---------------------------------------------------------------
---------------------------------------------------------------
A RRRR I TTTTTTTTT V V EEEEEE
A A R R I T V V E
A A RRRR I T V V EEE
AAAAAAA R R I T V V E
A A R R I T V EEEEEE 2K
---------------------------------------------------------------
---------------------------------------------------------------
USER'S GUIDE
Le Mans, first version: 1984
Last update: January 2000,
all previous 4 programs gathered in one,
greatly simplifying the use.
A. Le Bail
Universite du Maine
Laboratoire des Fluorures
ESA CNRS 6010
Avenue O. Messiaen
72085 Le Mans Cedex 9
FRANCE
E-mail :
either armel@fluo.univ-lemans.fr
or alb@cristal.org
or cristal@cybercable.tm.fr
or lebail@univ-lemans.fr
-----------------------------------------------------------------
-----------------------------------------------------------------
SHORT INTRODUCTION:
ARITVE is
a program for modelling amorphous structures by a Rietveld-type
refinement of the atomic coordinates and cell parameters taken
from a starting crystalline model. The data fitted are the S(2t)
interference functions (see § A). Up to three S functions can
be simulated simultaneously, either neutron or X-ray data or both
together with different wavelengths. The imaginary part of the
X-ray diffusion factors is correctly treated (I hope). A non-zero
S(0) cannot be reproduced.
From the crystalline model, a powder diffraction pattern is
generated by the summation of all contributing reflections (hkl)
with a Gaussian profile shape. The FWHM (Full Width at Half Maximum)
angular dependence may follow a microstrain-type description
(variation as tan(theta)), but the instrumental resolution and
eventually a size effect are simply taken into account by a
Cagliotti-type expression.
Not every model will give a "good" fit by using this method. Only
"good" model(s) will give "good" fit(s). An eventually "good" model
may be soon detected : low Rp (Reliability on profile, according to
the Rietveld original definition) is obtained BEFORE refinement.
It is recommended not to try models inconsistent with already known
structural aspects (distances, coordinations, density...).
ARITVE can only improve (by refinement) a yet "good" model.
Test files are provided that correspond to the fit of glassy SiO2
neutron and X-ray data published in :
"Modelling the silica glass structure by the Rietveld method,"
A. Le Bail, J. Non-Cryst. Solids, in press (1995).
However, the fit does not correspond to the final result in the
P212121 space group but to an intermediate result with higher
symmetry (P213).
Note that ARITVE is not
a convivial program easy to manipulate. Do not expect to get a
result before some weeks on your own problem although you may
install and apply the program on the test data in a few minutes.
Some french text may have been forgotten somewhere. This package
has not been extensively tested by many users, bugs are possible.
-----------------------------------------------------------------
-----------------------------------------------------------------
CONTENT
A- How to run the program: definitions of files and parameters
B- Strategy
B-1- The model
B-2- An order to respect for parameter refinement
C- What can be expected from ARITVE
D- How ARITVE could be eventually improved
E- Final recommendations
F- BIBLIOGRAPHY - REFERENCES
G- LIST OF CAUSES OF MAJOR PROBLEMS
-----------------------------------------------------------------
-----------------------------------------------------------------
A- How to run the program: definitions of files and parameters
Explanations are given corresponding to the example of the
test files:
ARCAR.DAT data file with commands for structure refinement
ASIO2N.NOR neutron interference function of SiO2
ASIO2X.NOR X-ray interference function of SiO2
ARITVE.ZIP contains also :
ARITVE.EXE : the axecutable for Windows 95/NT/98
ARITVE.F : the Fortran source code
LICENSE.HTML : the GNU General Public License
ARITVE.TXT : this manual
I the general case, one .dat file with instructions will be
needed as well as up to 3 .nor files containing interference
functions.
The data included in the SiO2 .NOR files are from:
J.H. KONNERT & J. KARLE, Acta Cryst. A29, 702 (1973)
for X-rays
and P.A.V. JOHNSON, A.C. WRIGHT & R.N. SINCLAIR, J.
Non-Cryst. Solids, 58, 109 (1983).
for neutrons
The starting model is that of the Carnegieite structure:
F.W. BARTH & E. POSNJAK, Z. fur Kristallogr. 8, 376-385 (1932).
