\(\newcommand{\AA}{\unicode{x212B}}\)

14.1. XAFS Functions: Overview and Naming Conventions

As with most Larch functions, each of the XAFS functions is designed to be able to act on arbitrary arrays of data to allow maximum flexibility in processing data. In addition, many of the Larch XAFS functions can write out several output values, including both scalars and arrays. While flexible, this could get rather cumbersome, and mean that you would generally have to keep track of a large set of related arrays.

14.1.1. Naming conventions for XAFS arrays

In Larch, it is most natural to put related data into a Group. This is often natural as data read in from a file is already held in a Group. If a set of XAFS data is held within a Group, then the Group members named energy and k and chi can be assumed to have the same standard meaning for all groups of XAFS Data. To make this most common usage easy, all of the XAFS functions follow a couple conventions for working more easily with Groups that can work on arbitrary arrays of data, but assume that they will write output values to a Group. In addition, the XAFS functions can usually just be given a Group that follows the expected XAFS naming convention. This is not rigidly enforced, and is not exclusive (that is, you can add data with other names), but following the expected naming conventions will make processing XAFS data much easier.

The naming convention define a set of expected names and meaning for data arrays and scalars within a Group. This is summarized in the Table of Conventional Names for an XAFS Group below.

Table of Conventional Names for an XAFS Group These are the standard names for arrays and scalars for various data associated with XAFS, including FEFF calculations. Given are the name, the physical quantity described, and the name of function that will generate this value.

name	type	meaning	created by
energy	array	\(E\) in eV	original data
mu	array	\(\mu\)	original data
e0	scalar	\(E_0\)	pre_edge(), find_e0()
edge_step	scalar	\(\Delta \mu\)	pre_edge()
dmude	array	\(d\mu/dE\)	pre_edge()
norm	array	normalized \(\mu\)	pre_edge()
flat	array	flattened \(\mu\)	pre_edge()
pre_edge	array	pre-edge curve	pre_edge()
post_edge	array	normalization curve	pre_edge()
bkg	array	\(\mu_0(E)\)	autobk()
chie	array	\(\chi(E)\)	autobk()
k	array	\(k\)	autobk()
chi	array	\(\chi(k)\)	autobk()
kwin	array	\(\Omega(k)\)	xftf(), ftwindow()
r	array	\(R\)	xftf()
chir	array	\(\chi(R)\) (complex)	xftf()
chir_mag	array	\(|\chi(R)|\)	xftf()
chir_re	array	\(\rm Re[\chi(R)]\)	xftf()
chir_im	array	\(\rm Im[\chi(R)]\)	xftf()
rwin	array	\(\Omega(R)\)	xftr()
q	array	\(q\)	xftr()
chiq	array	\(\chi(q)\) (complex)	xftr()
chiq_mag	array	\(|\chi(q)|\)	xftr()
chiq_re	array	\(\rm Re[\chi(q)]\)	xftr()
chiq_im	array	\(\rm Im[\chi(q)]\)	xftr()

where \(q\), \(\chi(q)\), and so on indicates back-transformed \(k\).

The XAFS functions encourage following this convention, in that they are consistent in wanting \(\chi(k)\) to be represented by the two arrays GROUP.k and GROUP.chi

14.1.1.1. group argument and _sys.xafsGroup

The XAFS functions need to write outputs to some group – there are simply too many outputs to return and expect you to manage. To better accomodate this, all functions take a group argument, which is used as the group into which results are written. This gives a convenient way to manage the results of the different analysis steps, but gets tedious to provide this argument repeatedly when working with a particular data set.

For XAFS analysis, there is also a special group, _sys.xafsGroup that is used as the default group to write outputs to if no group argument is supplied. In addition, when an explicit group argument is given, _sys.xafsGroup is set to this group. In short, the _sys.xafsGroup will be used as the “current, default group”. This means that when working with a set of XAFS data all contained within a single group (which is expected to be the normal case), the group argument does not need to be typed repeatedly.

Because this uses a global group in the Larch interpreter, this convention works from with the Larch language, but does not work from plain Python unless an instance of a Larch session is passed into the larch.xafs function using the _larch argument.

14.1.1.2. First Argument Group convention

Since the XAFS functions need to write outputs to some group and will generally work with groups that contain data following Table of Conventional Names for an XAFS Group, most of the XAFS functions follow what is called the First Argument Group convention. This convention gives a simple approach when working with groups of XAFS data and it is worth understanding and using this for most work with the XAFS work. This convention is built on the _sys.xafsGroup convention discussed above but is even easier to use.

