source: sasmodels/explore/realspace.py @ 226473d

core_shell_microgelsmagnetic_modelticket-1257-vesicle-productticket_1156ticket_1265_superballticket_822_more_unit_tests
Last change on this file since 226473d was 226473d, checked in by Paul Kienzle <pkienzle@…>, 7 years ago

add Iqxy to realspace explorer

  • Property mode set to 100644
File size: 18.3 KB
RevLine 
[cfa28d3]1from __future__ import division, print_function
2
3import time
4from copy import copy
5
6import numpy as np
7from numpy import pi, radians, sin, cos, sqrt
8from numpy.random import poisson, uniform
[a1c32c2]9from numpy.polynomial.legendre import leggauss
[cfa28d3]10from scipy.integrate import simps
11from scipy.special import j1 as J1
12
13# Definition of rotation matrices comes from wikipedia:
14#    https://en.wikipedia.org/wiki/Rotation_matrix#Basic_rotations
15def Rx(angle):
16    """Construct a matrix to rotate points about *x* by *angle* degrees."""
17    a = radians(angle)
18    R = [[1, 0, 0],
19         [0, +cos(a), -sin(a)],
20         [0, +sin(a), +cos(a)]]
21    return np.matrix(R)
22
23def Ry(angle):
24    """Construct a matrix to rotate points about *y* by *angle* degrees."""
25    a = radians(angle)
26    R = [[+cos(a), 0, +sin(a)],
27         [0, 1, 0],
28         [-sin(a), 0, +cos(a)]]
29    return np.matrix(R)
30
31def Rz(angle):
32    """Construct a matrix to rotate points about *z* by *angle* degrees."""
33    a = radians(angle)
34    R = [[+cos(a), -sin(a), 0],
35         [+sin(a), +cos(a), 0],
36         [0, 0, 1]]
37    return np.matrix(R)
38
39def rotation(theta, phi, psi):
40    """
41    Apply the jitter transform to a set of points.
42
43    Points are stored in a 3 x n numpy matrix, not a numpy array or tuple.
44    """
45    return Rx(phi)*Ry(theta)*Rz(psi)
46
[226473d]47def apply_view(points, view):
48    """
49    Apply the view transform (theta, phi, psi) to a set of points.
50
51    Points are stored in a 3 x n numpy array.
52
53    View angles are in degrees.
54    """
55    theta, phi, psi = view
56    return np.asarray((Rz(phi)*Ry(theta)*Rz(psi))*np.matrix(points.T)).T
57
58
59def invert_view(qx, qy, view):
60    """
61    Return (qa, qb, qc) for the (theta, phi, psi) view angle at detector
62    pixel (qx, qy).
63
64    View angles are in degrees.
65    """
66    theta, phi, psi = view
67    q = np.vstack((qx.flatten(), qy.flatten(), 0*qx.flatten()))
68    return np.asarray((Rz(-psi)*Ry(-theta)*Rz(-phi))*np.matrix(q))
69
70
[cfa28d3]71class Shape:
72    rotation = np.matrix([[1., 0, 0], [0, 1, 0], [0, 0, 1]])
73    center = np.array([0., 0., 0.])[:, None]
74    r_max = None
75
76    def volume(self):
77        # type: () -> float
78        raise NotImplementedError()
79
80    def sample(self, density):
81        # type: (float) -> np.ndarray[N], np.ndarray[N, 3]
82        raise NotImplementedError()
83
84    def rotate(self, theta, phi, psi):
85        self.rotation = rotation(theta, phi, psi) * self.rotation
86        return self
87
88    def shift(self, x, y, z):
89        self.center = self.center + np.array([x, y, z])[:, None]
90        return self
91
92    def _adjust(self, points):
93        points = np.asarray(self.rotation * np.matrix(points.T)) + self.center
94        return points.T
95
96    def r_bins(self, q, over_sampling=1, r_step=0.):
97        r_max = min(2 * pi / q[0], self.r_max)
98        if r_step == 0.:
99            r_step = 2 * pi / q[-1] / over_sampling
100        #r_step = 0.01
101        return np.arange(r_step, r_max, r_step)
102
103class Composite(Shape):
104    def __init__(self, shapes, center=(0, 0, 0), orientation=(0, 0, 0)):
105        self.shapes = shapes
106        self.rotate(*orientation)
107        self.shift(*center)
108
109        # Find the worst case distance between any two points amongst a set
110        # of shapes independent of orientation.  This could easily be a
111        # factor of two worse than necessary, e.g., a pair of thin rods
112        # end-to-end vs the same pair side-by-side.