But the starting coordinates and cell parameters are from
the (may be false) description of a high-cristoballite form
with the carnegieite model:
T.F.W BARTH, American Journal of Science, Serie 5,
23, 350-356 (1932).
(see also Wyckoff, Crystal Structures, Volume I).
WARNING : These "interference functions" are the so called
S(2t)
i.e. the equivalent of
S(Q)
with a contant 2*theta step (the diffracted
intensity, properly corrected, divided by
<f**2>). Here, the test files are given at
an arbitrary scale (multiplied by an
arbitrary constant).
ASIO2N.NOR is given there, the first line must contain three
values :
nbmes 2t Step unformatted
nbmes = number of intensity data
2t = starting two theta angle
Step = constant step in two theta degrees
Then the nbmes intensities are given (unformatted)
230 0.00 0.40
0.00000E+00 0.65620E+03 0.68291E+03 0.73245E+03 0.80835E+03
0.93140E+03 0.11210E+04 0.13651E+04 0.16390E+04 0.19135E+04
......
0.57189E+04 0.57198E+04 0.57212E+04 0.57232E+04 0.57258E+04
0.57290E+04 0.57327E+04 0.57370E+04 0.57418E+04 0.57474E+04
0.57535E+04 0.57602E+04 0.57674E+04 0.57753E+04
The same for ASIO2X.NOR :
230 0. 0.4
0. 93. 232. 421. 639. 833. 1042. 1255. 1461. 1674.
1890. 2106. 2321. 2558. 2800. 3054. 3368. 3760. 4287. 4959.
5787. 6738. 7703. 8556. 9142. 9386. 9251. 8808. 8203. 7549.
......
5195. 5195. 5202. 5215. 5230. 5254. 5296. 5350. 5401. 5469.
5538. 5620. 5694. 5793. 5890. 5984. 6059. 6133. 6206. 6268.
6335. 6398. 6457. 6500. 6542. 6566. 6578. 6589. 6596. 6583.
the ARITVE program is started by
ARITVE
entry file (no extension)?? ARCAR
The .DAT file must be prepared as follows:
The test file ARCAR.DAT:
line ! CONTENT
1. ! IF AMORPHOUS SiO2 WAS TYPE CARNEGIEITE P213
2. ! 90 2 13 1 0 0 0 0 0 0
3. ! SiO2-N
3. ! SiO2-X
4. ! 1 4
5. ! 4 2
6. ! 0.5000 0.7093
7. ! 0.5 0.8 0.8 0.05
8. ! P 21 3
9.1 ! 0.4149 0.5803
9.1 ! 0.0704 0.0060
9.2 ! 6.2915 2.4386 3.0353 32.3337 1.9891 0.6785 1.5410
81.6937 1.1407 0.0817
9.2 ! 3.0485 13.2771 2.2868 5.7011 1.5463 0.3239 0.867
32.9089 0.2508 0.0106
10.1 ! SI1
10.2 ! 1 0.25500 0.25500 0.25550 0.04000
10.1 ! SI2
10.2 ! 1 -0.00800 -0.00800 -0.00800 0.04000
10.1 ! O1
10.2 ! 2 0.12500 0.12500 0.12500 0.04000
10.1 ! O2
10.2 ! 2 0.66000 0.66000 0.06200 0.12000
11. ! 60.97914 313235.4 5000.0 106958.6
11. ! 48.11535 566242.9 10000.0 279894.0
12. ! 7.16000 7.16000 7.16000 90.00000 90.00000 90.00000
13.1 ! 2
13.2 ! 0 440
13.2 ! 9200 17500
13.1 ! 2
13.2 ! 0 600
13.2 ! 9200 17500
14. ! 1.0
14. ! 1.0
15. ! asio2n.nor
15. ! asio2x.nor
16. ! 81 81 81 0
16. ! 91 91 91 0
16. ! 101 101 101 0
16. ! 111 121 131 0
17. ! 11
17. ! 21
18. ! 31 0 41
18. ! 51 0 61
19. ! 71 71 71 0 0 0
