LG4X-V2: lmfit GUI for XPS

Planned features/Improvements

• redesign the parameter table so that the limits are placed next to the corresponding parameters, redesign clutty GUI #16 e.g. by using tabs for parameters, limits, results etc. (Add Tab indicator for the user as a reminder that limits were set)

• add tooltip info to GUI #18

• rework the export files, so that additional informations such as fwhm and areas are exported in readable format as well

• check lmfit version and catch UserWarning for independent vars as described in #10

• Pause/Interrupt fit button #5

• rewrite the Readme to explain the features introduced in LG4X-V2

• Undo button (log of last parameter sets, as suggested in #13, introduced on [dev branch] (https://github.com/Julian-Hochhaus/LG4X-V2/tree/undo-fct))

• Export fit parameters as readable table to be able to use them in e.g. a presentation

• Introduce 'Clear all' button for clearing all parameters/limits etc.

• remove fwhm's and area calculation from the usermodels and instead calculate them after fitting based on the ModelResult() parameters to be able to use error propagation #27

• introduce relative parameter for gamma and coster-kronig factor

• save the active background parameter checkbox to the parameter-files.

• keep the programm running if an error occurs in the fitting procedure

  • error handling introduced for errors during file import, parameter import, saving parameters, exporting results (Thanks to @Hexanders)

• show only relevant parameter for the used model and grey the others out #17

• add asymmetry ratio

• add limits for all parameters

• introduce class/function which converts .dat export/internal preset array to lmfit parameter set and vice versa

Introduction

LG4X-V2 is based on the great work of Hideki NAKAJIMA who developed the software LG4X.

LG4X provides a graphical user interface for XPS curve fitting analysis based on the lmfit package, which is the non-linear least-square minimization method on python platform. LG4X facilitates the curve fitting analysis for python beginners. LG4X was developed on Python 3, and PyQt5 was used for its graphical interface design. Shirley and Tougaard iterated methods are implemented as a supplementary code for XPS background subtraction. LG4X tidies up all fitting parameters with their bound conditions in table forms. Fitting parameters can be imported and exported as a preset file before and after analysis to streamline the fitting procedures. Fitting results are also exported as a text for parameters and csv file for spectral data. In addition, LG4X simulates the curve without importing data and evaluates the initial parameters over the data plot prior to optimization.

Methods

Installation

Download and install Python 3 and additional packages.

brew install python3

If you want to include the background into the fit-model, the so-called active background method as described i.e. by Alberto Herrera-Gomez: The active background method in XPS data peak-fitting , you need to install the latest development version of lmfit by:

git clone https://github.com/lmfit/lmfit-py.git

python setup.py install

Otherwise, install lmfit via pip:

pip3 install lmfit

Additionally, you need to install:

pip3 install pandas

pip3 install matplotlib

pip3 install PyQt5

The OS dependence of installation of python, pip, and brew is described in the link.

Update package if necessary

pip3 install --upgrade lmfit

Miniconda3

If you have Miniconda3, you can create the environment to install lmfit from conda-forge. Below is an example for environment name vpy3.9 on python version 3.9 (YouTube video).

conda config --add channels conda-forge

conda config --set channel_priority strict

conda create -n vpy3.9 python=3.9

conda activate vpy3.9

conda install lmfit

conda install matplotlib

conda install pandas

python main.python

Supplementary codes for XPS analysis

xpspy.py should be located in the same directory as main.py for XPS energy range selection for background (BG) subtraction in Shirley and Tougaard methods, which are taken from codes by Kane O'Donnell and James Mudd.

vamas.py and vamas_export.py are also necessary for importing ISO VAMAS format file. vamas.py is a modifed class of VAMAS format from Kane O'Donnell.

periodictable.py and periodictableui.py are the periodic table window to identify the peak elements. The codes are based on and revised from clusterid.

elements.py and elementdata.py are the class for peak energy and sensitivity used in the priodic table above. The codes are based on and revised from clusterid.

