Added ABACUS MD interface (#208)

* add ABACUS MD interface.

* Update system.py

* Add files via upload

README.md

@@ -80,6 +80,8 @@ The `System` or `LabeledSystem` can be constructed from the following file forma
 | Amber   | multi       | True         | True    | LabeledSystem | 'amber/md'           |
 | Amber/sqm | sqm.out   | False        | False   | System        | 'sqm/out'            |
 | Gromacs | gro         | True         | False   | System        | 'gromacs/gro'        |
+| ABACUS  | STRU        | False        | True    | LabeledSystem | 'abacus/scf'         |
+| ABACUS  | cif         | True         | True    | LabeledSystem | 'abacus/md'          |
 
 
 The Class `dpdata.MultiSystems`  can read data  from a dir which may contains many files of different systems, or from single xyz file which contains different systems.

dpdata/abacus/md.py

@@ -0,0 +1,190 @@
+import os,sys
+import numpy as np
+from .scf import ry2ev, kbar2evperang3, get_block, get_geometry_in, get_cell, get_coords
+
+# Read in geometries from an ABACUS MD trajectory.
+# The atomic coordinates are read in from generated files in OUT.XXXX.
+# Energies, forces
+# IMPORTANT: the program defaultly takes STRU input file as standard cell information,
+# therefore the direct and cartesan coordinates read could be different from the ones in 
+# the output cif files!!!
+# It is highly recommanded to use ORTHOGANAL coordinates in STRU file if you wish to get
+# same coordinates in both dpdata and output cif files. 
+
+def get_path_out(fname, inlines):
+    # This function is different from the same-name function in scf.py.
+    # This function returns OUT.XXXX's base directory.
+    path_out = os.path.join(fname, "OUT.ABACUS/")
+    for line in inlines:
+        if  len(line)>0 and "suffix" in line and "suffix"==line.split()[0]:
+           suffix = line.split()[1]
+           path_out = os.path.join(fname, "OUT.%s/" % suffix)
+           break
+    return path_out
+
+def get_coord_dump_freq(inlines):
+    for line in inlines:
+        if  len(line)>0 and "md_dumpmdfred" in line and "md_dumpmdfred" == line.split()[0]:
+            return int(line.split()[1])
+    return 1
+
+# set up a cell according to cell info in cif file.
+# maybe useful later
+'''
+def setup_cell(a, b, c, alpha, beta, gamma):
+    cell = np.zeros(3, 3)
+    cell[0, 0] = a
+    cell[1, 0] = b*np.cos(gamma/180*np.pi)
+    cell[1, 1] = b*np.sin(gamma/180*np.pi)
+    cell[2, 0] = c*np.cos(beta/180*np.pi)
+    cell[2, 1] = c*(b*np.cos(alpha/180*np.pi) - cell[1, 0]*np.cos(beta/180*np.pi))/cell[1, 1]
+    cell[2, 2] = np.sqrt(c**2 - cell[2, 0]**2 - cell[2, 1]**2)
+    return cell
+'''
+
+def get_single_coord_from_cif(pos_file, atom_names, natoms, cell):
+    assert(len(atom_names) == len(natoms))
+    nele = len(atom_names)
+    total_natoms = sum(natoms)
+    coord = np.zeros([total_natoms, 3])
+    a = 0
+    b = 0
+    c = 0
+    alpha = 0
+    beta = 0
+    gamma = 0
+    with open(pos_file, "r") as fp:
+        lines = fp.read().split("\n")
+    for line in lines:
+        if "_cell_length_a" in line:
+            a = float(line.split()[1])
+        if "_cell_length_b" in line:
+            b = float(line.split()[1])
+        if "_cell_length_c" in line:
+            c = float(line.split()[1])  
+        if "_cell_angle_alpha" in line:
+            alpha = float(line.split()[1])
+        if "_cell_angle_beta" in line:
+            beta = float(line.split()[1])
+        if "_cell_angle_gamma" in line:
+            gamma = float(line.split()[1])
+    assert(a > 0 and b > 0 and c > 0 and alpha > 0 and beta > 0 and gamma > 0)
+    #cell = setup_cell(a, b, c, alpha, beta, gamma)
+    coord_lines = get_block(lines=lines, keyword="_atom_site_fract_z", skip=0, nlines = total_natoms)
+
+    ia_idx = 0
+    for it in range(nele):
+        for ia in range(natoms[it]):
+            coord_line = coord_lines[ia_idx].split()
+            assert(coord_line[0] == atom_names[it])
+            coord[ia_idx, 0] = float(coord_line[1])
+            coord[ia_idx, 1] = float(coord_line[2])
+            coord[ia_idx, 2] = float(coord_line[3])
+            ia_idx+=1
+    coord = np.matmul(coord, cell)
+    # important! Coordinates are converted to Cartesian coordinate.
+    return coord
+
+
+def get_coords_from_cif(ndump, dump_freq, atom_names, natoms, types, path_out, cell):
