Debian Codegraph @

EPW/ZG_displacement/README

raw
1 Marios Zacharias [1,2] & Feliciano Giustino [3,4], Feb 2020
2
3 [1] Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4–6, D-14195 Berlin, Germany
4 [2] Department of Mechanical and Materials Science Engineering, Cyprus University of Technology,
5 P.O. Box 50329, 3603 Limassol, Cyprus
6 [3] Oden Institute for Computational Engineering and Sciences, The University of Texas at Austin,
7 Austin, Texas 78712, USA
8 [4] Department of Physics, The University of Texas at Austin, Austin, Texas 78712, USA
9 ---------------------------------------------------------------------------------------------------
10
11 If you use ZG.x please do cite the following work:
12
13 [1] Marios Zacharias and Feliciano Giustino, Phys. Rev. Research 2, 013357, (2020).
14 [2] Marios Zacharias and Feliciano Giustino, Phys. Rev. B 94, 075125, (2016).
15
16 ---------------------------------------------------------------------------------------------------
17
18 Acknowledgement: We thank Hyungjun Lee, Oden Institute for Computational Engineering and Sciences,
19 for help packaging this release.
20
21 ---------------------------------------------------------------------------------------------------
22 Instructions for the construction of the "ZG-displacement" (Zacharias-Giustino) following
23 Eq. (2) of Phys. Rev. Research 2, 013357, 2020. The approach for generating the ZG-displacement
24 is based on the generalized theory of the special displacement method as in
25 Phys. Rev. Research 2, 013357, (2020).
26
27 STEPS for generating the "ZG-displacement" for the calculation of temperature-dependent properties of solids:
28
29 0. Compile the version of Quantum Espresso 6.6.0; pw.x and ph.x are required to run ZG.x.
30 Go to "/path_to_your_espresso/EPW/ZG_displacement/src" and "make".
31 This code exploits some main routines of matdyn.x. In particular, the code
32 takes advantage of the phonon interpolation scheme implemented for matdyn.x. This is
33 important, for example, in taking into account the LO-TO splitting in polar materials.
34 The necessary input files are the interatomic force constants "file.fc" and "ZG.in".
35 See for example files in the "example/silicon/inputs" folder.
36
37 1. Run a scf calculation: for example "/path_to_your_espresso/bin/pw.x < si.scf.in > si.scf.out"
38 for the silicon unit cell.
39
40 2. Run a phonon calculation: for example "/path_to_your_espresso/bin/ph.x < si.ph.in > si.ph.out"
41 for a relatively "dense" q-grid to obtain a converged phonon dispersion.
42
43 3. Run "/path_to_your_espresso/bin/q2r.x <q2r.in> q2r.out".
44 The input "q2r.in" has the standard format of Quantum Espresso for
45 calculating the interatomic force constant matrix elements to be used to construct the
46 "ZG-configuration.dat" file.
47
48 4. Calculate the phonon dispersion with "matdyn.x" and make sure that your phonon dispersion is
49 correct (compare to other literature results). This is not a necessary step for the construction of the
50 "ZG-configuration". However, you need to ensure that the phonons, to be used for the construction
51 of the "ZG-configuration", are those you expect. For a polar material, LO-TO splitting should be
52 included.
53
54 5. Now one needs to decide on the size of the supercell configuration to be used for calculating
55 temperature dependent properties. For help, please see the example folder by "tar -xvf example.tar.gz".
56 In file "example/silicon/ZG_displacement_generation/inputs/ZG_444.in" we show the example for
57 constructing a 4x4x4 ZG-configuration. One could potentially generate any supercell size
58 by simply changing "dimx","dimy","dimz", and the list of q-points (optional, see below).
59 "ZG.in" has the standard format as a "matdyn.in" file for Quantum Espresso.
60 Here we use the following input parameters:
61 ---------------------------------------------------------------------------------------
62 i) "ZG_conf" : Logical flag that enables the creation of the ZG-displacement.
63 (default .true.)
64 "T" : Real number indicating the temperature at which the calculations will be performed.
65 "T" essentially defines the amplitude of the normal coordinates.
66 (default 0.00)
67 "dimx","dimy","dimz" : Integers corresponding to the dimensionality of the supercell i.e.:
68 size of supercell will be [dimx * a(1), dimy * a(2), dimz * a(3)],
69 where a(1), a(2), a(3) are the lattice vectors of the unit cell used
70 to compute the phonons.