While the XAFS functions can take arrays of data as the first two arguments most work will have these arrays in a single group with array names that follows the conventions above. As an example, the most general use of the autobk() function takes an array of energy as the first argument, an array of mu as the second argument, and supplying an output group for placing all the arrays and data calculated within the function. That is, the most general use would look like:

autobk(energy, mu, group=dat, rbkg=1, ....)

Of course, most usage will actually want to use energy and mu arrays from the same group, and use that group as the output group, so that all data stays contained within the same group. That would make the call above look like:

autobk(dat.energy, dat.mu, group=dat, rbkg=1, ....)

where the group name dat is repeated three times.

The First Argument Group convention allows this to be written as:

autobk(dat, rbkg=1, ....)

That is, as long as the Group dat follows the XAFS naming conventions (for autobk() that it has an energy array named energy and absorbance array named mu) the two forms above are equivalent. All the XAFS functions follow this convention and use a consistent set of attribute names (see Table of Conventional Names for an XAFS Group). This convention nearly makes the Larch XAFS routines into object-oriented, or in this case Group oriented, set of functions that interact in a coherent and predictable way on an XAFS dataset.

14.1.2. Plotting Macros for XAFS

XAFS analysis often uses several different standard views of the data arrays for \(\mu(E)\), \(\chi(k)\), and \(\chi(R)\). Larch’s plotting capabilities provide wide flexibility in how plots can be done. While that flexibility can be useful in general, within the narrow scope of plotting XAFS data, being able to easily create consistent plots with reasonable defaults produces results that are easier to digest and understand.

The macros described here attempt to provide that functionality of easy-to-use standard plotting macros. In particular, they automatically handle typesetting the labels for the plot axes in a consistent manner, and assign consistent labels to the different curves shown. The results are also easily extended, so that you can add curves, annotations, etc. Many of the examples in the following sections in this chapter make use of these macros.

14.1.2.1. plot_mu()

plot_mu(dgroup, norm=False, deriv=False, show_pre=False, show_post=False, show_e0=False, emin=None, emax=None, label=None, new=True, win=1)

Plot \(\mu(E)\) for an XAFS data group in various forms

Parameters:

• dgroup – group of XAFS data after pre_edge() results (see note below)

• norm – bool whether to show normalized data [False]

• deriv – bool whether to show derivative of XAFS data [False]

• show_pre – bool whether to show pre-edge curve [False]

• show_post – bool whether to show post-edge curve [False]

• show_e0 – bool whether to show E0 [False]

• show_deriv – bool whether to show deriv together with mu [False]

• emin – min energy to show, relative to E0 [None, start of data]

• emax – max energy to show, relative to E0 [None, end of data]

• label – string for label [None: ‘mu’, ‘dmu/dE’, or ‘mu norm’]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

The input data group must have the following attributes: energy, mu, norm, e0, pre_edge, edge_step, filename

14.1.2.2. plot_bkg()

plot_bkg(dgroup, norm=True, emin=None, emax=None, show_e0=False, label=None, new=True, win=1)

Plot \(\mu(E)\) and background \(\mu_0(E)\) for XAFS data group

Parameters:

• dgroup – group of XAFS data after autobk() results (see note below)

• norm – bool whether to show normalized data [True]

• emin – min energy to show, relative to \(E_0\) [None, start of data]

• emax – max energy to show, relative to \(E_0\) [None, end of data]

• show_e0 – bool whether to show E0 [False]

• label – string for label [None: ‘mu’]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

The input data group must have the following attributes: energy, mu, bkg, norm, e0, pre_edge, edge_step, filename

14.1.2.3. plot_chik()

plot_chik(dgroup, kweight=None, kmax=None, show_window=True, label=None, new=True, win=1)

Plot k-weighted \(\chi(k)\) for XAFS data group

Parameters:

• dgroup – group of XAFS data after autobk() results (see note below)

• kweight – k-weighting for plot [read from last xftf(), or 0]

• kmax – max k to show [None, end of data]

• show_window – bool whether to also plot k-window [True]

• label – string for label [None: ‘chi’]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

The input data group must have the following attributes: k, chi, kwin, filename.

14.1.2.4. plot_chir()

plot_chir(dgroup, show_mag=True, show_real=False, show_imag=False, rmax=None, label=True, new=True, win=1)

Plot \(\chi(R)\) for XAFS data group

Parameters:

• dgroup – group of XAFS data after xftf() results (see note below)

• show_mag – bool whether to plot \(|\chi(R)|\) [True]

• show_real – bool whether to plot \(Re[\chi(R)]\) [False]

• show_imag – bool whether to plot \(Im[\chi(R)]\) [False]

• rmax – max R to show [None, end of data]

• label – string for label [None: ‘chir’]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

The input data group must have the following attributes: r, chir_mag, chir_im, chir_re, kweight, filename

14.1.2.5. plot_chifit()

plot_chifit(dataset, kmin=0, kmax=None, rmax=None, show_mag=True, show_real=False, show_imag=False, new=True, win=1)

Plot k-weighted \(\chi(k)\) and \(\chi(R)\) for fit to feffit dataset

Parameters:

• dataset – feffit dataset, after running feffit().

