def test_convert_points_to_sph(self):
"""Test convert random points to spherical based on atomic structure."""
rad_pts = np.array([0.1, 0.5, 1])
rad_wts = np.array([0.3, 0.4, 0.3])
rad_grid = OneDGrid(rad_pts, rad_wts)
center = np.random.rand(3)
atgrid = AtomGrid(rad_grid, degrees=[7], center=center)
points = np.random.rand(100, 3)
calc_sph = atgrid.convert_cart_to_sph(points)
# compute ref result
ref_coor = points - center
# radius
r = np.linalg.norm(ref_coor, axis=-1)
# azimuthal
theta = np.arctan2(ref_coor[:, 1], ref_coor[:, 0])
# polar
phi = np.arccos(ref_coor[:, 2] / r)
assert_allclose(np.stack([r, theta, phi]).T, calc_sph)
assert_equal(calc_sph.shape, (100, 3))
# test single point
point = np.random.rand(3)
calc_sph = atgrid.convert_cart_to_sph(point)
ref_coor = point - center
r = np.linalg.norm(ref_coor)
theta = np.arctan2(ref_coor[1], ref_coor[0])
phi = np.arccos(ref_coor[2] / r)
assert_allclose(np.array([r, theta, phi]).reshape(-1, 3), calc_sph)
def test_poisson_solve_mtr_cmpl(self):
"""Test solve poisson equation and interpolate the result."""
oned = GaussChebyshev(50)
btf = BeckeRTransform(1e-7, 1.5)
rad = btf.transform_1d_grid(oned)
l_max = 7
atgrid = AtomGrid(rad, degrees=[l_max])
value_array = self.helper_func_gauss(atgrid.points)
p_0 = atgrid.integrate(value_array)
# test density sum up to np.pi**(3 / 2)
assert_allclose(p_0, np.pi**1.5, atol=1e-4)
sph_coor = atgrid.convert_cart_to_sph()[:, 1:3]
spls_mt = Poisson._proj_sph_value(
atgrid.rgrid,
sph_coor,
l_max // 2,
value_array,
atgrid.weights,
atgrid.indices,
)
ibtf = InverseRTransform(btf)
linsp = np.linspace(-1, 0.99, 50)
bound = p_0 * np.sqrt(4 * np.pi)
pois_mtr = Poisson.solve_poisson(spls_mt, linsp, bound, tfm=ibtf)
assert pois_mtr.shape == (7, 4)
near_rg_pts = np.array([1e-2, 0.1, 0.2, 0.3, 0.5, 0.7, 1.0, 1.2])
near_tf_pts = ibtf.transform(near_rg_pts)
ref_short_res = [
6.28286, # 0.01
6.26219, # 0.1
6.20029, # 0.2
6.09956, # 0.3
5.79652, # 0.5
5.3916, # 0.7
4.69236, # 1.0
4.22403, # 1.2
]
for i, j in enumerate(near_tf_pts):
assert_almost_equal(
Poisson.interpolate_radial(pois_mtr, j, 0, True) /
near_rg_pts[i],
ref_short_res[i] * np.sqrt(4 * np.pi),
decimal=3,
)
matrix_result = Poisson.interpolate_radial(pois_mtr, j)
assert_almost_equal(
matrix_result[0, 0] / near_rg_pts[i],
ref_short_res[i] * np.sqrt(4 * np.pi),
decimal=3,
)
# test interpolate with sph
result = Poisson.interpolate(pois_mtr, j, np.random.rand(5),
np.random.rand(5))
assert_allclose(result / near_rg_pts[i] - ref_short_res[i],
np.zeros(5),
atol=1e-3)
def test_poisson_proj(self):
"""Test the project function."""
oned = GaussChebyshev(30)
btf = BeckeRTransform(0.0001, 1.5)
rad = btf.transform_1d_grid(oned)
l_max = 7
atgrid = AtomGrid(rad, degrees=[l_max])
value_array = self.helper_func_gauss(atgrid.points)
spl_res = spline_with_atomic_grid(atgrid, value_array)
# test for random, r, theta, phi
for _ in range(20):
r = np.random.rand(1)[0] * 2
theta = np.random.rand(10) * 3.14
phi = np.random.rand(10) * 3.14
result = interpolate(spl_res, r, theta, phi)
x = r * np.sin(phi) * np.cos(theta)
y = r * np.sin(phi) * np.sin(theta)
z = r * np.cos(phi)
result_ref = self.helper_func_gauss(np.array([x, y, z]).T)
# assert similar value less than 1e-4 discrepancy
assert_allclose(result, result_ref, atol=1e-4)
sph_coor = atgrid.convert_cart_to_sph()[:, 1:3]
spls_mt = Poisson._proj_sph_value(
atgrid.rgrid,
sph_coor,
l_max // 2,
value_array,
atgrid.weights,
atgrid.indices,
)
# test each point is the same
for _ in range(20):
r = np.random.rand(1)[0] * 2
# random spherical point
int_res = np.zeros((l_max, l_max // 2 + 1), dtype=float)
for j in range(l_max // 2 + 1):
for i in range(-j, j + 1):
int_res[i, j] += spls_mt[i, j](r)
assert_allclose(spl_res(r), int_res)
def spline_with_atomic_grid(atgrid: AtomGrid, func_vals: np.ndarray):
r"""
Return spline to interpolate radial components wrt to expansion in real spherical harmonics.
