [],
198
- "symprec": [],
206
+ "tol": [],
199
207
  }
200
208
  iter_redrep = 0
201
209
  time_redrep = 0
202
210
 
203
- symprec_set = self.redrep_symprecs
204
- for symprec in symprec_set:
205
- if isinstance(symprec, float):
206
- input_symprec = [symprec] * 3
211
+ tol_set = self.tol_redpos
212
+ for tol in tol_set:
207
213
  t_start = time.time()
208
214
  reducer = PositionRepReducer(
209
215
  axis,
210
216
  positions,
211
217
  elements,
212
218
  spg_number,
213
- symprec=input_symprec,
219
+ tol=tol,
214
220
  normalize_axis=self.normalize_axis,
215
221
  verbose=self.verbose,
216
222
  )
@@ -241,12 +247,12 @@ class StructRepGenerator:
241
247
  all_reps["metric_tensors"].append(metric_tensor)
242
248
  all_reps["axis_abc"].append(axis_abc)
243
249
  all_reps["positions"].append(all_posrep)
244
- all_reps["symprec"].append(symprec)
250
+ all_reps["tol"].append(tol)
245
251
  if unique_metric or unique_pos:
246
252
  reduced_reps["metric_tensors"].append(metric_tensor)
247
253
  reduced_reps["axis_abc"].append(axis_abc)
248
254
  reduced_reps["positions"].append(red_posrep)
249
- reduced_reps["symprec"].append(symprec)
255
+ reduced_reps["tol"].append(tol)
250
256
 
