User:Geek3/hydrogen-cloud

From Wikimedia Commons, the free media repository
Jump to navigation Jump to search
Image created with hydrogen-cloud

hydrogen-cloud is a python script that creates 3D images of hydrogen wavefunctions, depicting the probability density as cloud opacity. For solid 3D wavefunctions with finite cutoff probability density, see my other script hydrogen.

About hydrogen-cloud[edit]

hydrogen-cloud uses the SciPy library to compute the analytic functions of atomic single-electron orbitals. Images are written with the Python Imaging Library.

Code[edit]

The code can be executed with a python 3 interpreter.

#!/usr/bin/python3
# -*- coding: utf8 -*-

'''
Copyright 2018 by Geek3, https://commons.wikimedia.org/wiki/User:Geek3
Licensed under the GNU General Public License 3.0 or later.
'''

from PIL import Image
import numpy as np
import scipy.optimize as op
import scipy.linalg as la
from scipy.special import eval_genlaguerre, lpmv
import time
from math import *


def hls_to_rgb(h, l, s):
    '''
    vectorized function is much faster than point-wise colorsys.hls_to_rgb()
    https://en.wikipedia.org/wiki/HSL_and_HSV#From_HSL
    '''
    shape = np.shape(h)
    h = np.ravel(h) % 1.
    l = np.clip(np.ravel(l), 0., 1.)
    s = np.clip(np.ravel(s), 0., 1.)
    h6 = h * 6.
    h6i = h6.astype(np.int)
    C = (1. - np.fabs(2. * l - 1.)) * s
    X = C * (1. - np.fabs(h6 % 2. - 1.))
    m = l - 0.5 * C
    RGB = np.zeros((len(h), 3))
    RGB[:,0] = np.choose(h6i, [C, X, 0., 0., X, C]) + m
    RGB[:,1] = np.choose(h6i, [X, C, C, X, 0., 0.]) + m
    RGB[:,2] = np.choose(h6i, [0., 0., X, C, C, X]) + m
    return RGB.reshape(shape + (3,))


def Rnl(n, l, r):
    '''
    radial part of the wavefunction. r may be an array
    '''
    rho = np.abs(r) * (2 / n)
    L = eval_genlaguerre(n - l - 1, 2 * l + 1, rho)
    p = factorial(n + l) / factorial(n - l - 1)
    return sqrt((2/n)**3 / (2*n*p)) * np.exp(-rho/2) * rho**l * L


def psinlm(n, l, m, x, y, z):
    '''
    hydrogen atom wavefunction. x, y, z may be arrays of equal length
    fast evaluation, avoiding trigonometric functions alltogether
    '''
    assert n >= 1 and l >= 0 and l < n and m >= -l and m <= l
    xxyy = x * x + y * y
    rxy = np.sqrt(xxyy)
    r = np.sqrt(xxyy + z * z)
    R = Rnl(n, l, r)
    
    N = sqrt((2 * l + 1) * factorial(l - m) / (factorial(l + m) * 4 * pi))
    costheta = np.clip(z / r, -1, 1)
    # compute sph_harm() without using trigonometric functions
    Ylm = N * lpmv(m, l, costheta) * ((x + 1j * y) / rxy) ** m
    return R * Ylm


def vnorm(x):
    d = sqrt(np.sum(x * x))
    if d != 0.: return np.array(x) / d
    return np.array(x)


def rtp_to_xyz(rtp):
    st, ct = sin(rtp[1]), cos(rtp[1])
    sp, cp = sin(rtp[2]), cos(rtp[2])
    return rtp[0] * np.array([cp * st, sp * st, ct])


def xyz_to_rtp(xyz):
    r = la.norm(xyz)
    t = atan2(hypot(xyz[0], xyz[1]), xyz[2])
    p = atan2(xyz[1], xyz[0])
    return np.array([r, t, p])


def phong_brightness(vlight, vsurf, vview, ambient=0.35,
        diffuse=0.4, diffuse_dark = 0.2, specular=0.23, specularity=13.):
    '''
    phong shading, vsurf and vview may be arrays of vectors
    vlight: vector of light direction
    vsurf: vectors of surface normal vectors (or density gradients)
    vview: vectors of viewing direction
    '''
    N = vsurf.shape[1]
    vlight /= la.norm(vlight)
    vsurf /= la.norm(vsurf, axis=0)
    vview /= la.norm(vview, axis=0)
    prod_light_surf = np.dot(vlight, vsurf)
    vreflect = 2. * prod_light_surf * vsurf - np.tile(vlight, (N, 1)).T
    vreflect /= la.norm(vreflect, axis=0)
    amb = ambient * np.ones(N)
    diffuse_frac = np.dot(vlight, vsurf)
    dif = diffuse * diffuse_frac
    dif[diffuse_frac < 0.] = diffuse_dark * diffuse_frac[diffuse_frac < 0.]
    spec = specular * np.maximum(0., np.sum(vview *  vreflect, 0))**specularity
    # no specular reflection towards the inside of a surface or density gradient
    spec[prod_light_surf < 0.] = 0.
    lightness = np.clip(amb + dif + spec, 0, 1)
    return lightness


