from sans.models.BaseComponent import BaseComponent from math import exp, sqrt from numpy import power from scipy.special import erf max_level_n = 7 class UnifiedPowerRgModel(BaseComponent): """ This model is based on Exponential/Power-law fit method developed by G. Beaucage """ def __init__(self, multfactor=1): BaseComponent.__init__(self) """ :param multfactor: number of levels in the model, assumes 0<= level# <=5. """ ## Setting model name model description self.name = "UnifiedPowerRg" self.description="""Multiple Levels of Unified Exponential/Power-law Method. Up to Level 6 is provided. Note; the additional Level 0 is an inverse linear function, i.e., y = scale/x + background. The Level N is defined as y = background + scale * Sum(1..N)[G_i*exp(-x^2*Rg_i^2/3) + B_i/x^(power_i)*(erf(x*Rg_i/sqrt(6))^(3*power_i))]. Ref: G. Beaucage (1995). J. Appl. Cryst., vol. 28, p717-728. G. Beaucage (1996). J. Appl. Cryst., vol. 29, p134-146. """ self.level_num = multfactor ## Define parameters self.params = {} ## Parameter details [units, min, max] self.details = {} # non-fittable parameters self.non_fittable = [] # list of function in order of the function number self.fun_list = self._get_func_list() ## dispersion self._set_dispersion() ## Define parameters self._set_params() ## Parameter details [units, min, max] self._set_details() #list of parameter that can be fitted self._set_fixed_params() ## functional multiplicity of the model self.multiplicity_info = [max_level_n,"Level No.:",[],[]] def _unifiedpowerrg(self,x): """ Scattering function :param x: q value(s) :return answer: output of the function """ # common parameters for the model functions bkg = self.params['background'] scale = self.params['scale'] l_num = self.level_num # set default output answer = 0.0 # Set constant on lebel zero (special case) if l_num == 0: answer = scale / x + bkg return answer # rearrange the parameters for the given label no. for ind in range(1,l_num+1): # get exp term exp_now = exp(-power(x*self.params['Rg%s'% ind],2)/3.0) # get erf term erf_now = erf(x*self.params['Rg%s'% ind]/sqrt(6.0)) # get power term pow_now = power((erf_now*erf_now*erf_now/x),self.params['power%s'% ind]) # get next exp term only if it exists try: exp_next = exp(-power(x*self.params['Rg%s'% (ind+1)],2)/3.0) except: exp_next = 1.0 # get to the calculation answer += self.params['G%s'% ind]*exp_now + self.params['B%s'% ind]* \ exp_next * pow_now # take care of the singular point if x == 0.0: answer = 0.0 for ind in range(1,l_num+1): answer += self.params['G%s'% ind] # get scaled answer *= scale # add background answer += bkg return answer def _set_dispersion(self): """ model dispersions """ ##set dispersion from model self.dispersion = {} def _set_params(self): """ Concatenate the parameters of the model to create this model parameters """ # common parameters for the model functions self.params['background'] = 0.0 self.params['scale'] = 1.0 l_num = self.level_num # rearrange the parameters for the given label no. for ind in range(0,l_num+1): if ind == 0: continue # multiple factor for higher labels mult = 1.0 mul_pow = 1.0 if ind != l_num: mult = 10.0 * 4.0/3.0 mul_pow = 2.0 # Set reasonably define default values that consistent # w/NIST for label #1 self.params['G%s'% ind] = 0.3*mult*pow(10, \ (l_num+1 - float('%s'% ind))) self.params['Rg%s'% ind] = 21.0/mult*pow(10, \ (l_num - float('%s'% ind))) self.params['B%s'% ind] = 6e-03/mult*pow(10, \ -(l_num+1 - float('%s'% ind))) self.params['power%s'% ind] = 2.0 * mul_pow def _set_details(self): """ Concatenate details of the original model to create this model details """ # common parameters for the model functions self.details['background'] = ['[1/cm]', None, None] self.details['scale'] = ['', None, None] # rearrange the parameters for the given label no. for ind in range(0,self.level_num+1): if ind == 0: continue self.details['G%s'% ind] = ['[1/(cm.sr)]', None, None] self.details['Rg%s'% ind] = ['[A]', None, None] self.details['B%s'% ind] = ['[1/(cm.sr)]', None, None] self.details['power%s'% ind] = ['', None, None] def _get_func_list(self): """ Get the list of functions in each cases """ func_list = {} return func_list def getProfile(self): """ Get SLD profile : return: None, No SLD profile supporting for this model """ return None def setParam(self, name, value): """ Set the value of a model parameter : param name: name of the parameter : param value: value of the parameter """ # set param to new model self._setParamHelper(name, value) def _setParamHelper(self, name, value): """ Helper function to setParam """ # Look for standard parameter for item in self.params.keys(): if item.lower()==name.lower(): self.params[item] = value return raise ValueError, "Model does not contain parameter %s" % name def _set_fixed_params(self): """ Fill the self.fixed list with the model fixed list """ pass def run(self, x = 0.0): """ Evaluate the model : param x: input q-value (float or [float, float] as [r, theta]) : return: (DAB value) """ if x.__class__.__name__ == 'list': # Take absolute value of Q, since this model is really meant to # be defined in 1D for a given length of Q #qx = math.fabs(x[0]*math.cos(x[1])) #qy = math.fabs(x[0]*math.sin(x[1])) return self._unifiedpowerrg(x) elif x.__class__.__name__ == 'tuple': raise ValueError, "Tuples are not allowed as input to BaseComponent models" else: return self._unifiedpowerrg(x) return self._unifiedpowerrg(x) def runXY(self, x = 0.0): """ Evaluate the model : param x: input q-value (float or [float, float] as [qx, qy]) : return: DAB value """ if x.__class__.__name__ == 'list': q = math.sqrt(x[0]**2 + x[1]**2) return self._unifiedpowerrg(x) elif x.__class__.__name__ == 'tuple': raise ValueError, "Tuples are not allowed as input to BaseComponent models" else: return self._unifiedpowerrg(x) def calculate_ER(self): """ """ # Not implemented!!! pass def set_dispersion(self, parameter, dispersion): """ Set the dispersion object for a model parameter : param parameter: name of the parameter [string] :dispersion: dispersion object of type DispersionModel """ pass