surface-construct 0.12.4__tar.gz → 0.12.6__tar.gz

{surface_construct-0.12.4/surface_construct.egg-info → surface_construct-0.12.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: surface_construct
-Version: 0.12.4
+Version: 0.12.6
 Summary: Surface construction and surface reaction sampling tools.
 Home-page: https://gitee.com/pjren/surface_construct/
 Author: ren

{surface_construct-0.12.4 → surface_construct-0.12.6}/setup.py

@@ -15,7 +15,7 @@ install_requires = [
 
 setup(
     name='surface_construct',
-    version='0.12.4',
+    version='0.12.6',
     packages=find_packages(),
     url='https://gitee.com/pjren/surface_construct/',
     license='GPL',

{surface_construct-0.12.4 → surface_construct-0.12.6}/surface_construct/sg_sampler.py

@@ -145,6 +145,14 @@ class RandomSGSampler(SGSamplerBase):
         else:
             self.seed = None
 
+    @property
+    def _population(self):
+        """
+        排除已经采过的 index
+        :return:
+        """
+        return [i for i in range(self._pop_size) if i not in self.sg_obj.sample_idx]
+
     def _samples(self, size, **kwargs):
         idx = random.sample(self._population, size)
         return idx
@@ -157,11 +165,20 @@ class MaxSigmaSGSampler(SGSamplerBase):
     def _samples(self, size, **kwargs):
         if 'energy' in self.sg_obj.grid_property:
             # 如果已经读入了一些能量，则返回误差最大的点
-            sigma_array = self.sg_obj.grid_property['energy']
+            sigma_array = self.sg_obj.grid_property_sigma['energy']
             sigma0 = sigma_array.max()
-            idx_list = np.argwhere(sigma_array <= sigma0-0.1).flatten().tolist()
-            idx = random.sample(idx_list, size)
-            return idx
+            if len(sigma_array) < size:
+                raise ValueError("Number of points should greater than sample!")
+            elif len(sigma_array) == size:
+                return np.asarray(range(len(sigma_array)))
+            else:
+                delta = 0.01
+                idx_list = []
+                while len(idx_list) >= size:
+                    idx_list = np.argwhere(sigma_array > sigma0-delta).flatten().tolist()
+                    delta += 0.01
+                idx = random.sample(idx_list, size)
+                return idx
         else:
             raise "No energy for all population, pls do initial sampling first!"
 
@@ -173,6 +190,10 @@ class MinEnergySGSampler(SGSamplerBase):
     def _samples(self, size, **kwargs):
         if 'energy' in self.sg_obj.grid_property:
             E_array = self.sg_obj.grid_property['energy']
+            if len(E_array) < size:
+                raise ValueError("Number of points should greater than sample!")
+            elif len(E_array) == size:
+                return np.asarray(range(len(E_array)))
             # 如果已经读入了一些能量，则返回能量最低的点 (<0.1eV 以内，然后随机选一个)
             E0 = E_array.min()
             idx_list = np.argwhere(E_array <= E0+0.1).flatten().tolist()

{surface_construct-0.12.4 → surface_construct-0.12.6}/surface_construct/structures/surface_grid.py

@@ -770,13 +770,19 @@ class SurfaceGrid:
         :return:
         """
         if self._vector_interval is None:
-            nsample = 100
+            nsample = min(100, len(self.points)-1)  # 如果 points 被过滤了，就需要修改了。
             rng = np.random.default_rng()
-            idx_0 = rng.choice(range(len(self.points)-1), size=nsample)
+            if nsample < 10:
+                comb = itertools.combinations(range(len(self.points)-1), nsample)
+                idx_0 = rng.choice(list(comb))
+            else:
+                idx_0 = rng.choice(range(len(self.points)-1), size=nsample)
             idx_1 = idx_0+1
-            idx = np.asarray([[i,j] for i,j in zip(idx_0, idx_1) if (i not in idx_1 and j not in idx_0)])
+            idx = np.asarray([[i,j] for i,j in zip(idx_0, idx_1)])
             d_grid = np.linalg.norm(self.points[idx[:, 0]] - self.points[idx[:, 1]], axis=1)
-            idx = idx[d_grid<1.2 * self.interval]
+            tol = 1.2 * self.interval
+            if np.any(d_grid<tol):
+                idx = idx[d_grid<tol]
             d_vector = np.linalg.norm(self.vector[idx[:,0]]-self.vector[idx[:,1]], axis=1).mean()
             k = d_vector / self.interval
             self._vector_interval = np.min(k)
@@ -899,6 +905,11 @@ class SurfaceGrid:
         self.sample_dct[idx]['calculated'] = True
 
