Installation Instructions for GANDR-4.4, 28 November 2011
Below are detailed instructions for updating and compiling
the GANDR source codes, installing the GANDR master PENDF
and EXFOR libraries on the user's platform, initializing the
dedicated disk space file space used in updating the GANDR
parameter and covariance library, and running a suite of 17
test problems.
For additional information, please visit the GANDR web site,
www-nds.iaea.org/gandr. Included there is a six-volume report
describing the goals and methodology of the GANDR system, as
well as the actual files required for local installation of
the system. Also see materials referenced in the Bibliography,
Section L. below.
A. General preparations
------------------------
The following comments assume that the user has access to a
modern computer in the desktop workstation class or better,
running under the Unix operating system, preferably in the
Bash shell.
Installation of the full system and the associated library of
data files requires about 2.2 gigabytes of free disk space in
the /usr directory. The system is designed to read and write
repeatedly to and from data directories "usr2" and "usr3",
which should ideally reside on separate disk drives, each
having at least 36 gigabytes of free space.
In previous GANDR releases the "usr2" and "usr3" disk drives
were mounted at /usr2 and /usr3, respectively. In the present
Fedora 13 implementation of GANDR, the corresponding drives
are mounted at /media/_usr2 and /media/_usr3, respectively.
Everything required to install GANDR-4.4 and to run the GANDR
suite of 17 test problems is contained in the following four
files:
g44readme.txt (this file, 47280 bytes)
g44install77.tar.gz (? bytes)
g44installgf.tar.gz (5939330 bytes)
g4files.tar.gz (280284201 bytes)
The file g44install77.tar.gz installs GANDR-4.4 using the g77
Fortran-77 compiler, whereas the file g44installgf.tar.gz
installs GANDR-4.4 using the gfortran compiler. Only one of
these two files is required for a full installation of GANDR.
Note: The codes have been extensively modified in going from
GANDR-4.0 to GANDR-4.4. However, the data files containing
EXFOR and processed ENDF information have not changed. If the
g4files.tar.gz file have already been installed on the local
computer system, it is not necessary to repeat this step in
moving up to the GANDR-4.4 versions of the programs.
The three needed files can be accessed from CD or other media,
or directly from the GANDR web site. Downloading the large file
g4files.tar.gz via the web requires from 10 minutes to over an
hour, depending on the speed of connection.
B. Instructions for Installing, Updating and Compilation
---------------------------------------------------------
To begin the installation, choose an arbitrary working lo-
cation on the /usr directory as the GANDR working directory,
for example, /usr/work.
This directory will be the primary location for executable
programs and will be the home directory for associated
directories such as /usr/work/upd (for code updating
operations), /usr/work/loop (for management of the GANDR
energy grid structures), /usr/work/sens (for multigroup
sensitivity calculations), /usr/work/0022.9010.74000 (a
typical program-generated directory name for storing code
output).
Place a copy of the desired installer in the working
directory, and unpack it using the usual Unix commands.
gunzip g44install77.tar.gz
tar xvf g44install77.tar
OR
gunzip g44installgf.tar.gz
tar xvf g44installgf.tar
Next, change directory to /usr/work/upd. After the tar
command has been executed, this directory should contain
the following files:
gup.f
initialize.k
src
up0
up1
upalt2008
upalt2010
update.k
Next, compile the update program GUP
g77 -o gup.x gup.f
OR
gfortran -o gup.x gup.f
and then execute GUP in the "compile all" mode.
cp up0 upn
gup.x
The only instruction contained in the up0 file, "*cpl all",
generates the current versions of the following files, complete
with line references in columns 74-84 that can be used in
performing additional updates.
abuns.f
cfix.f
cleanx.f
doall.f
dometa.f
gabrow.f
gacont.f
gapost.f
gaprep.f
gaupd.f
gup.f
inita.f
initc2.f
initc3.f
meta.f
modcov.f
names.f
pert32.f
semove.f
sepost.f
seprep.f
spel.f
spex.f
zottvl.f
Most, but not all, of these generated files are Fortran files.
(GUP adds the ".f" suffix to all files extracted from src
regardless of file content.)
These files will evolve over time in response to future code updates.
For this reason, it may be useful to make "permanent" copies of the
more important of these "originals". For example,
cp gaprep.f gaprep.f.orig
cp zottvl.f zottvl.f.orig
cp gapost.f gapost.f.orig
At this time, implement any desired custom features or make needed
corrections by editing the file up1 with a text editor, and then
re-running GUP. For example, one can create a unique name for the
local version of GANDR.
cp up1 upn
gup.x
For help with the updating procedure, type
more /usr/work/upd/gup.f
C. Compiling the GANDR Codes
-----------------------------
Once the up1 file is finalized in the manner described above,
the supplied Unix script update.k can be invoked to update and
compile the entire suite of GANDR programs.
cp up1 upn
update.k
Note that, within this script, vi is invoked to generate the
necessary "reversed" version of zottvl.f (named zottvl_rev.f)
prior to actual compilation. One can safely ignore the Vim
warning message that is issued, i.e.,
Vim: Warning: Input is not from a terminal
The resulting executable files are written to directory /usr/work.
D. Installation of the GANDR PENDF and EXFOR Data Directories
--------------------------------------------------------------
As mentioned above, if the GANDR-4.0 data files contained on
g4files.tar.gz were NOT installed as part of an earlier GANDR
installation process, it it necessary to install these files at
this time. This process requires approximately 2.2 gigabytes
of free space on the /usr directory. To proceed now with this
stage of the implementation, first create the data directory
/usr/gandr/zott.
Next copy or download the large file g4files.tar.gz to directory
/usr/gandr/zott and change to that directory. Unpack the tar
file with the command
gunzip g4files.tar.gz
The un-zipped tar file occupies slightly over 1081 megabytes.
