Chapter 5  Advanced usage of WHIZARD

• Feynman diagram selections
• Restrictions on intermediate states
• User-defined spectra and structure functions
• User-defined cuts
• Reweighting matrix elements
• Command-line options
• Tables and file management
• WHIZARD as a subroutine library
• Interface of the WHIZARD module

In the preceding chapters we have described the use of WHIZARD as a stand-alone program. As such, it is able to calculate total cross sections and to generate events on the partonic level, or even on the hadronic level if fragmentation is enabled. The events can be analyzed online (on the partonic level) or written to file (on the partonic or on the hadronic level).

This already covers many real-life applications of a Monte-Carlo integration and event-generation package. For the theorist, it may be sufficient to calculate a total cross section with a given set of parameters and to plot differential distributions. For an experimental simulation, the generated event files can be used for further processing.

However, one encounters situations where a more elaborate framework is desired. For instance, one would like to consider multiple sets of cuts, energies and parameter sets. For each set, the input file has to be modified, the program has to be run, and the results have to be stored somewhere in order to be compared later. This can easily become cumbersome, and it is desirable to have a generic approach. Another common problem is the selection of a certain subclass of Feynman diagrams contributing to a given process, either for speeding up the calculation by dropping irrelevant terms, or for getting a better understanding of the relative weight of interfering subprocesses. Some possibilities are described in the following sections.

5.1  Feynman diagram selections

While the matrix element generators called by WHIZARD, by default, generate an amplitude which corresponds to the complete set of Feynman diagrams which connect a given initial and final state, there are problems where only a subset is needed. Therefore, WHIZARD has support for diagram selections. Since O’Mega does not use Feynman diagrams, this option is available only when using the CompHEP and MadGraph matrix element generators. (O’Mega, however, supports restrictions on the internal lines of an amplitude. This is explained below in Sec. 5.2.) Since the CompHEP and MadGraph matrix element generators are considered to be deprecated features we describe Feynman diagram selection here only briefly.

The selection must be made before the final executable is generated. In the configuration file, the process must be marked accordingly by the letter “s” in the last column, for instance:

# Processes
# Tag           In      Out             Method  Option
#=====================================================
cc3             e1,E1   e2,N2,u,D       chep    s

Note that this makes sense only for the methods chep and mad, as mentioned before. Next, the diagram numbers to be selected have to be determined. To this end, for each process with this option a diagram selection file has to be put into the configuration directory.

Since the user usually does not know diagram numbers in advance, it is possible to do an empty run of the matrix element generators first. If the command

make diagrams

is issued, matrix element generation is started as usual, but the Fortran95 source code is not actually generated (at least, with CompHEP: for the other generators it makes little difference). Instead, after this command is completed, you will find PostScript files with Feynman diagrams in graphical representation in the subdirectory processes-src. View or print them and note, for each process, the diagrams you are interested in. In the subdirectory conf you should now find, for each process, a file with the extension .sel, which contains one entry for each diagram. Initially, all entries are set 0 (false). As a consequence, if the file is not modified, a second WHIZARD run would fail since all diagrams are deleted initially. Change all entries you wish to enable into T or 1 (true). If you now rerun WHIZARD to generate the actual source code (you might need to touch the file whizard.prc in the conf subdirectory in order to force re-generation of the process under consideration)

make proc

or even

make prg install

WHIZARD should read the selection files and take into account only the specified subsets of diagrams. You may control this by looking at the PostScript output this second pass will produce: it should contain only the Feynman diagrams you have selected.

Needless to say, this feature should be used with great care. By violating electroweak or even electromagnetic gauge invariance, the results may be off by far, since delicate cancellations may have been spoiled. As a rule, never delete diagrams in any gauge other than unitarity gauge (in the CompHEP) case, and remember which diagrams have been selected (i.e., save the PostScript file somewhere) when storing results, since the output of the WHIZARD executable itself does not contain any hint that only a subset of Feynman diagrams have been taken into account. In many cases the same effect can also be achieved in a simpler way by putting some parameter to extreme values. For instance, instead of removing Higgs diagrams by hand, one may set the Higgs mass to a very large value — the results will then be valid in any gauge.

