Abstract. WinXAS is a new X-ray absorption spectroscopy (XAS) data analysis program. It runs under the operation system MS-Windows and offers several unique features. It has a user-friendly graphical environment and is capable of processing a variety of data formats. It contains a number of useful numerical algorithms beyond those used in standard XAS analysis and offers a simple interface to the ab-initio theoretical code FEFF. The availability of fast macros in WinXAS makes it particularly useful for on-line data examination at a synchrotron radiation facility during XAS experiments as well as analysis of multiple-scan data such as those from time-resolved experiments.

Since the user community is growing rapidly and beamtime at today’s synchrotron radiation facilities becomes more and more precious, every researcher should seek to use this time as efficient as possible. In the case of X-ray absorption spectroscopy (XAS), experimental stations as well as theoretical calculation codes have been improved to a large extend over the past years. Nevertheless, the availability of use-friendly, graphical data analysis programs does not seem to catch up with this evolution.
WinXAS is a new graphical XAS data reduction program running on IBM compatible personal computers (PC) under the operation system MS-Windows. It is aimed to satisfy the requirements of a user-friendly analysis tool, capable of performing all basic XAS data reduction steps in order to help the researcher decide whether more advanced codes need to be employed. Moreover, its easy-to-use and partially automated data reduction procedures (macros), make WinXAS especially suited to be used on line at an early stage of a XAS experiment. Thus, information about the progress of an on-going experiment can be obtained and more often than not, wasting an entire block of beamtime can be avoided by such a tentative data analysis. Additionally, a PC based analysis program seems to be particularly adequate since users might want to take their own computers with them and not rely on analysis software that may or may not be available at the beamline. In addition to its numerical and XAS specific algorithms, WinXAS contains unique features for the analysis of data obtained from energy-dispersive XAS (DXAS) experiments. Investigations that result in a large number of consecutive spectra, such as time-resolved experiments, can benefit from WinXAS, too. Its macro capabilities allow to record single reduction steps and to employ them in a serial treatment to each spectra out of a large data set. Where such an approach is feasible, efficiency can be enhanced and calculation time can be decreased substantially.
In the following, the variety of numerical procedures (XAS related as well as others) and graphical routines that comprise the basic modules of WinXAS are described, together with some general information on system requirements and data format.

WinXAS is running under the operation system Microsoft Windows 95 or later on IBM compatible personal computers. Using a Pentium processor and SVGA monitor resolution is recommended. WinXAS is embedded in the well known MS-Windows Graphical User Interface and takes advantage of its common device drivers and shared facilities. Owing to this general features, the appearance of most MS-Windows programs is similar with respect to their basic functions and getting started with WinXAS for XAS data reduction should be fairly simple even for novice users. A two-button pointing device, such as a PC mouse or a track ball, is used extensively not only in order to select single menu entries but also, for instance, to input values directly from several dialog boxes and to change function parameters interactively. Experienced users, however, will appreciate the numerous keyboard accelerators and short cuts in order to speed-up data processing.
Currently, the latest WinXAS version can be obtained from this site on the world wide web: (http://www.winxas.de). The program is distributed together with an online assistant, a manual, and a graphical visualization program for molecules and clusters (RasMol v2.6 by R. Sayers, 1995).

Since every existing XAS beamline seems to favor its own data format for experimental spectra, hardly one data reduction program can be expected to serve in all cases. Therefore, users are often confronted with the unloved task to transform their data from one format into another. For that purpose, the file format of WinXAS has been chosen very simple and the file-read procedure has been kept as open and flexible as possible. Multiple column ASCII files are automatically recognized and the user is prompted to enter the corresponding column number for photon energy as well as I0 and I1 in order to calculate the absorption spectrum. Therefore, a separate conversion program may turn out to be unnecessary in many cases. However, where such a program is needed its encoding should be feasible with a minimum of time and effort. WinXAS stores data in a binary format as default in order to save storage space and access time. Again, this format has been kept very simple and is well documented. In order to transfer data between applications, spectra can be saved in a two-column ASCII format.
Certainly one of the outstanding features of WinXAS is its user friendly graphical interface. Unlike other line oriented analysis programs, WinXAS offers a full visualization of results and intermediate steps during the data reduction. Up to four spectra can be processed simultaneously, each in its own graphic window, and plots from each window can be superimposed and compared at any time. Moreover, each of the four graphic window contains four internal data arrays that are used to hold and display intermediates and results. Thus, every data reduction step can easily be made undone and input and output of each analysis procedure are available in the same window for comparison. Furthermore, WinXAS plots can readily be exported to all common word processors for documentation using the MS-Windows Clipboard. Additionally, graphics can be printed directly from WinXAS employing any printer that is supported by MS-Windows.