def draw_orbital(nlm, w=200, fname=None, density=None, camera_phi=radians(-90),
        camera_theta=radians(60), light_phi=radians(30), light_theta=0.7,
        angle_of_view=atan(4 / 3), view_center=[0,0,0], zoom=None):
    '''
    creates a pixel graphic of an orbital.
    nlm: either quantum numbers [n, l, m] or a list [[n1, l1, m1, ampl1], ...]
    '''
    
    # shortcut for wavefunction with given parameters
    if type(nlm[0]) == int:
        n, l, m = nlm
        nmax = n
        def psi(x, y, z):
            return psinlm(n, l, m, x, y, z)
    else:
        # mix of different eigenfunctions
        # nlm = [[n1, l1, m1, amplitude1], [n2, l2, m2, amplitude2], ...]
        nmax = max([nlma[0] for nlma in nlm])
        def psi(x, y, z):
            return np.sum([a * psinlm(n, l, m, x, y, z) for n,l,m,a in nlm],
                          axis=0)
    
    if zoom is None:
        zoom = 1 / sqrt(1.0 + 0.6 / nmax**2)
    bohr_radii_per_halfwidth = 2.5 * nmax**2 / zoom
    h = w # image size
    unit = (w / 2) / bohr_radii_per_halfwidth
    
    # density0 scales overall cloud density
    if density is not None:
        density0 = density
    else:
        density0 = 12. * nmax**4
    
    view_center = np.array(view_center)
    
    # camera location
    camera = view_center + rtp_to_xyz([bohr_radii_per_halfwidth *
        sqrt(1 + (h / w)**2) / tan(radians(angle_of_view) / 2),
        camera_theta, camera_phi])
    
    # light source
    dlight = rtp_to_xyz([1.0, light_theta, light_phi])
    vm = view_center - camera
    z0 = np.array([0, 0, 1.0])
    
    # image plane axes
    image_z = vnorm(vm)
    image_y = vnorm(z0 - np.dot(z0, image_z) * image_z)
    image_x = vnorm(np.cross(image_z, z0))
    
    # draw
    im = Image.new('RGBA', (w, h))
    for ny in range(h):
        for nx in range(w):
            x, y = (nx - 0.5*(w-1)) / unit, (0.5*(h-1) - ny) / unit
            p2 = view_center + x * image_x + y * image_y
            
            # Computation of psi along line of sight has to be vectorized
            # with numpy for reasonable performance.
            Npoints = 501
            tmin = -6 * nmax**2
            tmax = 6 * nmax**2
            
            # Use sinh distribution function which samples denser at small
            # radii where the wavefunction has more features.
            tlin = np.linspace(asinh(tmin), asinh(tmax), Npoints)
            tlist = np.sinh(tlin)
            dt = np.cosh(tlin) * (tlin[1] - tlin[0])
            ray_vec = vnorm(p2 - camera)
            points = p2.reshape((3,1)) + np.outer(ray_vec, tlist)
            # Calls of psi are expensive (around 50 microseconds)
            # So combine the density and gradient calls.
            d_grad = 1e-3
            points_grad = np.tile(points, (1, 4))
            points_grad[0,Npoints:2*Npoints] += d_grad
            points_grad[1,Npoints*2:3*Npoints] += d_grad
            points_grad[2,Npoints*3:4*Npoints] += d_grad
            
            psi_array = psi(points_grad[0], points_grad[1], points_grad[2])
            psi2_array = np.abs(psi_array)**2
            
            density_array = psi2_array[:Npoints]
            # compute density gradient at each point
            grad = (psi2_array[Npoints:4*Npoints].reshape(3, -1)
                    - density_array) / d_grad
            grad_norm = np.sqrt(np.sum(grad*grad, axis=0))
            grad /= np.maximum(grad_norm, np.finfo(np.float).eps)
            
            # normalize the gradient
            rr = np.sum(points * points, axis=0)
            gradnorm_rpsi = grad_norm * rr * nmax**3
            
            gradient_scale = 25.
            grad_rel = 1.0 - np.exp(-gradient_scale * gradnorm_rpsi)
            
