DPL94 derivatives

From relax wiki
Jump to: navigation, search

These are the first and second partial derivatives of the equations of the DPL94 relaxation dispersion model.

Equation

[math] \mathrm{R}_{1\rho}= \mathrm{R}_1\cos^2\theta + \left( \mathrm{R}_{1\rho}{´} + \frac{\Phi_\textrm{ex} \textrm{k}_\textrm{ex}}{\textrm{k}_\textrm{ex}^2 + \omega_\textrm{e}^2} \right) \sin^2\theta [/math]

Jacobian

sympy

from sympy import *
 
# In contrast to other Computer Algebra Systems, in SymPy you have to declare symbolic variables explicitly:
R1 = Symbol('R1')
theta = Symbol('theta')
R1rho_p = Symbol('R1rho_p')
phi_ex = Symbol('phi_ex')
kex = Symbol('kex')
we = Symbol('we')
 
# Define function
f = R1 * cos(theta)**2 + (R1rho_p + ( (phi_ex * kex) / (kex**2 + we**2) ) ) * sin(theta)**2
 
print("Now calculate the Jacobian. The partial derivative matrix.\n")
print("Jacobian is m rows with function derivatives and n columns of parameters.")
 
d_f_d_R1 = diff(f, R1)
d_f_d_theta = diff(f, theta)
d_f_d_R1rho_p = diff(f, R1rho_p)
d_f_d_phi_ex = diff(f, phi_ex)
d_f_d_kex = diff(f, kex)
d_f_d_we = diff(f, we)
 
print("""Form the Jacobian matrix by:
------------------------------------------------------------------------------
from numpy import array, cos, sin, pi, transpose
 
R1 = 1.1
theta = pi / 4
R1rho_p = 10.
phi_ex = 1100.
kex = 2200.
we = 3300.
 
d_f_d_R1 = %s
d_f_d_theta = %s
d_f_d_R1rho_p = %s
d_f_d_phi_ex = %s
d_f_d_kex = %s
d_f_d_we = %s
jacobian_matrix = transpose(array( [d_f_d_R1 , d_f_d_theta, d_f_d_R1rho_p, d_f_d_phi_ex, d_f_d_kex, d_f_d_we] ) )
 
print jacobian_matrix
------------------------------------------------------------------------------
""" % (d_f_d_R1, d_f_d_theta, d_f_d_R1rho_p, d_f_d_phi_ex, d_f_d_kex, d_f_d_we) )
 
#### Method 2
# http://docs.sympy.org/0.7.2/modules/matrices/matrices.html
 
# The vectorial function.
X = Matrix([R1 * cos(theta)**2 + (R1rho_p + ( (phi_ex * kex) / (kex**2 + we**2) ) ) * sin(theta)**2])
# What to derive for.
Y = Matrix([R1, theta, R1rho_p, phi_ex, kex, we])
 
# Make the Jacobian
Jacobian = X.jacobian(Y)
 
jac_string = str(Jacobian)
jac_string_arr = jac_string.replace("Matrix", "array")
 
print("""Form the Jacobian matrix by:
------------------------------------------------------------------------------
from numpy import array, cos, sin, pi, transpose
 
R1 = 1.1
theta = pi / 4
R1rho_p = 10.
phi_ex = 1100.
kex = 2200.
we = 3300.
 
jacobian_matrix_2 = %s
 
print jacobian_matrix_2
------------------------------------------------------------------------------
""" % (jac_string_arr) )

output

output is

from numpy import array, cos, sin, pi, transpose
 
R1 = 1.1
theta = pi / 4
R1rho_p = 10.
phi_ex = 1100.
kex = 2200.
we = 3300.
 
d_f_d_R1 = cos(theta)**2
d_f_d_theta = -2*R1*sin(theta)*cos(theta) + 2*(R1rho_p + kex*phi_ex/(kex**2 + we**2))*sin(theta)*cos(theta)
d_f_d_R1rho_p = sin(theta)**2
d_f_d_phi_ex = kex*sin(theta)**2/(kex**2 + we**2)
d_f_d_kex = (-2*kex**2*phi_ex/(kex**2 + we**2)**2 + phi_ex/(kex**2 + we**2))*sin(theta)**2
d_f_d_we = -2*kex*phi_ex*we*sin(theta)**2/(kex**2 + we**2)**2
jacobian_matrix = transpose(array( [d_f_d_R1 , d_f_d_theta, d_f_d_R1rho_p, d_f_d_phi_ex, d_f_d_kex, d_f_d_we] ) )
 
print jacobian_matrix
 
R1 = 1.1
theta = pi / 4
R1rho_p = 10.
phi_ex = 1100.
kex = 2200.
we = 3300.
 
