\(\renewcommand\AA{\unicode{x212B}}\)

BayesQuasi v1

Summary

This algorithm runs the Fortran QLines programs which fits a Delta function of amplitude 0 and Lorentzians of amplitude A(j) and HWHM W(j) where j=1,2,3. The whole function is then convolved with the resolution function.

Properties

Name

Direction

Type

Default

Description

Program

Input

string

QL

The type of program to run (either QL or QSe). Allowed values: [‘QL’, ‘QSe’]

SampleWorkspace

Input

MatrixWorkspace

Mandatory

Name of the Sample input Workspace

ResolutionWorkspace

Input

MatrixWorkspace

Mandatory

Name of the resolution input Workspace

ResNormWorkspace

Input

WorkspaceGroup

Name of the ResNorm input Workspace

MinRange

Input

number

-0.2

The start of the fit range. Default=-0.2

MaxRange

Input

number

0.2

The end of the fit range. Default=0.2

SampleBins

Input

number

1

The number of sample bins

ResolutionBins

Input

number

1

The number of resolution bins

Elastic

Input

boolean

True

Fit option for using the elastic peak

Background

Input

string

Flat

Fit option for the type of background. Allowed values: [‘Sloping’, ‘Flat’, ‘Zero’]

FixedWidth

Input

boolean

True

Fit option for using FixedWidth

UseResNorm

Input

boolean

False

fit option for using ResNorm

WidthFile

Input

string

The name of the fixedWidth file

Loop

Input

boolean

True

Switch Sequential fit On/Off

OutputWorkspaceFit

Output

WorkspaceGroup

Mandatory

The name of the fit output workspaces

OutputWorkspaceResult

Output

MatrixWorkspace

Mandatory

The name of the result output workspaces

OutputWorkspaceProb

Output

MatrixWorkspace

The name of the probability output workspaces

Description

This algorithm can only be run on windows due to f2py support and the underlying fortran code

The model that is being fitted is that of a delta-function (elastic component) of amplitude A(0) and Lorentzians of amplitude A(j) and HWHM W(j) where j=1,2,3. The whole function is then convolved with the resolution function. The -function and Lorentzians are intrinsically normalised to unity so that the amplitudes represent their integrated areas.

For a Lorentzian, the Fourier transform does the conversion: \(1/(x^{2}+\delta^{2}) \Leftrightarrow exp[-2\pi(\delta k)]\). If x is identified with energy E and \(2\pi k\) with \(t/\hbar\) where t is time then: \(1/[E^{2}+(\hbar / \tau)^{2}] \Leftrightarrow exp[-t/\tau]\) and \(\sigma\) is identified with \(\hbar / \tau.\) The program estimates the quasielastic components of each of the groups of spectra and requires the resolution file and optionally the normalisation file created by ResNorm.

For a Stretched Exponential, the choice of several Lorentzians is replaced with a single function with the shape : \(\psi\beta(x) \Leftrightarrow exp[-2\pi(\sigma k)\beta]\). This, in the energy to time FT transformation, is \(\psi\beta(E) \Leftrightarrow exp[-(t/\tau)\beta]\). So \(\sigma\) is identified with \((2\pi)\beta\hbar/\tau\). The model that is fitted is that of an elastic component and the stretched exponential and the program gives the best estimate for the \(\beta\) parameter and the width for each group of spectra.

Usage

Example - BayesQuasi

# Load in test data
sampleWs = Load('irs26176_graphite002_red.nxs')
resWs = Load('irs26173_graphite002_red.nxs')

# Run BayesQuasi algorithm
fit_ws, result_ws, prob_ws = BayesQuasi(Program='QL', SampleWorkspace=sampleWs, ResolutionWorkspace=resWs,
                                    MinRange=-0.547607, MaxRange=0.543216, SampleBins=1, ResolutionBins=1,
                                    Elastic=False, Background='Sloping', FixedWidth=False, UseResNorm=False,
                                    WidthFile='', Loop=True)

Categories: AlgorithmIndex | Workflow\MIDAS

Source

Python: BayesQuasi.py