source: sasmodels/explore/multiscat.py @ e309e23

core_shell_microgelsmagnetic_modelticket-1257-vesicle-productticket_1156ticket_1265_superballticket_822_more_unit_tests
Last change on this file since e309e23 was e309e23, checked in by Paul Kienzle <pkienzle@…>, 6 years ago

multiscat: improved plots

  • Property mode set to 100644
File size: 11.6 KB
RevLine 
[ae91fce]1#!/usr/bin/env python
2r"""
3Multiple scattering calculator
4
5Calculate multiple scattering using 2D FFT convolution.
6
7Usage:
8
[a09d55d]9    -p, --probability: the scattering probability
[ae91fce]10    -q, --qmax: that max q that you care about
11    -w, --window: the extension window (q is calculated for qmax*window)
12    -n, --nq: the number of mesh points (dq = qmax*window/nq)
13    -r, --random: generate a random parameter set
14    -2, --2d: perform the calculation for an oriented pattern
15    model_name
16    model_par=value ...
17
18Assume the probability of scattering is $p$. After each scattering event,
19$1-p$ neutrons will leave the system and go to the detector, and the remaining
20$p$ will scatter again.
21
[a09d55d]22Let the scattering probability for $n$ scattering event at $q$ be $f_n(q)$,
[ae91fce]23where
[a09d55d]24.. math:: f_1(q) = \frac{I_1(q)}{\int I_1(q) dq}
[ae91fce]25for $I_1(q)$, the single scattering from the system. After two scattering
[a09d55d]26events, the scattering probability will be the convolution of the first
27scattering and itself, or $f_2(q) = (f_1*f_1)(q)$.  After $n$ events it will be
28$f_n(q) = (f_1 * \cdots * f_1)(q)$.  The total scattering is calculated
29as the weighted sum of $f_k$, with weights following the Poisson distribution
30.. math:: P(k; \lambda) = \frac{\lambda^k e^{-\lambda}}{k!}
31for $\lambda$ determined by the total thickness divided by the mean
32free path between scattering, giving
33.. math::
34    I(q) = \sum_{k=0}^\infty P(k; \lambda) f_k(q)
35The $k=0$ term is ignored since it involves no scattering.
36We cut the series when cumulative probability is less than cutoff $C=99\%$,
37which is $\max n$ such that
[ae91fce]38.. math::
[a09d55d]39    \frac{\sum_{k=1}^n \frac{P(k; \lambda)}{1 - P(0; \lambda)} < C
[ae91fce]40
41Using the convolution theorem, where
42$F = \mathcal{F}(f)$ is the Fourier transform,
[a09d55d]43.. math:: f * g = \mathcal{F}^{-1}\{\mathcal{F}\{f\} \cdot \mathcal{F}\{g\}\}
[ae91fce]44so
[a09d55d]45.. math:: f * \ldots * f = \mathcal{F}^{-1}\{ F^n \}
[ae91fce]46Since the Fourier transform is a linear operator, we can move the polynomial
47expression for the convolution into the transform, giving
48.. math::
[a09d55d]49    I(q) = \mathcal{F}^{-1}\left\{ \sum_{k=1}^{n} P(k; \lambda) F^k \right\}
50In the dilute limit $L \rightarrow 0$ only the $k=1$ term is active,
51and so
52.. math::
53    P(1; \lambda) = \lambda e{-\lambda} = \int I_1(q) dq
54therefore we compute
55.. math::
56    I(q) = \int I_1(q) dq \mathcal{F}^{-1}\left\{
57        \sum_{l=1}^{n} \frac{P(k; \lambda)}{P(1; \lambda))} F^k \right\}
[ae91fce]58
[a09d55d]59For speed we may use the fast fourier transform with a power of two.
60The resulting $I(q)$ will be linearly spaced and likely heavily oversampled.
61The usual pinhole or slit resolution calculation can performed from these
62calculated values.