251
257
  self.iter_redrep += iter_redrep
252
258
  self.time_redrep += time_redrep
@@ -99,9 +99,15 @@ class UniqueStructIdentifier:
99
99
  pos_tol=self.pos_tol,
100
100
  )
101
101
  elif self.assign_method == "pymatgen":
102
- st1 = self.pymat.polymlp_to_pymat_st(uniq_str.original_polymlp_st)
103
- st2 = self.pymat.polymlp_to_pymat_st(
104
- struct_reps.original_polymlp_st
102
+ st1 = self.pymat.parameter_to_pymat_st(
103
+ uniq_str.original_axis,
104
+ uniq_str.original_positions,
105
+ uniq_str.original_elements,
106
+ )
107
+ st2 = self.pymat.parameter_to_pymat_st(
108
+ struct_reps.original_axis,
109
+ struct_reps.original_positions,
110
+ struct_reps.original_elements,
105
111
  )
106
112
  judge_duplicate = self.pymat.match_str(st1, st2)
107
113
  else:
@@ -180,8 +186,8 @@ def identify_unique_structure(
180
186
  pos_tol: float = 0.01,
181
187
  assign_method: Literal["redrep", "greedy", "HK", "pymatgen"] = "redrep",
182
188
  keep_unique: bool = False,
183
- primitive_symprecs: list[float] = [1e-5, 1e-4, 1e-3, 1e-2],
184
- redrep_symprecs: list[float] = [1e-4, 1e-2, 1e-1],
189
+ tol_sm: list[float] = [1e-2],
190
+ tol_redpos: list[float] = [1e-4, 1e-2, 1e-1],
185
191
  normalize_axis: bool = False,
186
192
  num_process: int = -1,
187
193
  backend: str = "loky",
@@ -215,8 +221,8 @@ def identify_unique_structure(
215
221
  if verbose:
216
222
  print(" - Converting reduced crystal structure representation...")
217
223
  generator = StructRepGenerator(
218
- primitive_symprecs=primitive_symprecs,
219
- redrep_symprecs=redrep_symprecs,
224
+ tol_sm=tol_sm,
225
+ tol_redpos=tol_redpos,
220
226
  normalize_axis=normalize_axis,
221
227
  )
222
228
  redreps_list = generator.generate_struct_reps_parallel(
@@ -0,0 +1,203 @@
1
+ import numpy as np
2
+
3
+
4
+ def niggli_reduce(
5
+ axis: np.ndarray,
6
+ positions: np.ndarray,
7
+ tol: float = 1e-3,
8
+ ):
9
+ """Niggli reduction."""
10
+ axis = np.array(axis)
11
+ positions = np.array(positions)
12
+
13
+ axis_niggli, P = get_niggli_reduced_cell(
14
+ axis,
15
+ tol=tol,
16
+ )
17
+ pos = positions @ np.linalg.inv(P)
18
+ pos %= 1.0
19
+
20
+ assert abs(round(np.linalg.det(P))) == 1
21
+ return axis_niggli, pos
22
+
23
+
24
+ def get_niggli_reduced_cell(axis, tol: float = 1e-5):
25
+ """
26
+ Get the Niggli-reduced cell using a numerically stable algorithm.
27
+
28
+ References
29
+ ----------
30
+ 1. I. Krivy and B. Gruber, Acta Cryst. A 32, 297-298 (1976).
31
+ 2. R. W. Grosse-Kunstleve, N. K. Sauter, and P. D. Adams, Acta Cryst. A 60, 1-6 (2004).
32
+
33
+ Parameters
34
+ ----------
35
+ axis : (3, 3) array_like
36
+ Lattice vectors defining the unit cell in Cartesian coordinates.
37
+ Each row represents one lattice vector (a, b, or c).
38
+ Equivalent to np.array([[a1, a2, a3], [b1, b2, b3], [c1, c2, c3]]).
39
+ tol : float, default=1e-5
40
+ Numerical tolerance for the reduction procedure.
41
+
42
+ Returns
43
+ -------
44
+ axis_niggli : (3, 3) ndarray
45
+ Niggli-reduced lattice vectors, with each row representing one lattice vector.
46
+ """
47
+
48
+ def get_G_components(G):
49
+ return (
50
+ G[0, 0],
51
+ G[1, 1],
52
+ G[2, 2],
53
+ 2 * G[1, 2],
54
+ 2 * G[0, 2],
55
+ 2 * G[0, 1],
56
+ )
57
+
58
+ def convert_G_and_P(C, G, P):
59
+ G = C.T @ G @ C
60
+ P = P @ C
61
+ return G, P
62
+
63
+ G = np.dot(axis, axis.T)
64
+ P = np.eye(3, dtype=int)
65
+
66
+ norm_a = G[0, 0] ** 0.5
67
+ norm_b = G[1, 1] ** 0.5
68
+ norm_c = G[2, 2] ** 0.5
69
+ e_aa = tol * 2 * norm_a
70
+ e_bb = tol * 2 * norm_b
71
+ e_cc = tol * 2 * norm_c
72
+ e_bc = tol * (norm_b + norm_c)
73
+ e_ac = tol * (norm_a + norm_c)
74
+ e_ab = tol * (norm_a + norm_b)
75
+
76
+ iteration = 100
77
+ for n_iter in range(iteration):
78
+ if n_iter == iteration - 1:
79
+ raise RuntimeError("Niggli reduction did not converge.")
80
+
81
+ aa, bb, cc, bc, ac, ab = get_G_components(G)
82
+
83
+ # A1: Main condition: aa <= bb; Special condition: |bc| < |ac|
84
+ if bb + e_aa + e_bb < aa or (
85
+ abs(aa - bb) < e_aa + e_bb and abs(bc) > abs(ac) + e_bc + e_ac
86
+ ):
87
+ C = np.array([[0, -1, 0], [-1, 0, 0], [0, 0, -1]])
88
+ G, P = convert_G_and_P(C, G, P)
89
+ aa, bb, cc, bc, ac, ab = get_G_components(G)
90
+
91
+ # A2: Main condition: bb <= cc; Special condition: |ac| < |ab|
92
+ if cc + e_bb + e_cc < bb or (
93
+ abs(bb - cc) < e_bb + e_cc and abs(ac) > abs(ab) + e_ac + e_ab
94
+ ):
95
+ C = np.array([[-1, 0, 0], [0, 0, -1], [0, -1, 0]])
96
+ G, P = convert_G_and_P(C, G, P)
97
+ continue