Now enter
tar xvf g4files.tar
Type "du" to check that the necessary PENDF and EXFOR data
structures have been created and filled. The result should be
very similar to what is listed in the final section below. If
all is well, g4files.tar can now be deleted. A second "du"
command should show 1148700 kilobytes of disk usage in the
/usr/gandr/zott directory.
E. Initializing the Cross Material Covariance Arrays
-----------------------------------------------------
As discussed above, executing the GANDR system in full generality
requires 36 gigabytes of free space on both the _usr2 and _usr3
directories. This is where the the 91000-element GANDR parameter
array and its full covariance matrix reside. The covariance
matrix is stored in triangular form, so it contains 91000x91001/2
elements. Two copies are retained at any given time, with the
pre-update parameters and covariances on _usr2 and the post-
update parameters and covariances on _usr3, or vice versa. (The
roles of _usr2 and _usr3 are continually interchanged as the
evaluation process moves forward.)
It it is recommended to initialize the GANDR data directories for
the fully general, multi-material, case at this time. This will
permit, for example, the execution of the multi-material test
problem based on the incorporation of the International Neutron
Cross Section Standards Evaluation. The assumed non-informative
prior consists of unit variances (100 percent uncertainty) on
the diagonal, a fully correlated covariance component with
magnitude 1.E-12 (0.0001 percent uncertainty) within a given
material, and a covariance magnitude 1.E-16 (0.000001 percent)
correlated between the parameters of different materials.
To see what is meant by a "material" in the GANDR context, ex-
amine the file names.f, which assigns specific names to the 130
allowed GANDR materials. This list of materials can be modified,
at the beginning of a series of GANDR analyses, by adding an
appropriate update "ident" to upn, as discussed above.
The material-naming convention (which controls the file-naming
convention) is a simple 5-character coding of Z and A. For
example, H-1 is material 01001 and H-2 is 01002. Files with
names beginning "00" and "99" are included in this official list
(and therefore are included in the GANDR data structures) to
accommodate the extension of data assessment studies to include
fission products and resonance absorbers, respectively, as
discussed in Volume 1 of the GANDR report.
It is the purpose of the included programs INITC2 and INITC3
to initialize all same-material and cross-material covariance
matrices, using the assumed non-informative prior, and to write
these covariances on a set of files having standardized GANDR
names.
To begin this process, change directory to /usr/work/upd. Then
compile the two codes as follows:
g77 -o initc2.x initc2.f
g77 -o initc3.x initc3.f
OR
gfortran -o initc2.x initc2.f
gfortran -o initc3.x initc3.f
Also at this time, make a second copy of the names file,
cp names.f names
Next, create directories /media/_usr2/gandr/caa and
/media/_usr3/gandr/caa. If they already exist, it is good
practice to delete the entire contents /media/_usr2/gandr/caa
and /media/_usr3/gandr/caa before continuing.
We can now perform the (approximately 90-minute) process of
creating non-informative priors for all combinations of
materials and writing them on the two caa directories. To
do so now, type
initialize.k
Progress in the initialization can be monitored on the console
and by viewing the contents of the text file, initialize.out.
Using the GANDR standard material names, the cross-material
covariances between H-1 and H-2, for example, will be written
on the file 01001.01002 in the two mentioned data directories.
If the directories _usr2 and _usr3 have been partitioned to
contain 36 gigabytes, then, at this point, they should be 98%
to 99% full. This high utilization of the available space
does not prevent the efficient operation of the GANDR programs.
F. Executing the First GANDR Test Problem, GANDRLOOP
-----------------------------------------------------
In the GANDR code system, there is a fixed limit of 700
parameters per material. Even though the GANDR coding logic
suppresses zero parameter values, for some materials it is
necessary to modify the otherwise fixed energy grid used
for the parameter evaluation process in order to satisfy the
700-parameter limit.
Test problem GANDRLOOP shows how the grid structures can be
defined, in a single job submission, to satisfy this limit
for all 103 of the GANDR general purpose materials. To
execute this test problem, change directory to /usr/work/loop
and type
gandrloop > gandrloop.screen (Test Problem 1)
Re-direction of the normal output from the computer screen
to a file allows one to do a comparison of the output with
that generated on the GANDR host computer. For example,
diff gandrloop.screen.orig gandrloop.screen | more
Note that we have chosen to use the same grid structure for
all isotopes of a given element, achieved by imposing the
coarsest required structure on all isotopes of that element
(compare B-10 and B-11). This facilitates the analysis of
elemental data, as discussed below in Section I. The in-
cluded file nalst is an extract of the GAPREP output file
from this run, gaprep.out.orig. This file contains the
important "bottom line" regarding the actual number of GANDR
parameters for each material at the conclusion of this run.
Once the GANDRLOOP script has been run, the default value of
GAPREP input parameter igrid (igrid=0) can be used in all
subsequent job submissions. In this case, the energy grids
and reaction definitions set by GANDRLOOP will be employed
from then on.
To learn how to modify the script to achieve a different set
of material-dependent grids and reaction definitions, consult
the GAPREP input instructions. These input instructions can
be viewed conveniently with the following UNIX command:
more /usr/work/upd/src
A short introductory GANDR test problem is discussed in
considerable detail in Volume 3 of the GANDR documentation
(see Section L below.) This test problem demonstrates a
simple case of updating an evaluation of the total cross
section of W-184 using information from three differential
measurements of the total stored in the EXFOR experimental
nuclear reaction data database. To run this problem, change
directory to /usr/work, and then type
ganvol3 > ganvol3.screen (Test Problem 1a)
A detailed explanation of the graphics and printed output is
contained in Volume 3, so will not be repeated here. Again,
a diff command can be used to highlight any differences that
have shown up in the local implementation.
G. Executing the Five GANDR Test Problems from Volume 4
--------------------------------------------------------
A set of five semi-realistic test problems is introduced and
discussed in Vol. 4 of the GANDR report. In this section,
we indicate how these problems can be run in the current
GANDR environment. In addition, eight test problems, based
on incorporating information from the International Neutron
Cross Section Standards Evaluation, are described below in
Section H.