113        distances = [((s1.r_max + s2.r_max)/2
114                      + sqrt(np.sum((s1.center - s2.center)**2)))
115                     for s1 in shapes
116                     for s2 in shapes]
117        self.r_max = max(distances + [s.r_max for s in shapes])
118
119    def volume(self):
120        return sum(shape.volume() for shape in self.shapes)
121
122    def sample(self, density):
123        values, points = zip(*(shape.sample(density) for shape in self.shapes))
124        return np.hstack(values), self._adjust(np.vstack(points))
125
126class Box(Shape):
127    def __init__(self, a, b, c,
128                 value, center=(0, 0, 0), orientation=(0, 0, 0)):
129        self.value = np.asarray(value)
130        self.rotate(*orientation)
131        self.shift(*center)
132        self.a, self.b, self.c = a, b, c
133        self._scale = np.array([a/2, b/2, c/2])[None, :]
134        self.r_max = sqrt(a**2 + b**2 + c**2)
135
136    def volume(self):
137        return self.a*self.b*self.c
138
139    def sample(self, density):
140        num_points = poisson(density*self.a*self.b*self.c)
141        points = self._scale*uniform(-1, 1, size=(num_points, 3))
142        values = self.value.repeat(points.shape[0])
143        return values, self._adjust(points)
144
145class EllipticalCylinder(Shape):
146    def __init__(self, ra, rb, length,
147                 value, center=(0, 0, 0), orientation=(0, 0, 0)):
148        self.value = np.asarray(value)
149        self.rotate(*orientation)
150        self.shift(*center)
151        self.ra, self.rb, self.length = ra, rb, length
152        self._scale = np.array([ra, rb, length/2])[None, :]
153        self.r_max = sqrt(4*max(ra, rb)**2 + length**2)
154
155    def volume(self):
156        return pi*self.ra*self.rb*self.length
157
158    def sample(self, density):
159        # density of the bounding box
160        num_points = poisson(density*4*self.ra*self.rb*self.length)
161        points = uniform(-1, 1, size=(num_points, 3))
162        radius = points[:, 0]**2 + points[:, 1]**2
163        points = self._scale*points[radius <= 1]
164        values = self.value.repeat(points.shape[0])
165        return values, self._adjust(points)
166
167class TriaxialEllipsoid(Shape):
168    def __init__(self, ra, rb, rc,
169                 value, center=(0, 0, 0), orientation=(0, 0, 0)):
170        self.value = np.asarray(value)
171        self.rotate(*orientation)
172        self.shift(*center)
173        self.ra, self.rb, self.rc = ra, rb, rc
174        self._scale = np.array([ra, rb, rc])[None, :]
175        self.r_max = 2*max(ra, rb, rc)
176
177    def volume(self):
178        return 4*pi/3 * self.ra * self.rb * self.rc
179
180    def sample(self, density):
181        # randomly sample from a box of side length 2*r, excluding anything
182        # not in the ellipsoid
183        num_points = poisson(density*8*self.ra*self.rb*self.rc)
184        points = uniform(-1, 1, size=(num_points, 3))
185        radius = np.sum(points**2, axis=1)
186        points = self._scale*points[radius <= 1]
187        values = self.value.repeat(points.shape[0])
188        return values, self._adjust(points)
189
190class Helix(Shape):
191    def __init__(self, helix_radius, helix_pitch, tube_radius, tube_length,
192                 value, center=(0, 0, 0), orientation=(0, 0, 0)):
193        self.value = np.asarray(value)
194        self.rotate(*orientation)
195        self.shift(*center)
196        self.helix_radius, self.helix_pitch = helix_radius, helix_pitch
197        self.tube_radius, self.tube_length = tube_radius, tube_length
198        helix_length = helix_pitch * tube_length/sqrt(helix_radius**2 + helix_pitch**2)
199        self.r_max = sqrt((2*helix_radius + 2*tube_radius)*2
200                          + (helix_length + 2*tube_radius)**2)
201
202    def volume(self):
203        # small tube radius approximation; for larger tubes need to account
204        # for the fact that the inner length is much shorter than the outer
205        # length
206        return pi*self.tube_radius**2*self.tube_length
207
208    def points(self, density):
209        num_points = poisson(density*4*self.tube_radius**2*self.tube_length)
210        points = uniform(-1, 1, size=(num_points, 3))
211        radius = points[:, 0]**2 + points[:, 1]**2
212        points = points[radius <= 1]
213
214        # Based on math stackexchange answer by Jyrki Lahtonen
215        #     https://math.stackexchange.com/a/461637
216        # with helix along z rather than x [so tuples in answer are (z, x, y)]
217        # and with random points in the cross section (p1, p2) rather than
218        # uniform points on the surface (cos u, sin u).