98
+
99
+ ell = 0 if abs(bc) < e_bc else bc / abs(bc)
100
+ m = 0 if abs(ac) < e_ac else ac / abs(ac)
101
+ n = 0 if abs(ab) < e_ab else ab / abs(ab)
102
+
103
+ # A3: Main condition: bc > 0, ac > 0, ab > 0 (Type I)
104
+ if ell * m * n == 1:
105
+ i = -1 if ell == -1 else 1
106
+ j = -1 if m == -1 else 1
107
+ k = -1 if n == -1 else 1
108
+ C = np.diag((i, j, k))
109
+ G, P = convert_G_and_P(C, G, P)
110
+
111
+ # A4: Main condition: bc <= 0, ac <= 0, ab <= 0 (Type II)
112
+ if ell * m * n in (0, -1):
113
+ i = -1 if ell == 1 else 1
114
+ j = -1 if m == 1 else 1
115
+ k = -1 if n == 1 else 1
116
+ if i * j * k == -1:
117
+ # Keep the proper rotation
118
+ if n == 0:
119
+ k = -1
120
+ elif m == 0:
121
+ j = -1
122
+ elif ell == 0:
123
+ i = -1
124
+ C = np.diag((i, j, k))
125
+ G, P = convert_G_and_P(C, G, P)
126
+
127
+ aa, bb, cc, bc, ac, ab = get_G_components(G)
128
+
129
+ # A5: Main condition: |bc| <= bb;
130
+ # Special conditions: if bc = bb: ab/2 <= ac,
131
+ # elif bc = -bb: ab = 0 (otherwise, convert to Type I)
132
+ if (
133
+ abs(bc) > bb + e_bc + e_bb
134
+ or (abs(bc - bb) < e_bc + e_bb and ab - e_ab - e_ac > 2 * ac)
135
+ or (abs(bc + bb) < e_bc + e_bb and -e_ab > ab)
136
+ ):
137
+ # Basis transformation: c' = c - sign(bc) * b
138
+ # cc' = cc + bb - |bc|
139
+ # bc' = bc/2 - sign(bc) * bb
140
+ # ac' = ac/2 - sign(bc) * ab/2
141
+ # If bc = bb and ab > 2 * ac:
142
+ # The transformed cell is Type I and satisfies |ab'| / 2 <= |ac'|,
143
+ # because 2 * |ac'| = 2 * |ac - ab| > ab.
144
+ # If bc = -bb and ab < 0:
145
+ # ac' = ac + ab < 0 and bc' = bb > 0.
146
+ # The transformed cell is Type I.
147
+ # A6 and A7 follow by analogous arguments.
148
+ C = np.array([[1, 0, 0], [0, 1, -bc / abs(bc)], [0, 0, 1]])
149
+ G, P = convert_G_and_P(C, G, P)
150
+ continue
151
+
152
+ # A6: Main condition: |ac| <= aa;
153
+ # Special conditions: if ac = aa: ab/2 <= bc,
154
+ # elif ac = -aa: ab = 0 (otherwise, convert to Type I)
155
+ if (
156
+ abs(ac) > aa + e_aa + e_ac
157
+ or (abs(aa - ac) < e_aa + e_ac and ab - e_ab - e_bc > 2 * bc)
158
+ or (abs(aa + ac) < e_aa + e_ac and -e_ab > ab)
159
+ ):
160
+ # Basis transformation: c' = c - sign(ac) * a
161
+ # cc' = cc + aa - |ac|
162
+ # ac' = ac/2 - sign(ac) * aa
163
+ # bc' = bc/2 - sign(ac) * ab/2
164
+ C = np.array([[1, 0, -ac / abs(ac)], [0, 1, 0], [0, 0, 1]])
165
+ G, P = convert_G_and_P(C, G, P)
166
+ continue
167
+
168
+ # A7: Main condition: |ab| <= aa;
169
+ # Special conditions: if ab = aa: ac/2 <= bc,
170
+ # elif ab = -aa: ac = 0 (otherwise, convert to Type I)
171
+ if (
172
+ abs(ab) > aa + e_aa + e_ab
173
+ or (abs(aa - ab) < e_aa + e_ab and ac - e_ac - e_bc > 2 * bc)
174
+ or (abs(aa + ab) < e_aa + e_ab and -e_ac > ac)
175
+ ):
176
+ # Basis transformation: b' = b - sign(ab) * a
177
+ # bb' = bb + aa - |ab|
178
+ # ab' = ab/2 - sign(ab) * aa
179
+ # bc' = bc/2 - sign(ab) * ac/2
180
+ C = np.array([[1, -ab / abs(ab), 0], [0, 1, 0], [0, 0, 1]])
181
+ G, P = convert_G_and_P(C, G, P)
182
+ continue
183
+
184
+ # A8: Main condition: |bc| + |ac| + |ab| <= aa + bb (Type II);
185
+ # Special condition: if |bc| + |ac| + |ab| = aa + bb: aa <= |ac| + |ab/2|
186
+ # (otherwise, convert to Type I)
187
+ if -(e_bc + e_ac + e_ab + e_aa + e_bb) > bc + ac + ab + aa + bb or (
188
+ abs(bc + ac + ab + aa + bb)
189
+ < (e_bc + e_ac + e_ab + e_aa + e_bb)
190
+ < ab + (aa + ac) * 2
191
+ ):
192
+ # Basis transformation: c' = a + b + c
193
+ # cc' = cc + aa + bb + ab + ac + bc
194
+ # ac' = 2 * aa + ab + ac (special condition: ac' > 0)
195
+ # bc' = 2 * bb + ab + bc (special condition: bc' < 0)
196
+ C = np.array([[1, 0, 1], [0, 1, 1], [0, 0, 1]])
197
+ G, P = convert_G_and_P(C, G, P)
198
+ continue
199
+
200
+ break
201
+
202
+ axis_niggli = P.T @ axis
203
+ return axis_niggli, P.T
@@ -1,16 +1,4 @@
1
- """Recover a primitive-cell transformation from translational symmetry.
2
-
3
- Convention:
4
- axis stores lattice vectors as rows.
5
- L_primitive = inv(M) @ L_original
6
- frac_primitive = frac_original @ M
7
-
8
- The implementation is intended for structures whose translational symmetry is
9
- exact up to a user-specified Cartesian tolerance. It finds candidate
10
- translations by mapping one reference atom to atoms of the same elements,
11
- checks each candidate against the complete structure, and reconstructs M using
12
- an exact integer row-lattice HNF calculation.
13
- """
1
+ """Get a primitive cell from translational symmetry."""
14
2
 