Start LG4X

python3 main.py

Testing and developing environment

• Python 3.9.5
• asteval==0.9.23
• lmfit==1.0.2
• matplotlib==3.4.3
• numpy==1.20.3
• pandas==1.3.2
• PyQt5==5.15.4
• scipy==1.6.3
• uncertainties==3.1.5

Usage

1. Import data
   • Import csv, text, or vamas (.vms/.npl) file format.
   • All csv and text files in a directory.
   • Choose data from file list if it was already imported.
2. Setup background and peak parameters with their types
   • Select energy range of spectrum for optimization.
   • Setup initial BG parameters including polynomial coefficients.
   • Setup peak model and its parameters.
   • Increase and decrease the number of peaks.
   • Load a preset file if available.
3. Evaluate parameters
   • Plot the curves without optimization.
   • Simulate the curves based on the range and peaks if no data file is selected in the File list.
4. Fit curve
   • Adjust parameters and bounds until they become converged
5. Export results
   • Export csv file for curves
   • Export text file for parameters
   • Save parameters as a preset for next analysis

Export csv file for curves

The exported .csv file contains the raw data as well as all fitted components: Thereby, the first and second column contain the input data, respectively the x and y data. The third column contains the intensity data minus the background. In the forth column, the sum over all components is given, if you wish to plot your data without background, this would be the sum curve over all components you wish to plot. In column five and six, the background and the polynomial background are given. Important to note that the background already contains the polynomial background, the polynomial background in column six is only given because LG4X-V2 adds a polynomial background to each fit, if the parameters pg_i are not fixed to 0. In the seventh column, the sum curve over all components and backgrounds is given, that's the sum curve you wish to plot if you are using the raw intensities including the background to present your data. The following columns contain the components of the fit model.

Home directory to import data

You can change the HOME directory in the main.py edited in a way below. # makes a line comment out.

# Home directory

self.filePath = QtCore.QDir.homePath()

# self.filePath = '/Users/hidekinakajima/Desktop/WFH2021_2/lg4x/LG4X-master/Python/'

Citing

https://doi.org/10.5281/zenodo.3901523

Video

YouTube: Introduction of LG4X

YouTube: Installation of LG4X in miniconda3 environment

Database reference

X-ray data booklet for binding energy

• http://xdb.lbl.gov/

"Hartree-Slater subshell photoionization cross-sections at 1254 and 1487 eV" J. H. Scofield, Journal of Electron Spectroscopy and Related Phenomena, 8129-137 (1976).

• http://dx.doi.org/10.1016/0368-2048(76)80015-1
• https://a-x-s.org/research/cross-sections/

"Calculated Auger yields and sensitivity factors for KLL-NOO transitions with 1-10 kV primary beams" S. Mroczkowski and D. Lichtman, J. Vac. Sci. Technol. A 3, 1860 (1985).

• http://dx.doi.org/10.1116/1.572933
• http://www.materialinterface.com/wp-content/uploads/2014/11/Calculated-AES-yields-Matl-Interface.pdf

(Electron beam energy at 1, 3, 5, and 10 keV for relative cross section and derivative factors)

Interface design

An initial gui concept is taken from here and there.

Buttons

Evaluate

You can evaluate the fitting parameters without fitting optimization on data spectrum. If you do not select the data, it works as simulation mode in the range you specify in BG table (x_min, x_max, the number of data points pt).

Fit

You can optimize the fitting parameters by least-square method, and parameters in the table are updated after optimization.

Export

LG4X exports fitting results in two different files. One is a text file for fitting conditions _fit.txt, the other is a csv format file for data _fit_csv, which is saved at the same directory of the former file.

Add and rem peak

You can add and remove peak at the end of column from the Fit table.

Drop-down menus

Importing data

LG4X imports csv format or tab separated text files. A data file should contain two columns. First column is energy and second column is spectral intensity. LG4X skips first row, because it is typically used for column names. Example data files are available in Example. Energy and instensiy are calibrated in the Excel XPS macro (EX3ms) prior to the analysis for convenience. The method of energy calibration is discussed in the link.

VAMAS file format can also be imported in LG4X by decomposing a VAMAS file into the tab separated text files based on the block and sample idenfitifers. Exported tab separated text files are available in the same directory as the VAMAS file. You can just use LG4X to convert the VAMAS file into tab separated text files for the other program you prefer. Note that the binding energy scale is automatically created from VAMAS for XPS and UPS data.

Imported data is displayed in the figure and listed in the file list. You can also open the directory to import all csv and text files in the file list.

File list

Imported file path is added in the list. You can choose the path to import a data file again from the list once you import the data file. Fitting parameters are loaded from Fitting preset menu below.

Fitting preset

Fitting condition can be created in the BG and Fit tables. From fitting preset drop-down menu, you can create the most simple single-peak preset from New. If you have a preset previously saved, you can load a preset file, which will be listed in the Fitting preset. Typical conditions for XPS C1s and C K edge are also available from the list as examples. A preset filename is ended with _pars.dat, and parameters include in the preset file as a list in the following way.