[ae91fce]63"""
64
[a09d55d]65from __future__ import print_function, division
[ae91fce]66
67import argparse
68import time
69
70import numpy as np
[a09d55d]71from scipy.special import gamma
[ae91fce]72
73from sasmodels import core
74from sasmodels import compare
75from sasmodels import resolution2d
76from sasmodels.resolution import Resolution, bin_edges
77from sasmodels.data import empty_data1D, empty_data2D, plot_data
78from sasmodels.direct_model import call_kernel
79
80
81class MultipleScattering(Resolution):
[a09d55d]82    def __init__(self, qmax, nq, probability, is2d, window=2, power=0):
[ae91fce]83        self.qmax = qmax
84        self.nq = nq
85        self.power = power
[a09d55d]86        self.probability = probability
[ae91fce]87        self.is2d = is2d
88        self.window = window
89
90        q_range = qmax * window
91        q = np.linspace(-q_range, q_range, nq)
92        qx, qy = np.meshgrid(q, q)
93
94        if is2d:
95            q_calc = [qx.flatten(), qy.flatten()]
96        else:
97            q_range_corners = np.sqrt(2.) * q_range
98            nq_corners = int(np.sqrt(2.) * nq/2)
99            q_corners = np.linspace(0, q_range_corners, nq_corners+1)[1:]
100            q_calc = [q_corners]
101            self._qxy = np.sqrt(qx**2 + qy**2)
102            self._edges = bin_edges(q_corners)
103            self._norm = np.histogram(self._qxy, bins=self._edges)[0]
104
105        self.q_calc = q_calc
106        self.q_range = q_range
107
[a09d55d]108    def apply(self, theory, background=0.0, outfile="", plot=False):
109        #t0 = time.time()
[ae91fce]110        if self.is2d:
111            Iq_calc = theory
112        else:
113            q_corners = self.q_calc[0]
114            Iq_calc = np.interp(self._qxy, q_corners, theory)
115        Iq_calc = Iq_calc.reshape(self.nq, self.nq)
116        #plotxy(Iq_calc); import pylab; pylab.figure()
117        if self.power > 0:
[a09d55d]118            Iqxy = scattering_power(Iq_calc, self.power)
119            powers = []
[ae91fce]120        else:
[a09d55d]121            Iqxy, powers = multiple_scattering(
122                Iq_calc, self.probability,
123                coverage=0.99,
124                return_powers=(outfile!="") or plot,
125                )
126        #print("multiple scattering calc time", time.time()-t0)
127
[ae91fce]128        #plotxy(Iqxy); import pylab; pylab.figure()
129        if self.is2d:
[a09d55d]130            if outfile and powers:
131                data = np.vstack([Ipower.flatten() for Ipower in powers]).T
132                np.savetxt(outfile + "_powers.txt", data)
133            if outfile:
134                data = np.vstack(Iq_calc).T
135                np.savetxt(outfile + ".txt", data)
136            if plot:
[ae91fce]137                import pylab
138                plotxy(Iq_calc)
139                pylab.title("single scattering")
140                pylab.figure()
141
[a09d55d]142            return Iqxy + background
[ae91fce]143        else:
144            # circular average, no anti-aliasing
145            Iq = np.histogram(self._qxy, bins=self._edges, weights=Iqxy)[0]/self._norm
[a09d55d]146            if powers:
147                Iq_powers = [np.histogram(self._qxy, bins=self._edges, weights=Ipower)[0]/self._norm
148                             for Ipower in powers]#  or powers[:5] for the first five
149
150            if outfile:
151                data = np.vstack([q_corners, theory, Iq]).T
152                np.savetxt(outfile + ".txt", data)
153            if outfile and powers:
154                # circular average, no anti-aliasing for individual powers
155                data = np.vstack([q_corners] + Iq_powers).T
156                np.savetxt(outfile + "_powers.txt", data)
157            if plot:
[ae91fce]158                import pylab
[a09d55d]159                plotxy(Iqxy)
160                pylab.title("multiple scattering")
161                pylab.figure()
162            if plot and powers:
163                import pylab
[e309e23]164                L = -np.log(1-self.probability)
165                pylab.loglog(q_corners, Iq, label="total %g"%self.probability)
166                for n, Ipower in enumerate(Iq_powers):
167                    k = n+1
168                    w = L**(k-1)/gamma(k+1)
169                    pylab.loglog(q_corners, w*Ipower, label="scattering**%d"%k)
[ae91fce]170                pylab.legend()
171                pylab.figure()
172
[a09d55d]173            return q_corners, Iq + background
[ae91fce]174
[a09d55d]175def scattering_power(Iq, n):
[54a5024]176    """
177    Calculate the nth scattering power as a distribution.  To get the
178    weighted contribution, scale by $\lambda^k e^{-\lambda}/k!$.
179    """
[a09d55d]180    scale = np.sum(Iq)
181    F = _forward_fft(Iq/scale)
[54a5024]182    result = _inverse_fft(F**n)
[a09d55d]183    return result
184
185def multiple_scattering(Iq, p, coverage=0.99, return_powers=False):
186    """
187    Compute multiple scattering for I(q) given scattering probability p.
188
189    Given a probability p of scattering with the thickness, the expected
190    number of scattering events, $\lambda$ is $-\log(1 - p)$, giving a
[54a5024]191    Poisson weighted sum of single, double, triple, etc. scattering patterns.
192    The number of patterns used is based on coverage (default 99%).
[a09d55d]193    """
194    L = -np.log(1-p)
195    num_scatter = truncated_poisson_invcdf(coverage, L)
[ae91fce]196
197    # Compute multiple scattering via convolution.