-----------------------------------------------
| ------------------------------------------- |
| | PARAMETER DEFINITIONS | |
| ------------------------------------------- |
------------------------------------------------
LINE 1 : TEXT --> TITLE FORMAT 20A4
---------
LINE 2 : NCYCLE,NPAT,MAXS,LG,IF1 to IF6 FREE FORMAT
---------
NCYCLE --> NUMBER OF REFINEMENT CYCLES
NPAT --> NUMBER OF PATTERNS (MAXIMUM 3)
MAXS --> NUMBER OF REFINED PARAMETERS
LG --> CODE TO SEE EVENTUALLY Iobs AND Icalc
ON THE SCREEN: IF 0 --> YOU SEE THEM
IF 1 --> YOU DO NOT
INTEREST : ESTIMATE THE SCALE FACTOR
WHEN THE REFINEMENT IS STARTING
AND ALSO THE FWHM PARAMETERS
IF1 to IF6 --> Codes for optional output files
IFn=0 no output
IFn=1 output
IF1 : *.out file, contains contributing hkl to each point
IF2 : bidon.hkl, the list of hkl and multiplicity
IF3 : *.pre, intermediate file before refinement
IF4 : bidon.imp, intermediate file
IF5 : bidon2.imp, intermediate file
IF6 : *.lt4 contains observed and calculated data
LINE 3 : PTEXT --> ONE TITLE FOR EACH PATTERN FORMAT 4A4
---------
=====> NPAT LINES (max NPAT = 3)
LINE 4 : KXR(1) KXR(2) KXR(NPAT) FREE FORMAT
---------
=====> NPAT VALUES ON ONE LINE
KXR --> CODE FOR THE PATTERN TYPE:
1 : NEUTRONS
4 : X-RAYS
LINE 5 : NA,KL FREE FORMAT
---------
STRUCTURE INDICATORS :
NA --> THE NUMBER OF INDEPENDENT ATOMS CONTAINED
IN THE ASYMMETRIC UNIT (Max = 50)
KL --> THE NUMBER OF ATOMS HAVING DIFFERENT
DIFFUSION FACTOR, EITHER NEUTRONS OR X-RAYS
(Max = 4)
LINE 6 : DLABDA(1) .... DLABDA(NPAT) FREE FORMAT
---------
WAVELENGTHS IN ANGSTROMS FOR EACH PATTERN
NPAT VALUES ON ONE LINE
LINE 7 : RELAXC,RELAXB,RELAXS,RELAXH
---------
RELAXATION FACTORS APPLIED BY MULTIPLICATION
ON THE SHIFTS AFTER EACH REFINEMENT CYCLE
RELAXC --> CONCERNS ATOMIC COORDINATES
RELAXB --> UNUSED
RELAXS --> CONCERNS SCALE FACTORS AND OCCUPANCY FACTORS
RELAXH --> CONCERNS CELL PARAMETERS AND FWHM
LINE 8. : Space group
----------
Spacegroups are defined as in Lazy-Pulverix :
KLAUS YVON, WOLFGANG JEITSCHKO AND ERWIN PARTHE
J.APPL.CRYST. (1977), 10, P 73-74
S P A C E G R O U P SYMBOLS
DO N O T INCLUDE THE STAR PRECEEDING SOME OF THE SYMBOLS.
THE STAR INDICATES CENTROSYMMETRIC SPACE GROUPS WHICH HAVE
BEEN DESCRIBED WITH SEVERAL SETTINGS. THE PROGRAM GENERATES
ONLY THE SETTING WITH THE CENTRE OF SYMMETRY AT THE ORIGIN OF
THE UNIT CELL.
W A R N I N G A SYMBOL THAT DOES NOT FIGURE IN THIS LIST
MAY YIELD WRONG EQUIPOINTS.
TRICLINIC
P 1 P -1
MONOCLINIC
P 2 P 21 C 2 P M P C
C M C C P 2/M P 21/M C 2/M
P 2/C P 21/C P 21/N C 2/C
THE POINT POSITIONS GENERATED FROM THESE SYMBOLS CORRESPOND TO
THE MONOCLINIC SETTING WITH B AS UNIQUE AXIS (ALPHA=GAMMA=90.)