     def fit(self, key='energy'):
+        if len(self.points) == len(self.calculated_sample):
+            self.grid_property[key] = [self.sample_dct[idx][key] for idx in range(len(self.points))]
+            self.grid_property_sigma[key] = [0.] * len(self.points)
+            return
+
         from sklearn.gaussian_process import GaussianProcessRegressor
         from sklearn.gaussian_process.kernels import RBF, ConstantKernel, WhiteKernel
         from sklearn.preprocessing import StandardScaler
@@ -909,12 +920,14 @@ class SurfaceGrid:
         scaler.fit(y)
         y_scaled = scaler.transform(y).T[0]
         # RBF Kernel
-        length_scale_grid = 1.0  # 实空间的length scale, 1 angstrom
-        length_scale_vector = length_scale_grid * self.vector_interval
-        length_scale_bounds = (length_scale_vector/2.0, length_scale_vector*2.0)
-        # length_scale_bounds = 'fixed'
-        # length_scale = [length_scale_vector] * self._vector_dim  # 向量空间的 length_scale, anisotropic
-        length_scale = length_scale_vector  # 向量空间的 length_scale, isotropic
+        try:
+            length_scale_grid = 1.0  # 实空间的length scale, 1 angstrom
+            length_scale_vector = length_scale_grid * self.vector_interval
+            length_scale_bounds = (length_scale_vector/10.0, length_scale_vector*10.0)
+            length_scale = length_scale_vector  # 向量空间的 length_scale, isotropic
+        except IndexError:
+            length_scale = 1.0
+            length_scale_bounds = (1e-3, 1e3)
         rbf_kernel = RBF(length_scale=length_scale, length_scale_bounds=length_scale_bounds)
         # noise kernel
         noise_level_dict = {
@@ -928,7 +941,7 @@ class SurfaceGrid:
         noise_level_norm = np.abs(noise_level / scaler.scale_[0])  # 标准化缩放
         white_kernel = WhiteKernel(noise_level=noise_level_norm, noise_level_bounds='fixed')
         # 总的 kernel
-        constant_kernel = ConstantKernel(constant_value=1.0, constant_value_bounds='fixed')
+        constant_kernel = ConstantKernel(constant_value=1.0, constant_value_bounds=(0.1,10))
         kernel = constant_kernel * rbf_kernel + white_kernel
         gp = GaussianProcessRegressor(kernel=kernel, n_restarts_optimizer=9, alpha=noise_level_norm*0.01)
         print(f"Kernel parameters before fit:{kernel})")

{surface_construct-0.12.4 → surface_construct-0.12.6}/surface_construct/tasks/__init__.py

@@ -12,4 +12,5 @@
 4. 分析和作图
 """
 from .sitesampling import SurfaceSiteSampleTask
-from .afm import AFMTask
+from .afm import AFMTask
+from .vipsitetask import VIPSiteTask

{surface_construct-0.12.4 → surface_construct-0.12.6}/surface_construct/tasks/sitesampling.py

@@ -53,7 +53,6 @@ class SurfaceSiteSampleTask(TaskBase):
             self.sg_obj = combiner.sg_obj
             if self.sg_obj.points is None:
                 self.sg_obj.initialize()
-            self.to_pkl()  # 保存当前状态
 
         if self.sampler is None:
             self.sampler = [{'surface':'KeyPointSGSampler'}]  # 默认关键点采样
@@ -63,8 +62,6 @@ class SurfaceSiteSampleTask(TaskBase):
         self.grid_idx = None
         self.stime = datetime.datetime.now()
 
-        self.print_task_info()
-
     def print_task_info(self):
         msg = [f"Starting task {self.__class__.__name__} at "
                f"{datetime.datetime.ctime(self.stime)}.\n",
@@ -192,8 +189,9 @@ class SurfaceSiteSampleTask(TaskBase):
                             self.irun(grid_idx, fit=False)
                 else:
                     for i in range(size):
-                        grid_idx = sg_sampler_obj(self.sg_obj).samples(size=1)
-                        self.irun(grid_idx, fit=True)
+                        grid_idx = sg_sampler_obj(self.sg_obj).samples(size=1)[0]
+                        self.irun(grid_idx)
+                    self.sg_obj.fit()
             elif type(sg_sampler) in (tuple, list):
                 msg = f"  Appending sampling with {size} points."
                 self.log(msg+'\n')