219        a, R = self.tube_radius, self.helix_radius
220        h = self.helix_pitch
221        scale = 1/sqrt(R**2 + h**2)
222        t = points[:, 3] * (self.tube_length * scale/2)
223        cos_t, sin_t = cos(t), sin(t)
224
225        # rx = R*cos_t
226        # ry = R*sin_t
227        # rz = h*t
228        # nx = -a * cos_t * points[:, 1]
229        # ny = -a * sin_t * points[:, 1]
230        # nz = 0
231        # bx = (a * h/scale) * sin_t * points[:, 2]
232        # by = (-a * h/scale) * cos_t * points[:, 2]
233        # bz = a*R/scale
234        # x = rx + nx + bx
235        # y = ry + ny + by
236        # z = rz + nz + bz
237        u, v = (R - a*points[:, 1]), (a * h/scale)*points[:, 2]
238        x = u * cos_t + v * sin_t
239        y = u * sin_t - v * cos_t
240        z = a*R/scale + h * t
241
242        points = np.hstack((x, y, z))
243        values = self.value.repeat(points.shape[0])
244        return values, self._adjust(points)
245
[226473d]246def calc_Iqxy(qx, qy, rho, points, volume=1, view=(0, 0, 0)):
247    x, y, z = points.T
248    qx, qy = np.broadcast_arrays(qx, qy)
249    qa, qb, qc = invert_view(qx, qy, view)
250    rho, volume = np.broadcast_arrays(rho, volume)
251    values = rho*volume
252
253    # I(q) = |sum V(r) rho(r) e^(1j q.r)|^2 / sum V(r)
254    Iq = [abs(np.sum(values*np.exp(1j*(qa_k*x + qb_k*y + qc_k*z))))**2
255            for qa_k, qb_k, qc_k in zip(qa.flat, qb.flat, qc.flat)]
256    return np.array(Iq).reshape(qx.shape) / np.sum(volume)
257
[cfa28d3]258def _calc_Pr_nonuniform(r, rho, points):
259    # Make Pr a little be bigger than necessary so that only distances
260    # min < d < max end up in Pr
261    n_max = len(r)+1
262    extended_Pr = np.zeros(n_max+1, 'd')
263    # r refers to bin centers; find corresponding bin edges
264    bins = bin_edges(r)
265    t_next = time.clock() + 3
266    for k, rho_k in enumerate(rho[:-1]):
267        distance = sqrt(np.sum((points[k] - points[k+1:])**2, axis=1))
268        weights = rho_k * rho[k+1:]
269        index = np.searchsorted(bins, distance)
270        # Note: indices may be duplicated, so "Pr[index] += w" will not work!!
