version 0.0.49
This commit is contained in:
125
PyPI/src/guan/band_structures_and_wave_functions.py
Normal file
125
PyPI/src/guan/band_structures_and_wave_functions.py
Normal file
@@ -0,0 +1,125 @@
|
||||
# Guan is an open-source python package developed and maintained by https://www.guanjihuan.com/about. The primary location of this package is on website https://py.guanjihuan.com.
|
||||
|
||||
# band_structures_and_wave_functions
|
||||
|
||||
## band structures
|
||||
|
||||
import numpy as np
|
||||
import cmath
|
||||
|
||||
def calculate_eigenvalue(hamiltonian):
|
||||
if np.array(hamiltonian).shape==():
|
||||
eigenvalue = np.real(hamiltonian)
|
||||
else:
|
||||
eigenvalue, eigenvector = np.linalg.eigh(hamiltonian)
|
||||
return eigenvalue
|
||||
|
||||
def calculate_eigenvalue_with_one_parameter(x_array, hamiltonian_function, print_show=0):
|
||||
dim_x = np.array(x_array).shape[0]
|
||||
i0 = 0
|
||||
if np.array(hamiltonian_function(0)).shape==():
|
||||
eigenvalue_array = np.zeros((dim_x, 1))
|
||||
for x0 in x_array:
|
||||
hamiltonian = hamiltonian_function(x0)
|
||||
eigenvalue_array[i0, 0] = np.real(hamiltonian)
|
||||
i0 += 1
|
||||
else:
|
||||
dim = np.array(hamiltonian_function(0)).shape[0]
|
||||
eigenvalue_array = np.zeros((dim_x, dim))
|
||||
for x0 in x_array:
|
||||
if print_show==1:
|
||||
print(x0)
|
||||
hamiltonian = hamiltonian_function(x0)
|
||||
eigenvalue, eigenvector = np.linalg.eigh(hamiltonian)
|
||||
eigenvalue_array[i0, :] = eigenvalue
|
||||
i0 += 1
|
||||
return eigenvalue_array
|
||||
|
||||
def calculate_eigenvalue_with_two_parameters(x_array, y_array, hamiltonian_function, print_show=0, print_show_more=0):
|
||||
dim_x = np.array(x_array).shape[0]
|
||||
dim_y = np.array(y_array).shape[0]
|
||||
if np.array(hamiltonian_function(0,0)).shape==():
|
||||
eigenvalue_array = np.zeros((dim_y, dim_x, 1))
|
||||
i0 = 0
|
||||
for y0 in y_array:
|
||||
j0 = 0
|
||||
for x0 in x_array:
|
||||
hamiltonian = hamiltonian_function(x0, y0)
|
||||
eigenvalue_array[i0, j0, 0] = np.real(hamiltonian)
|
||||
j0 += 1
|
||||
i0 += 1
|
||||
else:
|
||||
dim = np.array(hamiltonian_function(0, 0)).shape[0]
|
||||
eigenvalue_array = np.zeros((dim_y, dim_x, dim))
|
||||
i0 = 0
|
||||
for y0 in y_array:
|
||||
j0 = 0
|
||||
if print_show==1:
|
||||
print(y0)
|
||||
for x0 in x_array:
|
||||
if print_show_more==1:
|
||||
print(x0)
|
||||
hamiltonian = hamiltonian_function(x0, y0)
|
||||
eigenvalue, eigenvector = np.linalg.eigh(hamiltonian)
|
||||
eigenvalue_array[i0, j0, :] = eigenvalue
|
||||
j0 += 1
|
||||
i0 += 1
|
||||
return eigenvalue_array
|
||||
|
||||
## wave functions
|
||||
|
||||
def calculate_eigenvector(hamiltonian):
|
||||
eigenvalue, eigenvector = np.linalg.eigh(hamiltonian)
|
||||
return eigenvector
|
||||
|
||||
## find vector with the same gauge
|
||||
|
||||
def find_vector_with_the_same_gauge_with_binary_search(vector_target, vector_ref, show_error=1, show_times=0, show_phase=0, n_test=10001, precision=1e-6):
|
||||
phase_1_pre = 0
|
||||
phase_2_pre = np.pi
|
||||
for i0 in range(n_test):
|
||||
test_1 = np.sum(np.abs(vector_target*cmath.exp(1j*phase_1_pre) - vector_ref))
|
||||
test_2 = np.sum(np.abs(vector_target*cmath.exp(1j*phase_2_pre) - vector_ref))
|
||||
if test_1 < precision:
|
||||
phase = phase_1_pre
|
||||
if show_times==1:
|
||||
print('Binary search times=', i0)
|
||||
break
|
||||
if i0 == n_test-1:
|
||||
phase = phase_1_pre
|
||||
if show_error==1:
|
||||
print('Gauge not found with binary search times=', i0)
|
||||
if test_1 < test_2:
|
||||
if i0 == 0:
|
||||
phase_1 = phase_1_pre-(phase_2_pre-phase_1_pre)/2
|
||||
phase_2 = phase_1_pre+(phase_2_pre-phase_1_pre)/2
|
||||
else:
|
||||
phase_1 = phase_1_pre
|
||||
phase_2 = phase_1_pre+(phase_2_pre-phase_1_pre)/2
|
||||
else:
|
||||
if i0 == 0:
|
||||
phase_1 = phase_2_pre-(phase_2_pre-phase_1_pre)/2
|
||||
phase_2 = phase_2_pre+(phase_2_pre-phase_1_pre)/2
|
||||
else:
|
||||
phase_1 = phase_2_pre-(phase_2_pre-phase_1_pre)/2
|
||||
phase_2 = phase_2_pre
|
||||
phase_1_pre = phase_1
|
||||
phase_2_pre = phase_2
|
||||
vector_target = vector_target*cmath.exp(1j*phase)
|
||||
if show_phase==1:
|
||||
print('Phase=', phase)
|
||||
return vector_target
|
||||
|
||||
def find_vector_with_fixed_gauge_by_making_one_component_real(vector, precision=0.005, index=None):
|
||||
if index == None:
|
||||
index = np.argmax(np.abs(vector))
|
||||
sign_pre = np.sign(np.imag(vector[index]))
|
||||
for phase in np.arange(0, 2*np.pi, precision):
|
||||
sign = np.sign(np.imag(vector[index]*cmath.exp(1j*phase)))
|
||||
if np.abs(np.imag(vector[index]*cmath.exp(1j*phase))) < 1e-9 or sign == -sign_pre:
|
||||
break
|
||||
sign_pre = sign
|
||||
vector = vector*cmath.exp(1j*phase)
|
||||
if np.real(vector[index]) < 0:
|
||||
vector = -vector
|
||||
return vector
|
Reference in New Issue
Block a user