Data Processing – First
Peak-Fitting – Second
Philosophy for Peak-fitting and Data Processing
Our philosophy is to collect spectra under analysis conditions that are practical, readily reproduced, and typically used in laboratories that use monochromatic X-ray sources and work under real world practical analysis conditions. We have assumed that the most XPS laboratories need practical reference spectra and will not spend the time or money to produce and to analyze pure, clean surfaces under ultimate energy resolution conditions. For practical reasons we used the C (1s) spectra from the naturally formed layer of adventitious hydrocarbons because that signal is the “de facto” standard for charge referencing insulating materials.
The spectral data contained within this database are designed to assist engineers, scientists, analysts, theoreticians, and teachers who use XPS on an everyday basis under practical working conditions. We believe that these spectra will help XPS users to analyze industrial problems, gather reference data, perform basic research, test theories, and teach others. These spectra are designed to be practical tools for everyday use and were obtained under practical working conditions. No attempt was made to produce research grade spectra, but many of the spectra are actually research grade spectra because of the self-consistent methods used.
If the element is part of insulating chemical compound, then a C (1s) spectrum is included to allow the user to correct for sample charging or remove charge correction. All narrow scan spectra are peak-fitted to reveal FWHM, peak asymmetry, and peak separation for spin-orbit pairs. The strong signals observed in the wide scan survey spectra are labeled and tabulated together with rough BE values of those strong signals. The details of the experimental protocol used to produce each these spectra are provided in the “Data Collection and Analysis Details” section. All pure elements, except for Silicon, Selenium and single crystals, were ion etched prior to analysis.
We opted to collect a fast survey and fast chemical state spectra and then repeat with longer scans to check for X-ray induced degradation. In general, it is not useful to ion etch any chemical compound before analyzing it. After collecting a useful set of chemical state spectra, then you can ion etch the surface, but you should only remove the adventitious carbon layer which is at most 50 ang thick. If you ion etch more than that, then you risk changing the chemistry of your surface.
If you have one of the new Argon ion cluster guns, then you can more safely remove the adventitious carbon and adsorbed water, but again be careful especially if you have a one-of-a-kind sample which is usually very valuable.
Data Processing before Peak-fitting
Data processing must be done before any peak-fitting because the fundamental nature of a spectrum is changed by “Data Processing” routines. If you have a software that saves all processing steps to memory and provides Un-Do and Re-Do routines, then you test and reset as you need.
Typical Structure and Contamination of an Old Native Oxide or your Sample
Typical Structure and Contamination of a Freshly Exposed Bulk (eg silicon wafer)
This contamination can occur even if the sample was cleaved in the load-lock or a special chamber because that chamber, just like the analysis chamber, always has residual UHV gases (H2, CO, CH4, H2O) that are very difficult to remove. When a sample is cleaved and moved to the analysis chamber with 10-10 torr pressure, then you have 2-4 hours to finish your analyses until 2 monolayers of UHV gases have adsorbed onto the surface of your cleaved sample.
→ Periodic Table
Five (5) Data Processing Types
Five (5) Types of Data Processing that must be done before Peak-fitting
- Charge Correction (Referencing)
- Data-Point (spectrum) Smoothing
- Data Point Interpolation (interspersing more data points)
- Remove spin-orbit coupled peak
- Remove (subtract) background
Data Processing Type #1: Charge Correction (Referencing)
WARNINGS:
Do not charge shift native oxides because they very often have differential charging regions.
Float the native oxide, and then charge shift all of the resulting spectra.
Advice
We prefer to define the C (1s) BE of the hydrocarbon moiety in the adventitious carbon layer as being 285.0 eV. (Others prefer 284.8 eV or 284.6 eV or another BE.)
Data Processing Type #2: Data-Point / Spectrum Smoothing
Data Processing Type #3: Data Point Interpolation
Data Processing Type #4: Remove spin-orbit coupled peak
Data Processing Type #5: True Removal (subtract) of background
(This is not adding, fitting, or overlaying a background onto the spectrum)
Peak-fitting Assistance
The four (4) critical variables for peak-fitting are:
- FWHM to be used for all peaks (1st try)
- BEs
- Number of Peaks
- Peak-shape (% Gaussian and % Lorentzian)
After those variables, we must choose the best (2nd try):
- Background shape
- Background endpoints and averaging
- Constraints for FWHM
- Constraints for BE
Choices for FWHM of Peaks from Insulators
FWHM (eV) for Peak-fitting Chemical Compound Peaks (not metal or conductor peaks)
Choice for FWHM of Peaks from Conductors
FWHM (eV) for Peak-fitting Pure Metal or Pure Element Peaks (not Insulator peaks)
Sequence for Peak-fitting
Peak-fitting Flow Chart we Use for Peak-fitting
Peak-fitting Spectra
Peak-Fitting (Curve-Fitting) High Energy Resolution Spectra
Peak-fitting was performed by using the software that we produced and which is based on the SSI S-Probe software. This software allows the user to control the full width at half maxima (FWHM) value of any peak, the binding energy (BE) of any peak, peak areas, the ratio of two peak areas, the energy difference between two peak maxima, the shape of a peak as a sum-function of Gaussian and Lorentzian peak shapes in any peak, and the percentage of asymmetry in any peak.