jacobian_matrix_2 = array([[cos(theta)**2, -2*R1*sin(theta)*cos(theta) + 2*(R1rho_p + kex*phi_ex/(kex**2 + we**2))*sin(theta)*cos(theta), 
sin(theta)**2, kex*sin(theta)**2/(kex**2 + we**2), (-2*kex**2*phi_ex/(kex**2 + we**2)**2 + phi_ex/(kex**2 + we**2))*sin(theta)**2, -2*kex*phi_ex*we*sin(theta)**2/(kex**2 + we**2)**2]])
 
print jacobian_matrix_2

Hessian

sympy

from sympy import *
 
# In contrast to other Computer Algebra Systems, in SymPy you have to declare symbolic variables explicitly:
R1 = Symbol('R1')
theta = Symbol('theta')
R1rho_p = Symbol('R1rho_p')
phi_ex = Symbol('phi_ex')
kex = Symbol('kex')
we = Symbol('we')
 
# Define function
f = R1 * cos(theta)**2 + (R1rho_p + ( (phi_ex * kex) / (kex**2 + we**2) ) ) * sin(theta)**2
 
print("Now calculate the Hessian.\n")
 
# Define symbols to derive for.
syms = [R1, theta, R1rho_p, phi_ex, kex, we]
 
# Do the hessian.
hess = hessian(f, syms)
 
hess_string = str(hess)
hess_string_arr = hess_string.replace("Matrix", "array")
 
print("""Form the Hessian matrix by:
------------------------------------------------------------------------------
from numpy import array, cos, sin, pi
 
R1 = 1.1
theta = pi / 4
R1rho_p = 10.
phi_ex = 1100.
kex = 2200.
we = 3300.
 
hessian_matrix = %s
 
print hessian_matrix
------------------------------------------------------------------------------
""" % (hess_string_arr) )

output

output is

from numpy import array, cos, sin, pi
 
R1 = 1.1
theta = pi / 4
R1rho_p = 10.
phi_ex = 1100.
kex = 2200.
we = 3300.
 
hessian_matrix = array([[0, -2*sin(theta)*cos(theta), 0, 0, 0, 0], 
[-2*sin(theta)*cos(theta), 2*R1*sin(theta)**2 - 2*R1*cos(theta)**2 - 2*(R1rho_p + kex*phi_ex/(kex**2 + we**2))*sin(theta)**2 + 2*(R1rho_p + kex*phi_ex/(kex**2 + we**2))*cos(theta)**2, 2*sin(theta)*cos(theta), 
2*kex*sin(theta)*cos(theta)/(kex**2 + we**2), 2*(-2*kex**2*phi_ex/(kex**2 + we**2)**2 + phi_ex/(kex**2 + we**2))*sin(theta)*cos(theta), -4*kex*phi_ex*we*sin(theta)*cos(theta)/(kex**2 + we**2)**2], 
[0, 2*sin(theta)*cos(theta), 0, 0, 0, 0], [0, 2*kex*sin(theta)*cos(theta)/(kex**2 + we**2), 0, 0, -2*kex**2*sin(theta)**2/(kex**2 + we**2)**2 + sin(theta)**2/(kex**2 + we**2), -2*kex*we*sin(theta)**2/(kex**2 + we**2)**2], 
[0, 2*(-2*kex**2*phi_ex/(kex**2 + we**2)**2 + phi_ex/(kex**2 + we**2))*sin(theta)*cos(theta), 0, -2*kex**2*sin(theta)**2/(kex**2 + we**2)**2 + sin(theta)**2/(kex**2 + we**2), 
(8*kex**3*phi_ex/(kex**2 + we**2)**3 - 6*kex*phi_ex/(kex**2 + we**2)**2)*sin(theta)**2, (8*kex**2*phi_ex*we/(kex**2 + we**2)**3 - 2*phi_ex*we/(kex**2 + we**2)**2)*sin(theta)**2], 
[0, -4*kex*phi_ex*we*sin(theta)*cos(theta)/(kex**2 + we**2)**2, 0, -2*kex*we*sin(theta)**2/(kex**2 + we**2)**2, 
(8*kex**2*phi_ex*we/(kex**2 + we**2)**3 - 2*phi_ex*we/(kex**2 + we**2)**2)*sin(theta)**2, 8*kex*phi_ex*we**2*sin(theta)**2/(kex**2 + we**2)**3 - 2*kex*phi_ex*sin(theta)**2/(kex**2 + we**2)**2]])
 
print hessian_matrix

See also