5.2  Restrictions on intermediate states

As an alternative (and often more convenient) method for selecting Feynman diagrams, the O’Mega interface supports restrictions on internal lines.¹ As an example, the following specification selects the resonant WW diagrams, analogous to the previous example:

# Processes
# Tag           In      Out             Method  Option
#=====================================================
cc3             e1,E1   e1,N1,u,D       omega   r: 3+4~W-&&5+6~W+

Simultaneous requirements for a process are concatenated by && (AND). Note that spaces in the restrictions option (’r’) might lead to a failure of the scripts triggering the O’Mega matrix element generation.

The following examples select only diagrams with an s-channel and t-channel γ^*, respectively:

cc_gamma_s      e1,E1   e1,N1,u,D       omega   r: 3+4+5+6~A
cc_gamma_t      e1,E1   e1,N1,u,D       omega   r: 1+3~A

The in- and out-particles are numbered consecutively. The tilde selects only diagrams where the specified combination of momenta corresponds to the propagator for the given particle. A question mark instead of a particle name stands for all possible propagators with the specified momentum combination. These particle names have to follow the O’Mega notation; they are not translated by WHIZARD as in the process specification. For colored intermediate states, WHIZARD takes care automatically that the intermediate particle can appear in different color flow combinations (in WHIZARD 2 this is handled directly by O’Mega itself in the matrix element generation). It is important to keep in mind that such a selection typically violates gauge invariance, since the intermediate particle is not forced on shell. (An on-shell option is also present in O’Mega, but not yet supported by WHIZARD.)

The || (OR) relation is also possible, and sometimes needed, as in this example:

ww_or_zz   e1,E1   e1,N1,E1,n1    omega   r: 3+4~W- && 5+6~W+ || 3+5~Z && 4+6~Z

AND and OR relations can be grouped by parantheses (). Note that combinations of AND and OR as implemented in the setup of O’Mega in WHIZARD 1 could lead to logical inconsistencies in the selections of intermediate states, so use this feature with great care. In WHIZARD 2 this has therefore been replaced by the option to use lists of intermediate states.

Currently, WHIZARD does not interpret or use such restrictions for its phase space setup, so the possible speedup is constrained to the matrix element evaluation. Nevertheless, this feature can be very useful. Such an option is foreseen for a future release of WHIZARD 2, however.

5.3  User-defined spectra and structure functions

It is possible to insert arbitrary beam spectra into the WHIZARD code. If the corresponding code is included in the user module, it can be called from the main program. To do this, in the whizard-src subdirectory, copy the file user.f90.default to user.f90 and edit the appropriate functions defined within this module according to your needs. The code you have written will be compiled instead of the default code.

The comments in the user.f90.default source should guide you. There is the possibility to define either single-beam (factorized) or correlated spectra. The particle codes which have been specified in the beam_input block of the input file can be accessed in order to select between different spectra for different beams.

In the input file, for each beam there is a master switch USER_spectrum_on which determines whether the user code is called. If no user code has been written, by default a uniform x spectrum between 0 and 1 is inserted which could be used for testing purposes. Furthermore, two parameters can be specified: USER_spectrum_mode, an integer which the user function can access to select between different spectra for the same particle, and USER_spectrum_sqrts which by default is the total c.m. energy, but in principle can be used to transfer any real constant parameter.

Apart from user-defined spectra, user-defined structure functions can be inserted. They are handled in an exactly analogous way (with spectrum replaced by strfun in parameter names). The only difference is that user spectra are applied before any built-in spectra and structure functions (like beamstrahlung, PDFs, etc.), while user structure functions are applied after built in ones, i.e., they are called for the partonic initial state. For applications, think of a photon collider spectrum for the initial beams as opposed to the scattering of W bosons emitted from quarks.