71 (default 0, 0, 0)
72 "atm_zg(1), etc.." : String describing the element of each atomic species
73 (default "Element")
74 "synch" : Logical flag that enables the synchronization of the modes.
75 (default .false.)
76 "nloops" : Integer for the number of loops the algorithm needs to
77 go through for finding the optimum configuration. The algorithm
78 generates a set of "+,-,+,-" signs and its possible permutations,
79 trying to minimize the error coming from the coupling of modes with
80 the same q-wavevector but at different branch. For a finite supercell
81 size the order of using the "+,-,+,-" set and its permutations is
82 important giving different results. Therefore the algorithm checks
83 the combination that brings the error lower than a threshold.
84 (default 15000)
85 "compute_error" : Logical flag: if set to .true. allows the code to find the optimal ZG configuration
86 by minimizing the error based on the "threshold" flag (see below). Set it
87 to .false. if speed up is required. This is useful when (i) large supercell
88 sizes are considered for which the error is minimized by the first set of signs,
89 (ii) only single phonon displacements are of interest (see below)
90 (default .true.)
91 "threshold" : Real number indicating the error at which the algorithm stops while it's
92 looking for possible combinations of signs. Once this limit is reached,
93 the ZG-displacement is constructed. The threshold is usually chosen
94 to be less than 5% of the diagonal terms, i.e. those terms that contribute
95 to the calculation of temperature-dependent properties.
96 (default 0.05)
97 "incl_qA" : Logical flag, to decide whether to include phonon modes in set A or not.
98 (default .true.)
99 "single_phonon_displ": Logical flag that allows to displace the nuclei along single phonon modes.
100 Use output configurations to compute electron-phonon matrix elements with a direct
101 supercell calculation. Set the displacement to the zero point by "T = 0".
102 This finite displacement should carry precisely the effect of diagonal elements of [g(q)+g(-q)].
103 Output files: "single_phonon-displacements.dat" and "single_phonon-velocities.dat".
104 (default .false.)
105 "q_external" : Logical flag that allows the use of a q-point list specified by the user in the input file.
106 If .false. the q-point list is specified by the supercell dimensions dimx, dimy, and dimz.
107 If .true. the q-point list must be provided by the user (see "qlist_AB.txt").
108 (default .false.)
109 "qlist_AB.txt" : This file contains the external q-list in crystal coordinates as in the "ZG_444.in" example,
110 after the input flags. It corresponds to the q-points commensurate to the supercell size.
111 Only one of the q-point time-reversal partners is kept for the construction of the
112 ZG-displacement. The calculations, for the moment, assume systems with time-reversal symmetry.
113 For the generation of the "qlist_AB.txt" set the q-gird in file
114 "example/silicon/input/qlist.in" and run "../../../src/create_qlist.x < qlist.in > qlist.out".
115 One can modify the "create_qlist.f90" to generate a different path for consecutive q-points.
116 Paste the output of "qlist_AB.txt" to "ZG.in" after namelist &input. Set the flag
117 q_external = .true. for the code to read the list.
118
119 ii) To generate the ZG-displacement run "/path_to_your_espresso/bin/ZG.x <ZG_444.in> ZG_444.out".
120 This generates three output files: the "equil_pos.txt", "ZG-configuration.dat" and "ZG-velocities.dat".
121 The first file has the equilibrium coordinates of the nuclei and the second has the optimum set of nuclear coordinates
122 that define the ZG-displacement for a particular temperature and supercell size.
123 The third one has the ZG-velocities or momenta of the nuclei generated in the same spirit with the ZG displacement.
124 The outputs for a supercell size of 4x4x4 are in the folder "example/silicon/output/".
125 iii) The calculation of the ZG-displacement should usually takes a few seconds to few minutes with one processor.
126
127 6. VERY IMPORTANT NOTE: It is perfectly reasonable to find different ZG-displacements / ZG-configurations,
128 since the modes obtained by diagonalizing the dynamical matrix can differ by a phase factor
129 (or a unitary matrix in case of degeneracy) if the processor, or compiler, or libraries have changed.
130 The eigenvalues of course should remain the same in all cases. The synchronization routine should align the sign
131 of the modes with respect to a reference mode, but the sign of this reference depends on the machine,
132 and degeneracy is not taken into account. The best way to check the validity of your configuration is
133 by comparing the anisotropic displacement tensor with the exact values, both obtained at the end of the
134 "ZG-configuration.dat". Reducing the value of "threshold" should bring the anisotropic displacement tensor
135 closer to the exact values.
