Bad Peak-fits Explained

 

 

 

Bad Peak-fits compared to Good Peak-fits

These spectra are included in Spectra-Base #1.  The spectra are used here to produce a bad peak-fit example, to compare it to a “good” peak-fit example.

None of the good peak-fit examples have been optimized to produce “best” peak-fits.

 

 

Explanation
Why Bad ?

Bad Peak-fit Example

 Good Peak-fit Example Explanation
Why Good ?
Ni (2p3) – Ni metal foil – ion etched Ni (2p3) – Ni metal foil – ion etched clean Ni (2p3) – Ni metal foil – ion etched clean
Because the sample is conductive, we should use a Shirley type background. Shirley is commonly used, but is not necessarily the optimum background for all peak-fits. Conductors normally require some peak asymmetry due to valence-core interactions. The sample was ion etched and has only a trace of “adsorbed” oxygen so it is impossible to have many metal-oxide peaks. It is very difficult to identify so many different metal oxide species even if the surface really has much more oxygen.  Based on the survey spectrum the adsorbed oxygen for this sample is <5 atom%. Shirley background was used because nickel is conductive. Asymmetry was applied to the main peak because sample is conductive. To avoid over-fitting the data, only two more peaks were added.  If we know more about the material, then we might add another peak.  But we must explain and identify each peak. Peak “C” has been attributed to a 2-electron process.
C (1s) – PET (Mylar) film – IPA cleaned  C (1s) – PET (Mylar) film – IPA cleaned  C (1s) – PET (Mylar) film – IPA cleaned 
This peak-fit uses the same FWHM for all peaks which is reasonable as a first try. FWHM are usually in the 1.0 to 1.6 eV range for chemical compounds. Chi-Square is a little high. Baseline region shows a good fit, but the tops of B and C show small but significant gaps. Peak B (30% L) was given a larger amount of Lorentzian tail than Peaks A and C (20% L) to try to make Chi-square smaller. A linear background was used.

A clean surface of PET has 3 different chemical states (A,B,C) for Carbon. The empirical ratio of these 3 peak areas is 3: 1: 1. The peak-fit shown here has a 55: 23: 19.2 ratio that reduces to 2.9 : 1.2: 1.0. The difference in the ratios is small, but the question is: Why?

 Polymers usually do not have any contamination. But, here is an exception.  By adding and constraining a small new hydrocarbon peak, and a small alcohol peak, the peak-fit produced the expected empirical ratio 3: 1: 1 shown in the C (1s) spectrum at the right.  Peaks D and E are due to a single pi->pi* shake-up.

Be sure not to use too small or too large a FWHM for compounds, A small FWHM for conductors is OK.
C (1s) – PET (Mylar) film – IPA cleaned  C (1s) – PET (Mylar) film – cleaned with IPA C (1s) – PET (Mylar) film – cleaned with IPA
This peak-fit has a small Chi-square which indicates a good peak-fit. Baseline region  and the tops of B and C peaks have good fits, no gaps. All peaks use a 80:20 Gaussian:Lorentzian Sum peak shape ratio which is normal for similar chemical states. Variation in the FWHM values is <20% which is a little higher than normal 10%. This indicates a difference in chemical states. The background is linear.  This looks like a good peak-fit.  However, we know that the material is PET (Mylar). For this reason the data-analyst must constrain the peak area ratios of peaks A, B, and C to produce the empirical 3:1:1 ratio of PET. Polymers usually do not have any contamination. But, here is an exception.  By adding and constraining a small new hydrocarbon peak, and a small alcohol peak, the peak-fit produced the expected empirical ratio 3: 1: 1 shown in the C (1s) spectrum at the right.  Peaks D and E are due to a single pi->pi* shake-up.
O (1s) – Native Silicon Oxide – as recd O (1s) – Native Silicon Oxide – as received O (1s) – Native Silicon Oxide – as received
This peak-fit has 5 synthetic peaks each having a FWHM of 1.00. The chemical shifts are all the same, ~0.7 eV. Chi-square is small, <4. The sample is a native oxide of silicon. In general, FWHM for compounds range from 1.0-1.6 eV. O (1s) FWHM are normally 1.5-1.8 eV.  Because the FWHMs are 1.0 eV, this fit is not good. There is no chemical reason for Silicon to have 5 different types of SiOx. Native Oxides usually form mainly the most stable and most oxidized from of the element. If the sample was contaminated, you need to identify chemical state of all 5 peaks. This peak-fit has 2 synthetic peaks each having a FWHM of 1.7 eV, typical of O (1s). Chi-Square is ~4. Because this is a native oxide, there can be an oxide and a hydroxide or a 2nd type of oxide. We need to think what chemical states are possible. Using many narrow peaks does not match reality,
O (1s) – Native Silicon Oxide – as recd O (1s) – Native Silicon Oxide – as received O (1s) – Native Silicon Oxide – as received
This peak-fit has 3 peaks each having FWHM of 1.7 eV. Chi-Square is ~5. OK.  Chemical shifts are ~0.3 eV. The presence of 3 peaks means there are 3 different chemical states for Oxygen attached to Silicon.  This is possible. But chemical shifts are rarely only 0.3 eV apart.  It is possible, but we need good reference BEs to use such small shifts. This peak-fit has 3 peaks each having a FWHM of 1.6 eV, typical of O (1s). Chi-Square is ~2. Great! Because this is a native oxide, there can be an oxide and a hydroxide or a 2nd type of oxide. We need to be able to assign good states.
Ce (3d) – CeO2 dried in air from slurry Ce (3d) – CeO2 dried in air from slurry Ce (3d) – CeO2 dried in air from slurry
Using basic peak-fitting logic, the data-analysis would count the number of peaks and shoulders and use that as a starting point. That approach produces peaks with large FWHM (see spectrum at right). If we think all FWHM should be similar, then we add more peaks to produce the spectrum at the left.