+    total_natoms = sum(natoms)
+    #cell = np.zeros(ndump, 3, 3)
+    coords = np.zeros([ndump, total_natoms, 3])
+    pos_file = os.path.join(path_out, "STRU_READIN_ADJUST.cif")
+    # frame 0 file is different from any other frames
+    coords[0] = get_single_coord_from_cif(pos_file, atom_names, natoms, cell)
+    for dump_idx in range(1, ndump):
+        pos_file = os.path.join(path_out, "md_pos_%d.cif" %(dump_idx*dump_freq))
+        #print("dump_idx = %s" %dump_idx)
+        coords[dump_idx] = get_single_coord_from_cif(pos_file, atom_names, natoms, cell)
+    return coords
+
+def get_energy_force_stress(outlines, inlines, dump_freq, ndump, natoms, atom_names):
+    stress = None
+    total_natoms = sum(natoms)
+    for line in inlines:
+        if len(line)>0 and "stress" in line and "stress" == line.split()[0] and "1" == line.split()[1]:
+            stress = np.zeros([ndump, 3, 3])
+            break
+    if type(stress) != np.ndarray:
+        print("The ABACUS program has no stress output. Stress will not be read.")
+    nenergy = 0
+    nforce = 0
+    nstress = 0
+    energy = np.zeros(ndump)
+    force = np.zeros([ndump, total_natoms, 3])
+
+    for line_idx, line in enumerate(outlines):
+        if "final etot is" in line:
+            if nenergy%dump_freq == 0:
+                energy[int(nenergy/dump_freq)] = float(line.split()[-2])
+            nenergy+=1
+        if "TOTAL-FORCE (eV/Angstrom)" in line:
+            for iatom in range(0, total_natoms):
+                force_line = outlines[line_idx+5+iatom]
+                atom_force = [float(i) for i in force_line.split()[1:]]
+                assert(len(atom_force) == 3)
+                atom_force = np.array(atom_force)
+                if nforce%dump_freq == 0:
+                    force[int(nforce/dump_freq), iatom] = atom_force
+            nforce+=1
+            assert(nforce==nenergy)
+        if "TOTAL-STRESS (KBAR)" in line:
+            for idx in range(0, 3):
+                stress_line = outlines[line_idx+4+idx]
+                single_stress = [float(i) for i in stress_line.split()]
+                if len(single_stress) != 3:
+                    print(single_stress)
+                assert(len(single_stress) == 3)
+                single_stress = np.array(single_stress)
+                if nstress%dump_freq == 0:
+                    stress[int(nstress/dump_freq), idx] = single_stress
+            nstress+=1
+            assert(nstress==nforce)
+    if type(stress) == np.ndarray:
+        stress *= kbar2evperang3
+    return energy, force, stress
+
+
+def get_frame (fname):
+    if type(fname) == str:
+        # if the input parameter is only one string, it is assumed that it is the 
+        # base directory containing INPUT file;
+        path_in = os.path.join(fname, "INPUT")
+    else:
+        raise RuntimeError('invalid input')    
+    with open(path_in, 'r') as fp:
+        inlines = fp.read().split('\n')
+    geometry_path_in = get_geometry_in(fname, inlines) # base dir of STRU
+    path_out = get_path_out(fname, inlines) 
+
+    with open(geometry_path_in, 'r') as fp:
+        geometry_inlines = fp.read().split('\n')
+    celldm, cell = get_cell(geometry_inlines) 
+    atom_names, natoms, types, coords = get_coords(celldm, cell, geometry_inlines, inlines) 
+    # This coords is not to be used.
+    dump_freq = get_coord_dump_freq(inlines = inlines)
+    ndump = int(os.popen("ls --l %s | grep 'md_pos_' | wc -l" %path_out).readlines()[0])
+    # number of dumped geometry files
+    coords = get_coords_from_cif(ndump, dump_freq, atom_names, natoms, types, path_out, cell)
+
+    # TODO: Read in energies, forces and pressures.
+    with open(os.path.join(path_out, "running_md.log"), 'r') as fp:
+        outlines = fp.read().split('\n')
+    energy, force, stress = get_energy_force_stress(outlines, inlines, dump_freq, ndump, natoms, atom_names)
+    if type(stress) == np.ndarray:
+        stress *= np.linalg.det(cell)
+    data = {}
+    data['atom_names'] = atom_names
+    data['atom_numbs'] = natoms
+    data['atom_types'] = types
+    data['cells'] = np.zeros([ndump, 3, 3])
+    for idx in range(ndump):
+        data['cells'][:, :, :] = cell
+    data['coords'] = coords
+    data['energies'] = energy
+    data['forces'] = force
+    data['virials'] = stress
+    data['orig'] = np.zeros(3)
+
+    return data