By empirically peak-fitting the spectra from large sets of closely related materials in a trial and error method and analyzing the trends, it was possible to recognize several fundamental peak-shape and peak-fitting parameters for pure elements, binary oxides, polymers, and semiconductors. We used those empirical results to guide our efforts to peak-fit many of the spectra which had complicated peak shapes. In some cases we used the theoretical ratio of spin-orbit coupled signals to assist the peak-fitting of some spectra and also the energy interval between spin-orbit coupled signals that were derived from pure element spectra. No attempt was made to fit the spectra in accordance with theoretical expectations or calculations.
A reduced “chi-squared” value, which indicates the goodness of a peak-fit, was used to determine if a peak-fit was reasonable or not. Based on practical experience a “chi-squared” value between 1 and 2 implies a relatively good peak-fit. A “chi-squared” value between 2 and 4 implies that the fit has not yet been optimized. A “chi- squared” value larger than 4 implies that one or more signals may be missing from the peak-fit effort.
A Shirley-type baseline was used for all peak-fits. Peak shapes for the main XPS signals obtained from chemical compounds (e.g. oxides, halides, etc.) were typically optimized by using a Gaussian:Lorentzian ratio between 80:20 and 90:10. For pure metals, the Gaussian:Lorentzian ratio for the main XPS signals was normally between 50:50 and 70:30. The main XPS signals for semi-conductor materials usually required a Gaussian:Lorentzian peak-shape between 70:30 and 80:20.
From the peak-fitting of the binary oxides, we have observed that FWHM for the C (1s), O (1s) and the main metal signal from the binary oxide are usually in range 1.0-1.4 eV. This trend helped us to decide if we had good charge compensation.
The peak-shape that is most common is an 80:20 ratio of Gaussian:Lorentzian shapes. For polymers, we used 90:10. For true insulators that have a significant band-gap after the peak, we used a linear background. A Shirley background gives nearly the same result. If the sample was conductive, then we used a fully iterated Shirley background.
The FWHM of all peaks was expected to be nearly the same unless there is a significant change in polarity. The number of peaks was kept at a minimum. The FWHM values are based on the FWHM shown in our tables of FWHM.
When spin-orbit splitting is present, we used the theoretical ratio in place of the Scofield cross-section ratios.
When differential charging was detected, a peak was added for peak-fitting, and removed after the fit was complete.
Peak-Fitting Parameters
to Choose, Select or Test (for 1st iteration)
After “Data Processing and Charge Shifting” all chemical state spectra, then you must choose the Peak-fitting Parameters for your spectra:
- Background Type (BG)
- Linear (almost flat)
- Shirley (Sherwood, Proctor)
- Tougaard
- Background Endpoints
- Number of Expected Peaks
- Binding Energies (BE) of obvious Peak Maxima that seem Correct
- Full Width at Half of Maximum (Peak-width) – FWHM – based on FWHM tables
- Metal FWHM range = 0.4 to 1.0 eV (usually 0.6-0.8 eV)
- Metal Oxide FWHM range = 1.1 to 2.0 eV (usually 1.3-1.6 eV)
- Use same FWHM unless you have reference FWHM
- Gaussian-Lorentzian Peak-shape %s
- Typical G:L for Inorganics (80:20)
- Typical G:L for Polymers (90:10)
- G:L ratio increases to 50:50 at higher BE
- Peak Asymmetry % (Doniach-Sunjic) for conductors
- Typically 15%
- Peak Area Ratios (Theoretical)
- for “p” orbitals: 2:1
- for “d” orbitals: 6:4
- for “f” orbitals: 8:6
- Peak Area Constraints for Peak Area Ratios or Empirical Chemistry Ratios
- BE Constraints (when correct BE is known)
- FWHM Constraints (when FWHM control is essential)
- Difference in BE Constraint (difference is known)
- Chi-Square that is acceptable (<4 for low count rate data. <15 for very high count rate data.
- Decide if Differential Charging Tails are Present or Absent – Low BE side
- Peak-fit tail and then delete that peak
Choices for Baseline / Background Type and Endpoint Range
Choose FWHM for First Peak at Lowest BE
Use same FWHM for all peaks for first peak-fit
If one peak is much wider and symmetrical then use a FWHM that is 2X wider just for that peak (use constraint?)
In general, pure metal peaks are 2X more narrow than non-conductive chemical compound peaks (insulators).
Metal FWHM = 0.9 eV, Corresponding Metal Oxide FWHM = 1.7 eV
In general, the largest FWHM for Insulators is 1.8 eV. A few are slightly larger (eg 2.0 eV).
Examples of FWHM For Chemical Compounds and Insulators
Examples of FWHM For Pure Metals and Conductive Materials
Decide Total Number of Peaks in Spectrum – Use Minimum
Three (3) Peak-fitting Examples (A-C)
Difficulty Levels: 1-3
Example A: Level 1 – Peak-fitting of Single Chemical State Spectrum
Example B: Level 2 – Peak-fitting of Complicated Chemical State Spectrum
Example C: Level 3 – Peak-fitting Complicated Spectrum with Constraints