• kmin – min k to show [0]

• kmax – max k to show [None, end of data]

• rmax – max R to show [None, end of data]

• show_mag – bool whether to plot \(|chi(R)|\) [True]

• show_real – bool whether to plot \(Re[`chi(R)]\) [False]

• show_imag – bool whether to plot \(Im[\chi(R)]\) [False]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

14.1.2.6. plot_path_k()

plot_path_k(dataset, ipath, kmin=0, kmax=None, offset=0, label=None, new=False, win=1, **kws)

Plot k-weighted \(\chi(k)\) for a single Path of a feffit dataset

Parameters:

• dataset – feffit dataset, after running feffit()

• ipath – index of path, starting count at 0 [0]

• kmin – min k to show [0]

• kmax – max k to show [None, end of data]

• offset – vertical offset to use for plot [0]

• label – path label [‘path I’]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

• kws – additional keyword arguments are passed to plot()

14.1.2.7. plot_path_r()

plot_path_r(dataset, ipath, rmax=None, offset=0, label=None, show_mag=True, show_real=False, show_imag=True, new=False, win=1, **kws)

Plot \(\chi(R)\) for a single Path of a feffit dataset

Parameters:

• dataset – feffit dataset, after running feffit()

• ipath – index of path, starting count at 0 [0]

• kmax – max k to show [None, end of data]

• offset – vertical offset to use for plot [0]

• label – path label [‘path I’]

• show_mag – bool whether to plot \(|\chi(R)|\) [True]

• show_real – bool whether to plot \(Re[\chi(R)]\) [False]

• show_imag – bool whether to plot \(Im[\chi(R)]\) [False]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

• kws – additional keyword arguments are passed to plot()

14.1.2.8. plot_paths_k()

plot_paths_k(dataset, offset=-1, kmin=0, kmax=None, new=True, win=1, **kws):

Plot k-weighted chi(k) for model and all paths of a feffit dataset

Parameters:

• dataset – feffit dataset, after running feffit()

• kmin – min k to show [0]

• kmax – max k to show [None, end of data]

• offset – vertical offset to use for paths for plot [-1]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

• kws – additional keyword arguments are passed to plot()

14.1.2.9. plot_paths_r()

plot_paths_r(dataset, offset=-0.5, rmax=None, show_mag=True, show_real=False, show_imag=False, new=True, win=1, **kws):

Plot \(\chi(R)\) for model and all paths of a feffit dataset

Parameters:

• dataset – feffit dataset, after running func:feffit

• offset – vertical offset to use for paths for plot [-0.5]

• rmax – max R to show [None, end of data]

• show_mag – bool whether to plot \(|\chi(R)|\) [T``rue``]

• show_real – bool whether to plot \(Re[\chi(R)]\) [False]

• show_imag – bool whether to plot \(Im[\chi(R)]\) [False]

• new – bool whether to start a new plot [True]

• win – integer plot window to use [1]

• kws – additional keyword arguments are passed to plot()

14.1.2.10. plot_prepeaks_baseline()

plot_prepeaks_baseline(dgroup, subtract_baseline=False, show_fitrange=True, show_peakrange=True, win=1, **kws):

Plot pre-edge peaks and baseline fit, as from pre_edge_baseline() or Larix.

Parameters:

• dgroup – data group, after running pre_edge_baseline()

• subtract_baseline – bool whether to subtract baseline for plot

• show_fitrange – bool whether to show fit range as vertical bars

• show_peakrange – bool whether to show pre-edge peak range with markers

• win – integer plot window to use [1]

• kws – additional keyword arguments are passed to plot()

The dgroup group must have a prepeaks subgroup.

14.1.2.11. plot_prepeaks_fit()

plot_prepeaks_fit(dgroup, show_init=False, subtract_baseline=False, show_residual=False, win=1, **kws):

Plot pre-edge peaks and fit, as in the Larix GUI

Parameters:

• dgroup – data group, after running pre-edge peak fit.

• show_init – bool whether to show initial model, before fitting

• subtract_baseline – bool whether to subtract baseline for plot

• show_residual – bool whether to show residual as a stacked plot.