271        extended_Pr += np.bincount(index, weights, n_max+1)
272        t = time.clock()
273        if t > t_next:
274            t_next = t + 3
275            print("processing %d of %d"%(k, len(rho)-1))
276    Pr = extended_Pr[1:-1]
277    return Pr / Pr.max()
278
279def calc_Pr(r, rho, points):
280    # P(r) with uniform steps in r is 3x faster; check if we are uniform
281    # before continuing
282    if np.max(np.abs(np.diff(r) - r[0])) > r[0]*0.01:
283        return _calc_Pr_nonuniform(r, rho, points)
284
285    # Make Pr a little be bigger than necessary so that only distances
286    # min < d < max end up in Pr
287    n_max = len(r)
288    extended_Pr = np.zeros(n_max+1, 'd')
289    t0 = time.clock()
290    t_next = t0 + 3
291    row_zero = np.zeros(len(rho), 'i')
292    for k, rho_k in enumerate(rho[:-1]):
293        distances = sqrt(np.sum((points[k] - points[k+1:])**2, axis=1))
294        weights = rho_k * rho[k+1:]
295        index = np.minimum(np.asarray(distances/r[0], 'i'), n_max)
296        # Note: indices may be duplicated, so "Pr[index] += w" will not work!!
297        extended_Pr += np.bincount(index, weights, n_max+1)
298        t = time.clock()
299        if t > t_next:
300            t_next = t + 3
301            print("processing %d of %d"%(k, len(rho)-1))
302    #print("time py:", time.clock() - t0)
303    Pr = extended_Pr[:-1]
304    # Make Pr independent of sampling density.  The factor of 2 comes because
305    # we are only accumulating the upper triangular distances.
306    #Pr = Pr * 2 / len(rho)**2
307    return Pr / Pr.max()
308
309    # Can get an additional 2x by going to C.  Cuda/OpenCL will allow even
310    # more speedup, though still bounded by the n^2 cose.
311    """
312void pdfcalc(int n, const double *pts, const double *rho,
313             int nPr, double *Pr, double rstep)
314{
315  int i,j;
316
317  for (i=0; i<n-2; i++) {
318    for (j=i+1; j<=n-1; j++) {
319      const double dxx=pts[3*i]-pts[3*j];
320      const double dyy=pts[3*i+1]-pts[3*j+1];
321      const double dzz=pts[3*i+2]-pts[3*j+2];
322      const double d=sqrt(dxx*dxx+dyy*dyy+dzz*dzz);
323      const int k=rint(d/rstep);
324      if (k < nPr) Pr[k]+=rho[i]*rho[j];
325    }
326  }
327}
328"""
329
330def j0(x):
331    return np.sinc(x/np.pi)
332
333def calc_Iq(q, r, Pr):
334    Iq = np.array([simps(Pr * j0(qk*r), r) for qk in q])
335    #Iq = np.array([np.trapz(Pr * j0(qk*r), r) for qk in q])
336    Iq /= Iq[0]
337    return Iq
338
339# NOTE: copied from sasmodels/resolution.py
340def bin_edges(x):
341    """
342    Determine bin edges from bin centers, assuming that edges are centered
343    between the bins.
344
345    Note: this uses the arithmetic mean, which may not be appropriate for
346    log-scaled data.
347    """
348    if len(x) < 2 or (np.diff(x) < 0).any():
349        raise ValueError("Expected bins to be an increasing set")
350    edges = np.hstack([
351        x[0]  - 0.5*(x[1]  - x[0]),  # first point minus half first interval
352        0.5*(x[1:] + x[:-1]),        # mid points of all central intervals
353        x[-1] + 0.5*(x[-1] - x[-2]), # last point plus half last interval
354        ])
355    return edges
356
357def plot_calc(r, Pr, q, Iq, theory=None):
358    import matplotlib.pyplot as plt
359    plt.subplot(211)
360    plt.plot(r, Pr, '-', label="Pr")
361    plt.xlabel('r (A)')
362    plt.ylabel('Pr (1/A^2)')
363    plt.subplot(212)
364    plt.loglog(q, Iq, '-', label='from Pr')
365    plt.xlabel('q (1/A')
366    plt.ylabel('Iq')
367    if theory is not None:
368        plt.loglog(theory[0], theory[1], '-', label='analytic')
369        plt.legend()
370
[226473d]371def plot_calc_2d(qx, qy, Iqxy, theory=None):
372    import matplotlib.pyplot as plt
373    qx, qy = bin_edges(qx), bin_edges(qy)
374    #qx, qy = np.meshgrid(qx, qy)
375    if theory is not None:
376        plt.subplot(121)
377    plt.pcolormesh(qx, qy, np.log10(Iqxy))
378    plt.xlabel('qx (1/A)')
379    plt.ylabel('qy (1/A)')
380    if theory is not None:
381        plt.subplot(122)
382        plt.pcolormesh(qx, qy, np.log10(theory))
383        plt.xlabel('qx (1/A)')
384
[cfa28d3]385def plot_points(rho, points):
386    import mpl_toolkits.mplot3d
387    import matplotlib.pyplot as plt
388
389    ax = plt.axes(projection='3d')
390    try:
391        ax.axis('square')
392    except Exception:
393        pass
394    n = len(points)
395    index = np.random.choice(n, size=1000) if n > 1000 else slice(None, None)
396    ax.scatter(points[index, 0], points[index, 1], points[index, 2], c=rho[index])
397    #low, high = points.min(axis=0), points.max(axis=0)
398    #ax.axis([low[0], high[0], low[1], high[1], low[2], high[2]])
399    ax.autoscale(True)
400
401def sas_2J1x_x(x):
402    with np.errstate(all='ignore'):
403        retvalue = 2*J1(x)/x
404    retvalue[x == 0] = 1.
