Professional Documents
Culture Documents
Gaussian-Lorentizian sum-XPS Complete Approach
Gaussian-Lorentizian sum-XPS Complete Approach
net/publication/263965482
CITATIONS READS
10 9,862
1 author:
Matthew R Linford
Brigham Young University - Provo Main Campus
431 PUBLICATIONS 7,183 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Matthew R Linford on 16 July 2014.
I
n this column I’ll be talking about a nearly unavoidable, and ian-Lorentzian sum function (GLS), and the Gaussian-Lorent-
certainly indispensible, part of X-ray photoelectron spec- zian product (GLP) function. We’ll discuss these three functions
troscopy (XPS): peak fitting of narrow scans. Sherwood in this article, and then move on to the impulse function, which
aptly observed that peak fitting in XPS is commonly practiced I believe is an interesting piece of mathematics that any scientist
because forms of the same element in different oxidation states1 or engineer should be familiar with.
often yield signals that are separated by about as much as their
line widths.1-2 Crist has observed that important decisions in the XPS Peak Fitting with Voigt and Gaussian-Lorentzian
laboratory and in industry are regularly based on the results from Sum and Product Functions
peak fitting.3 I have found all of this to be true. The most widely
In peak fitting an XPS narrow scan, one generally selects a
used functions for fitting peaks in XPS narrow scans are based on
baseline first followed by a series of peaks, where each peak will,
Gaussian and Lorentzian functions, and three such functions are
in general, represent a single chemical state (oxidation state1) of
regularly considered for this purpose: the Voigt function, which
an element. The most common functions chosen to represent
is the convolution of a Gaussian function with a Lorentzian func-
symmetric XPS signals are the Voigt function, which is the con-
tion (we considered convolution in the last column4), the Gauss-
volution of a Gaussian function with a Lorentzian function, the
Figure 1. Graph of the Gaussian function in Equation 1 with parame- Figure 2. Graph of the Lorentzian function in Equation 2 with param-
ters h = 1, E = 0, and F = 1. eters h = 1, E = 0, and F = 1.
Notice that the first and second terms in this function are the
Gaussian and Lorentzian functions in Equations 1 and 2, weight-
ed by a mixing parameter, m. That is, for m = 0, Equation 6
reduces to Equation 1, a Gaussian, and for m = 1, Equation 6 re-
Figure 5. Convolution of the triangle function with itself (T(x)*T(x)) duces to Equation 2, a Lorentzian. Obviously, by varying m from
forming a function that appears rather Gaussian-like.
0 to 1 we obtain a series of functions that have more Gaussian or
predecessor (see Figure 4). And finally, note that the convolution more Lorentzian character. It should also be clear that m could be
of a Gaussian with a second Gaussian is yet an additional Gauss- a fitting parameter in an algorithm used to fit XPS narrow scans.
ian, which will be broader than either of the original Gaussians. Figure 6 shows the GLS function (Equation 6) for three values
So all of this is to show the reasonableness of stating that when of m: m = 0 (the blue line), which again is a Gaussian, m = 1 (the
a set of photoelectrons, which may inherently have a Lorent- green line), which is again a Lorentzian, and m = 0.5 (the red
zian line shape, is excited by X-rays that have some line width line), which runs more or less between the other two functions.
and shape, and also perturbed by both a spectrometer and the Figure 7 goes a little further, zooming in on the region where the
broadening mechanisms discussed above, we would expect the Gaussian and Lorentzian functions differ most for m = 0, 0.1,
resulting signal to have at least some Gaussian character. Thus, 0.3, 0.5, 0.7, and 1. It is clear that the GLS function varies in a
it is common for the components of XPS narrow scans, espe- reasonable way between a pure Gaussian and a pure Lorentzian
cially when monochromatic sources are used, to be modeled as function.
peaks with both Gaussian and Lorentzian character. The amount Now let’s compare the GLS function to the GLP function.
of Gaussian or Lorentzian character assigned to these peaks de- This latter function is defined in Equation 7. Clearly, from its
pends on the factors mentioned above. For example, it has been name, we expect that we will be dealing with the product of a
recommended that the C 1s signals from polymers be modeled Gaussian function and a Lorentzian function, which is indeed
as 100% Gaussians or 90:10 Gaussian:Lorentzian mixes.3 This the case. That is, m = 0 in Equation 7 results in a pure Gaussian
seems reasonable. Polymers will often be in rather heteroge- (Equation 1), and m = 1 in Equation 7 yields a pure Lorentzian
neous environments – polymer chains can often be approximat- (Equation 2).
ed as random coils.