The efficiency of the program in the presence of user-defined spectra or structure functions depends on the way they are implemented by the user. It is straightforward to insert the corresponding functions in a literal way as functions f(x) where x is the beam energy fraction. However, if f is strongly peaked this might lead to poor performance. In many cases one rather should apply a variable transformation

where x=g(y). The user interface allows for this, where the input parameters are identified with y and the output parameters x=g(y) can be different. In that case, the spectral function f(x)=f(g(y)) has to be multiplied by the Jacobian dg/dy within the structure function code.

The implementation can be checked by making use of the “test” matrix elements (Sec. 4.1.2). You may set up a placeholder process like

  sftest   e1,E1   A,A    test

where the final state is massless and remove all cuts (in the example, set default_Q_cut=0 or define an empty entry in the cut file). The matrix element is set constant, and the would-be cross section for the process is the integral of the structure function

which is usually known. Furthermore, the energy distribution of the final state should reproduce the structure function shape.

In the current implementation, beam polarization (if any) cannot be accessed or modified by the user spectra or structure functions. Beam polarization will be kept unchanged if the in- and out-particle of the corresponding splitting coincide (as for ISR), and it will be ignored in the opposite case (as for PDFs).

5.4  User-defined cuts

In many applications the built-in capabilities for cuts are not sufficient. If cuts are defined using the standard interface, all cuts are composed such that an event is only retained if it passes all cuts (logical and). There can be several windows in one cut variable, but beyond that it is not possible to compose cuts by a logical or. Furthermore, you may be missing your favorite cut model in the list of predefined variables. Note that WHIZARD 2 revised this model and makes the definition of almost arbitrary functions for cuts on (combinations of) particles possible.

User-defined cut algorithms can easily be placed in the user module which comes with the distribution, the same module where user-defined spectra can be coded. To activate the extra code, the parameter user_cut_mode in the input file should be set to a nonzero integer. Then, the new cut function will be applied throughout program execution in addition to the cut conditions defined in the standard way. Within the cut function, the numerical value of user_cut_mode is used to select one of several cut modes defined by the user. (To disable the default cuts, which may be desired in this case, write an empty process entry in whizard.cut1.)

The cut function takes an array of momenta (this type and the functions operating on it is defined in the module lorentz.f90) and results in the logical variable accept. The default behavior is to unconditionally set accept=.true., but user code would change that, deciding whether to accept the event based on the momentum configuration and the particle codes.

As an example, let us implement a cut such that an event is accepted if there is a photon with energy larger than 50 GeV. (This is not possible by the built-in cut capabilities, since for a generic process we do not know which photon to check.) To this end, we first make a copy of the standard (no-op) user file, which then can be changed:

cd whizard-src
cp user.f90.default user.f90

The file user.f90 then has to be edited. We take the section which refers to the user cut definition, which originally reads

! This minimal cut function accepts everything

  function cut (p, code, mode) result (accept)
    type(four_momentum), dimension(:), intent(in) :: p
    integer, dimension(:), intent(in) :: code
    integer, intent(in) :: mode
    logical :: accept
    select case (mode)
    case default
       accept = .true.
    end select
  end function cut

and change it as follows:

  ! This cut function accepts an event only if there is a photon with 
!   energy larger than 50 GeV

  function cut (p, code, mode) result (accept)
    type(four_momentum), dimension(:), intent(in) :: p
    integer, dimension(:), intent(in) :: code
    integer, intent(in) :: mode
    logical :: accept
    integer :: i
    select case (mode)
    case (1)
       accept = .false.
       do i = 1, size (code)
          if (code(i) == PHOTON .and. energy (p(i)) > 50) then
             accept = .true.
          end if
       end do
    case default
       accept = .true.
    end select
  end function cut

We have inserted a loop over all outgoing particles, such that the code can be applied to an arbitrary process. For each particle i, its PDG code is checked against the predefined code PHOTON (which is defined as 22 in particles.f90), and the function energy is evaluated for the corresponding momentum p(i). This function, which takes a variable of type four_momentum as argument and returns a real number of default kind (i.e., double precision), is defined in lorentz.f90.