To execute the test problems from Vol. 4, change directory to
/usr/work. In that directory, there are the following five
shell scripts:
gandrab > gandrab.screen (Test Problem 2)
gandrseq > gandrseq.screen (Test Problem 3)
gandrnon > gandrnon.screen (Test Problem 4)
gandrfiss > gandrfiss.screen (Test Problem 5)
gandrbig > gandrbig.screen (Test Problem 6)
Execution of these test problems normally requires, in
/usr/work, in addition to the codes discussed above, an
executable copy of NJOY99.336, or its equivalent, with the
file name "xnjoy". Only the PLOTR and VIEWR modules are
invoked in these five problems. (Note, however, that
Test Problems 13 and 17 below invoke the MODER and GROUPR
modules of NJOY.)
Users without access to NJOY must supply their own
graphics processor (and to supply a small interface code
to translate the print commands apppearing on the file
plotr.in into appropriate input for the local plotting
code. Alternatively, one can set iprint equal to -1 or
-3 in all input to the GAPREP code to suppress graphics
output and thereby eliminate references to NJOY/PLOTR.
This will not affect the desired updating of the GANDR
library.
If one has not employed output re-direction as shown above,
one can get a very short comparison of the local "printed"
output to the computer screen with that on the GANDR host
by typing
tail -23 gandrab.screen.orig
To resubmit one of the scripts with modifications, check the
GAPREP input instructions and make appropriate changes to the
script. It is safe to experiment with different combinations
of target and reaction data. At this stage, input errors are
unlikely to cause any permanent harm, such as loss of one's
investment in the time consuming initialization process,
because these five test problems affect only the within-
material parameters and covariances. These quantities can be
re-initialized to standard initial values by GAPREP with
inopt=0 at any time.
For further discussion of the relevance and interpretation of
the output from the five test problems, see Vol. 4.
H. Incorporating the International Standards Evaluation
--------------------------------------------------------
This distribution also provides support for an important,
additional set of eight test problems. These problems
demonstrate how to
(a) initiate a global data assessment by analyzing EXFOR
data for the total cross sections for six materials, namely
Li-6, B-10, Au-197, U-235, U-238 and Pu-239, and
(b) continue the assessment with the incorporation of recent
data from the International Standards Evaluation for the same
set of six materials.
To perform the total cross section portion of this analysis,
change directory to /usr/work and execute the following scripts
in succession:
gandrli6 (Test Problem 7)
gandrb10 (Test Problem 8)
gandrau (Test Problem 9)
gandru5 (Test Problem 10)
gandru8 (Test Problem 11)
gandrpu9 (Test Problem 12)
These are relatively time-consuming runs because of the large
volume of total cross section data available for these important
materials. The longest, gandrpu9, required around 90 minutes
of execution time on a Dell Precision 530 workstation with a Xeon
processor.
Note: Since these 6 test problems consume a total of around
4.5 hours of computer time, the user may wish to automate these
submissions. To all 6 of these problems, with plots suppressed
in order to minimize required user interaction, simply type:
prestan.k
With appropriate changes to the above six scripts, one can ex-
tend this analysis of the EXFOR total cross section data to the
remaining 97 general purpose materials in the GANDR library.
For the next test problem, change directory to /usr/work/sens
and execute the script
seprepall.k (Test Problem 13)
Ordinarily, the execution of Test Problem 13 requires the
presence of an executable version of NJOY (containing the
GROUPR and MODER modules) to be present in the working
directory. If NJOY is not available, Test Problem 13 can
be skipped and the user can proceed to Test Problem 14,
provided only that the files mgbse.dat.orig and sepost.in.dat
are present in the working directory. If these two data
files are present, then change directory to /usr/work, and
type the following commands:
cp mgbse.dat.orig mgbse.dat
cp sepost.in.orig sepost.in
Then proceed to Test Problem 14 below.
The script seprepall.k invokes the recently developed programs
SEPREP and SEPOST four times in order to calculate the
sensitivity of the group averaged data for the heavy element
standards (as presented in the published standards evaluation)
to the point parameters of the GANDR library. To understand
the input files, such as seprep.in.03006, consult the SEPREP
input instructions,
more /usr/work/upd/seprep.f
As presently coded, it is assumed that the same multigroup
structure is used for all materials presented as a mixture of
point-energy and group averaged data, as is the case in the
International Neutron Cross Section Standards Evaluation.
Note that the script for Test Problem 13 invokes the MODER
and GROUPR modules of NJOY.
The next test problem is the actual update of the GANDR library
to incorporate the information from the standards evaluation.
A detailed discussion of this calculation can be found in a
paper in Nuclear Data Sheets, Volume 109. See the Bibliography
below for details. To proceed, change directory to /usr/work
and type
gandrst (Test Problem 14)
This script invokes gaprep in the recently developed multi-
material mode (see subroutine MULTI) and then invokes the GANDR
statistical engine ZOTTVL to do a library update to incorporate
the information from the entire standards evaluation in a
single execution. This script treats the entire published
standards evaluation as a single correlated measurement of 926
data quantities.
I. User-supplied Initial Covariances for Multi-Isotope Elements
--------------------------------------------------------------
For some time GANDR has had the capability of accepting a user-
supplied covariance matrix for the 700 parameters of a single
material. Begining with GANDR-4.3, it is possible for the user to
specify the initial covariances for a range of materials, including
all cross-material covariances. These covariances are written by
the user to the directory /usr/gandr/zott/modcov in 10E13.6 ASCII
format, prior to the actual GANDR processing.
The small code modcov.f is supplied with this distribution as an
aid in this initialization process. It must be modified by the user
to access the covariance information that is to be written into the
modcov directory. If no such changes are made, the code writes the
standard non-informative prior to the current range of materials.
The file naming convention in this directory is the same as employed
in the GANDR (binary) library in the directory _usr2/gandr/caa.