dpdata/abacus/scf.py

@@ -174,6 +174,6 @@ def get_frame (fname):
     # print("virial = ", data['virials'])
     return data
 
-if __name__ == "__main__":
-    path = "/home/lrx/work/12_ABACUS_dpgen_interface/dpdata/dpdata/tests/abacus.scf"
-    data = get_frame(path)
+#if __name__ == "__main__":
+#    path = "/home/lrx/work/12_ABACUS_dpgen_interface/dpdata/dpdata/tests/abacus.scf"
+#    data = get_frame(path)

dpdata/plugins/abacus.py

@@ -1,10 +1,19 @@
 import dpdata.abacus.scf
+import dpdata.abacus.md
 from dpdata.format import Format
 
 
 @Format.register("abacus/scf")
 @Format.register("abacus/pw/scf")
 class AbacusSCFFormat(Format):
-    @Format.post("rot_lower_triangular")
+    #@Format.post("rot_lower_triangular")
     def from_labeled_system(self, file_name, **kwargs):
         return dpdata.abacus.scf.get_frame(file_name)
+
+@Format.register("abacus/md")
+@Format.register("abacus/pw/md")
+@Format.register("abacus/lcao/md")
+class AbacusMDFormat(Format):
+    #@Format.post("rot_lower_triangular")
+    def from_labeled_system(self, file_name, **kwargs):
+        return dpdata.abacus.md.get_frame(file_name)

dpdata/system.py

@@ -77,7 +77,8 @@ def __init__ (self,
                 - ``vasp/poscar``: vasp POSCAR
                 - ``qe/cp/traj``: Quantum Espresso CP trajectory files. should have: file_name+'.in' and file_name+'.pos'
                 - ``qe/pw/scf``: Quantum Espresso PW single point calculations. Both input and output files are required. If file_name is a string, it denotes the output file name. Input file name is obtained by replacing 'out' by 'in' from file_name. Or file_name is a list, with the first element being the input file name and the second element being the output filename.
-                - ``abacus/scf``: ABACUS plane wave scf. The directory containing INPUT file is required. 
+                - ``abacus/scf``: ABACUS pw/lcao scf. The directory containing INPUT file is required. 
+                - ``abacus/md``: ABACUS pw/lcao MD. The directory containing INPUT file is required. 
                 - ``siesta/output``: siesta SCF output file
                 - ``siesta/aimd_output``: siesta aimd output file
                 - ``pwmat/atom.config``: pwmat atom.config