[a09d55d]198    scale = np.sum(Iq)
199    F = _forward_fft(Iq/scale)
200    coeffs = [L**(k-1)/gamma(k+1) for k in range(1, num_scatter+1)]
201    multiple_scattering = F * np.polyval(coeffs[::-1], F)
202    result = scale * _inverse_fft(multiple_scattering)
203
204    if return_powers:
205        powers = [scale * _inverse_fft(F**k) for k in range(1, num_scatter+1)]
206    else:
207        powers = []
208
209    return result, powers
210
211def truncated_poisson_invcdf(p, L):
212    """
213    Return smallest k such that cdf(k; L) > p from the truncated Poisson
214    probability excluding k=0
215    """
216    # pmf(k; L) = L**k * exp(-L) / (k! * (1-exp(-L))
217    cdf = 0
218    pmf = -np.exp(-L) / np.expm1(-L)
219    k = 0
220    while cdf < p:
221        k += 1
222        pmf *= L/k
223        cdf += pmf
224    return k
225
226def _forward_fft(Iq):
227    # Prepare padded array and forward transform
228    nq = Iq.shape[0]
[ae91fce]229    half_nq = nq//2
230    frame = np.zeros((2*nq, 2*nq))
[a09d55d]231    frame[:half_nq, :half_nq] = Iq[half_nq:, half_nq:]
232    frame[-half_nq:, :half_nq] = Iq[:half_nq, half_nq:]
233    frame[:half_nq, -half_nq:] = Iq[half_nq:, :half_nq]
234    frame[-half_nq:, -half_nq:] = Iq[:half_nq, :half_nq]
235    fourier_frame = np.fft.fft2(frame)
236    return fourier_frame
237
238def _inverse_fft(fourier_frame):
239    # Invert the transform and recover the transformed data
240    nq = fourier_frame.shape[0]//2
[ae91fce]241    half_nq = nq//2
[a09d55d]242    frame = np.fft.ifft2(fourier_frame).real
243    Iq = np.empty((nq, nq))
244    Iq[half_nq:, half_nq:] = frame[:half_nq, :half_nq]
245    Iq[:half_nq, half_nq:] = frame[-half_nq:, :half_nq]
246    Iq[half_nq:, :half_nq] = frame[:half_nq, -half_nq:]
247    Iq[:half_nq, :half_nq] = frame[-half_nq:, -half_nq:]
248    return Iq
[ae91fce]249
250def parse_pars(model, opts):
251    # type: (ModelInfo, argparse.Namespace) -> Dict[str, float]
252
253    seed = np.random.randint(1000000) if opts.random and opts.seed < 0 else opts.seed
254    compare_opts = {
255        'info': (model.info, model.info),
256        'use_demo': False,
257        'seed': seed,
258        'mono': True,
259        'magnetic': False,
260        'values': opts.pars,
261        'show_pars': True,
262        'is2d': opts.is2d,
263    }
264    pars, pars2 = compare.parse_pars(compare_opts)
265    return pars
266
267
268def main():
269    parser = argparse.ArgumentParser(
270        description="Compute multiple scattering",
271        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
272        )
[a09d55d]273    parser.add_argument('-k', '--power', type=int, default=0, help="show pattern for nth scattering")
274    parser.add_argument('-p', '--probability', type=float, default=0.1, help="scattering probability")
[ae91fce]275    parser.add_argument('-n', '--nq', type=int, default=1024, help='number of mesh points')
276    parser.add_argument('-q', '--qmax', type=float, default=0.5, help='max q')
277    parser.add_argument('-w', '--window', type=float, default=2.0, help='q calc = q max * window')
278    parser.add_argument('-2', '--2d', dest='is2d', action='store_true', help='oriented sample')
279    parser.add_argument('-s', '--seed', default=-1, help='random pars with given seed')
280    parser.add_argument('-r', '--random', action='store_true', help='random pars with random seed')
[a09d55d]281    parser.add_argument('-o', '--outfile', type=str, default="", help='random pars with random seed')
[049e02d]282    parser.add_argument('model', type=str, help='sas model name such as cylinder')
283    parser.add_argument('pars', type=str, nargs='*', help='model parameters such as radius=30')
[ae91fce]284    opts = parser.parse_args()
285    assert opts.nq%2 == 0, "require even # points"
286
287    model = core.load_model(opts.model)
288    pars = parse_pars(model, opts)
[a09d55d]289    res = MultipleScattering(opts.qmax, opts.nq, opts.probability, opts.is2d,
[ae91fce]290                             window=opts.window, power=opts.power)
291    kernel = model.make_kernel(res.q_calc)
292    #print(pars)
293    bg = pars.get('background', 0.0)
294    pars['background'] = 0.0
295    Iq_calc = call_kernel(kernel, pars)
[a09d55d]296    Iq = res.apply(Iq_calc, background=bg, outfile=opts.outfile, plot=True)
[ae91fce]297    plotxy(Iq)
298    import pylab
299    if opts.power > 0:
300        pylab.title('scattering power %d'%opts.power)
301    else:
[a09d55d]302        pylab.title('multiple scattering with fraction %g'%opts.probability)
[ae91fce]303    pylab.show()
304
305def plotxy(Iq):
306    import pylab
307    if isinstance(Iq, tuple):
308        q, Iq = Iq
309        pylab.loglog(q, Iq)
310    else:
311        data = Iq+0.
312        data[Iq <= 0] = np.min(Iq[Iq>0])/2
313        pylab.imshow(np.log10(data))
314    #import pylab; pylab.show()
315
316if __name__ == "__main__":
317    main()
Note: See TracBrowser for help on using the repository browser.