Namely, a file with a name like 74182 is the self-covariance for
the 700 paramters of W-182, while 74182.74183 is the cross-material
parameter covariances between W-182 and W-183.
In all cases, the covariances are written to /usr/gandr/zott/modcov
as square 700-by-700 matrices. If na is less than 700, as it very
often is, then it is good practice to set the undefined covariances
to zero.
To see an example of how the modcov.f code interacts with the other
GANDR codes, change directory to /usr/work and type
gandrmoc (Test Problem 15)
In this test problem, note that modcov.x is executed twice. The
first use of modcov.x is to set the ASCII "user-suppled" covar-
iances in /usr/gandr/zott/modcov equal to the standard GANDR non-
informative prior. As can be seen by the results of the first
"diff" command, the adjusted data on gapost.out are identical to
all printed digits to the results of using the built-in non-
informative covariance intialization.
The second invocation of modcov.x produces an initial set of
covariances which contain, as a test, a single, non-standard
off-diagonal element, equal to 0.999999, in position cov2(363,363)
for each of the "user-supplied" cross-material covariance matrices.
This strong correlation drives the solution in a way such that
the final adjusted value of parameter 363 for each of the four
materials are all close to the same value (0.93457), which is the
expected result. This result can be verified by inspection of the
results of the final "diff" command. Nearby parameters are also
affected to a minor extent by this artificial modification,
because of cross parameter correlations induced by correlations
in the experimental data.
J. Analysis of EXFOR data for isotopic and elemental tungsten
--------------------------------------------------------------
Beginning with version GANDR-4.0, it is possible to analyze data
for elements and mixtures. Also incorporated in this version
is a general cleanup of the user interface. For example, all
user input is now read from standard input unit, eliminating
the previous use of the file names.gaprep as an auxiliary source
of user input instructions. Another change is that the user
explicitly controls the mode of calculation by means of the input
parameter imode. The value imode=12, for example, selects the
"classic" mode, as employed in the first 12 test problems.
The value imode=10 selects the "standards" mode, as illustrated
in Test Problem 14. Finally, imode=11 selects the "mixture"
mode, wherein data for an element or chemical mixture can be
analyzed. Test Problem 15 illustrates the use of this new
feature. To execute this problem, type
gandrmix (Test Problem 16)
To clarify the differing input requirements for the three
calculational modes, a summary is included at the end of the
gaprep input instructions. To view just this summary, type
head -787 /usr/work/upd/src
Note on Card 7 in the full input instructions that, when
processing data for an element or mixture, it is necessary to
supply externally an appropriate PENDF file for the desired
element or mixture. For the common engineering elements
having more than one isotope in the GANDR general purpose
library, this preparatory work has already been done. The
Master PENDF file supplied with this distribution includes
elemental PENDFs, produced with the MIXR module of NJOY for
H, Li, N, Si, Ar, Ti, Cr, Fe, Ni, Cu, W, Rh, Hg, Pb, and U,
with names of the form /usr/gandr/zott/pendf/82000.
K. Sensitivity of General Group Data to the GANDR Parameters
-------------------------------------------------------------
SEMOVE is a Fortran-77 computer program for computing the
derivatives of processed multigroup nuclear data with respect
to the individual parameters of the GANDR library.
Development of the SEMOVE program was made possible by the
financial support from the OECD Nuclear Energy Agency. This
support was offered in order to facilitate the incorporation
of information from neutron benchmark integral experiments
into the GANDR evaluation process.
The basic premise of the SEMOVE program is that a user has
chosen a multigroup processing program such as NJOY and has
developed a script that allows the calculation of all nuclear
data of interest in his selected group structure and with
his selected weight function. The file njg included in
this distribution is an example of such a user-supplied script.
The SEMOVE program is currently restricted to cases where the
final calculation is performed with the GROUPR module, with the
results being written to a GROUPR format (GENDF) output file,
as is the case with the script njg. There are no other
restrictions placed on the type or complexity of the processed data.
A distinctive feature of the GANDR project is that the fundamental
data uncertainties are assumed to reside, not in the ENDF evaluations,
but in the parameters of the GANDR library. To interface the GANDR
library with normal sensitivity tools, and thereby to permit the GANDR
library to be improved by exploiting the information content of
accurate integral data, we need to be able to calculate the changes
in the multigroup cross sections dW that result from changes in the
GANDR parameters dA,
dW = R dA (1)
From R, one can directly obtain the multigroup cross section covariance
matrix,
D(W) = R D(A) R* (2)
Here we see that a system for determining the sensitivity matrix R plays
a role within the GANDR project that is similar to an ENDF covariance
processing program.
The estimation of R is not trivial, because the working cross section
set W does not have a simple one-to-one correspondence to the GANDR
parameters A. The reasons for this are as follows: First, the GANDR
parameters define uncertainty components that are continuous functions
of energy, defined in relation to a universal set of energy nodes. The
multigroup data, on the other hand, are energy-averaged quantities,
with the averages being performed over a group structure of the user's
choice.
Also, in the case of angular distributions, the GANDR parameters refer
to Legendre moments in the CM system (for the same reasons that CM
system is used for elastic and inelastic scattering in the ENDF formats),
while the Legendre moments of group data refer to angular distributions
in the LAB system. Finally, group cross sections for resonance materials
are self-shielded through a complex procedure that typically involves
the use of both NJOY and TRANSX.
The availability of D(W) in factored form, as it appears above in Eq.
(2), means that one can directly relate the uncertainties in the
parameters of the GANDR library to the uncertainty of calculated
quantities like Z,
D(Z) = S D(W) S* = S R D(A) R* S*, or
D(Z) = T D(A) T*,
where T = S R is the "complete" sensitivity matrix relating the change
in a GANDR parameter to changes in a set of integral data. An important
consequence of having available the complete sensitivity matrix, is that
we can directly update the GANDR library, based on the results of
integral experiments, using the usual method of generalized least
squares, as codified, for example, in the GANDR program ZOTTVL.