When WHIZARD is now (re-)compiled and installed,

make whizard install

the user code becomes part of the program. To activate it, the integration_input in the whizard.in file has to specify a nonzero user-cut mode, i.e.,

  &integration_input
  ...
  user_cut_mode = 1
  / 

since in the function defined above, the mode 1 refers to the special treatment of photons. When the program is started with this setting, you will see the message

  ! Activating additional preselection cuts in user.f90: Mode # 1

in the screen output, and the cross section and event distribution should reflect the new cut condition.

5.5  Reweighting matrix elements

Beyond the possibility to apply arbitrary cuts, the user might be interested in reweighting events. One should distinguish two different situations:

1. A given event sample is rescanned, recalculating the matrix element values event by event with some modification applied. The resulting events carry a weight which is given by the ratio of the new matrix element values compared to the old ones. The event selection is still based on the unmodified values.

   To achieve this, one should first generate a reference sample in the usual way (adaptation and event generation) which results in the raw event file whizard.evx. Then, one may change the input parameters as desired, set the parameter recalculate=T, and start a second run. This will have two effects: First, the parameters read_grids and read_events are assumed also to be set (even if the user has not modified them), together with read_grids_force and read_events_force.² Thus, the grids and events of the first run are reused. Second, when the event sample is read, the particle momenta are left unchanged, but the matrix element values and the quantum number probabilities are recalculated, using the new input data.

   The change in the matrix element values will be taken into account in the internal analysis: The integral and histograms will be calculated using the ratio of new and old matrix element values as event weights. If the results of the reference run have been stored (one should rename output files to prevent them from being overwritten), the results can easily be compared.

   To make further use of the reweighted events, one can write the events to file (using the option write_events=T) in “long” (ASCII) format: write_events_format=3. In addition to the particle codes, helicities and momenta, the file will contain the (recalculated) matrix element value and the ratio of new and old values for each event. Again, if several parameter sets are to be compared, one should make sure to rename the output file whizard.evt after each run to prevent overwriting.

   In any case, one has to be careful. While it does not harm to change a coupling constant by a small amount (e.g., an anomalous coupling) and rescan the events, modifying particle masses and cuts may have undesired side-effects since the momentum configurations are left intact. For instance, with new mass assignments particles may no longer be on-shell, such that delicate gauge cancellations in the matrix elements are invalid, resulting in unphysical effects.

2. The matrix element function itself is modified by a user-defined weight, a function of the outgoing momenta. The weight function is applied throughout adaptation and event generation, so the integral and the event selection depend on it.

   A natural application is the calculation of integrated observables other than the total cross section, e.g. moments of the angular distribution. One might also think of radiative corrections in “K-factor” form: a function of the momenta which multiplies the lowest-order matrix element.³ The corresponding code can be inserted in the weight routine in the user.f90 module. Similar to the cut routine (Sec.5.4), it takes the array of particle momenta and particle codes to generate a single number weight. In the default routine, weight is always equal to one. The user may change this to an arbitrary function, provided it is nonnegative in the whole accessible phase space.⁴ The user-defined weight function is then activated by setting user_weight_mode nonzero in the input file. The numerical value can be used to select between different reweighting modes.

It is perfectly possible to combine the two approaches: One may define one or more reweighting functions in the user module. As a reference run, events are generated with user_weight_mode=0. Then, both recalculate is set and the mode is set nonzero in a second run. A user-defined cut function can be treated analogously.