15
3
  from dataclasses import dataclass
16
4
  from itertools import product
@@ -27,12 +15,12 @@ from sympy.matrices.normalforms import hermite_normal_form
27
15
  class PrimitiveCell:
28
16
  """Result of primitive-cell transformation recovery."""
29
17
 
30
- M: np.ndarray # integer 3x3 matrix
31
- index: int # |det(M)|, number of original cells in primitive cell
32
- translations: np.ndarray # valid translations in original fractional coordinates
33
18
  axis: np.ndarray # inv(M) @ original lattice matrix
34
19
  positions: np.ndarray # unique sites in primitive-cell fractional coordinates
35
20
  elements: np.ndarray # elements corresponding to primitive_positions
21
+ M: np.ndarray # integer 3x3 matrix
22
+ index: int # |det(M)|, number of original cells in primitive cell
23
+ translations: np.ndarray # valid translations in original fractional coordinates
36
24
  mapped_positions: (
37
25
  np.ndarray
38
26
  ) # all original sites transformed by M, wrapped to [0, 1)
@@ -43,9 +31,8 @@ def get_primitive_cell(
43
31
  positions: np.ndarray,
44
32
  elements: np.ndarray | list[str],
45
33
  tol: float = 1e-2,
46
- translation_tol: float | None = None,
47
34
  integer_tol: float = 1e-2,
48
- verbose: bool = True,
35
+ verbose: bool = False,
49
36
  ) -> PrimitiveCell:
50
37
  """Find a primitive-cell transformation from lattice and atom arrays.
51
38
 
@@ -59,8 +46,6 @@ def get_primitive_cell(
59
46
  Length-N sequence identifying each atom elements (e.g. ``["Si", "Si"]``).
60
47
  tol
61
48
  Cartesian distance tolerance in angstrom for atom matching.
62
- translation_tol
63
- Cartesian tolerance for deduplicating candidate translations.
64
49
  integer_tol
65
50
  Tolerance for checking that ``index * translation`` is integer-valued.
66
51
 
@@ -72,20 +57,12 @@ def get_primitive_cell(
72
57
 
73
58
  Notes
74
59
  -----
75
- - This implementation assumes the input structure is translationally
76
- periodic without changing elements.
77
- - For disordered/partial-occupancy sites, convert to an explicit ordered
78
- representation before using this function.
79
60
  - For strongly skewed, unreduced lattices, the +/-1 image replication used
80
61
  by the KD-tree may be insufficient; reduce the lattice first or replace
81
62
  the image search with an exact nearest-image routine.
82
63
  """
83
64
  if tol <= 0:
84
65
  raise ValueError("tol must be positive.")
85
- if translation_tol is None:
86
- translation_tol = tol
87
- if translation_tol <= 0:
88
- raise ValueError("translation_tol must be positive.")
89
66
 
