@@ -89,7 +89,7 @@ Within the AO representation, the Hartree-Fock equations reduce to the Roothaan-
89
89
90
90
.. math ::
91
91
92
- \mathbf {F}^{\alpha } \mathbf {C}^{\alpha } = \bf {\varepsilon }^{\alpha } \mathbf {S} \mathbf {C}^{\alpha } \;,
92
+ \mathbf {F}^{\alpha } \mathbf {C}^{\alpha } = \boldsymbol {\varepsilon }^{\alpha } \mathbf {S} \mathbf {C}^{\alpha } \;,
93
93
94
94
where
95
95
@@ -112,6 +112,119 @@ For example, the ground-state electronic energy is expressed as
112
112
113
113
Periodic boundary conditions
114
114
----------------------------
115
+ PySCF also alows the user to perform SCF calculations for solids.
116
+ With crystalline Gaussian-type AOs as the underlying single-partial basis (see :numref: `theory_pbc_gto `),
117
+ the molecular SCF code can be easily adapted to the cases where periodic boundary conditions (PBCs)
118
+ are applied. Instead of solving only one set of Roothaan-Hall or Pople-Nesbet equtions for molecules,
119
+ it is now necessary to solve them for each k point for solids:
120
+
121
+ .. math ::
122
+
123
+ \mathbf {F}(\mathbf {k}) \mathbf {C}(\mathbf {k}) = \boldsymbol {\varepsilon }(\mathbf {k}) \mathbf {S}(\mathbf {k}) \mathbf {C}(\mathbf {k}) \;,
124
+
125
+ where the Fock matrix is defined (within the restricted formalism) as
126
+
127
+ .. math ::
128
+
129
+ \mathbf {F}(\mathbf {k}) = \mathbf {T}(\mathbf {k}) + \mathbf {V}^{\rm PP}(\mathbf {k})
130
+ +\mathbf {J}(\mathbf {k}) - \frac {1 }{2 } \mathbf {K}(\mathbf {k}) + \mathbf {V}^{L+J}(\mathbf {k}) \;.
131
+
132
+ Here, :math: `\mathbf {V}^{\rm PP}` denotes the pseudopotential contribution and
133
+ :math: `\mathbf {V}^{L+J}` deals with the divergence of local pseudopotential and Hartree potential (see below).
134
+
135
+ The one-electron overlap, kinetic energy, and local pseudopotential integrals
136
+ are evaluated through numerical integrations on the real-space grid according to
137
+
138
+ .. math ::
139
+
140
+ S_{\mu\nu }(\mathbf {k}) = \int _\Omega d\mathbf {r} \phi _{\mu \mathbf {k}}^{*}(\mathbf {r}) \phi _{\nu \mathbf {k}}(\mathbf {r}) \;,
141
+
142
+ .. math ::
143
+
144
+ T_{\mu\nu }(\mathbf {k}) = -\frac {1 }{2 } \int _\Omega d\mathbf {r} \phi _{\mu \mathbf {k}}^{*}(\mathbf {r})
145
+ \boldsymbol {\nabla }_{\mathbf {r}}^2 \phi _{\nu \mathbf {k}}(\mathbf {r}) \;,
146
+
147
+ and
148
+
149
+ .. math ::
150
+
151
+ V_{\mu\nu }^{\rm L-PP}(\mathbf {k}) = \int _\Omega d\mathbf {r} \phi _{\mu \mathbf {k}}^{*}(\mathbf {r})
152
+ v^{\rm L-PP}(\mathbf {r}) \phi _{\nu \mathbf {k}}(\mathbf {r}) \;,
153
+
154
+ where :math: `\Omega ` labels the unit cell volume.
155
+ The non-local part of the pseudopotential is computed in the reciprocal space:
156
+
157
+ .. math ::
158
+
159
+ V_{\mu\nu }^{\rm NL-PP}(\mathbf {k}) = \Omega \sum _{\mathbf {G},\mathbf {G}'} \phi _{\mu \mathbf {k}}^{*}(\mathbf {G})
160
+ v^{\rm NL-PP}(\mathbf {k}+\mathbf {G}, \mathbf {k}+\mathbf {G}') \phi _{\nu \mathbf {k}}(\mathbf {G}') \;,
161
+
162
+ where
163
+
164
+ .. math ::
165
+
166
+ v^{\rm NL-PP}(\mathbf {k}+\mathbf {G}, \mathbf {k}+\mathbf {G}') = \frac {1 }{\Omega } \int d\mathbf {r} \int d\mathbf {r}'
167
+ e^{-i(\mathbf {k}+\mathbf {G})\cdot \mathbf {r}} v^{\rm NL-PP}(\mathbf {r},\mathbf {r}')
168
+ e^{ i(\mathbf {k}+\mathbf {G}^{'})\cdot \mathbf {r}'} \;.