In considering the practicality of the discrete-difference approach to
the calculation of R, it is important to observe that the discrete-
difference calculation does NOT have to be repeated for every new
integral experiment. For the large class of problems for which
resonance self-shielding is relatively unimportant, once the parameter
sensitivities R are calculated for a single nuclide, those sensitivities
can be re-used for any variation in geometry, material composition or
source description. The neutronic multigroup sensitivities S will
change, of course, in the usual way. The conversion of each of the
new S matrices (S1, S2, S3, etc.) back to the complete sensitivity T
would require only a new matrix multiplication with the same R matrix,
T1 = S1 R
T2 = S2 R
T3 = S3 R
and so on.
We now turn to the details of the calculation of the parameter
sensitivity matrix R. As we have noted, the relationship between the
vectors A and W is complicated. These complications may make it
impractical to develop a procedure for calculating R analytically. For
this reason, in program SEMOVE the current approach to the calculation
of the matrix R is based on discrete differences.
The SEMOVE program assumes the existence of reference ENDF and PENDF
files, such as are normally supplied to the GROUPR module for multigroup
processing. The SEMOVE program performs a lengthy series of discrete
alterations. In each alteration, one of the (up to 700) GANDR parameters
for the material of interest is either increased or (in the case of odd-
numbered Legendre coefficients) decreased by 1%.
The GANDR parameters are defined such that perturbing one parameter
leads to an altered cross section table (in a PENDF file) or an altered
distribution (in an ENDF file), but not both. This might be called the
"exclusive or" property of the parameters. To keep track of the numerous
altered files, a specific data directory structure is employed. For
example, the file stored at the location 03007/endf/136 is the ENDF file
produced as the result of perturbing GANDR parameter number 136 for Li-7.
(This particular parameter, in turn, controls the P-1 component of
elastic scattering in the vicinity of 10 MeV.) By comparison, the file
03007/pendf/440 is the PENDF file resulting from perturbing GANDR
parameter number 440 for Li-7, which controls the file-3 cross section
for mt=102 in the vicinity of 1 eV. Because of the "exclusive or"
property, files 03007/pendf/136 and 03007/endf/440 would be identical
to the unperturbed PENDF and ENDF files so are not produced explicitly.
The SEMOVE code can, with a single execution, populate the reserved
data directories with the perturbed ENDF and PENDF files required to
calculate all parameter derivatives for a selected nuclide. An
illustration of how this works in practice is provided below in the
"Sample Problem" section.
The coding logic of SEMOVE builds on the capability of the MODER
module of NJOY to read an ENDF, PENDF or GENDF (group-ENDF) file in
either ASCII or NJOY blocked-binary format and then to write out a copy
of the same information in either ASCII or blocked-binary format. In
this process every significant nuclear datum passes through the core of
the computer. This controlled flow of all data through the core is
exploited n the SEMOVE program in applying the desired perturbations.
The heart of SEMOVE is 1400 lines of new programming that implements the
desired perturbation function. This programming is overlaid on a code
base provided by the 99.259 version of MODER. The complete program also
includes a number of utility routines from the NJOY module, such as
LISTIO and MOREIO, which perform the actual data reads and writes.
Numerous modifications have been made to the included MODER and NJOY
subroutines, some major and some minor. It is intended that the entire
4600 lines of the resulting SEMOVE code will be updated and maintained
as a single unified program.
Although it does not pertain to the coding of SEMOVE, it is worth noting
that the main NJOY program, njoy99.336, required a minor update before
preparing the sample problem. The update, ident UPBRO, changes the grid
thinning logic in BROADR so as to preserve the fixed GANDR energy grid
under Doppler broadening. The Master PENDF library supplied with this
distribution, was prepared using this update. The update instruction set
UPBRO is included with this distribution (in the semove directory).
Sample Problem
--------------
Included in this distribution is the complete input for a SEMOVE
sample problem in which the desired sensitivity matrix R is computed
for targets of Li-7 and C-nat. To execute this test problem, first
change directory to /usr/work/semove, and invoke the script for
the test problem.
sem2.k (Test Problem 17)
Note of caution: this is a large problem that runs to completion in
around 6 hours on a modern workstation. For a much faster-running
test, see Test Problem 16a below.
The script sem2.k performs two calls to the much larger script semove.k,
which contains the main input for this two-material sample problem.
In this run, the GROUPR module of NJOY is executed approximately 1000
times to provide the necessary input data for the requested derivative
calculations, performed by the SEMOVE program by discrete differences.
The output derivatives are written to a dedicated set of directories.
For convenience, tar files 03007.dgdif.tar and 06000.dgdif.tar,
containing all of the output derivatives for the two-material sample
problem, are also included as part of this distribution.
For a faster running demonstration of the semove program, one can
use the script sem387.k which is just the script semove.k, modified
by changing "/" to "1 387/" in line 37. Instead of entering sem2.k,
just type the following,
sem387.k 600 06000 (Test Problem 17a)
This shorter version computes the derivatives of the multigroup data for
C-nat to a single GANDR parameter, instead of all parameters. The
running time is reduced, correspondingly, from 6 hours to less than a
a minute.
After running either the long or short versions of Test Problem 16,
an explicit display of the definitions of all GANDR parameters for all
materials that have been processed can be found in the corresponding
"edit" files, 03007/edit.txt and/or 06000/edit.txt.