136
137 7. The ZG-displacement can be employed for the calculation of any temperature-dependent
138 observable that is described by the Fermi Golden rule. The procedure is exactly the same as for
139 the calculation of the observable obtained with the nuclei clamped at their equilibrium positions,
140 but now, instead, using the ZG-displacement (supercell) that incorporates automatically
141 the electron-phonon interaction and quantum zero-point motion.
142
143 8. The ZG-displacement can also be employed for the calculation of full temperature-dependent
144 band structures using the unfolding technique for supercell calculations.
145
146 ----------------------------------------------------------------------------------------------------------
147
148 Example using the ZG-displacement
149
150 Here we show a simple example on how to use the ZG-displacement for calculating the temperature dependence
151 of the JDOS and eventually extract the zero-point renormalization and temperature dependence of the band gap.
152
153 Steps:
154
155 1. Once you generate the ZG-configuation from previous step, then you can run your first SCF calculation
156 with the special displacement method. Use the atomic positions in the file "ZG-configuration.dat" and
157 prepare your "scf.in" file. Make sure your lattice constants are comensurate with the supercell size.
158 For example, check the input files in the directory 'example/silicon/JDOS/inputs/333/ZG_0K'.
159
160 2. Run an scf calculation and then prepare an "nscf.in" to run a non-self-consistent (nscf) calculation on a denser grid.
161 For example, check the input files in the directory 'example/silicon/JDOS/inputs/333/ZG_0K'.
162
163 3. Once you obtain the ".xml" file from the nscf calculation, run "commands.sh" as in the directory
164 'example/silicon/JDOS/inputs/333/ZG_0K'. This will generate the file "data_one_column.dat"
165 with all eigenvalues calculated for each k-point using the ZG-configuration. Note that in
166 the file "commands.sh", only "NR+44" should be changed for a different run. 44 indicates how many
167 lines containing the eigenvalues exist, after the string match "eigenvalues size" in the ".xml" file.
168
169 4. Use the output file "data_one_column.dat" to calculate the joint-density of states (JDOS) via the executable
170 "JDOS_Gaus.x" located in the "src/JDOS" folder. Command: "/path_to/JDOS_Gaus.x <JDOS_Gaus.in > JDOS_Gaus.out".
171 For extracting the band gap from the joint-density of states follow the procedure in Ref.[Phys. Rev. B 94, 075125, (2016)].
172
173 5. Compare your results with the data in the directory 'example/silicon/JDOS/outputs/333'.
174 Gnuplot commands are aslo given to facilitate comparison.
175
176 6. In the examples you will also find ZG-configurations generated by setting the flag "incl_qA = .true."
177 and the flag "synch = .false.". The names of the ZG files end with "...qA.dat" and "...nosym.dat", respectively.
178 Outputs of the "scf", "nscf" and "JDOS" are also provided. Comparion of the results obtained with different settings
179 shows the importance of exluding the qA modes, which break the degeneracy at high-symmetry points,
180 in small supercell structures. For example, the valence band top of silicon is threefold degenerated. A general rule is
181 to keep "synch = .true." and "incl_qA = .false.", unless there is a particular reason to change them.
182
183 7. Examples for the 3x3x3 and 5x5x5 Si ZG-configurations are provided. In the "outputs" directory the calculated
184 zero-point renormalization (ZPR) for each case is given in the gnuplot files. For a 5x5x5 Si ZG-configuration the
185 calculated ZPR is 55 meV, very close to the converged value of 57 meV obtained using an 8x8x8 Si ZG-configuration
186 in Ref. [Phys. Rev. B 94, 075125, (2016)].
187
188 --------------------------------------------------------------------------------------------------------------
189
190 Other capabilities of ZG.x:
191
192 1. ZG.x can also be used to generate displaced configurations along single phonon modes.
193 Please see "single_phonon_displ" flag.
194 Find an example in:
195 "example/silicon/Displacement_along_single_phonon_modes/inputs/".
196 To run the example use:
197 "/path_to_your_espresso/bin/ZG.x <ZG_444.in> ZG_444.out".
198 The output files of interest are:
199 "single_phonon-displacements.dat" and "single_phonon-velocities.dat".
200
201 --------------------------------------------------------------------------------------------------------------
202
203 NOW YOU CAN ENJOY GENERATING SPECIAL DISPLACENTS AND CALCULATE FINITE TEMPERATURE EFFECTS.