tests/abacus.md/INPUT

@@ -0,0 +1,30 @@
+INPUT_PARAMETERS
+#Parameters	(General)
+suffix          autotest
+pseudo_dir		./
+ntype			1	
+nbands			8
+calculation     md
+read_file_dir   ./
+
+#Parameters (Accuracy)
+ecutwfc			20
+niter			20
+
+basis_type		pw
+nstep           21 # number of MD/relaxation steps
+
+stress          1
+stress_thr      1e-6
+force           1
+force_thr_ev    1.0e-3
+
+ks_solver       cg
+mixing_type     pulay
+mixing_beta     0.7
+
+md_mdtype        1 # 0 for NVE; 1 for NVT with Nose Hoover; 2 for NVT with velocity rescaling
+md_tfirst        10 # temperature, unit: K
+md_dt            1 # unit: fs
+md_rstmd         0 # 1 for restart
+md_dumpmdfred    5 

tests/abacus.md/OUT.autotest/STRU_READIN_ADJUST.cif

@@ -0,0 +1,21 @@
+data_test
+
+_audit_creation_method generated by ABACUS
+
+_cell_length_a 3.81668
+_cell_length_b 3.81668
+_cell_length_c 3.81668
+_cell_angle_alpha 60
+_cell_angle_beta 60
+_cell_angle_gamma 60
+
+_symmetry_space_group_name_H-M 
+_symmetry_Int_Tables_number 
+
+loop_
+_atom_site_label
+_atom_site_fract_x
+_atom_site_fract_y
+_atom_site_fract_z
+Si 0 0 0
+Si 0.245 0.237 0.265

tests/abacus.md/OUT.autotest/md_pos_1.cif

@@ -0,0 +1,21 @@
+data_test
+
+_audit_creation_method generated by ABACUS
+
+_cell_length_a 3.81668
+_cell_length_b 3.81668
+_cell_length_c 3.81668
+_cell_angle_alpha 60
+_cell_angle_beta 60
+_cell_angle_gamma 60
+
+_symmetry_space_group_name_H-M 
+_symmetry_Int_Tables_number 
+
+loop_
+_atom_site_label
+_atom_site_fract_x
+_atom_site_fract_y
+_atom_site_fract_z
+Si 0.00106888 0.99849 0.000247409
+Si 0.245535 0.237394 0.262886

tests/abacus.md/OUT.autotest/md_pos_10.cif

@@ -0,0 +1,21 @@
+data_test
+
+_audit_creation_method generated by ABACUS
+
+_cell_length_a 3.81668
+_cell_length_b 3.81668
+_cell_length_c 3.81668
+_cell_angle_alpha 60
+_cell_angle_beta 60
+_cell_angle_gamma 60
+
+_symmetry_space_group_name_H-M 
+_symmetry_Int_Tables_number 
+
+loop_
+_atom_site_label
+_atom_site_fract_x
+_atom_site_fract_y
+_atom_site_fract_z
+Si 0.00711478 0.993904 0.9916
+Si 0.253316 0.23197 0.255312

tests/abacus.md/OUT.autotest/md_pos_15.cif

@@ -0,0 +1,21 @@
+data_test
+
+_audit_creation_method generated by ABACUS
+
+_cell_length_a 3.81668
+_cell_length_b 3.81668
+_cell_length_c 3.81668
+_cell_angle_alpha 60
+_cell_angle_beta 60
+_cell_angle_gamma 60
+
+_symmetry_space_group_name_H-M 
+_symmetry_Int_Tables_number 
+
+loop_
+_atom_site_label
+_atom_site_fract_x
+_atom_site_fract_y
+_atom_site_fract_z
+Si 0.00722617 0.979891 0.000667301
+Si 0.260608 0.240666 0.237553

tests/abacus.md/OUT.autotest/md_pos_20.cif

@@ -0,0 +1,21 @@
+data_test
+
+_audit_creation_method generated by ABACUS
+
+_cell_length_a 3.81668
+_cell_length_b 3.81668
+_cell_length_c 3.81668
+_cell_angle_alpha 60
+_cell_angle_beta 60
+_cell_angle_gamma 60
+
+_symmetry_space_group_name_H-M 
+_symmetry_Int_Tables_number 
+
+loop_
+_atom_site_label
+_atom_site_fract_x
+_atom_site_fract_y
+_atom_site_fract_z
+Si 0.0136053 0.987589 0.984304
+Si 0.261345 0.227707 0.245529