The main output of the SEMOVE program is the full matrix R introduced
above, or (in the shorter version) one column of that matrix. Recalling
Eq. (1),
dW = R dA
The desired matrix R is stored in the directory dgdif two levels below the
working directory. To illustrate the use of the matrix R in detail, con-
sider that one wishes to view the derivatives of the 175-group multigroup
data for the mt=51 reaction in C-nat with respect to GANDR parameter number
387. Put another way, we wish to view the 387-th column of the matrix R.
(From the file 06000/edit.txt, we see that parameter 387 controls the
P-1 component of inelastic scattering in the CM system in the vicinity of
7.94 MeV.) To view the derivatives, after running either version of the
sample problem within the directory /usr/work/semove, type the following:
cd 06000
cd dgdif
gunzip 387
more 387
Alternatively one can, without actually running the test problem, extract
the data from the relevant tar file as follows:
tar xvf 06000.dgdif.tar 387.gz
gunzip 387
more 387
For example, we may wish to examine the derivatives of all of the Legendre
components of the inelastic scattering cross section, in energy groups
144-156, for the first inelastic level. To accomplish this, perform the
following searches in "more":
/mf = 6
/mt = 51
This brings one to the material listed below. To interpret this output data,
first note that the multigroup transfer matrix for this reaction has, in
this case, 7 Legendre moments in the LAB system (mpol = 7). Next, note that,
at this energy and for this reaction, only infinitely dilute data are
present on the GENDF file (nsig = 1). These data, like all others in this
sample problem are Doppler broadened to 300 Kelvin.
==================================================================
mat = 600
mf = 6
mt = 51
xmult = -100.
mpol = 7
nsig = 1
temp = 300 grp = 144
Legendre order = 0
0.000000E+00
Legendre order = 1
0.000000E+00
Legendre order = 2
0.000000E+00
Legendre order = 3
0.000000E+00
Legendre order = 4
0.000000E+00
Legendre order = 5
0.000000E+00
Legendre order = 6
0.000000E+00
...
...
temp = 300 grp = 154
Legendre order = 0
-2.040000E-06
Legendre order = 1
-1.871248E-02
Legendre order = 2
-3.130690E-03
Legendre order = 3
-2.477200E-04
Legendre order = 4
-9.510000E-06
Legendre order = 5
-7.500000E-07
Legendre order = 6
2.000000E-07
temp = 300 grp = 155 <<<<<<<<<<<<
Legendre order = 0
2.110000E-06
Legendre order = 1
-2.839568E-02
Legendre order = 2
-4.573850E-03
Legendre order = 3
-3.497500E-04
Legendre order = 4
-1.312000E-05
Legendre order = 5
-2.000000E-08
Legendre order = 6
-3.400000E-07
temp = 300 grp = 156
Legendre order = 0
-1.900000E-07
Legendre order = 1
9.069290E-03
Legendre order = 2
1.407200E-03
Legendre order = 3
1.040500E-04
Legendre order = 4
3.900000E-06
Legendre order = 5
2.900000E-07
Legendre order = 6
-3.000000E-08
...
...
A particularly interesting group is group 155 of the Vitamin-J 175
group structure employed in this test problem. This group covers the
energy range range from 7.79 to 8.19 MeV, directly in the center of
the area that experiences the maximum impact of variations in GANDR
parameter 387. This group is highlighted with "left arrows" above.
First note that the P-0 component has nearly zero sensitivity. This
is the expected result since, by orthogonality, changing the P-1 CM
component of the inelastic scattering angular distribution cannot affect
the integrated cross section, because this cross section is the same
as the P-0 component in either the CM or LAB systems.
Also, as expected, changing one component in the CM has a noticeable
affect on all other components (other than P-0) in the LAB system.
Finally, note that the magnitude and sign of the several P-n
derivatives are fully consistent with expectations. At 7.94 MeV,
the cross section for this reaction is 0.35 barns, and the P-1
normalized coefficient in the CM is -0.119, the product being -0.0417
barns. This figure sets an upper limit to the derivatives of interest.
L. GANDR Bibliography
--------------------------------------------------------------
D.W. Muir, The Global Assessment of Nuclear Data Requirements
(GANDR Project). Volume 1. Project Overview. Available online
at www-nds.iaea.org/gandr/docs.html.
D.W. Muir, "Global Assessment of Nuclear Data Requirements
(GANDR Project) "Volume 2. The ZOTTVL Program. Available online.
EXFOR Systems Manual, Brookhaven National Laboratory Report
BNL-NCS-63330-04/01, compiled and edited by V. McLane, April 2001.
D.W. Muir, The Global Assessment of Nuclear Data Requirements
(GANDR Project). Volume 5. Preparation of the EXFOR Master Library
for the GANDR Project. Available online.
"International Evaluation of Neutron Cross Section Standards,"
STI/PUB/1291, International Atomic Energy Agency, Vienna, Austria,
November 2007. Available online at
www-pub.iaea.org/MTCD/publications/PDF/Pub1291_web.pdf.
D.W. Muir, The Global Assessment of Nuclear Data Requirements
(GANDR Project). Volume 6. Multigroup Sensitivities. Available
online.