The Ce (3d) signal, like many rare earth signals, has multiplet splittings due to interactions of unpaired electrons. The questions is:  How many peaks and what FWHM values should be used? FWHM for Multiplet Splittings is not well documented. CeO2 is often contaminated with low levels of Ce2O3. Theoreticians and experts have produced the plots shown at the far right. This is just a fraction of the info available on Multiplet Splitting in Rare Earths.

 Here are two peak-fits with assignments to guide your peak-fits of Ce (3d).
Si (2p) – Native Oxide of Silicon – as recd Si (2p) – Native Oxide of Silicon – as recd Si (2p) – Native Oxide of Silicon – as recd
The “1s, 2s, 3s, 4s, and 5s” peaks do NOT have spin-orbit splitting, but all other XPS peaks (p, d, f) do have spin-orbit splitting that produces a doublet of peaks. The energy difference between the two peaks can be very small (0.1-1.0 eV) and may not be visible, while others have energy differences ranging from 1.1 to 100 eV. The Si (2p) signal has a doublet with an energy difference of ~0.6 eV. Peaks A and B are due to the (2p3) and (2p1) peaks.  The peaks C, D, and E were peak-fit with single peaks which is a common practice, but is wrong. The spectrum at the right shows the correct style for peak-fitting the Si (2p) signal. Metal oxide FWHM can be 2-3x wider than the FWHM of the pure metal as shown here.