405    return retvalue
406
407def sas_3j1x_x(x):
408    """return 3*j1(x)/x"""
409    with np.errstate(all='ignore'):
410        retvalue = 3*(sin(x) - x*cos(x))/x**3
411    retvalue[x == 0.] = 1.
412    return retvalue
413
414def cylinder_Iq(q, radius, length):
[a1c32c2]415    z, w = leggauss(76)
416    cos_alpha = (z+1)/2
417    sin_alpha = sqrt(1.0 - cos_alpha**2)
[cfa28d3]418    Iq = np.empty_like(q)
419    for k, qk in enumerate(q):
[a1c32c2]420        qab, qc = qk*sin_alpha, qk*cos_alpha
421        Fq = sas_2J1x_x(qab*radius) * j0(qc*length/2)
422        Iq[k] = np.sum(w*Fq**2)
[cfa28d3]423    Iq = Iq/Iq[0]
424    return Iq
425
[226473d]426def cylinder_Iqxy(qx, qy, radius, length, view=(0, 0, 0)):
427    qa, qb, qc = invert_view(qx, qy, view)
428    qab = np.sqrt(qa**2 + qb**2)
429    Fq = sas_2J1x_x(qab*radius) * j0(qc*length/2)
430    Iq = Fq**2
431    return Iq.reshape(qx.shape)
432
[cfa28d3]433def sphere_Iq(q, radius):
434    Iq = sas_3j1x_x(q*radius)**2
435    return Iq/Iq[0]
436
437def csbox_Iq(q, a, b, c, da, db, dc, slda, sldb, sldc, sld_core):
[a1c32c2]438    z, w = leggauss(76)
439
[cfa28d3]440    sld_solvent = 0
441    overlapping = False
442    dr0 = sld_core - sld_solvent
443    drA, drB, drC = slda-sld_solvent, sldb-sld_solvent, sldc-sld_solvent
444    tA, tB, tC = a + 2*da, b + 2*db, c + 2*dc
445
[a1c32c2]446    outer_sum = np.zeros_like(q)
447    for cos_alpha, outer_w in zip((z+1)/2, w):
[cfa28d3]448        sin_alpha = sqrt(1.0-cos_alpha*cos_alpha)
449        qc = q*cos_alpha
450        siC = c*j0(c*qc/2)
451        siCt = tC*j0(tC*qc/2)
[a1c32c2]452        inner_sum = np.zeros_like(q)
453        for beta, inner_w in zip((z + 1)*pi/4, w):
[cfa28d3]454            qa, qb = q*sin_alpha*sin(beta), q*sin_alpha*cos(beta)
455            siA = a*j0(a*qa/2)
456            siB = b*j0(b*qb/2)
457            siAt = tA*j0(tA*qa/2)
458            siBt = tB*j0(tB*qb/2)
459            if overlapping:
[a1c32c2]460                Fq = (dr0*siA*siB*siC
461                      + drA*(siAt-siA)*siB*siC
462                      + drB*siAt*(siBt-siB)*siC
463                      + drC*siAt*siBt*(siCt-siC))
[cfa28d3]464            else:
[a1c32c2]465                Fq = (dr0*siA*siB*siC
466                      + drA*(siAt-siA)*siB*siC
467                      + drB*siA*(siBt-siB)*siC
468                      + drC*siA*siB*(siCt-siC))
469            inner_sum += inner_w * Fq**2
470        outer_sum += outer_w * inner_sum
471    Iq = outer_sum / 4  # = outer*um*zm*8.0/(4.0*M_PI)
[cfa28d3]472    return Iq/Iq[0]
473
474def check_shape(shape, fn=None):
475    rho_solvent = 0