The numerical library implemented in WinXAS contains a selection of common and useful algorithms in order to calculated, for instance, integration, derivatives, root determination, Fourier transformation, smoothing, least-squares refinement of a variety of profile functions, and more. All algorithms can be employed independent from an on-going XAS analysis at every step of the reduction procedure allowing the largest possible flexibility in the data analysis. Each numerical procedure is documented either in the WinXAS manual or the public source it has been taken from and details such as constraints, limitations, and accuracy can be obtained from the cited literature (Press et al., 1989; Engeln-Müllges & Uhlig, 1996). In addition to these numerical algorithms, WinXAS contains several procedures in order to conveniently manipulate entire data sets as well as single data points. Both abscissa and ordinate of a spectrum can be shifted or multiplied by a constant and sections from one spectrum can be copied to another one (cut and paste; for instance, in order to remove steps in the absorption spectrum). Single data points can be changed or deleted just by using the mouse pointer, for example, in order to remove so-called ‘glitches’ (in addition to the more advanced procedures implemented into WinXAS for this purpose).
Besides experimental data, each of the WinXAS graphic windows contains an explicit numerical function. This function can be comprised out of a variety of build-in profile functions (Gaussian, Lorentzian, Voigt, trigonometrical functions, polynomials, etc.). Not only can profile function and XAS spectrum be superimposed, WinXAS also allows to add or multiply profile function and data as well as to apply the profile function to the abscissa or ordinate of the spectrum. In this way, difference files can be calculated and spectra can be normalized. Most of the numerical algorithms can also be applied to the profile function (i.e. integration, differentiation, roots determination, etc.).
Besides basic mathematical combination, profile functions can be refined to experimental data using a standard least-squares algorithm. WinXAS employs the Levenberg-Marquardt algorithm, which is known to converge fast and stable for reasonable initial parameters. Furthermore, WinXAS offers several schemes in order to correlate or constrain function parameters during the refinement (i.e. fixed, equal to each other, basic math operations). In addition to the refined function parameters, the output of a least-squares refinement includes standard deviations, correlation matrices, and the graphical residual. Moreover, in case of a XAS refinement, the Fourier transform of the residual as well as the experimental spectrum and the theoretical XAS function are displayed.

Not all scientific questions addressed by today’s biologists, chemists, and physicists using XAS require a full multiple scattering ab-initio approach. On the other hand, basically every XAS data analysis begins with a visualization of experimental data and basic reduction steps such as energy calibration, glitch removal, and background subtraction. During this first analysis, results need to be properly visualized in order to enable a comparison between different experiments, samples, or reaction conditions. Only on the basis of this tentative data reduction, one can decide in favor of a single scattering analysis using either experimental or theoretical phases, or a full multiple scattering treatment.
Since WinXAS does not consist of separate analysis modules, all numerical procedures and graphical features can be used at every stage of a XAS data analysis. Therefore, WinXAS exceeds the step-by-step approach of elder line-orientated data reduction programs. Furthermore, data reduction features such as smoothing, convolution, XANES profile fitting, linear combination XANES fit using experimental spectra or difference file technique (Vaarkamp et al., 1993), extend the implemented standard XAS capabilities. All algorithms can be employed on an easy-to-use basis and often the most suitable procedure can be chosen out of several different routines. Moreover, already performed reduction steps can easily be made undone and different reduction procedures can be applied to the same spectrum and the respective results can be compared afterwards. Figure 1 displays the WinXAS user interface and four different examples for its XAS data analysis capabilities. WinXAS does not imply one particular way to analyze XAS spectra, but offers the largest possible flexibility. Nevertheless, certain recommendations for the reduction of XAS spectra have been implemented (Teo, 1986; Koningsberger & Prins, 1988). All necessary steps for conventional XAS data reduction can be carried out (background subtraction (Victoreen, one or two polynomial), E0 determination (1st and 2nd derivative, XANES profile function fitting), atomic absorption m0 (cubic spline fitting, polynomial fit) , Fast Fourier Transform (several windows, variable step width in R space)). Furthermore, experimental phase shifts and backscattering amplitudes obtained from reference compounds, as well as theoretical phases and amplitudes can be used in order to perform single-scattering plane wave XAS refinements. Higher cumulants can be added to the standard XAS formula to account for a non-Gaussian distribution in the thermal disorder. XAS refinements can be carried out in k space and in R space.
During all these steps, a proper visualization of results and their dependence on the various parameters is provided. The atomic absorption (m0-fit) window, for instance, not only displays the spectrum and the refined cubic spline, but also the resulting c(k), the first derivative of the cubic spline and the Fourier transformed c(k). Therefore, the user has a maximum of information at hand on how changing the number of nodes for the spline, the weighting scheme, or the section of the spectrum affects the resulting c(k) and its Fourier transform.
As an additional feature, every data reductions step in WinXAS can be recorded in a macro. A complete XAS analysis, containing background correction, m0 fitting, Fourier transform and XAS refinement can basically be carried out by hitting one single button. After recording, these macros can be applied conveniently to a large number of spectra. Although one certainly has to be careful with this sort of automated analysis, it may be a fast and handy way for many opportunities, for instance, a preliminary data analysis at the beamline.