Because the (2p3) and (2p1) peaks are only 0.6 eV apart, we do not see any peak-tops or shoulders to guide our peak-fitting effort. We must rely on our knowledge of chemistry and XPS.
Fe (2p) – Bituminous Coal – Freshly exposed bulk Fe (2p) – Bituminous Coal – Freshly exposed bulk Fe (2p) – Bituminous Coal – Freshly exposed bulk
A sample of freshly exposed Bituminous Coal was found to have very low levels of Fe, Si, and Al.  This Fe (2p) spectrum from that sample has a lot of noise making it very difficult to locate peaks or shoulders for peak-fitting. Iron is often heavily oxidized so the first peak should correspond to an Iron Oxide species. Fe2O3 has an unpaired valence electron, so Fe2O3 has multiplet splitting peaks. The data-analyst added 3 more peaks to fit the Fe (2p3) signal. After adding a series of peaks that align with peaks found in Fe2O3, the results are too noisy and should not be trusted. The Fe (2p) spectrum at the left has enough data-points to use smoothing to try to locate meaningful signals. The peak-fit shown here look useful, but the data-analyst should go back and collect more data for better S/N if the data is important.
Ti (2p) – TiO2 single crystal – Anatase – Freshly exposed  Ti (2p) – TiO2 single crystal – Anatase – Freshly exposed  Ti (2p) – TiO2 single crystal – Anatase – Freshly exposed 
Ti (2p) has a spin-orbit pair of peaks. Based on theory, the ratio of the 2 peaks is 2: 1.  Based on the Scofield SFs the ratio is 1.94: 1.00. The Relative Area in this peak-fit is: 2.23: 1.00, which is different from theory by >20%. For this reason, this peak-fit is not so useful.  This difference means there is more information to be discovered.  A problem in this peak-fit is that the background does not include the 2 peaks located at 472 and 477 eV. A Shirley background was used, but it does not include those 2 peaks. To properly analyze this spectrum, the spectrum needs to be extended to 500 or 520 eV. In this peak-fit the relative % of the (2p3) and (2p1) peaks is 2.0, same as theory. To achieve this ratio, two additional peaks (C, D) were added, but no attempt was made to identify the chemical state or nature of those peaks. The chi-square of this fit is 50% larger than the other. Why? If we zoom the Y axis, we might understand.
S (2p) – Sulfur chip – Freshly exposed bulk S (2p) – Sulfur chip – Freshly exposed bulk S (2p) – Sulfur chip – Freshly exposed bulk
S (2p) has a spin-orbit pair. Other spectra show the energy difference is 1.2 eV for the (2p3) – (2p1) pair. The peak-fit on the left has not constrained the energy difference and has not forced the spin-orbit peak areas to be 2.0: 1.0 (theory). FWHM are similar except for one. The Chi-square value is small which indicates a good fit. Should we stop because the Chi-square is a small number? The spectrum at the right shows the result of applying the true binding energy difference and also the theoretical peak-area ratio. After constraining the energy difference to be 1.2 for peaks A and C, and also B and D, the fit is improved. By also constraining the peak area ratios to be 2:1, the fit is a bit better. The FWHM should have also been constrained, but still the Chi-square was smaller than the uncontrolled peak-fit.
Y (3d) – Y2O3 powder – as received Y (3d) – Y2O3 powder – as received Y (3d) – Y2O3 powder – as received
This Y (3d) spectrum should show a spin-orbit pair of peaks, but instead shows a broad signal that could be fit with 3-6 peaks, but the data-analyst had some idea of what the spectrum should look like from previous analyses of Y2O3. There are only two obvious features – a peak max and a shoulder.  The data-analyst looked at the O (1s) spectrum (shown below) to try to understand the Y (3d) spectrum.  By looking at the right side of both spectra, the analyst decided there was a charge control problem. After pressing the sample into a pellet and adjusting the XY positions of the flood gun the spectra at the right were measured from the same material. This is the Y (3d) spectrum measured after pressing the loose powder into a pellet and after adjusting the XY position of the flood gun and measuring C (1s) of polypropylene.