D.W. Muir, A. Mengoni, and I. Kodeli, "Integration of the
International Standards Evaluation into a Global Data Assessment,"
Nuclear Data Sheets 109 (2008) 2874-2879. Available from Science
Direct online at doi:10.1016/j.nes.2008.11.032
M. Result of executing "du" command in directory /usr/gandr/zott
----------------------------------------------------------------
8280 ./list
1156 ./aref
512 ./thresh
468 ./cont
228 ./xexfor
13768 ./exfor
69228 ./modcov
8 ./xlinpl/01001
8 ./xlinpl/01002
8 ./xlinpl/02003
8 ./xlinpl/02004
20 ./xlinpl/03006
8 ./xlinpl/03007
8 ./xlinpl/04009
24 ./xlinpl/05010
8 ./xlinpl/05011
4 ./xlinpl/06000/6000
12 ./xlinpl/06000
8 ./xlinpl/07014
8 ./xlinpl/07015
8 ./xlinpl/08016
8 ./xlinpl/09019
8 ./xlinpl/11023
8 ./xlinpl/12024
8 ./xlinpl/13027
8 ./xlinpl/14028
8 ./xlinpl/14029
8 ./xlinpl/14030
8 ./xlinpl/15031
8 ./xlinpl/16032
8 ./xlinpl/17035
8 ./xlinpl/18040
8 ./xlinpl/19039
8 ./xlinpl/20040
8 ./xlinpl/21045
8 ./xlinpl/22046
8 ./xlinpl/22047
8 ./xlinpl/22048
8 ./xlinpl/22049
8 ./xlinpl/22050
8 ./xlinpl/23000
8 ./xlinpl/24050
8 ./xlinpl/24052
8 ./xlinpl/24053
8 ./xlinpl/24054
8 ./xlinpl/25055
8 ./xlinpl/26054
8 ./xlinpl/26056
8 ./xlinpl/26057
8 ./xlinpl/26058
8 ./xlinpl/27059
8 ./xlinpl/28058
8 ./xlinpl/28060
8 ./xlinpl/28061
8 ./xlinpl/28062
8 ./xlinpl/29063
8 ./xlinpl/29065
8 ./xlinpl/30000
8 ./xlinpl/33075
8 ./xlinpl/39089
8 ./xlinpl/41093
8 ./xlinpl/53127
8 ./xlinpl/55133
8 ./xlinpl/73181
12 ./xlinpl/74182
12 ./xlinpl/74183
12 ./xlinpl/74184
12 ./xlinpl/74186
16 ./xlinpl/79197
8 ./xlinpl/82206
8 ./xlinpl/82207
8 ./xlinpl/82208
8 ./xlinpl/83209
8 ./xlinpl/90229
8 ./xlinpl/90230
12 ./xlinpl/90232
8 ./xlinpl/91231
8 ./xlinpl/91233
8 ./xlinpl/92232
8 ./xlinpl/92233
8 ./xlinpl/92234
16 ./xlinpl/92235
8 ./xlinpl/92236
8 ./xlinpl/92237
20 ./xlinpl/92238
8 ./xlinpl/93236
8 ./xlinpl/93237
8 ./xlinpl/93238
8 ./xlinpl/93239
8 ./xlinpl/94236
8 ./xlinpl/94238
16 ./xlinpl/94239
8 ./xlinpl/94240
8 ./xlinpl/94241
8 ./xlinpl/94242
8 ./xlinpl/94244
8 ./xlinpl/95241
8 ./xlinpl/95242
8 ./xlinpl/9542M
8 ./xlinpl/95243
8 ./xlinpl/96242
8 ./xlinpl/96243
8 ./xlinpl/96244
8 ./xlinpl/96245
8 ./xlinpl/96246
8 ./xlinpl/96247
8 ./xlinpl/96248
8 ./xlinpl/97249
8 ./xlinpl/98250
8 ./xlinpl/98251
8 ./xlinpl/98252
4 ./xlinpl/03000
4 ./xlinpl/04000
4 ./xlinpl/05000
4 ./xlinpl/07000
4 ./xlinpl/08000
4 ./xlinpl/09000
4 ./xlinpl/10000
4 ./xlinpl/11000
4 ./xlinpl/12000
4 ./xlinpl/13000
4 ./xlinpl/14000
4 ./xlinpl/15000
4 ./xlinpl/16000
4 ./xlinpl/17000
4 ./xlinpl/18000
4 ./xlinpl/19000
4 ./xlinpl/20000
4 ./xlinpl/21000
4 ./xlinpl/22000
4 ./xlinpl/24000
4 ./xlinpl/25000
4 ./xlinpl/26000
4 ./xlinpl/27000
4 ./xlinpl/28000
4 ./xlinpl/29000
4 ./xlinpl/31000
4 ./xlinpl/32000
4 ./xlinpl/33000
4 ./xlinpl/34000
4 ./xlinpl/35000
4 ./xlinpl/36000
4 ./xlinpl/37000
4 ./xlinpl/38000
4 ./xlinpl/39000
4 ./xlinpl/40000
4 ./xlinpl/41000
4 ./xlinpl/42000
4 ./xlinpl/43000
4 ./xlinpl/44000
4 ./xlinpl/45000
4 ./xlinpl/46000
4 ./xlinpl/47000
4 ./xlinpl/48000
4 ./xlinpl/49000
4 ./xlinpl/50000
4 ./xlinpl/51000
4 ./xlinpl/52000
4 ./xlinpl/53000
4 ./xlinpl/54000
4 ./xlinpl/55000
4 ./xlinpl/56000
4 ./xlinpl/57000
4 ./xlinpl/58000
4 ./xlinpl/59000
4 ./xlinpl/60000
4 ./xlinpl/61000
4 ./xlinpl/62000
4 ./xlinpl/63000
4 ./xlinpl/64000
4 ./xlinpl/65000
4 ./xlinpl/66000
4 ./xlinpl/67000
4 ./xlinpl/68000
4 ./xlinpl/69000
4 ./xlinpl/70000
4 ./xlinpl/71000
4 ./xlinpl/72000
4 ./xlinpl/73000
8 ./xlinpl/74000
4 ./xlinpl/75000
4 ./xlinpl/76000
4 ./xlinpl/77000
4 ./xlinpl/78000
4 ./xlinpl/79000
4 ./xlinpl/80000
4 ./xlinpl/81000
4 ./xlinpl/82000
4 ./xlinpl/83000
4 ./xlinpl/84000
4 ./xlinpl/85000
4 ./xlinpl/86000
4 ./xlinpl/87000
4 ./xlinpl/88000
4 ./xlinpl/89000
4 ./xlinpl/90000