90
67
  axis = np.asarray(axis, dtype=float)
91
68
  positions = np.mod(np.asarray(positions, dtype=float), 1.0)
@@ -96,9 +73,6 @@ def get_primitive_cell(
96
73
  raise ValueError("positions must have shape (N, 3).")
97
74
  if len(elements) != len(positions):
98
75
  raise ValueError("elements and positions must have the same length.")
99
- n_sites = len(positions)
100
- if n_sites == 0:
101
- raise ValueError("Cannot analyze an empty structure.")
102
76
 
103
77
  # Select a reference atom from the least-populated elements to minimize the
104
78
  # number of translation candidates.
@@ -117,10 +91,14 @@ def get_primitive_cell(
117
91
  for translation in candidate_translations:
118
92
  shifted_positions = np.mod(positions + translation, 1.0)
119
93
  shifted_cart = shifted_positions @ axis
120
- if not _translation_is_valid(shifted_cart, elements, trees, tol):
94
+ is_valid, mean_translation = _translation_is_valid(
95
+ shifted_cart, elements, trees, tol
96
+ )
97
+ if not is_valid:
121
98
  continue
99
+ translation += mean_translation @ np.linalg.inv(axis)
122
100
  duplicate = any(
123
- _periodic_translation_distance(translation, old, axis) <= translation_tol
101
+ _periodic_translation_distance(translation, old, axis) <= tol
124
102
  for old in valid_translations
125
103
  )
126
104
  if not duplicate:
@@ -144,7 +122,7 @@ def get_primitive_cell(
144
122
  mapped_positions, elements, primitive_lattice, tol
145
123
  )
146
124
 
147
- expected_n_sites = n_sites // index
125
+ expected_n_sites = len(positions) // index
148
126
  if len(primitive_positions) != expected_n_sites:
149
127
  raise RuntimeError(
150
128
  "The recovered primitive cell has an unexpected number of unique sites: "
@@ -153,12 +131,12 @@ def get_primitive_cell(
153
131
  )
154
132
 
155
133
  return PrimitiveCell(
156
- M=M,
157
- index=index,
158
- translations=translations,
159
134
  axis=primitive_lattice,
160
135
  positions=primitive_positions,
161
136
  elements=primitive_elements,
137
+ M=M,
138
+ index=index,
139
+ translations=translations,
162
140
  mapped_positions=mapped_positions,
163
141
  )
164
142
 
@@ -188,19 +166,20 @@ def _translation_is_valid(
188
166
  elements: np.ndarray,
189
167
  trees: dict[str, tuple[cKDTree, np.ndarray]],
190
168
  tol: float,
191
- ) -> bool:
169
+ ) -> tuple[bool, np.ndarray]:
192
170
  """Check whether a translation maps all sites onto same-elements sites."""
171
+ translations = []
193
172
  for element in np.unique(elements):
194
173
  target_ids = np.flatnonzero(elements == element)
195
174
  tree, image_to_atom_id = trees[str(element)]
196
175
  target_count = len(target_ids)
176
+
197
177
  candidate_rows: list[int] = []
198
178
  candidate_cols: list[int] = []
199
-
200
179
  neighbor_lists = tree.query_ball_point(shifted_cart[target_ids], r=tol)
201
180
  for shifted_atom_id, match_image_idxs in enumerate(neighbor_lists):
202
181
  if not match_image_idxs:
203
- return False
182
+ return False, np.array([])
204
183
  match_atom_ids = np.unique(
205
184
  image_to_atom_id[np.asarray(match_image_idxs, dtype=int)]
206
185
  )
@@ -216,8 +195,14 @@ def _translation_is_valid(
216
195
  )
217
196
  matching = maximum_bipartite_matching(graph, perm_type="column")
218
197
  if matching.shape[0] != target_count or np.any(matching < 0):
219
- return False
220
- return True
198
+ return False, np.array([])
199
+
200
+ matched_cart = shifted_cart[target_ids]
201
+ target_cart = shifted_cart[target_ids]
202
+ translations.extend(target_cart[matching] - matched_cart)
203
+
204
+ mean_translation = np.mean(translations, axis=0)
205
+ return True, mean_translation
221
206
 
222
207
 
223
208
  def _periodic_translation_distance(