5.6  Command-line options

WHIZARD allows to specify options on the command line. Options are available both in long form (double dashes) and in short form (a single dash). For instance, the following command executes WHIZARD in subdirectory tmp for the process ee_nnh, without reading an input file:

./whizard -nd tmp --process_input ’process_id = "ee_nnh"’

5.6.1  General options

These options are executed by WHIZARD before any files are read or written

If a working directory different from the default one is chosen by the -d option, one has to make sure that either a phase space file (.phs or .phx) or the model definition file (.mdl) is present in the chosen directory. Otherwise, one could use command-line options to point to files in the initial directory (or anywhere else), for instance:

./whizard --directory tmp --integration_input ’read_model_file="./whizard"’

When a --directory setting is in effect, it can be overriden by filenames that begin with “/” (absolute file name) or with “./” (filename relative to the initial directory). Therefore, in the example the model file whizard.mdl will be looked for in the directory where the command was executed, while all other files will be looked for in or written to the subdirectory tmp.

5.6.2  Options for setting parameters

Parameters can be set not just in the input file, but also on the command line. The command-line parameters are read after all input files have been read; they override settings in the input file(s). Thus, one can set up a generic input file whizard.in where the constant parameters are defined (or no input file at all) and use the command line to run WHIZARD for, e.g., a specific process, beam energy and Higgs boson mass:

./whizard -p ’process_id="ee_nnh" sqrts=800’ -P ’mH = 130’

Note that the option arguments are interpreted verbatim as if they were contained in the input file, so string values have to be enclosed by extra quotes.

5.7  Tables and file management

The WHIZARD program, after it has been compiled with a specific list of processes, may be viewed as a kind of filter: It takes as input

• the model definition file whizard.mdl,
• the parameter definition file whizard.in, and optionally
• the cut definition file whizard.cut1
• and the cut/analysis definition file whizard.cut5,

and transforms the information contained within into output files, namely

• the integration results and diagnostics, collected in the file whizard.out,
• the integration grids, usable to bypass further time-consuming adaptation runs (whizard.grb), and optionally
• an event file usable for re-running the program (whizard.evx);
• an event file for further processing by external programs (whizard.evt);
• and a set of histograms (whizard-plots.dat) which can be converted to PostScript format (whizard-plots.ps), or alternatively read in by an analysis package like ROOT.

For a second WHIZARD run, the input files have to be edited, and the output files will be overwritten if they have not been moved to a different location. Instead of doing this by hand, one could write a script that executes a loop where in each iteration the relevant line in the input file is modified, the program is run, and the output files are renamed or moved such that they do not get overwritten. Here is such a script:

#!/usr/bin/perl
# Loop over the total energy, write a table and store the results
# in subdirectories

$sqrts_min = 350;  $sqrts_max = 800;  $sqrts_step = 50;

# These files will be preserved
@outfiles = ("whizard.out", "whizard.grb", "whizard.evt");

# This will become the table of results
$resultfile = "whizard.out";
$table = sprintf(" %-10s  %-20s  %-20s\n", "sqrts [GeV]", 
                 "Total cross section [fb]", "Integration error [fb]");

# Main loop
for ($sqrts=$sqrts_min; $sqrts<=$sqrts_max; $sqrts=$sqrts+$sqrts_step) {
  
    # Read the input file template, modify the sqrts entry, and write the 
    # new input file
    $infile = "whizard.in.0";  $outfile = "whizard.in";
    open(INFILE, $infile) or die "Can't open $infile for reading";
    open(OUTFILE, ">$outfile") or die "Can't open $outfile for writing";
    while (<INFILE>) {
        s/sqrts = \d+/sqrts = $sqrts/;
        print OUTFILE;
    }
    close(OUTFILE);
    close(INFILE);

    # Run WHIZARD
    print "\n**** sqrts = $sqrts ****\n";
    $stat = system("./whizard");
    $stat==0 or die "WHIZARD run exited abnormally";
    