            # Color hue is determined by the phase of psi
            phases = np.angle(psi_array[:Npoints])
            
            # Lightness is determined by density gradient with phong shading
            phongs = phong_brightness(dlight, -grad,
                                    np.tile(camera, (Npoints, 1)).T - points)
            # Put ambient brightness where gradient is low
            ambient = 0.35
            lightness = phongs * grad_rel + ambient * (1. - grad_rel)
            
            colors = 256. * hls_to_rgb((phases/(2*pi) - 1/3) % 1.,
                                       lightness, np.ones_like(lightness))
            
            # Sum up the colors along viewing direction
            dens_integral = np.cumsum(dt * density_array)
            # This is the https://en.wikipedia.org/wiki/Beer-Lambert_law
            opacity = 1. - np.exp(-density0 * dens_integral)
            weights = np.copy(opacity)
            weights[1:] -= opacity[:-1]
            total_opacity = opacity[-1]
            
            if total_opacity <= 0.:
                total_color = [0.,0.,0.]
                total_opacity = 0.
            else:
                weights /= total_opacity
                total_color = np.dot(weights, colors)
            
            rgba = np.concatenate((total_color, (256. * total_opacity,)))
            
            rgba_tuple = tuple(np.clip(rgba, 0, 255).astype('uint8'))
            im.putpixel((nx,ny), rgba_tuple)
        
        # print status
        outstr = ' row ' + str(ny+1) + ' of ' + str(h) + ' complete'
        print('\b{0}{1}'.format(outstr, '\b' * len(outstr)), end='', flush=True)
    
    if fname is None:
        fname = 'hydrogen_n' + str(n) + '_l' + str(l) + '_m' + str(m) + '.png'
    else:
        if len(fname) < 4 or fname[-4] != '.':
            fname += '.png'
    im.save(fname, optimize=1)
    print('image written to', fname)


def main():
    imgsize = 100
    for n in range(1, 5+1):
        for l in range(0, n):
            for m in range(-l, l+1):
                fname = 'atomic-orbital-cloud_n{}_l{}_m{}.png'.format(n, l, m)
                print(fname)
                nlm = [n, l, m]
                draw_orbital(nlm, w=imgsize, fname=fname)


    for n in range(2, 5+1):
        if n >= 2:
            fname = 'atomic-orbital-cloud_n{}_px.png'.format(n)
            print(fname)
            nlm = [[n, 1, 1, -sqrt(0.5)], [n, 1, -1, +sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_py.png'.format(n)
            print(fname)
            nlm = [[n, 1, 1, 1j*sqrt(0.5)], [n, 1, -1, 1j*sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)

        if n >= 3:
            fname = 'atomic-orbital-cloud_n{}_dxz.png'.format(n)
            print(fname)
            nlm = [[n, 2, 1, -sqrt(0.5)], [n, 2, -1, +sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_dyz.png'.format(n)
            print(fname)
            nlm = [[n, 2, 1, 1j*sqrt(0.5)], [n, 2, -1, 1j*sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_dx^2-y^2.png'.format(n)
            print(fname)
            nlm = [[n, 2, 2, +sqrt(0.5)], [n, 2, -2, +sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_dxy.png'.format(n)
            print(fname)
            nlm = [[n, 2, 2, -1j*sqrt(0.5)], [n, 2, -2, 1j*sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)

        if n >= 4:
            fname = 'atomic-orbital-cloud_n{}_fxz^2.png'.format(n)
            print(fname)
            nlm = [[n, 3, 1, +sqrt(0.5)], [n, 3, -1, -sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_fyz^2.png'.format(n)
            print(fname)
            nlm = [[n, 3, 1, -1j*sqrt(0.5)], [n, 3, -1, -1j*sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_fz(x^2-y^2).png'.format(n)
            print(fname)
            nlm = [[n, 3, 2, +sqrt(0.5)], [n, 3, -2, +sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_fxyz.png'.format(n)
            print(fname)
            nlm = [[n, 3, 2, -1j*sqrt(0.5)], [n, 3, -2, 1j*sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_fx(x^2-3y^2).png'.format(n)
            print(fname)
            nlm = [[n, 3, 3, -sqrt(0.5)], [n, 3, -3, +sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)
            
            fname = 'atomic-orbital-cloud_n{}_fy(y^2-3x^2).png'.format(n)
            print(fname)
            nlm = [[n, 3, 3, 1j*sqrt(0.5)], [n, 3, -3, 1j*sqrt(0.5)]]
            draw_orbital(nlm, w=imgsize, fname=fname)

main()

Requirements[edit]

hydrogen-cloud is a python script using numpy arrays. Hence you need the following programs on your computer:

Usage[edit]

Add a command which orbital to create at the end of the source code:

draw_orbital([n, l, m])

The script can be executed with Python.

Images[edit]

See Wikimedia Commons search for a list of created images.