4 ./xlinpl/91000
4 ./xlinpl/92000
4 ./xlinpl/93000
4 ./xlinpl/94000
4 ./xlinpl/95000
4 ./xlinpl/96000
4 ./xlinpl/97000
4 ./xlinpl/98000
4 ./xlinpl/01000
4 ./xlinpl/02000
1320 ./xlinpl
717904 ./pendf
3004 ./lst
156 ./nodes
305092 ./spex
2832 ./meta
8 ./exfor2/01001
8 ./exfor2/01002
8 ./exfor2/02003
8 ./exfor2/02004
128 ./exfor2/03006
8 ./exfor2/03007
8 ./exfor2/04009
112 ./exfor2/05010
8 ./exfor2/05011
8 ./exfor2/06000
8 ./exfor2/07014
8 ./exfor2/07015
8 ./exfor2/08016
8 ./exfor2/09019
8 ./exfor2/11023
8 ./exfor2/12024
8 ./exfor2/13027
8 ./exfor2/14028
8 ./exfor2/14029
8 ./exfor2/14030
8 ./exfor2/15031
8 ./exfor2/16032
8 ./exfor2/17035
8 ./exfor2/18040
8 ./exfor2/19039
8 ./exfor2/20040
8 ./exfor2/21045
8 ./exfor2/22046
8 ./exfor2/22047
8 ./exfor2/22048
8 ./exfor2/22049
8 ./exfor2/22050
8 ./exfor2/23000
8 ./exfor2/24050
8 ./exfor2/24052
8 ./exfor2/24053
8 ./exfor2/24054
8 ./exfor2/25055
8 ./exfor2/26054
8 ./exfor2/26056
8 ./exfor2/26057
8 ./exfor2/26058
8 ./exfor2/27059
8 ./exfor2/28058
8 ./exfor2/28060
8 ./exfor2/28061
8 ./exfor2/28062
8 ./exfor2/29063
8 ./exfor2/29065
8 ./exfor2/30000
8 ./exfor2/33075
8 ./exfor2/39089
8 ./exfor2/41093
8 ./exfor2/53127
8 ./exfor2/55133
8 ./exfor2/73181
48 ./exfor2/74182
64 ./exfor2/74183
72 ./exfor2/74184
12 ./exfor2/74186
140 ./exfor2/79197
8 ./exfor2/82206
8 ./exfor2/82207
8 ./exfor2/82208
8 ./exfor2/83209
8 ./exfor2/90229
8 ./exfor2/90230
32 ./exfor2/90232
8 ./exfor2/91231
8 ./exfor2/91233
8 ./exfor2/92232
8 ./exfor2/92233
8 ./exfor2/92234
208 ./exfor2/92235
8 ./exfor2/92236
8 ./exfor2/92237
188 ./exfor2/92238
8 ./exfor2/93236
8 ./exfor2/93237
8 ./exfor2/93238
8 ./exfor2/93239
8 ./exfor2/94236
8 ./exfor2/94238
224 ./exfor2/94239
8 ./exfor2/94240
8 ./exfor2/94241
8 ./exfor2/94242
8 ./exfor2/94244
8 ./exfor2/95241
8 ./exfor2/95242
8 ./exfor2/9542M
8 ./exfor2/95243
8 ./exfor2/96242
8 ./exfor2/96243
8 ./exfor2/96244
8 ./exfor2/96245
8 ./exfor2/96246
8 ./exfor2/96247
8 ./exfor2/96248
8 ./exfor2/97249
8 ./exfor2/98250
8 ./exfor2/98251
8 ./exfor2/98252
4 ./exfor2/01000
4 ./exfor2/02000
4 ./exfor2/03000
4 ./exfor2/04000
4 ./exfor2/05000
4 ./exfor2/07000
4 ./exfor2/08000
4 ./exfor2/09000
4 ./exfor2/10000
4 ./exfor2/11000
4 ./exfor2/12000
4 ./exfor2/13000
4 ./exfor2/14000
4 ./exfor2/15000
4 ./exfor2/16000
4 ./exfor2/17000
4 ./exfor2/18000
4 ./exfor2/19000
4 ./exfor2/20000
4 ./exfor2/21000
4 ./exfor2/22000
4 ./exfor2/24000
4 ./exfor2/25000
4 ./exfor2/26000
4 ./exfor2/27000
4 ./exfor2/28000
4 ./exfor2/29000
4 ./exfor2/31000
4 ./exfor2/32000
4 ./exfor2/33000
4 ./exfor2/34000
4 ./exfor2/35000
4 ./exfor2/36000
4 ./exfor2/37000
4 ./exfor2/38000
4 ./exfor2/39000
4 ./exfor2/40000
4 ./exfor2/41000
4 ./exfor2/42000
4 ./exfor2/43000
4 ./exfor2/44000
4 ./exfor2/45000
4 ./exfor2/46000
4 ./exfor2/47000
4 ./exfor2/48000
4 ./exfor2/49000
4 ./exfor2/50000
4 ./exfor2/51000
4 ./exfor2/52000
4 ./exfor2/53000
4 ./exfor2/54000
4 ./exfor2/55000
4 ./exfor2/56000
4 ./exfor2/57000
4 ./exfor2/58000
4 ./exfor2/59000
4 ./exfor2/60000
4 ./exfor2/61000
4 ./exfor2/62000
4 ./exfor2/63000
4 ./exfor2/64000
4 ./exfor2/65000
4 ./exfor2/66000
4 ./exfor2/67000
4 ./exfor2/68000
4 ./exfor2/69000
4 ./exfor2/70000
4 ./exfor2/71000
4 ./exfor2/72000
4 ./exfor2/73000
12 ./exfor2/74000
4 ./exfor2/75000
4 ./exfor2/76000
4 ./exfor2/77000
4 ./exfor2/78000
4 ./exfor2/79000
4 ./exfor2/80000
4 ./exfor2/81000
4 ./exfor2/82000
4 ./exfor2/83000
4 ./exfor2/84000
4 ./exfor2/85000
4 ./exfor2/86000
4 ./exfor2/87000
4 ./exfor2/88000
4 ./exfor2/89000
4 ./exfor2/90000
4 ./exfor2/91000
4 ./exfor2/92000
4 ./exfor2/93000
4 ./exfor2/94000
4 ./exfor2/95000
4 ./exfor2/96000
4 ./exfor2/97000
4 ./exfor2/98000
2376 ./exfor2
236 ./bars
22120 ./custom
1148700 .