For fixed r, a function :math:`f(r, \theta, \phi)` is projected onto the spherical
harmonic expansion
.. math::
f(r, \theta, \phi) = \sum_{l=0}^\infty \sum_{m=-l}^l \rho^{lm}(r) Y^m_l(\theta, \phi)
The coefficients :math:`\rho^{lm}(r)` are interpolated on the radial component using a cubic
spline for each consequent :math:`r` values, where one can evaluate :math:`f` on any set
of points.
Parameters
----------
atgrid : AtomGrid
Atomic grid which contains a radial grid for integrating on radial axis and
angular grid for integrating on spherical angles (the unit sphere).
func_vals : ndarray(N,)
Function values on each point on the atomic grid.
Returns
-------
scipy.CubicSpline
CubicSpline object for interpolating the coefficients :math:`\rho^{lm}(r)`
on the radial coordinate :math:`r` based on the spherical harmonic expansion
(for a fixed :math:`r`). The input of spline is array of :math:`N` points in
:math:`[0, \infty)` and the output is (N, M, L) matrix with (i, m, l) matrix
entries :math:`\rho^{lm}(r_i)`.
"""
l_max = atgrid.l_max // 2
# Convert grid points from Cartesian to Spherical coordinates
sph_coor = atgrid.convert_cart_to_sph()
# Construct all real spherical harmonics up to degree l_max on the grid points
r_sph = generate_real_sph_harms(l_max, sph_coor[:, 1], sph_coor[:, 2])
return spline_with_sph_harms(
r_sph, func_vals, atgrid.weights, atgrid.indices, atgrid.rgrid
)
def test_raises_errors(self):
"""Test proper error raises."""
oned = GaussChebyshev(50)
btf = BeckeRTransform(1e-7, 1.5)
rad = btf.transform_1d_grid(oned)
l_max = 7
atgrid = AtomGrid(rad, degrees=[l_max])
value_array = self.helper_func_gauss(atgrid.points)
p_0 = atgrid.integrate(value_array)
# test density sum up to np.pi**(3 / 2)
assert_allclose(p_0, np.pi**1.5, atol=1e-4)
sph_coor = atgrid.convert_cart_to_sph()
with self.assertRaises(ValueError):
Poisson._proj_sph_value(
atgrid.rgrid,
sph_coor,
l_max // 2,
value_array,
atgrid.weights,
atgrid.indices,
)
def test_poisson_solve(self):
"""Test the poisson solve function."""
oned = GaussChebyshev(30)
oned = GaussChebyshev(50)
btf = BeckeRTransform(1e-7, 1.5)
rad = btf.transform_1d_grid(oned)
l_max = 7
atgrid = AtomGrid(rad, degrees=[l_max])
value_array = self.helper_func_gauss(atgrid.points)
p_0 = atgrid.integrate(value_array)
# test density sum up to np.pi**(3 / 2)
assert_allclose(p_0, np.pi**1.5, atol=1e-4)
sph_coor = atgrid.convert_cart_to_sph()[:, 1:3]
spls_mt = Poisson._proj_sph_value(
atgrid.rgrid,
sph_coor,
l_max // 2,
value_array,
atgrid.weights,
atgrid.indices,
)
# test splines project fit gauss function well
def gauss(r):
return np.exp(-(r**2))
for _ in range(20):
coors = np.random.rand(10, 3)
r = np.linalg.norm(coors, axis=-1)
spl_0_0 = spls_mt[0, 0]
interp_v = spl_0_0(r)
ref_v = gauss(r) * np.sqrt(4 * np.pi)
# 0.28209479 is the value in spherical harmonic Z_0_0
assert_allclose(interp_v, ref_v, atol=1e-3)
ibtf = InverseRTransform(btf)
linsp = np.linspace(-1, 0.99, 50)
bound = p_0 * np.sqrt(4 * np.pi)
res_bv = Poisson.solve_poisson_bv(spls_mt[0, 0],
linsp,
bound,
tfm=ibtf)
near_rg_pts = np.array([1e-2, 0.1, 0.2, 0.3, 0.5, 0.7, 1.0, 1.2])
near_tf_pts = ibtf.transform(near_rg_pts)
long_rg_pts = np.array([2, 3, 4, 5, 6, 7, 8, 9, 10])
long_tf_pts = ibtf.transform(long_rg_pts)
short_res = res_bv(near_tf_pts)[0] / near_rg_pts / (2 * np.sqrt(np.pi))
long_res = res_bv(long_tf_pts)[0] / long_rg_pts / (2 * np.sqrt(np.pi))
# ref are calculated with mathemetical
# integrate[exp[-x^2 - y^2 - z^2] / sqrt[(x - a)^2 + y^2 +z^2], range]
ref_short_res = [
6.28286, # 0.01
6.26219, # 0.1
6.20029, # 0.2
6.09956, # 0.3
5.79652, # 0.5
5.3916, # 0.7
4.69236, # 1.0
4.22403, # 1.2
]
ref_long_res = [
2.77108, # 2
1.85601, # 3
1.39203, # 4
1.11362, # 5
0.92802, # 6
0.79544, # 7
0.69601, # 8
0.61867, # 9
0.55680, # 10
]
assert_allclose(short_res, ref_short_res, atol=5e-4)
assert_allclose(long_res, ref_long_res, atol=5e-4)
# solve same poisson equation with gauss directly
gauss_pts = btf.transform(linsp)
res_gs = Poisson.solve_poisson_bv(gauss, gauss_pts, p_0)
gs_int_short = res_gs(near_rg_pts)[0] / near_rg_pts
gs_int_long = res_gs(long_rg_pts)[0] / long_rg_pts
assert_allclose(gs_int_short, ref_short_res, 5e-4)
assert_allclose(gs_int_long, ref_long_res, 5e-4)
