gisaxs data analysis with bornagain

40
MLZ is a cooperation between: GISAXS data analysis with BornAgain Céline Durniak * , Jonathan Fisher, Marina Ganeva , Gennady Pospelov, Walter Van Herck, Joachim Wuttke Jülich Centre for Neutron Science (JCNS) at Heinz Maier-Leibnitz Zentrum (MLZ), Forschungszentrum Jülich GmbH, Garching, Germany * Now at: DMSC filial of European Spallation Source ESS AB, Copenhagen, Denmark

Upload: ngocong

Post on 02-Jan-2017

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: GISAXS data analysis with BornAgain

MLZ is a cooperation between:

GISAXS data analysis with BornAgain

Céline Durniak*, Jonathan Fisher, Marina Ganeva, Gennady Pospelov, Walter Van Herck, Joachim Wuttke

Jülich Centre for Neutron Science (JCNS) at Heinz Maier-Leibnitz Zentrum (MLZ), Forschungszentrum Jülich GmbH, Garching, Germany

*Now at: DMSC filial of European Spallation Source ESS AB, Copenhagen, Denmark

Page 2: GISAXS data analysis with BornAgain

Outline

l  Theoretical background

l  Introduction to BornAgain

l  Working with BornAgain I: Graphical User Interface

l  Working with BornAgain II: Python API

l  Fitting

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 2

Page 3: GISAXS data analysis with BornAgain

Theoretical background

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 3

Born Approximation DWBA

•  considers only incident and outgoing wave

•  multiple scattering is ignored

•  the differential cross section

J  accounts for reflection-refraction effects close to the critical angle

J  equations for the layer structure solved exactly

J  surface structure is treated as a small perturbation

Page 4: GISAXS data analysis with BornAgain

The form factor

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 4

Born DWBA The Born form factor is the Fourier transform of the shape of the particle

Example: spherical particle of radius R

Page 5: GISAXS data analysis with BornAgain

DWBA: interference between particles

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 5

The expectation value of the differential cross section is

where

Cylinder form factor Interference function Cylinder form factor with interference function

Page 6: GISAXS data analysis with BornAgain

Approximations for polydisperse structures

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 6

Decoupling approximation (DA) è No correlation between type and interparticle distances

Local Monodisperse Approximation (LMA) è Incoherent superposition of different domains, each with their own type and interference function

dσdΩ

q( ) = Id q( )+ S q( ) ⋅ Fα q( )α

2

dσdΩ

q( ) = Sα q( ) ⋅ Fα q( )2

α

Size-Space Correlation Approximation (SSCA) è Interparticle distance depends on sizes of the two particles considered

Page 7: GISAXS data analysis with BornAgain

Interface roughness: parameters

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 7

Roughness of the interface is described by: •  RMS roughness σ •  Hurst parameter H •  Lateral correlation length Lc

Additionally for multilayer: •  Cross-correlation length Lh

uncorrelated correlated

Lh = 0 Lh >> h

Page 8: GISAXS data analysis with BornAgain

DWBA limitations: gratings

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 8

V. Soltwisch et. al., arXiv:1509.02003v2

V. Soltwisch et. al., GISAS 2015 abstract booklet

Page 9: GISAXS data analysis with BornAgain

References

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 9

Theoretical background

l  BornAgain User Manual, 2016, http://bornagainproject.org/documentation

l  Gilles Renaud et. al., Surf. Sci. Rep. 64, 255 (2009)

l  Rémi Lazzari, J. Appl. Cryst. 35, 406-421 (2002)

Interface roughness

l  V. Holý and T. Baumbach, Phys. Rev. B 49, 10668 (1994)

l  Sinha, et al., Phys. Rev. B 38, 2297 (1988)

Books

l  Jens Als-Nielsen and Des McMorrow “Elements of modern x-ray physics”

l  Martin Schmidbauer “Diffuse Scattering from Self-Organized Mesoscopic Semiconductor Structures”

l  Ezquerra, T.A., et al. (Eds.), “Applications of Synchrotron Light to Scattering and Diffraction in Materials and Life Sciences”

Page 10: GISAXS data analysis with BornAgain

BornAgain framework l  Open-source multi-platform software project developed by scientific computing

group of MLZ (Garching, Germany) l  Simulation of grazing-incidence small-angle scattering for X-rays and neutrons l  Physics model is based on the Distorted Wave Born Approximation (DWBA) l  Development started in April, 2012. l  Project website http://bornagainproject.org

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 10

Page 11: GISAXS data analysis with BornAgain

Functionality overview

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 11

l  Fit of the experimental data l  Graphical user interface l  Different shapes of nanoparticles l  Size distributions (polydispersity) l  Nanoparticle assemblies l  Multilayer systems l  Different interference functions

l  Polarized neutrons l  2 types of detectors l  Detector resolution l  Beam divergence

Page 12: GISAXS data analysis with BornAgain

BornAgain: how to start

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 12

Page 13: GISAXS data analysis with BornAgain

Introduction bornagainproject.org

Page 14: GISAXS data analysis with BornAgain

Introduction bornagainproject.org

Page 15: GISAXS data analysis with BornAgain

Introduction bornagainproject.org

Page 16: GISAXS data analysis with BornAgain

Demo: BornAgain GUI define a sample

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 16

Page 17: GISAXS data analysis with BornAgain

Available form factors

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 17

Page 18: GISAXS data analysis with BornAgain

Complex shapes

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 18

Core shell particles Particles with size distribution With possibility to link parameters