    # Extract the integration results and add them to the table
    open(RESULTS, "grep 'integral =' $resultfile|")
        or die "Grep failed on $resultfile";
    while(<RESULTS>) { ($dummy, $equals, $integral) = split; }
    close(RESULTS);
    open(RESULTS, "grep 'error    =' $resultfile|")
        or die "Grep failed on $resultfile";
    while(<RESULTS>) { ($dummy, $equals, $error) = split; }
    close(RESULTS);
    $table .= sprintf(" %10g  %20g  %20g\n", 
                      $sqrts, $integral, $error);

    # Create a subdirectory and move the results there
    $dirname = "sqrts=$sqrts";
    mkdir($dirname, oct("0755")) or die "Can't mkdir $dirname";
    foreach $file(@outfiles) {
        system("mv", $file, $dirname);
    }

}  

print "\n*** Results:\n";
print $table;

print "*** End of WHIZARD runs\n";
exit 0;

The script assumes that the subdirectories do not yet exist, and that you have a template whizard.in.0 for the input file which contains an entry like sqrts = 0. The table is shown on screen after all runs have been completed.

This kind of task can be formalized further. The program whizwrap takes care of loops over parameters and provides a convenient management of output files. It is called from the command line; the kind of task to be performed by WHIZARD is specified in terms of command-line options. Note that within WHIZARD 2 these tasks can be easily steered from the input file by means of its powerful interpreter language SINDARIN. Cf. the manual of WHIZARD 2 for the details.

5.8  WHIZARD as a subroutine library

A more ambitious enterprise is the integration of WHIZARD into a larger framework which may also account for the subsequent steps of experimental simulation and analysis. Even if this is not desired, it may be useful to access WHIZARD by subroutine calls in wrapper programs written in Fortran, C, or other program languages.

To make this possible, WHIZARD is organized as a set of libraries (in the UNIX sense). When the configuration file (the process list) has been specified, instead of making a stand-alone executable by

make prg

one may generate the libraries only

make libs

which then can be linked with a user-supplied main program.

To facilitate this procedure, after executing make libs, the user will find the generated libraries in the subdirectory lib (except for system-supplied libraries from the CERNLIB), together with a script called whizard.ld in the bin subdirectory which links an arbitrary program with the libraries in correct order. Thus, the user may write his own Fortran 95 main program, call it, e.g., my_main.f90, compile it⁵ and execute

bin/whizard.ld -o my_whizard my_main.o

to generate his own version of WHIZARD.⁶

Here is a main program that performs the same task as the script shown in the preceding section:

! Loop over the total energy, write a table and store the results in
! separate output files.
program my_main

  use kinds       ! Defines the default real kind
  use whizard     ! Defines the WHIZARD subroutines
  implicit none   ! No implicit typing

  ! Parameters
  integer, parameter :: n_values = 10
  real(kind=default), parameter :: sqrts_min = 350, sqrts_max = 800

  ! Variables
  integer :: i
  character(len=14+3) :: filename

  ! Tables
  real(kind=default), dimension(n_values) :: &
      sqrts, integral, error, chi2, efficiency

  ! Main loop
  do i = 1, n_values

     ! Calculate energy, set filename accordingly
     sqrts(i) = ((i-1) * sqrts_max + (n_values-i) * sqrts_min) / (n_values-1)
     write(filename, "(A14,I3)")  "whizard_sqrts=", nint(sqrts(i))
     call whizard_set_filename (filename)

     ! Read input and override energy setting
     write (*,*)
     write (*, "(A,1x,I3,1x,A)") "*** sqrts =", nint(sqrts(i)), "***"
     call whizard_read_input
     whizard_input%sqrts = sqrts(i)