169
+
170
+ .. note ::
171
+ The way that the pseudopotential integrals are computed differs in different density fitting schemes and for different
172
+ pseudopotentials. Interested readers should refer to :numref: `theory_pbc_df ` and :numref: `theory_pbc_pp `.
173
+
174
+ The Coulomb and exchange matrices are defined similarly as
175
+
176
+ .. math ::
177
+
178
+ J_{\mu\nu }(\mathbf {k}) = \int _{\Omega } d\mathbf {r} \phi _{\mu \mathbf {k}}^{*}(\mathbf {r}) v_{\rm H}(\mathbf {r}) \phi _{\nu \mathbf {k}}(\mathbf {r}) \;,
179
+
180
+ and
181
+
182
+ .. math ::
183
+
184
+ K_{\mu\nu }(\mathbf {k}) = \int _{\Omega } d\mathbf {r} \int d\mathbf {r}' \phi _{\mu \mathbf {k}}^{*}(\mathbf {r})
185
+ \frac {\rho (\mathbf {r}, \mathbf {r}')}{|\mathbf {r}-\mathbf {r}'|} \phi _{\nu \mathbf {k}}(\mathbf {r}') \;.
186
+
187
+ Here :math: `v_{\rm H}` is the Hartree potential
188
+
189
+ .. math ::
190
+
191
+ v_{\rm H}(\mathbf {r}) = \frac {4 \pi }{\Omega } \sum _{\mathbf {G}\neq \mathbf {0 }} \frac {\rho (\mathbf {G})}{G^2 } e^{i\mathbf {G}\cdot \mathbf {r}} \;,
192
+
193
+ and :math: `\rho (\mathbf {r}, \mathbf {r}')` is the density matrix
194
+
195
+ .. math ::
196
+
197
+ \rho (\mathbf {r}, \mathbf {r}') = \sum _{\mathbf {k}} w_{\mathbf {k}} \sum _{\lambda\sigma } P_{\lambda\sigma }(\mathbf {k})
198
+ \phi _{\lambda \mathbf {k}}(\mathbf {r}) \phi _{\sigma \mathbf {k}}^{*}(\mathbf {r}') \;,
199
+
200
+ where :math: `w_{\mathbf {k}}` represents the weight of each k point.
201
+
202
+ Note that the local part of the pseudopotential and the Hartree potential diverge at :math: `G=0 `;
203
+ however, their sum is not, which leads to the :math: `V^{\rm L+J}` term (for charge neutral unit cell):
204
+
205
+ .. math ::
206
+
207
+ V_{\mu\nu }^{\rm L+J} (\mathbf {k}) = \frac {S_{\mu\nu }}{\Omega }
208
+ \int d\mathbf {r} \left (v^{\rm L-PP}(\mathbf {r}) + \sum _{\alpha } \frac {Z_{\alpha }e^2 }{r} \right ) \;.
209
+
210
+ .. note ::
211
+
212
+ For details about how to compute the Coulomb (:math: `\mathbf {J}`)
213
+ and exchange (:math: `\mathbf {K}`) integrals, see :numref: `theory_pbc_df `.
214
+
215
+ Finally, the total electronic energy differs from the molecular case only by a k-point summation:
216
+
217
+ .. math ::
218
+
219
+ E_{\rm HF} = \sum _{\mathbf {k}} w_{\mathbf {k}} E_{\rm HF}(\mathbf {k}) \;,
220
+
221
+ where
222
+
223
+ .. math ::
224
+
225
+ E_{\rm HF}(\mathbf {k}) = \frac {1 }{2 } \left \{ {\rm Tr}\left [\mathbf {h}(\mathbf {k}) (\mathbf {P}^{\alpha }(\mathbf {k})+\mathbf {P}^{\beta }(\mathbf {k}))\right ]
226
+ + {\rm Tr}\left [\mathbf {F}^{\alpha }(\mathbf {k}) \mathbf {P}^{\alpha }(\mathbf {k})\right ]
227
+ + {\rm Tr}\left [\mathbf {F}^{\beta }(\mathbf {k}) \mathbf {P}^{\beta }(\mathbf {k})\right ] \right \} \;.
115
228
116
229
117
230
Initial guess
0 commit comments