ORTHORHOMBIC
P 2 2 2 P 2 2 21 P 21 21 2 P 21 21 21 C 2 2 21
C 2 2 2 F 2 2 2 I 2 2 2 I 21 21 21 P M M 2
P M C 21 P C C 2 P M A 2 P C A 21 P N C 2
P M N 21 P B A 2 P N A 21 P N N 2 C M M 2
C M C 21 C C C 2 A M M 2 A B M 2 A M A 2
A B A 2 F M M 2 F D D 2 I M M 2 I B A 2
I M A 2 P M M M *P N N N P C C M *P B A N
P M M A P N N A P M N A P C C A P B A M
P C C N P B C M P N N M *P M M N P B C N
P B C A P N M A C M C M C M C A C M M M
C C C M C M M A *C C C A F M M M *F D D D
I M M M I B A M I B C A I M M A
TETRAGONAL
P 4 P 41 P 42 P 43 I 4
I 41 P -4 I -4 P 4/M P 42/M
*P 4/N *P 42/N I 4/M *I 41/A P 4 2 2
P 4 21 2 P 41 2 2 P 41 21 2 P 42 2 2 P 42 21 2
P 43 2 2 P 43 21 2 I 4 2 2 I 41 2 2 P 4 M M
P 4 B M P 42 C M P 42 N M P 4 C C P 4 N C
P 42 M C P 42 B C I 4 M M I 4 C M I 41 M D
I 41 C D P -4 2 M P -4 2 C P -4 21 M P -4 21 C
I -4 M 2 P -4 C 2 P -4 B 2 P -4 N 2 P -4 M 2
I -4 C 2 P -4 2 M I -4 2 D P 4/M M M P 4/M C C
*P 4/N B M *P 4/N N C P 4/M B M P 4/M N C *P 4/N M M
*P 4/N C C P 42/M M C P 42/M C M *P 42/N B C *P 42/N N M
P 42/M B C P 42/M N M *P 42/N M C *P 42/N C M I 4/M M M
I 4/M C M *I 41/A M D *I 41/A C D
TRIGONAL
P 3 P 31 P 32 R 3 P -3
R -3 P 3 1 2 P 3 2 1 P 31 1 2 P 31 2 1
P 32 1 2 P 32 2 1 R 3 2 P 3 M 1 P 3 1 M
P 3 C 1 P 3 1 C R 3 M R 3 C P -3 1 M
P -3 1 C P -3 M 1 P -3 C 1 R -3 M R -3 C
ALL R-SPACE GROUPS REFER TO THE HEXAGONAL SETTING
HEXAGONAL
P 6 P 61 P 65 P 62 P 64
P 63 P -6 P 6/M P 63/M P 6 2 2
P 61 2 2 P 65 2 2 P 62 2 2 P 64 2 2 P 63 2 2
P 6 M M P 6 C C P 63 C M P 63 M C P -6 M 2
P -6 C 2 P -6 2 M P -6 2 C P 6/M M M P 6/M C C
P 63/M C M P 63/M M C
CUBIC
P 2 3 F 2 3 I 2 3 P 21 3 I 21 3
P M 3 *P N 3 F M 3 *F D 3 I M 3
P A 3 I A 3 P 4 3 2 P 42 3 2 F 4 3 2
F 41 3 2 I 4 3 2 P 43 3 2 P 41 3 2 I 41 3 2
P -4 3 M F -4 3 M I -4 3 M P -4 3 N F -4 3 C
I -4 3 D P M 3 M *P N 3 N P M 3 N *P N 3 M
F M 3 M F M 3 C *F D 3 M *F D 3 C I M 3 M
I A 3 D
LINE 9.1 : KL VALUES (SEE LINE 5) OF
---------- - FERMI LENGTHS - NEUTRON CASE
OR OF - DELTA F" - X-RAY CASE
(IMAGINARY DISPERSION CORRECTION)
FOR EACH PATTERN (NPAT LINES)
FREE FORMAT
THE ORDER OF THE KL VALUES MUST BE CONSISTENT
WITH THAT GIVEN LATER
WHEN THE PATTERN IS A X-RAY PATTERN, A 9.2 LINE
MUST FOLLOW
LINE 9.2 : A1 B1 A2 B2 A3 B3 A4 B4 C DELTAF' FREE FORMAT
-----------
9 COEFFICIENTS FOR ANALYTICAL APPROXIMATION TO THE X-RAY
SCATTERING FACTORS FOLLOWED BY THE REAL PART OF THE
DISPERSION CORRECTION
--------------------------------------------
LINES 10.1 AND 10.2 :: ATOMIC PARAMETERS
--------------------------------------------
TO BE GIVEN NA-TIME (SEE LINE 5)
LINE 10.1 : IDENTIFICATION ASCII FOR THE N-IEME ATOM
------------
FORMAT A4
LINE 10.2 : NTYP,X,Y,Z,NOCCUP
------------
FREE FORMAT
NTYP --> ORDER NUMBER OF THE CORRESPONDING DIFFUSION FACTOR
X,Y,Z --> REDUCED ATOMIC COORDINATES
NOCCUP --> SITE OCCUPANCY
ALL VALUES CAN BE MULTIPLIED BY A CONSTANT,
THIS CAN BE TUNED ALSO BY THE SCALE FACTOR.