Particle compositions collection of particles with fixed inter-particle distance coherent interference

All can be rotated

Page 19: GISAXS data analysis with BornAgain

Very large particles

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 19

Large particles gives rise to a problem, known in communication theory as aliasing: Rapidly oscillating signal measured at fixed points shows up as slow sinusoid

In GISAS simulation Rapidly oscillating form factor of large particles leads to a significant variation of intensity over the detector bin. dx

dy

Page 20: GISAXS data analysis with BornAgain

Very large particles

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 20

analytical calculations Monte-Carlo integration

Small cylinders height = 10 nm radius = 20 nm

Large cylinders height = 1000 nm radius = 2000 nm

Page 21: GISAXS data analysis with BornAgain

Interference functions

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 21

o  1D lattice o  Radial paracrystal

a b

o  2D lattice o  2D paracrystal

Page 22: GISAXS data analysis with BornAgain

Demo: BornAgain GUI set up the instrument

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 22

Page 23: GISAXS data analysis with BornAgain

GISAXS Instrument: Beam Parameters

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 23

l  beam is defined via wavelength and incidence angles

l  it is possible to define beam divergence

beam divergence OFF beam divergence ON

Page 24: GISAXS data analysis with BornAgain

GISAXS Instrument: Detector

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 24

BornAgain supports two kinds of detectors: Spherical detector Rectangular detector

Detectors are defined by the number of bins and the accessible range

Page 25: GISAXS data analysis with BornAgain

Specular and Off-specular geometry

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 25

specular geometry

off-specular geometry

Page 26: GISAXS data analysis with BornAgain

Demo: BornAgain Python API

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 26

Page 27: GISAXS data analysis with BornAgain

Software validation

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 27

l  Validation against IsGISAXS

l  Self validation

In process

l  Validation against experimental data

Planned

l  We agreed with HipGisaxs team upon cross-validation of our packages (& possibly others)

Page 28: GISAXS data analysis with BornAgain

Validation against IsGISAXS

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 28

BornAgain results mostly coincide with IsGisaxs on numerical level

Page 29: GISAXS data analysis with BornAgain

Self Validation

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 29

Part of new BornAgain’s functionality can be validated via BornAgain itself o  Rotation machinery example o  Particle compositions example

o  Create box (30,20,6) o  RotateY by 90 degrees o  Compare with non-rotated box

(6,20,30) o  Scattering intensities should be

identical

o  Create particle composition from two hemi spheres

o  Assign same material to them o  Compare with normal full sphere, same

material, same radius o  Scattering intensities should be identical

Page 30: GISAXS data analysis with BornAgain

Demo: Fitting of experimental data

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 30

Page 31: GISAXS data analysis with BornAgain

Fitting in BornAgain

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 31

BornAgain

Page 32: GISAXS data analysis with BornAgain

Fitting in BornAgain: main features (1)

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 32

o  Variety of minimization algorithms

o  Possibility to fit every sample parameter or their combination

o  Various fit strategies (e.g. fix/release

parameters)

FitParameter(“par1”, 8.0*nm, limited(5.0, 15.0))

radius = fun1(par1); lattice_length = fun2(par1)

Page 33: GISAXS data analysis with BornAgain

Fitting in BornAgain: main features (2)

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 33

o  Organizing different minimization algorithms into the chain o  Genetic minimizer explores large parameter space, Levenberg-Marquardt finalize location of

minima o  Simultaneous fit of multiple datasets

o  Two or more experimental images obtained for different incident angles can be fitted with one sample model

o  Fitting along slices, masking certain areas of the detector image

o  Fitting along slices

Page 34: GISAXS data analysis with BornAgain

Fitting problems

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 34

Why?

1e+05 CPU seconds later … Minimizer: “J”

fitting

Page 35: GISAXS data analysis with BornAgain

Fitting problems

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 35

Possible reasons

l  An unreliable sample model

l  Large correlations between parameters

l  Very different scales of parameters involved in the calculation

l  Too many fit parameters

l  Multiple local minima

Page 36: GISAXS data analysis with BornAgain

Fitting problems: multiple local minima

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 36

Page 37: GISAXS data analysis with BornAgain

Fitting problems: multiple local minima

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 37

There you are!

Page 38: GISAXS data analysis with BornAgain

Fitting problems

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 38

Possible reasons

l  An unreliable sample model

l  Large correlations between parameters

l  Very different scales of parameters involved in the calculation

l  Too many fit parameters

l  Multiple local minima

Troubleshooting

l  Choose a small number of free fitting parameters

l  Eliminate redundant parameters

l  Provide a good initial guess for the fit parameters

Page 39: GISAXS data analysis with BornAgain

Announcement

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 39

1st BornAgain School and User Meeting

(a satellite of the GISAXS2016 workshop)

which will take place on 21-22 November 2016 at Heinz Maier-Leibnitz-Zentrum in Garching (near Munich), Germany.

Registration at https://webapps.frm2.tum.de/indico/event/34/ More information at http://bornagainproject.org

Page 40: GISAXS data analysis with BornAgain

Thank you for your attention!

15/06/16 M. Ganeva, GISAXS Tutorial, ALBA SR-ALD workshop 40

Contact:

http://bornagainproject.org

[email protected]

[email protected]