     ! Integrate, record result, generate events
     call whizard_integrate
     call whizard_get_results (i, integral(i), error(i), chi2(i), efficiency(i))
     call whizard_generate
     call whizard_end 

  end do

  ! Write table
  write (*,*)
  write (*, "(A)") "*** Results:"
  write (*, "(1x,A15,2x,A25,2x,A25)") "sqrts [GeV]", &
       "Total cross section [fb]", "Integration error [fb]"
  do i = 1, n_values
     write (*, "(1x,G15.5,2x,G25.5,2x,G25.5)") &
       sqrts(i), integral(i), error(i)
  end do
  write (*, "(A)") "*** End of WHIZARD runs"

end program my_main

The meaning of the individual subroutine calls is described in detail in the following section. Note that parameters in the input block can be accessed via the object whizard_input which is a module variable of the whizard module. More exactly, variables in the process_input, integration_input, simulation_input and diagnostics_input blocks can be accessed as above

whizard_input%sqrts = sqrts

while variables in the parameter_input block are accessed like

whizard_input%par%mh = 115are

and beam parameters are accessed like

whizard_input%beam(1)%PDF_nset = 43

5.9  Interface of the WHIZARD module

All subroutines described in this section are accessed via the module whizard, which is included in the library whizard.a. Therefore, a main program written in Fortran 95 has to contain the line

  use whizard

in the preamble.⁷ For other programming languages, the subroutines are accessed by a platform-dependent prefix. For instance, the NAG compiler preprends the string whizard_MP_ to all names in the module whizard, so the subroutine whizard_integrate is accessed by the name whizard_MP_whizard_integrate. The access to arguments in subroutine calls may also be platform-dependent.

5.9.1  Auxiliary routines

The following routines are needed only if some non-standard behavior of WHIZARD is intended. They closely match the command-line arguments described in Sec. 5.6.

whizard_set_directory

: Set a working directory different from the default one.

Invocation:

call whizard_set_directory (directory)

Arguments:

directory

: string (input parameter)

whizard_set_filename

: Set a specific filename to be used instead of whizard for all input and output files during this run. This is one way of avoiding files getting overwritten. If an input file with specific filename does not exist, the generic input files whizard.in etc. will be used. The output files will always get the specific name, if any.

Invocation:

call whizard_set_filename (filename)

Arguments:

filename

: string (input parameter)

whizard_set_input_file

: Define the name of the main input file. Needed only if it is different from the default. The extension .in will be appended.

Invocation:

call whizard_set_input_file (filename)

Arguments:

filename

: string (input parameter)

whizard_enable_input_file

: Enable/disable reading of the main input file. If the logical argument is .false., calls to whizard_read_input will not read any file. This makes sense if parameters are to be set by different means, in particular by calls to whizard_set_input.

Invocation:

call whizard_enable_input_file (flag)

Arguments:

flag

: logical true or false (input parameter)

whizard_set_input

: Set input parameters directly. Each argument is a string of parameter = value definitions in namelist format, as it would appear in the input file enclosed by “&xxx” and “/”, where xxx is the respective input block (e.g., process_input). The strings can be empty.

Invocation:

call whizard_set_input (process_input, integration_input,
simulation_input, diagnostics_input, parameter_input,
beam_input1, beam_input2)

Arguments:

process_input

: namelist string (input parameter)

integration_input

: namelist string (input parameter)

simulation_input

: namelist string (input parameter)

diagnostics_input

: namelist string (input parameter)

parameter_input

: namelist string (input parameter)

beam_input1

: namelist string (input parameter)

beam_input2

: namelist string (input parameter)

5.9.2  Standard routines

The following routines are called, directly or indirectly, in standard WHIZARD runs. The explicit interface allows for calling them directly without reference to the main program.

whizard_read_input

: Start a new instance of WHIZARD and terminate any pending run. Read the input file. If no specific filename has been given, this is called whizard.in. If this routine is not called explicitly, it will be called automatically by whizard_integrate. Calling it explicitly allows for overriding input data before they are fed into the integration pass.

Invocation:

call whizard_read_input

Arguments:

none.

whizard_number_processes

: Return the number of processes specified in the input file.