[*BG type index*, [*BG table parameters*], [*Fit table parameters*]]

Periodic table is available to identify the peak position and relative intensity based on XPS Al Kα excitation source (1486.6 eV). If you change the excitation energy hn and work function wf, the core-level and Auger peak energies are revised according to the equation below.

BE = hn - wf - KE

BE represents the binding energy, and KE kinetic energy. The database reference and example usage of periodic table are shown below. Refresh button enables us to display elements in the other dataset, and Clear button removes all elements.

BG types (Shirley BG to be shown as a default)

You can choose the BG type to be subtracted from the raw data as listed below. Shirley and Tougaard BG iteration functions are available from xpypy.py, which should be located with main.py. From lmfit built-in models, 3rd-order polynomial and 3 step functions are implemented. Fermi-Dirac (ThermalDistributionModel) is used for the Fermi edge fitting, and arctan and error functions (StepModel) for NEXAFS K edge BG. Polynomial function is added to the other BG models configured in the BG table, so polynomial parameters have to be taken into account for all BG optimization. You can turn off polynomial parameters by filling all zeros with turning on checkbox. Valence band maximum and secondary electron cutoff can be fitted with the 4th polynomial function for the density of states or edge jump at the onset.

No.	String	BG model	Parameters
0			x_min, x_max for fitting range, data points in simulation pt, hn, wf
1		Shirley BG	Initial cv, max iteration number it
2		Tougaard BG	B, C, C', D
3	pg	Polynomial BG	c0, c1, c2, c3
4	bg	Fermi-Dirac BG	amplitude, center, kt
5	bg	Arctan BG	amplitude, center, sigma
6	bg	Error BG	amplitude, center, sigma
7	bg	VBM/cutoff	center, d1, d2, d3, d4

Tables

BG table

You can specify the range for fitting region in the first row of BG table. Checkbox works as a constraint at the value beside. Range and polynomial rows are independent from the drop-down menu selection for background.

Fit table

All conditions are based on the lmfit built-in models listed in the Fit table. Peak models are listed below. For standard XPS analysis, amplitude ratios amp_ratio and peak differences ctr_diff can be setup from their referenced peak amp_ref and ctr_ref, respectively from drop-down menu in each column. The number of peaks can be varied with add peak and rem peak buttons. Checkbox can be used for either fixing values or bound conditions if you check beside value. Empty cells do not effect to the optimization.

No.	String	Peak model	Parameters
1	g	GaussianModel	amplitudeamp, centerctr, sigmasig
2	l	LorentzianModel	amplitude, center, sigma
3	v	VoigtModel	amplitude, center, sigma, gammagam
4	p	PseudoVoigtModel	amplitude, center, sigma, fractionfrac
5	e	ExponentialGaussianModel	amplitude, center, sigma, gamma
6	s	SkewedGaussianModel	amplitude, center, sigma, gamma
7	a	SkewedVoigtModel	amplitude, center, sigma, gamma, skew
8	b	BreitWignerModel	amplitude, center, sigma, q
9	n	LognormalModel	amplitude, center, sigma
10	d	DoniachModel	amplitude, center, sigma, gamma

Amplitude ratio and energy difference

XPS doublet peaks are splitted by the spin-orbit coupling based on the atomic theory. Spin-orbit interaction depends on the atomic element and its orbit. The energy separation of doublet corresponds to the spin-orbit constant. Amplitude ratio of doublet is based on the degeneracy (2j+1) of each total angular quantum number (j). LG4X constrains amp_ratio and ctr_diff from a reference peak amp_ref and ctr_ref selected by dropdown menus. For example, Ag3d has j=5/2 and 3/2, and their amplitude ratio corresponds to 3:2. You can setup second peak amplitude ratio by selecting the first peak at j=5/2 and amp_ratio = 0.67. This means that amplitude of second peak at j=3/2 is constrained by a factor of 0.67 against that of first peak. Peak difference parameter also works in a way that second peak position is away from first peak ctr_ref by ctr_diff = 6 eV as shown in the figure below.

Note that amplitude used in the lmfit package is equivalent to the peak area that is propoertional to the amount of element in analytical area and depth by XPS. The atomic ratio is evaluated by the peak area normalized by the sensitivity factor. The ratio of sensitivity factors on doublet peaks is the same as that in multiplicity, so the normalized peak area of one doublet peak is the same as that in other one.

A comprehensive review on XPS technique and analytical procedures is available in link below.

"X-ray photoelectron spectroscopy: Towards reliable binding energy referencing" G. Greczynski and L. Hultman, Progress in Materials Science 107, 100591 (2020).

Examples

You can find the VAMAS format data of various spectra from Spectroscopy Hub.