• win – integer plot window to use [1]

• kws – additional keyword arguments are passed to plot()

The dgroup group must have a peakfit_history subgroup. Currently, this is automatically generated only using the Larix GUI or scripts written (and possibly altered) by it.

14.1.3. Utility Functions for XAFS

Listed here are some general purpose functions for XAFS.

14.1.3.1. ktoe() and etok()

etok(energies)

Convert photo-electron energy in eV to wavenumber in \(\AA^{-1}\). energies can be a single number or array of numbers.

ktoe(wavenumbers)

Convert photo-electron wavenumber in \(\AA^{-1}\) to energy in eV. wavenumber can be a single number or array of numbers.

An example use would be to print out a table of energies and \(k\) values:

larch> kvals = linspace(0, 25, 26)
larch> evals = ktoe(kvals)
larch> for k,e in zip(kvals, evals)):
larch>      print " %5.1f 1/Ang ->  %8.2f eV" %(k , e)
larch> endfor
   0.0 1/Ang ->      0.00 eV
   1.0 1/Ang ->      3.81 eV
   2.0 1/Ang ->     15.24 eV
   3.0 1/Ang ->     34.29 eV
   4.0 1/Ang ->     60.96 eV
   5.0 1/Ang ->     95.25 eV
   6.0 1/Ang ->    137.16 eV
   7.0 1/Ang ->    186.69 eV
   8.0 1/Ang ->    243.84 eV
   9.0 1/Ang ->    308.61 eV
  10.0 1/Ang ->    381.00 eV
  11.0 1/Ang ->    461.01 eV
  12.0 1/Ang ->    548.64 eV
  13.0 1/Ang ->    643.89 eV
  14.0 1/Ang ->    746.76 eV
  15.0 1/Ang ->    857.25 eV
  16.0 1/Ang ->    975.36 eV
  17.0 1/Ang ->   1101.08 eV
  18.0 1/Ang ->   1234.43 eV
  19.0 1/Ang ->   1375.40 eV
  20.0 1/Ang ->   1523.99 eV
  21.0 1/Ang ->   1680.20 eV
  22.0 1/Ang ->   1844.03 eV
  23.0 1/Ang ->   2015.48 eV
  24.0 1/Ang ->   2194.55 eV
  25.0 1/Ang ->   2381.24 eV

14.1.3.2. estimate_noise()

estimate_noise(k, chi=None, group=None, rmin=15, rmax=30, ....)

Automatically estimate the noise level in a \(\chi(k)\) spectrum.

Parameters:

• k – 1-d array of \(k\)

• chi – 1-d array of \(\chi\)

• group – output group.

• rmin – minimum \(R\) value for noise estimate.

• rmax – maximum \(R\) value for noise estimate.

• kweight – exponent for weighting spectra by k**kweight [1]

• kmin – starting k for FT Window [0]

• kmax – ending k for FT Window [20]

• dk – tapering parameter for FT Window [4]

• dk2 – second tapering parameter for FT Window [None]

• window – name of window type [‘kaiser’]

• nfft – value to use for N_fft [2048].

• kstep – value to use for delta_k ( Ang^-1) [0.05]

The method uses an XAFS Fourier transform, and many of arguments (kmin, kmax, etc) are identical to those of xftf().

This function follows the First Argument Group convention with arrays named k and chi. The following outputs are written to the supplied group (or _sys.xafsGroup if group is not supplied):

attribute	meaning
epsilon_k	estimated noise level in \(\chi(k)\).
epsilon_r	estimated noise level in \(\chi(R)\).
kmax_suggest	suggested highest \(k\) value for which \(|\chi(k)| > \epsilon_k\)

This method uses the high-R portion of \(\chi(R)\) (between rmin and rmax) as a measure of the noise level in the \(\chi(R)\) data and uses Parseval’s theorem to convert this noise level to that in \(\chi(k)\). This method implicitly assumes that there is no signal in the high-R portion of the spectrum, and that the noise in the spectrum is “white” (independent of \(R\)) . Each of these assumptions can be legitimately questioned. Then again, making the assertion that these assumptions are invalid and disregarding the estimated noise level here would require knowledge of the noise in an XAFS spectrum that most users do not have. At the very least, this estimate should be be interpreted as a minimal estimate of the noise level in \(\chi(k)\).

The estimate for the output value kmax_suggest has a tendency to be pessimistic in how far out the \(\chi(k)\) data goes before being dominated by noise, but has the advantage of being an impartial measure of data quality. It is particularly pessimistic for extremely good data. Then again, considering that the estimate for \(\epsilon\) is probably too small, the estimate may not be that bad.