WinXAS contains a simple interface to the ab-initio multiple scattering XAS code FEFF by J. Rehr (Rehr et al., 1992). In the recent past, FEFF has proven to be a reliable and user-friendly theoretical XAS code. Nevertheless, owing to its independence on a particular operating system, FEFF lacks any graphical visualization of its input or output data. Although powerful numerical tools in addition to FEFF are available (Stern et al., 1995), they basically suffer from the same conditions. Since calculation speed on modern PCs became competitive to larger systems, many users may intend to run FEFF calculations on their desktop PC. For these users, WinXAS is meant to fill the gap of a suitable visualization program. Running FEFF on a PC normally involves setting up the proper input files in an editor, executing FEFF and afterwards changing to a graphic program in order to display the calculated XAS function. With WinXAS, these steps can be carried out conveniently in one program’s environment. After FEFF or ATOMS input files have been set up in a WinXAS edit window, the programs can be launched directly from WinXAS and after the calculation is completed, the resulting c(k) is loaded and displayed instantaneously for further analysis or comparison with experimental data.
Furthermore, FEFF generated theoretical phases and amplitudes can be used in order to perform a XAS analysis. WinXAS allows to choose all necessary paths from the FEFF output file „FILES.DAT“ which contains the names of all single output files that belong to the scattering paths included into the calculation. The selected FEFF phases and amplitudes can be employed in a multiple-shell XAS simulations or refinements in both k-space and R-space. Moreover, single XAS function parameters can be changed interactively prior to a refinement. A corresponding dialog box allows to change each XAS parameter stepwise, while after each step the effect on the theoretical c(k) is displayed immediately. This procedure is also very much suited as an educational tool in order to visualize the dependence of the overall shape of a XAS c(k) on the various parameters. Additionally, WinXAS supports RasMol, a simple but powerful molecular structure viewing program (Sayle, 1995). FEFF input files are converted into three-column x,y,z-files that can be read from RasMol and the molecule or cluster for which a theoretical XAS function shall be calculated can be displayed in colors, rotated, and printed. The combination of WinXAS and RasMol does not only make a convenient tool, it can also help novice users to understand the concept of nearest neighbor shells, of different coordination geometry and distortions, and of linear multiple scattering paths for which a particularly strong contribution to the clusters XAS function can be expected.

The analysis of XAS data measured in an energy-dispersive mode (Hagelstein et al., 1989) requires some additional reduction steps compared to conventional XAS data analysis. In contrast to a standard double crystal monochromator, where the energy range is scanned stepwise by changing the Bragg angle of both crystals, the dispersive spectrometer applies a bent monochromator crystal. The absorption signal is then recorded on a position sensitive detector.
The major difference between conventional and energy-dispersive XAS data reduction lies in the energy calibration of the measured absorption spectra. In order to calibrate dispersive XAS spectra, a polynomial has to be calculated representing the energy-position correlation on the detector and converting the measured absorption spectra to photon energy. With WinXAS, the polynomial coefficients can be determined from a automated comparison of the measured metal foil with a calibrated metal foil spectrum obtained at a standard XAS beamline. If DXAS data and the reference metal foil spectrum are available in a comparable quality, this procedure usually results in a satisfactory energy calibration of the dispersive data. A more detailed description on using WinXAS for energy-dispersive XAS analysis can be found in (Ressler, 1995, 1997). Although WinXAS was intentionally written for the DXAS beamline at Hasylab, Hamburg, its functions are generalized enough in order to be used at any other DXAS station. Experimental data either can be used as measured or can easily be converted into a binary WinXAS format.

The analysis of XAS data obtained from time-resolved experiments can become a time consuming and tedious task due to the large number of consecutive absorption spectra. Standard XAS beamlines equipped with fast scanning options (Quick-EXAFS) or energy-dispersive XAS beamlines provide time resolution in the sub-second range and the number of spectra measured during one single time-resolved experiment can amount to several hundred. Such investigations demand data reduction software prepared to cope with large data sets.
WinXAS offers several features in order to treat the large amount of data obtained from time-resolved experiments. Considering the number of spectra from a single run, a data reduction of each spectrum by hand appears to be highly inefficient. Since many analysis steps have to be carried out similarly for each spectrum, efficient data treatment requires a possibility to apply a sequence of analysis steps in a serial procedure. WinXAS can record single data reduction steps into a macro and use this macro in order to analyze each individual spectrum out of a large data set. All numerical and XAS algorithms contained in WinXAS can be used this way. Additionally, several spectra can be averaged in order to improve data quality. Moreover, a serial data treatment can be performed at an early stage of a time-resolved experiment investigation. Despite the fact that WinXAS aims at time- resolved XAS investigations, it can be employed in a variety of experiments that produce a large number of consecutive spectra or patterns that require a serial treatment.