476    q = np.logspace(-3, 0, 200)
477    r = shape.r_bins(q, r_step=0.01)
478    sampling_density = 15000 / shape.volume()
479    rho, points = shape.sample(sampling_density)
480    Pr = calc_Pr(r, rho-rho_solvent, points)
481    Iq = calc_Iq(q, r, Pr)
482    theory = (q, fn(q)) if fn is not None else None
483
484    import pylab
485    #plot_points(rho, points); pylab.figure()
486    plot_calc(r, Pr, q, Iq, theory=theory)
487    pylab.show()
488
[226473d]489def check_shape_2d(shape, fn=None, view=(0, 0, 0)):
490    rho_solvent = 0
491    qx = qy = np.linspace(-1, 1, 100)
492    Qx, Qy = np.meshgrid(qx, qy)
493    sampling_density = 50000 / shape.volume()
494    rho, points = shape.sample(sampling_density)
495    Iqxy = calc_Iqxy(Qx, Qy, rho, points, view=view)
496    Iqxy += 0.001 * Iqxy.max()
497    theory = fn(Qx, Qy)+0.001 if fn is not None else None
498
499    import pylab
500    plot_calc_2d(qx, qy, Iqxy, theory=theory)
501    pylab.show()
502
[cfa28d3]503def check_cylinder(radius=25, length=125, rho=2.):
504    shape = EllipticalCylinder(radius, radius, length, rho)
505    fn = lambda q: cylinder_Iq(q, radius, length)
506    check_shape(shape, fn)
507
[226473d]508def check_cylinder_2d(radius=25, length=125, rho=2., view=(0, 0, 0)):
509    shape = EllipticalCylinder(radius, radius, length, rho)
510    fn = lambda qx, qy, view=view: cylinder_Iqxy(qx, qy, radius, length, view=view)
511    check_shape_2d(shape, fn, view=view)
512
[cfa28d3]513def check_sphere(radius=125, rho=2):
514    shape = TriaxialEllipsoid(radius, radius, radius, rho)
515    fn = lambda q: sphere_Iq(q, radius)
516    check_shape(shape, fn)
517
518def check_csbox(a=10, b=20, c=30, da=1, db=2, dc=3, slda=1, sldb=2, sldc=3, sld_core=4):
519    core = Box(a, b, c, sld_core)
520    side_a = Box(da, b, c, slda, center=((a+da)/2, 0, 0))
521    side_b = Box(a, db, c, sldb, center=(0, (b+db)/2, 0))
522    side_c = Box(a, b, dc, sldc, center=(0, 0, (c+dc)/2))
523    side_a2 = copy(side_a).shift(-a-da, 0, 0)
524    side_b2 = copy(side_b).shift(0, -b-db, 0)
525    side_c2 = copy(side_c).shift(0, 0, -c-dc)
526    shape = Composite((core, side_a, side_b, side_c, side_a2, side_b2, side_c2))
527    fn = lambda q: csbox_Iq(q, a, b, c, da, db, dc, slda, sldb, sldc, sld_core)
528    check_shape(shape, fn)
529
530if __name__ == "__main__":
531    check_cylinder(radius=10, length=40)
[226473d]532    #check_cylinder_2d(radius=10, length=40, view=(90,30,0))
[cfa28d3]533    #check_sphere()
534    #check_csbox()
535    #check_csbox(da=100, db=200, dc=300)
Note: See TracBrowser for help on using the repository browser.