Invocation:

call whizard_number_processes (number)

Arguments:

number

: integer (output parameter)

whizard_integrate

: Calculate the total cross section resp. read pre-generated results from file, as required by the input data. This routine reads the model definition file whizard.mdl, the cut configuration file whizard.cut1 and, if required, it reads/writes the grids whizard.grc and whizard.grb. If this routine is not called explicitly, it will be called automatically by the event generation routines.

Invocation:

call whizard_integrate

Arguments:

none.

whizard_get_results

: Read out the current integration results for the indicated process: Total cross section, error estimate, χ² value, reweighting efficiency estimate. If process_index is zero, return the total integral, error and efficiency average for all activated processes. The χ² value is zero in that case. If no integration has yet been performed, all returned numbers are zero. If the process_index is larger than the number of processes (as returned by whizard_number_processes), an error is issued.

The returned numbers are of the default real kind, which is double precision under normal circumstances. The default real kind may be accessed in a Fortran 95 program by a use kinds directive and including the kinds-src subdirectory in the module search path.

Invocation:

call whizard_get_results (process_index, integral, error, chi2, efficiency)

Arguments:

process_index

: integer (input variable)

integral

: default real (output variable)

error

: default real (output variable)

chi2

: default real (output variable)

efficiency

: default real (output variable)

whizard_generate

: Generate a fixed number of events resp. read them from file, as specified by the input data. This routine reads the cut/analysis configuration whizard.cut5 and reads/writes the event file(s) whizard.evt and whizard.evx, if required. Furthermore, it can write histogram data in whizard-plots.dat. This subroutine is equivalent to consecutive calls of whizard_events_begin, whizard_event (multiple times), and whizard_events_end (see below).

Invocation:

call whizard_generate

Arguments:

none

whizard_events_begin

: Prepare event generation for a n_calls events. It is assumed that the main program contains a loop with exactly n_calls to whizard_event. However, in case of unweighted events generation, this number is used only for opening a STDHEP format file, so the value is irrelevant if that option is not used.

Invocation:

call whizard_events_begin (n_calls)

Arguments:

n_calls

: 64-bit integer (input parameter)

whizard_event

: Generate a single (unweighted) event. The event is stored in the standard HEPEVT COMMON block. If requested by the input data, it is also written to file. To make use of the current event, the caller routine must be able to access HEPEVT.

Invocation:

call whizard_event

Arguments:

none

whizard_events_end

: Finalize event generation.

Invocation:

call whizard_events_end

Arguments:

none

whizard_end

: Finalize the WHIZARD run. Close all open files and deallocate memory.

Invocation:

call whizard_end

Arguments:

none

1

Similar restrictions are possible with the official CompHEP and MadGraph versions, but not supported by WHIZARD.

2

The latter are necessary because the procedure makes sense only if input data have changed, so the checksum will be invalid. This makes the user responsible for guaranteeing compatible input data in both runs; otherwise the program may crash.

3

NLO matrix elements are nontrivial, and WHIZARD may be not the appropriate program for dealing with them. An obvious limitation is that the user weight function does not take polarization into account. However, several NLO matrix elements have been interfaced to WHIZARD, and in WHIZARD 2 an automatic generation of NLO infrastructure is being prepared.

4

If used with recalculate=T, this limitation is absent, since reweighting will be applied after adaptation and event rejection. In that case, an unmodified event sample has to be generated first. Furthermore, in [19, 20] integrations of indefinite function have been performed with WHIZARD.

5

To access the module information, several subdirectories must be included in the module search path of the compiler. Inspect the setting of F90INCL in the Makefile in the whizard-src subdirectory.

6

This works for a Fortran 95 main program; the analogous steps for a main program in a different language will require modifications of whizard.ld, in particular the additional linking of the compiler-specific Fortran 95 runtime library.

7

See footnote 5.8.