surface_construct-0.12.6/surface_construct/tasks/vipsitetask.py

@@ -0,0 +1,74 @@
+from matplotlib import pyplot as plt
+from matplotlib.patches import Circle
+
+from .sitesampling import SurfaceSiteSampleTask
+
+class VIPSiteTask(SurfaceSiteSampleTask):
+    def __init__(self, exclude=None, include=None, **kwargs):
+        super().__init__(**kwargs)
+
+        vip_id = self.sg_obj.vip_id
+        grid_site_type, site_type_dict = self.sg_obj.get_grid_site_type()
+        if exclude is None and include is None:
+            msg = [f"Either exclude or include must be specified"]
+            msg += [f"Possible site types are {len(site_type_dict)} in total:"]
+            msg += [f"{k}: {v}" for k,v in site_type_dict.items() if type(k) in (int,)]
+            raise ValueError('\n'.join(msg))
+        elif exclude is not None and include is not None:
+            raise ValueError("Can not specified exclude and include at the same time")
+        elif exclude is not None and include is None:
+            filtered = [i for i in vip_id if site_type_dict[grid_site_type[i]] not in exclude]
+        else:
+            filtered = [i for i in vip_id if site_type_dict[grid_site_type[i]] in include]
+
+        points = self.sg_obj.points[filtered]
+        vector = self.sg_obj.vector[filtered]
+        raw_vector = self.sg_obj._raw_vector[filtered]
+        _ = self.sg_obj.vector_interval  # 此处call 一下，用于后面的 fit 中的参数
+        self.sg_obj.points = points
+        self.sg_obj.vector = vector
+        self.sg_obj._raw_vector = raw_vector
+        self.sg_obj.vectorgen = None
+
+        sampler_size = 0
+        new_sampler = []
+        for isampler in self.sampler:  # 不需要初始采样和key 点采样
+            if isampler['surface'] not in ('InitialSGSampler', 'KeyPointSGSampler'):
+                sampler_size += isampler.get('size', 0)
+                new_sampler.append(isampler)
+
+        # 如果 filter 之后的点小于等于 self.sampler['size'], 重新定义sampler
+        if sampler_size >= len(points):
+            self.sampler = [{
+                'size': len(points),  # 采样大小
+                'surface': "RandomSGSampler",  # 表面采样方法
+            }]
+        else:
+            self.sampler = new_sampler
+
+    def plot(self, figname='vip_site_heatmap.png'):
+        fig, ax = plt.subplots()
+        ax.set_aspect('equal')
+        # 画晶格
+        fpts = [[0, 0, 0], [1, 0, 0], [1, 1, 0], [0, 1, 0], [0, 0, 0]]  # 得到max_z处表面的分数坐标
+        pts = self.sg_obj.atoms.cell.cartesian_positions(fpts)[:, :2]  # 转为直角坐标，并取xy 两列
+        ax.plot(pts[:, 0], pts[:, 1], 'k-', linewidth=2)
+        # 画原子
+        cmap = plt.cm.jet
+        energies = self.sg_obj.grid_property['energy']
+        norm = plt.Normalize(vmin=min(energies), vmax=max(energies))
+        radius = 0.5
+        cell = self.sg_obj.atoms.cell
+        for point,e in zip(self.sg_obj.points,energies):
+            color = cmap(norm(e))
+            fx, fy = cell.scaled_positions(point)[:2]
+            px, py = cell.cartesian_positions([fx, fy, 0.])[:2]  # 适用于非正交盒子，将其平移到 xy 平面
+            circle = Circle((px, py), radius, facecolor=color,edgecolor='k', linewidth=1, alpha=0.8)
+            ax.add_patch(circle)
+        # 添加 colorbar
+        sm = plt.cm.ScalarMappable(cmap=cmap, norm=norm)
+        sm.set_array([])  # 需要调用 set_array 但不传数据
+        cbar = plt.colorbar(sm, ax=ax)
+        cbar.set_label('E/eV', fontsize=12)
+        plt.tight_layout()
+        fig.savefig(figname, bbox_inches='tight')

{surface_construct-0.12.4 → surface_construct-0.12.6/surface_construct.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: surface_construct
-Version: 0.12.4
+Version: 0.12.6
 Summary: Surface construction and surface reaction sampling tools.
 Home-page: https://gitee.com/pjren/surface_construct/
 Author: ren

{surface_construct-0.12.4 → surface_construct-0.12.6}/surface_construct.egg-info/SOURCES.txt

@@ -21,6 +21,7 @@ surface_construct/tasks/afm.py
 surface_construct/tasks/sitesampling.py
 surface_construct/tasks/taskbase.py
 surface_construct/tasks/terminations.py
+surface_construct/tasks/vipsitetask.py
 surface_construct/utils/__init__.py
 surface_construct/utils/atoms.py
 surface_construct/utils/geometry.py