(7) GLP(x;F,E,m,h) = h * exp – 4ln2(1-m) (x-E)
2
1
So we have established that many of the component peaks of *
2
F (x-E)2
XPS narrow scans will be best defined/fit by peaks that have both 1+4m
F2
Gaussian and Lorentzian character. The ‘purest’ way to handle
this problem is to use a Voigt function, which is the convolution
Figure 6. Graph of the GLS function (Equation 6) with parameters h = Figure 7. Graph of the GLS function (Equation 6) with parameters h
1, E = 0, and F = 1 for m = 0 (bottom, blue line), m = 0.5 (middle, red = 1, E = 0, and F = 1 for (from bottom to top) m = 0, 0.1, 0.3, 0.5, 0.7,
line), and m = 1 (top, yellow line). and 1.
So all of this sounds pretty good, and one might think at this made by convolving Gaussian and a Lorentzian functions with
point that the GLS and GLP functions would be more or less equal widths. All three of these functions had equal widths of
interchangeable. I you believe this, you may wish to continue 2. I’ve reproduced this plot in Figure 10, albeit with the peaks
reading. Figure 8 shows the GLP function for m = 0, m = 0.5, shifted to the origin for consistency with the other peaks in the
and m = 1. It is obvious here that while the m = 0.5 GLS func- other figures shown herein. The Voigt function (yellow line) is
tion was approximately in the middle of the pure Gaussian and the widest of the three functions. The red line just inside it is
Lorentzian functions (see Figure 6), the m = 0.5 GLP function is the GLS function. Obviously the GLS function approximates the
only slightly different from the pure Gaussian function. Things Voigt function pretty well. The GLP function is represented by
just don’t look right here, which causes one to ask the question: the blue line. It is less obvious that it is a good approximation to
Why would the m = 0.5 GLP function not look more like the m the Voigt function. I performed a citation search on the Hesse
= 0.5 GLS function? Here’s the answer. Notice that the Gaussian paper. While a number of groups have cited it as a reference to
function is rather well contained and goes to zero quite quickly. their choice of fitting functions for XPS narrow scans, I have not
As a result, when one multiplies a Lorentzian by a Gaussian, the been able to find a study that contradicts their results or that fur-
‘wings’ of the Lorentzian are multiplied by values that are pretty ther develops this topic, i.e., I assume here that anyone writing on
close to zero and as a result they essentially disappear. Figure 9 the topic would cite the Hesse paper. In 2003, Neal Fairley also
further illustrates this issue. It shows a zoom in view of the m = discussed this issue, noting, as did Hesse, the significant decrease
0.5 GLP function and also of the m = 0.9 GLP function. Here, of the Lorentzian’s tail when it is multiplied by a Gaussian.5a In
the graph of the m = 0.9 GLP function reconfirms that the GLP summary, for the reasons outlined in this article, and also in the
function is a less than ideal representation of something interme- Hesse and Fairley documents, we similarly favor the GLS over
diate between a pure Gaussian and a pure Lorentzian function, the GLP for peak fitting XPS narrow scans.
i.e., we’d expect that the m = 0.9 GLP function would not be sig-
nificantly different from a pure Lorentzian, but the m = 0.9 GLP
function in Figure 9 only extends about half way between the
pure Gaussian and Lorentzian functions, and it does not appear
to effectively represent the ‘tails’ of the Lorentzian.
Around 1980,6 the GLP function was recommended in the
literature for XPS narrow scan peak fitting. This suggestion ap-
pears to have been relatively influential. However, a more recent
paper in 2007 by Hesse, Streubel, and Szargan5c was unable
to verify the previous recommendation – in their analyses the
GLS performed better than the GLP. Their study also had a rath-
er pragmatic component it. They looked at the fitting functions
(Voigt, GLS, and GLP) that are available in different XPS soft-
ware packages. Unfortunately, not all of the functions are avail-
able in all of the programs, and one software package only had Figure 10. Graphs of the GLS function with m = 0.5 (red line), the
the GLP as an option. One must sometimes do the best one can GLP function with m = 0.5 (blue line), and the Voigt function (yellow
line). All three functions have widths of 2 and are centered at the origin.
with the tools available. Hesse and coworkers also plotted the
The Voigt function here is the convolution of a Gaussian function with
GLS with m = 0.5, the GLP with m = 0.5, and a Voigt function a width of 1.3 and a Lorentzian function with the same width.
!
0 for x >½ defined through a triangle function or a Gaussian function. For
(8) (x) = ½ for x = ½ most practical purposes, however, the definition of the impulse
1 for x <½ function does not matter. It can’t! The idea behind the impulse
function is that it becomes so narrow that it does not matter
Figure 3 is again a good representation of this function – con- which definition we choose! (We get the same answer from any
vince yourself that this is the case. Now, let’s see how we can representation of it.) We will begin by defining the impulse func-
manipulate (x). First, we can change its height by multiplying tion at the origin, but we will relax this constraint shortly and
it by some constant, h. That is, h (x) has the same definition as show that it can be shifted to any position along the x-axis. The
Equation 8, except that it will have a value of h for x < ½ and h/2 impulse function, (x), is a function that only exists in a limit.