This teaches us that ugly data is sometimes ugly because the instrument has a problem.
O (1s) – Y2O3 powder – as received O (1s) – Y2O3 powder – as received O (1s) – Y2O3 powder – as received
This O (1s) spectrum shows a broad signal that could be fit with 3-6 peaks, but the data-analyst had some idea of what the spectrum should look like from previous analyses of Y2O3. There are only two obvious features – a peak max and a shoulder.  By looking at the right side of both spectra, the analyst decided there was a charge control problem. After pressing the sample into a pellet and adjusting the XY positions of the flood gun the spectrum at the right was measured from the same material. This O (1s) spectrum shows 2 distinct peaks. There might be a small peak at ~533 eV and in-between the 2 big peaks. The wide peak at 532 eV might be due to water or hydroxide peaks, but they are not due to charging. The low BE region at ~528 eV is a small Gaussian shape which indicates no differential charging.
C (1s) – Poly-Propylene sheet – Fresh scrape C (1s) – Poly-Propylene sheet – Fresh scrape C (1s) – Poly-Propylene sheet – Fresh scrape
The peak at 283.7 eV could be assigned to a carbide of maybe some form of graphene. The peak at 285.0 eV could be assigned to a hydrocarbon. But this is a simple sample of Poly-propylene freshly scraped clean. This C (1s) spectrum is part of a study of aligning the electron flood gun. The small peak at 283.7 eV is due to differential charging because the flood gun was not properly aligned. After aligning the XY positions from 2.9-0.0 to 0.0–5.0, the small peak disappeared and the FWHM decreased.  The Chi-square value is reporting how good is the peak-fit. Chi-square does not reveal flood gun alignment problems or differential charging. This C (1s) spectrum was part of a test to see if the electron flood gun was or was not properly aligned. Polypropylene has only one symmetrical peak which makes it ideal for checking flood gun alignment.  The spectrum at the left is due to a mis-aligned flood gun. When a flood gun is misaligned, or not working correctly, spectra can have tails at low BE, extra large FWHM, or other issues.
C (1s) – Poly-Propylene sheet – Fresh scrape C (1s) – Poly-Propylene sheet – Fresh scrape C (1s) – Poly-Propylene sheet – Fresh scrape
The peak at 283.7 eV could be assigned to a carbide of maybe some form of graphene. The peak at 285.6 eV could be assigned to a hydrocarbon. The peak at 286.7 should be assigned to an alcohol or ether peak. But this is a simple sample of Poly-propylene freshly scraped clean. This C (1s) spectrum is part of a study of aligning the electron flood gun. The small peaks at 283.7 eV and 285.6 are due to differential charging because the flood gun is not properly aligned. After aligning the XY positions both small peaks disappeared and the FWHM decreased to an even smaller value.  Chi-square does not reveal flood gun alignment problems or differential charging. This C (1s) spectrum was part of a test to see if the electron flood gun was or was not properly aligned. Polypropylene has only one symmetrical peak which makes it ideal for checking flood gun alignment.  This alignment position is better than the one shown above because the FWHM is 0.1+ smaller which means more Energy resolution and more information.
Al (2p) – Bituminous coal – Freshly exposed bulk Al (2p) – Bituminous coal – Freshly exposed bulk Al (2p) – Bituminous coal – Freshly exposed bulk
C (1s) – Diamond – single crystal – as received C (1s) – Diamond – single crystal – as received C (1s) – Diamond – single crystal – as received
N (1s) – Bituminous coal – Freshly exposed bulk N (1s) – Bituminous coal – Freshly exposed bulk N (1s) – Bituminous coal – Freshly exposed bulk
O (1s) – Native silicon oxide – as received O (1s) – Native silicon oxide – as received O (1s) – Native silicon oxide – as received
C (1s) – PET (Mylar) film – as received C (1s) – PET (Mylar) film – as received C (1s) – PET (Mylar) – wiped with Acetone
C (1s) – PET (Mylar) film – as received C (1s) – PET (Mylar) film – as received C (1s) – PET (Mylar) film – as received
Cu (2p3) – CuO lump – as received from Russia Cu (2p3) – CuO lump – as received from Russia Cu (2p3) – CuO lump – as received from Russia
Ca (2p) – CaCO3 (calcite) – Freshly exposed natural crystal Ca (2p) – CaCO3 (calcite) – Freshly exposed natural crystal Ca (2p) – CaCO3 (calcite) – Freshly exposed natural crystal
S (2p) – Bituminous coal – freshly exposed bulk S (2p) – Bituminous coal – freshly exposed bulk S (2p) – Bituminous coal – freshly exposed bulk
Si (2p) – Bituminous coal – freshly exposed bulk Si (2p) – Bituminous coal – freshly exposed bulk Si (2p) – Bituminous coal – freshly exposed bulk
Ca (2p) – CaCO3 (calcite) – Freshly exposed bulk of natural crystal
O (1s) – CaCO3 (calcite) – Freshly exposed bulk of natural crystal
O (1s) – SiO2 – native silicon oxide – as received
O (1s) – SiO2 – native silicon oxide – as received
O (1s) – SiO2 – native silicon oxide – as received
Cr (2p) – Cr2O3 – Single crystal – freshly exposed bulk
Mg (2p) – MgSO4 – anhydrous powder – pressed onto plate
Mg (2s) – MgSO4 – anhydrous powder – pressed onto plate
Mg (2p) – MgSO4 – anhydrous powder – pressed onto plate
O (1s) – MgSO4 – anhydrous powder – pressed onto plate
Mg (1s) – MgSO4 – anhydrous powder – pressed onto plate
S (2s) – MgSO4 – anhydrous powder – pressed onto plate