WinXAS is a new XAS data analysis program for PCs running MS-Windows 95 or later. The most important feature of WinXAS is its graphical user interface where all functions and procedures can be selected using either a pointing device or single keyboard function keys. WinXAS allows up to four data files to be processed at a time and results can be superimposed and compared conveniently. Furthermore, WinXAS represents an excellent educational tool due to its immediate graphical response on changes in, for instance, XAS fit parameters, k weighting, number of nodes for cubic spline fits, and much more. Thus, WinXAS can help to gain insight into the interaction of atomic distances, disorder parameters, or number of shells, and their influence on the XAFS function c(k).
WinXAS contains all necessary numerical algorithms and procedures for standard XAS data analysis. Moreover, many additional techniques that exceed standard data reduction can be carried out, such as difference file technique, experimental XANES linear combination fits, XANES profile fitting, use of FEFF theoretical phases and amplitudes for XAS refinement. Nevertheless, WinXAS is meant to complement rather than substitute more advanced XAS data analysis programs and may not be adequate in all cases (for instance, full multiple scattering treatment using ab-initio theoretical phases and amplitudes requires additional XAS analysis codes).
Certainly an advantage of WinXAS is, that it can be employed at a very early stage of every XAS experiments. In many cases beamtime can be used much more efficiently if a preliminary data analysis is carried out on-line. Owing to its open and flexible data file structure, WinXAS can read a variety of different multiple column files without previous conversion.
Besides its numerical and standard XAS features, WinXAS contains several routines for the analysis of energy dispersive XAS data. Furthermore, WinXAS macro can be employed in a serial treatment to a large number of consecutive spectra. These data may be obtained from time-resolved experiments at either energy dispersive or quick scanning XAS beamlines. Such experiment like standard XAS studies benefit from WinXAS in terms of speed, efficiency and convenience.

Engeln-Müllges, G. & Uhlig, F. (1996) Numerical Algorithms with C, Springer, Berlin Hagelstein, M., Cunis, S., Frahm, R, Niemann, W. & Rabe, P (1989). Physica B 158, 324 Koningsberger, D. C. & Prins, R. (1988) X-ray Absorption Spectroscopy, Chemical Analysis 92, Wiley, New York
Press, W. H., Flannery, B. P., Teukolsky, S. A. & Vetterling, W. T. (1989). Numerical Recipes in C, Cambridge University Press
Rehr, J. J., Albers, R. C. & Zabinsky, S.I. (1992). Phys. Rev. Lett. 69, 3397
Ressler, T. (1995). Ph.D. Thesis, University of Hamburg, Germany
Ressler, T. (1997). J. Physique, in press
Sayle, R. (1995), RasMol v2.6-ucb is based on RasMol v2.6, Copyright© 1995, 1996 Roger Sayle, available at ‘http://mc2.cchem.berkeley.edu/Rasmol’ (Enhanced by the MultiCHEM Facility, University of California, Berkeley).
Stern, E.A., Newville, M., Ravel, B., Yacoby, Y. & Haskel, D. (1995). Physica B 208&209, 117
Teo, B. K. (1986). EXAFS : Basic Principles and Data Analysis, Springer, New York
Vaarkamp, M., Modica, F. S., Miller, J. T. & Koningsberger, D. C. (1993). J. Catal. 144, 611
Vaarkamp, M., Dring, I., Oldman, R. J., Stern, E. A. & Koningsberger, D. C. (1994). Phys. Rev. B 50, 7872

Figure 1. ‘Screen-shot’ of the WinXAS program for XAS data analysis showing the user interface as well as four graphic windows containing examples for WinXAS analysis capabilities.
Window I: ‘Quick-XAS’ data analysis of an as-measured absorption spectrum (A). The ‘Quick-XAS’ option represents an automated data reduction procedure. Experimental spectra can be analysed on- line by pressing one single button on the WinXAS interface. The algorithm yields the normalised and background corrected spectrum (B), the extracted XAS function c(k) (C) and the Fourier transformed c(k) (D).
Window II: Least-squares XANES fit. The experimental XANES (dotted) can be simulated by a linear combination of four reference compounds (a (0.1), b (0.21), c (0.23), and d (0.46));
Window III: 3D plot of diffraction data from a time-resolved experiment. In a subsequent analysis, WinXAS’s macro capabilities can be used to determine peak shifts and intensity variation with time;
Window IV: Two shell XAS refinement in R-space using FEFF theoretical phases and amplitudes. The plot shows FT magnitude and imaginary part for experimental (dotted) and theoretical (solid) curve as well as the magnitude for both single shells.