Least squares refinement from structure factors measured on
twinned crystals
Description:
-
Main program for least squares refinement of structure factors measured
from twinned crystals. Extinction corrections can be included and the
extinction parameters refined.
Geometric slack constraints can be included (see chaper 5 of the CCSL users
manual and the documentation of program BONDS)
Input:
-
The crystal data file must contain all cards needed for structure factor
calculations vis:
-
S cards
giving the symmetry.
A
C card
with the cell dimensions.
A cards
defining the atoms, their positions and isotropic.
temperature factors.
F cards
giving the scattering factors
R cards
describing the twinning:
- R TMAT followed by 9 numbers giving the twin matrices
- (one for each twin component)
R TPOP followed by numbers giving the fractional population
- of each twin in the order of the TMA
T cards
.
optionally:
-
T cards
giving anisotropic temperature factors
Additionally:
-
L cards
indicating which parameters should be refined
- (see the CCSL users' manual)
I cards
which may have the words words NCYC, CYCL1, PRIN,
MCOR and CONV.
- NCYC the number of least squares cycles (default 3)
CYC1 the number to be assigned to the first cycle (default 1)
PRIN frequency at which output lists are printed
- PRIN = 0 no printing
PRIN = 1 first cycle
PRIN = 2 last cycle
PRIN = 3 first and last cycles
PRIN = 4 every cycle
MCOR controls the correlations printed at tyhe end of the job
- (default +70)
MCOR = 0 whole correlation matrix printed
MCOR = -ve no printing
MCOR = +ve correlations above MCOR printed
CONV convergence criterion; LSQ cycling is terminated if the
- maximum shift/esd < CONV (default 0.01)
The data file which should be a list of hkl and the measured structure
factors as implied by the
L MODE card
.
Output:
-
The usual listing file reporting the data which have been read. This file
also contains the output listings requested by the
I PRIN card
as well as
the R-factors and chisqrd at the end of each least squares cycle.
A new crystal data file corresponding to the structure at the end of the
least squares refinement.
Running the program:
-
On running the program the user is asked for the name of the crystal data
file and the name of the file containing the structure factor data.
The value of chisqrd at the end of each cycle is reported on the
terminal.
After the last cycle user is asked for a name for the new crystal data file.
Calls:
- APSHSF
CALCTW
CENTRE
DFTWSF
ERRIN2
ERRMES
GEOMCO
GEOMIN
GEOMLS
INDFLO
LFCALC
LSETSF
MATCOR
MATINV
MATSET
MATSHF
MATTOT
MESS
NOPFIL
NWINSF
OPSYM
PARRUN
PARSSF
PREFIN
PRNCYC
RELATE
RFACS
RREFSF
SETFC
STLSSF
VARMAK
VARSSF
WGHTSF
Common blocks used:
-
- /DEPMAT/ to use all members
- /EXTN/ to use IEXTYP
- /FCAL/ to use FC FCMOD
- /IOUNIT/ to use LPT ITO LUNI
- /NEWOLD/ to use AMAXSH
- /NSYM/ to use CENTRC
- /OBSCAL/ to use OBS GCALC DIFF NOBS WT
- /POSNS/ to use NATOM
- /PRBLEM/ to use NFAM LF1SP
- /REFINE/ to use NCYC NCYC1 LASTCY ICYC IPRNT MAG CONV
- /SCLDAT/ to use ISCALE SCALE
- /SCRACH/ to use MESSAG
- /SLAKDA/ to use NSLAK
- /SLKGEO/ to use SLONLY
- /TWIN/ to use NTWIN
*** SFTLSQ modernised by PJB April 2001 ***
Classification:
General Least Squares Refinement . . . . . . . Main Program
Makes the calculated function and derivatives for twinned crystal
LSQ.
Arguments:
-
On entry H is a 3-sized array containing h,k,l
- SFCALC is a subroutine name used for the actual structure factor
calculation, like LFCALC
Prerequisite calls:
-
Structure factor data must have been set up in various COMMON blocks
by SYMOP RECIP ATOPOS, SETFOR, SETANI
LSETUP, PARSSF, VARMAK must have set up the Least Squares.
STLSSF must have read the
L cards
:
Description:
-
Sets GCALC, the calculated function corresponding to an OBS just read
Applies an extinction correction if there was an
E card
.
Also produces the vector DERIVV (LVARV long) of derivatives of this
function wrt each variable in turn.
The function GCALC is made up of three functions multiplied together.
We write G=P1*P2*P3 where P2 is a simple function of FCALC, P1 is
independent of FCALC (being Scale * overall itf), and P3 is the extinction
correction.(They are written in this way for ease of differentiation.)
If IREF=2, we are in fact comparing OBS squared with GCALC squared. This is
dealt with right at the end, after proceeding as though we were not.
The sequence of operations is:
- Get STHL= sin theta/lambda, SSQRD= STHL squared.
Clear derivatives to zero.
Form P1 and its derivatives.
Form P2: get FCALC and derivatives of mod FCALC wrt family 2 variables
Convert derivatives to be of P2, not just mod FCALC
Form P3: convert derivatives
Form GCALC and make all derivatives so that they are of GCALC
Adjust for GCALC squared or not squared if necessary
Notes:
-
The scale factors are used to refine the twin ratios. Not that they apply
to F rather than Fsqrd. This should prevent negative domain populations
Calls:
- GMPRD
GMZER
ISPABS
LATABS
TESTOV
VCTMOD
Called by:
- SFTLSQ
Common blocks used:
-
- /BRAGG/ to use STHL SSQRD
- /DERVAR/ to use all members
- /EXTN/ to use KDOMR KMOSC EXTCOR DEXDFQ DEXDRQ DEXDGQ
- /FCAL/ to use FCMOD FCDERS
- /OBSCAL/ to use GCALC
- /PRBLEM/ to use LVFST1 NVARF
- /REFINE/ to use IREF
- /SCLDAT/ to use ISCALE SCALE KSCALE
- /TWIN/ to use all members
*** CALCTW updated by PJB C134 5-Nov-2004 ***
Classification:
Specific Least Squares Refinement . . . . . . . Crystallographic
Called as a substitute for DEFALT out of SUBROUTINE VARMAK, giving
default fix/vary information for structure parameters.
Arguments:
-
On entry IFAM, IGEN, ISPC specify the parameter.
Description:
-
On exit DFTWSF is .TRUE. if parameter IFAM,IGEN,ISPC is by default varied
- and .FALSE. if fixed.
In general varies, but fixes scattering factor, and only varies an itf
if there is no corresponding atf. Only allows A*-F* to vary if there
are slack constraints.
Fixes SCAL 1 and varies all TPOP by default
Calls:
- DFTWSF
Called by:
- SFTLSQ
DFTWSF
Common blocks used:
-
- /ANISO/ to use IAPT
- /SLAKDA/ to use NSLAK
*** DFTWSF by PJB Apr 2000 ***
Classification:
Specific Least Squares Refinement . . . . . . . Setting Up
P. Jane Brown
e-mail: brown@ill.fr
Institut Laue Langevin,
Grenoble, FRANCE