{surface_construct-0.12.4 → surface_construct-0.12.6}/tests/test_task.py

@@ -3,6 +3,7 @@ import shutil
 from random import randint
 
 import ase.io
+import numpy as np
 import pytest
 from ase import Atom, Atoms
 from ase.constraints import FixAtoms
@@ -11,7 +12,7 @@ from lasp_ase.lasp import Lasp
 from surface_construct import SurfaceGrid
 from surface_construct import AdsGridCombiner
 from surface_construct import Adsorbate
-from surface_construct.tasks import SurfaceSiteSampleTask, AFMTask
+from surface_construct.tasks import SurfaceSiteSampleTask, AFMTask, VIPSiteTask
 from ase.optimize import LBFGS, BFGS
 
 
@@ -54,7 +55,8 @@ class TestTask:
                 'weight': (0.1, 0.45, 0.45),  # 表面采样方法的权重
             }  # 第二步采样
         ]
-        task_obj = SurfaceSiteSampleTask(combiner=ads_grid_comb, sampler=sampler, optimizer=BFGS)
+        task_obj = SurfaceSiteSampleTask(combiner=ads_grid_comb, sampler=sampler, optimizer=None)
+        task_obj.print_task_info()
         task_obj.run()
         print('Done')
 
@@ -81,7 +83,9 @@ class TestTask:
                 'weight': (0.4, 0.3, 0.3),  # 表面采样方法的权重
             }  # 第二步采样
         ]
-        task_obj = SurfaceSiteSampleTask(combiner=ads_grid_comb, sampler=sampler, optimizer=LBFGS)
+        task_obj = SurfaceSiteSampleTask(combiner=ads_grid_comb, sampler=sampler, optimizer=None)
+
+        task_obj.print_task_info()
         task_obj.run()
         print('Done')
 
@@ -108,7 +112,8 @@ class TestTask:
                 'weight': (0.4, 0.3, 0.3),  # 表面采样方法的权重
             }  # 第二步采样
         ]
-        task_obj = SurfaceSiteSampleTask(combiner=ads_grid_comb, sampler=sampler, optimizer=LBFGS)
+        task_obj = SurfaceSiteSampleTask(combiner=ads_grid_comb, sampler=sampler, optimizer=None)
+        task_obj.print_task_info()
         task_obj.run()
         print('Done')
 
@@ -171,6 +176,41 @@ class TestTask:
                 'weight': (0.5, 0.5),  # 表面采样方法的权重
             }  # 第二步采样
         ]
-        task_obj = AFMTask(combiner=com_obj, sampler=sampler, optimizer=LBFGS, nz=10) # nz 定义 z方向采多少样
+        task_obj = AFMTask(combiner=com_obj, sampler=sampler, optimizer=None, nz=3) # nz 定义 z方向采多少样
+        task_obj.print_task_info()
         task_obj.run()
-        print('Done')
+        print('Done')
+
+    def test_vipsite(self):
+        """
+        Water on CuO/Cu surface, only O top site
+        :return:
+        """
+        shutil.copyfile('../atoms_files/CuCHO_lasp.in', 'lasp.in')
+        shutil.copyfile('../atoms_files/CuCHO.pot', 'CuCHO.pot')
+        atoms = ase.io.read('../atoms_files/CuOx-Cu100-CONTCAR')
+        atoms.calc = Lasp()
+        ads_atoms = ase.Atoms('H2O',[[0.,0.,-1.],[0.,0.86,0.5],[0.,0.,0.]])
+        ads_obj = Adsorbate(ads_atoms,rtype='vdw_radii')
+        sg_obj = SurfaceGrid(atoms,rads=ads_obj.rads)
+        ads_grid_comb = AdsGridCombiner(sg_obj, ads_obj)
+        sampler =[
+            {
+                'surface': "RandomSGSampler",  # 第一步，随机表面采样方法。
+                'size': 2,  # 采样大小
+            },
+            {
+                'surface': ("MaxDiversitySGSampler", "MinEnergySGSampler", "MaxSigmaSGSampler"),  # 表面采样方法
+                'size': 4,  # 采样大小
+                'weight': (0.4, 0.3, 0.3),  # 表面采样方法的权重
+            }  # 第二步采样
+        ]
+        task_obj = VIPSiteTask(combiner=ads_grid_comb, sampler=sampler, optimizer=None, include=[(8,)])
+        task_obj.print_task_info()
+        task_obj.run()
+        task_obj.plot()
+        energies = task_obj.sg_obj.grid_property['energy']
+        np.savetxt('energies.csv', np.column_stack((sg_obj.points, energies)),
+                   header='x, y ,z , energy/eV',
+                   delimiter=',', fmt='%f')
+        print('Done')