Version 11 Changes

The following list summarizes, for reference, the major changes to XSPEC from version 10 to 11.

• Various changes and bug fixes to the user interface have been made

  $\rightarrow$

  interrupt (Ctrl-C) handling has been added for interactive sessions. Ctrl-C no longer terminates the program except during script execution.

  $\rightarrow$

  In version 10, XSPEC could execute any valid XSPEC/tcl command while processing user input from interactive commands such as newpar, or fakeit. This could leave the program in a peculiar undefined state (for example, if the user typed 'fit' instead of giving parameter values). This possibility has now been excluded.

  $\rightarrow$

  XSPEC will now correctly process multiple line tcl commands (e.g. for {set $i 1} {i < 5} {incr i} { ...) from the command prompt.

  $\rightarrow$

  Changes in script processing: XSPEC scripts should be invoked with @ rather than source, and should be invoked from the unix prompt with %xspec - <script> rather than %xspec <script>. The functionality is unchanged apart from the need to add an exit command to scripts if the program is to terminate after executing the script.

  $\rightarrow$

  A new command, tclout, is available to return xspec internal information to the tcl interface. The command does not require the $xs_return_result variable to be set. tclout writes its output to a Tcl variable called $xspec_tclout, which can then be manipulated using Tcl commands. The v10.0 method of returning variables to the tcl interface is still supported but is deprecated.

  $\rightarrow$

  The single character abbreviations for XSPEC commands have been withdrawn owing to many clashes with commands recognized by tcl's command interpreter. XSPEC commands must be at least two characters in length.

  $\rightarrow$

  The command renorm can no longer be abbreviated to ren, for similar reasons.

  $\rightarrow$

  During parameter setting, XSPEC now displays a text string prompt indicating the data group number (if there is more than one data group), model component, and parameter name.

  $\rightarrow$

  XSPEC now loads the Tk graphical toolkit library at startup. Scripts for XSPEC can therefore be written using tcl and tk commands. For more information about tcl/tk, see
  
  http://www.scriptics.com.

  $\rightarrow$

  XSPEC now sets up a directory, $XSPEC_HOME, in which the user can place tcl/tk scripts that are loaded on start up. This allows users to create their own composite command procedures in tcl/tk as augmented by XSPEC commands. Some examples are provided in the distribution: see also the online documentation at
  
  http://xspec.gsfc.nasa.gov/xanadu/xspec/tcl/tcl.html.
  
  $XSPEC_HOME defaults to $HOME/.xspec

• Changed default Photoelectric absorption cross-sections. The set of photoabsorption cross-sections used by default has been updated. Thus, XSPEC 11 will not in general reproduce results from earlier versions of the program. Those results can be duplicated, however, by resetting the cross section table using the new xsect command. See below for details.

• Y2K compliance. XSPEC had only one substantive Y2K related issue, which arose when creating data simulations (the fakeit command). This version produces fake data files in FITS format with a DATE keyword compliant with the FITS standard for files created after January 1, 2000.

• The build instructions have been completely rewritten. To build the XANADU package, extract the package from the tarfile, change to the BUILD_DIR directory, and issue the commands ./configure ./hmake ./hmake install
• The implementation of user models is now performed by the construction of a shared library. This will enable users to compile their own user model libraries and use them with a system-wide XSPEC installation rather than having to build and maintain a personal copy of the program. See Appendix C for detailed instructions.

• New commands and techniques:

  $\rightarrow$

  New extend command extends the energy range over which a model is calculated. This is useful for convolution models.

  $\rightarrow$

  Bayesian inference. The bayes command sets up Bayesian inference including specifying priors for model parameters. This is valid for data with background even when both source and background spectrum have too few counts to be in the gaussian regime.

  $\rightarrow$

  Included Akaike and Bayesian Information Criteria when running C statistic.

  $\rightarrow$

  Line identification. The identify command lists possible line identifications in the energy range requested.

  $\rightarrow$

  Genetic algorithm (genetic command). A genetic global optimization scheme has been added to the methods. This is slow - with three free parameters it requires 1000s of generations to find the minimum. The current population can be plotted using the plot genpop option.

  $\rightarrow$

  New ftest command calculates F-statistic and probability.

  $\rightarrow$

  Absorption cross-sections. xsect command changes the photoelectric absorption cross-sections used in all the absorption models with the exception of wabs. The options are the old Balucinska-Church & McCammon cross-sections, the same but with new He cross-section, and a compilation by Verner et al.

  $\rightarrow$

  Algebraic models. The mdefine command lets the user define a new model component as an algebraic expression. This new component is evaluated by an interpreter so will be slightly slower than installed models. However, any mdefine'd model component is likely to be relatively simple so the slower evaluation speed should not be significant.

  $\rightarrow$

  XSPEC now auto-loads Tcl scripts to run the commands addline and modid. The former adds lines to the current model in an optimum fashion. The latter attempts to find IDs for all line components in the current model.

  $\rightarrow$

  Uncertainties associated with the model. XSPEC v11 has the capability to include an uncertainty in the model. This is added in quadrature with the uncertainty in the data.

• Modified Commands:

  $\rightarrow$

  plot command has new option, plot genpop, as above.

  $\rightarrow$

  The default steppar option is "best" rather than "current".

  $\rightarrow$

  The error command allows the user to restart the calculation if a new minimum is found. This can be made an automatic option.

  $\rightarrow$

  setplot rebin can take a plot group argument to allow plots to be rebinned independently.

  $\rightarrow$

  For OGIP Type II files, the data command now accepts ranges and wildcards allowing the user to read multiple spectra with a single command, e.g. data file1{1-10} will read the first ten spectra in file1, while data file1{*} will read all of the spectra in file1. The limit to the number of datasets that can be read into the program has been increase from 100 to 500.

  $\rightarrow$

  Parameter linking now allows operations of the form newpar 4 = 3 * a + b where a and b are real numbers. Both / and - are also allowed, e.g. newpar 4 = 3 * 2/3 - 1.5

  $\rightarrow$

  show free option shows only free parameters.

  $\rightarrow$

  Convolution components now operate on model*area rather than just model. This allows pile-up to be modelled as a convolution component.

  $\rightarrow$

  Modified goodness so it works for spectra with background and for the chi-squared statistic provided the only source of variance is counting statistics.

• The following new models have been added:

  ${\bf +}$

  apec, vapec

  ${\bf +}$

  bexrav, bexriv

  ${\bf +}$

  bmc

  ${\bf +}$

  equil, vequil

  ${\bf +}$

  lorentz

  ${\bf +}$

  nei, vnei, gnei, vgnei

  ${\bf +}$

  npshock, pshock, vnpshock, vpshock

  ${\bf +}$

  redden

  ${\bf +}$

  sedov, vsedov

  ${\bf +}$

  srcut,sresc

  
  
  
• The following models have been removed:

  ${\bf -}$

  tsabs

  ${\bf -}$

  tita_a

• The description of the normalization for all collisional plasma models in terms of luminosity distance was incorrect. This has been changed to angular size distance and there should now be the right number of (1+z) factors.
• The model.dat and lmodel.dat files have been simplified and msetup made more robust. For the change to lmodel.dat see the example file in the release. lmodel.dat and the local model source files can be placed in any directory. The XSPEC make procedure uses the environment variable LMODDIR to find the directory containing local models.