for x = ½. Thus we see that we can easily manipulate the height Accordingly, let’s first consider the function:
of (x). Obviously, we could also multiply (x) by a negative
constant. (9) lim (x/ )
0
We can also shift (x) along the x axis. To do this, we recog-
nize that the center of (x) is where its argument is zero. Accord-
ingly, (x - c) must be a rectangle function that is centered at the
point c along the x-axis, i.e., if you solve x – c = 0, you get x =
c. To find the limits over which (x - c) is equal to one, we can
again go back to its definition in Equation 8 and set the argument
equal to x = ½. Thus, if x – c = ½ for (x - c), the limits over
which this function has a value of 1 are c ± ½. Thus, the function
3 (x + 4) should have a height of 3, be centered at the point x
= -4, and have a value of 1 between x = -4.5 and -3.5. Convince
yourself that this is the case and that this function is correctly
represented in Figure 11.
But there’s more. We can also change the width of this func-
tion. To do this, let’s consider the function (x/b), where b is a
constant. First, to find it’s center point, we again set its argument
equal to zero and solve for x: x/b = 0, x = 0. Obviously the func- Figure 12. Graph of 2 ((x+3)/4).
is not zero, i.e., from -0.5 to 0.5, which is 1. Accordingly, 1*1*3 to 0.125, it is unity within these limits, and we again do a little
= 3. That is, we have just approximated our integral as the area algebra to get:
under the green line in Figure 14. Compared to the true value of
(19) 4 0.125
–0.125
(x2+ 2x+3) dx
the integral (3.0833), this isn’t such a bad approximation.
Let’s continue the process. We’ll now take the next approxi- This integral evaluates to 3.0052. Again, our approximation to
mation of the impulse function in Figure 13, multiply it by x2 + this integral is given by the value of the rectangle function over
2x+3, and take the integral of the product of the two functions, the limits of integration: 4 (red line in Figure 16), multipled by
as follows: the approximation of the function x2+ 2x+3 at x = 0, i.e., f(0) = 3
(yellow line in Figure 16), which gives the green line in Figure
(16) 1
x/ 12 (x2+ 2x+3) dx
– (1/2) 16. We now multiply this height (12) by the width of the function
Of course our approximation to the impulse function is better (1/4) to give us the value of the approximated integral: 12/4 = 3.
here – it is narrower and taller than in the previous example. Let’s We are now even closer to the true value of the integral (3.0052)
do a similar analysis to the one we did before. The function (1/2) 1 than we were before.
x/ 21 (red line in Figure 15) is only nonzero from -0.25 to What if we continue the process, allowing the rectangle func-
0.25, so we will make these values our limits of integration. We tion to become infinitely narrow and infinitely high while keep-
recognize that x/ 12 has a value of unity over these limits, and ing its area at unity? That is, the width of the rectangle function,
so we get: , goes to zero because 0, while the height of the function
1/ goes to infinity as 0. In each case, we approximate the
(17) 2 –0.25
0.25
(x2+ 2x+3) dx
integral in Equation 13 by multiplying 1/ by by f(0), which
This definite integral has a value of: 3.0208. But we can again is what we just did for the three functions in Figure 13 that we
approximate the x2+ 2x+3 function (blue line) by its value at the applied to a simple polynomial. Thus, Equation 13 reduces to *
center of the (1/2)
1
x/ 12 rectangle function, i.e., at x = 0, which 1/ * f(0) = f(0). Clearly, this approximation becomes perfect in
again gives us f(0) = 3 (yellow line in Figure 15). Thus, the ap- the limit of 0. Mathematically, we write:
proximation of the integral (Equation 16) is the value of the prod-
(20) (x) f (x) dx = f (0)
uct of the two functions: 3*2 (the yellow and red lines in Figure –
15, which gives the green line), multiplied by the width of the This result is an example of the sifting property of the impulse
function: ½. That is, the approximation to the area in Equation function.
16 is 3*2*½ = 3. Notice that this approximation is the same as it As one might expect, the impulse function can be shifted to
was before, which is not a coincidence, and that we are now even any point along the x-axis. That is, (a - x) is an impulse function
closer to the true value of the integral (3.0208). centered at the point a on the x-axis. In other words, because
We’ll repeat the process for the third, and narrowest, approxi- the impulse function is only nonzero when its argument is zero,
mation of the impulse function in Figure 13: (1/4)
1
(x/41) = 4 (4x) we solve a – x = 0 to get x = a. Clearly then, if we integrate the
(the red line in Figure 16). Accordingly, we’ll multiply it by x2+ product of (a - x) and f(x), we will pick out the value of f(x) at
2x+3 and take the integral of the product of the two functions: the point where x = a, i.e., we will get back f(a):
(18) 1
(x/41) (x2+ 2x+3) dx (21) –
(a–x) f (x) dx = f (a)
– (1/4)
Thus we can pull out any value of a function by using an appro-
The 1
(x/ 41) rectangle function is only nonzero from -0.125
(1/4) priately shifted impulse function.