LINE 11. : SCALE U V W FREE FORMAT
-----------
===> FOUR VALUES FOR EACH PATTERN
SCALE --> SCALE FACTOR
U --> LINE-WIDTH FACTOR
V --> LINE-WIDTH FACTOR
W --> LINE-WIDTH FACTOR
FWHM == SQRT(U*Tg**2(theta) + V*Tg(theta) + W)
and the FWHM is the full width at half maximum in
(2-theta degrees)*100
LINE 12. : A,B,C,ALPHA,BETA,GAMMA FREE FORMAT
-----------
DIRECT CELL PARAMETERS
npat groups of lines 13.
LINE 13.1 : Nex = number of excluded zones Free Format
----------- for the nth pattern
LINE 13.2 : ilow, ihigh Free Format
----------- = the low and high limits
of the excluded zone
in degrees 2*theta*100
Nex lines 13.2 have to be given
If Nex=0 for the nth pattern, give no 13.2 line
Then, npat groups of lines 14. have to be given
LINE 14. : Iscale Free Format
----------- Iscale = a multiplicative factor
used to eventually modify
the intensities of the nth
interference function.
npat groups of line 15.
LINE 15. : name.nor 20A1
----------- = name of the nth datafile
LINES 16 TO 19 : CODES TELLING THE PROGRAM TO REFINE OR NOT THE PARAMETERS
--------------------------------------------------------------------------
APPEARING IN THE SAME ORDER AS THE PARAMETERS LINES
10.2, 11 AND 12
THE KEY OF THESE CODES IS GIVEN HERE:
CODE = [ M * 10 + ABS(SM) ] * SIGN(SM)
WHERE : M --> M-IEME REFINED PARAMETER
SM --> MULTIPLYING FACTOR APPLIED TO THE SHIFT
FOUND FOR THE M-IEME PARAMETER AT EACH
REFINEMENT CYCLE
EXAMPLES :
IF CODE = 101 , this is the
10th PARAMETER, THE SHIFT WILL
BE MULTIPLIED BY +1
IF CODE = 100.5 , this is the
10th PARAMETER, THE SHIFT WILL
BE MULTIPLIED BY +0.5
IF CODE = 0
THIS PARAMETER IS NOT REFINED
IT IS POSSIBLE TO ASSOCIATE SOME PARAMETERS IN A SIMPLE
WAY:
FOR INSTANCE X, Y=1/2+X, Z
THE CODES FOR X AND Y SHOULD BE THE SAME BECAUSE
THE SHIFTS ON X AND Y=1/2+X WILL BE THE SAME :
101 101 111
FOR INSTANCE X, Y=-X, Z
THE CODES FOR X AND Y COULD BE 101 AND -101
OTHER EXAMPLE: X, Y=2X, Z
THE CODES FOR X AND Y COULD BE 100.5 AND 101.
OR 101. AND 102.
BECAUSE THE SHIFT ON Y IS TWICE THAT ON X
THE ORDER FOR THE NUMBER OF THE PARAMETERS IS
UNIMPORTANT. iT IS JUST RECOMMENDED THAT THE MAXIMUM
NUMBER IS EQUAL TO MAXS (LINE 2)
FORMAT ALWAYS FREE FOR THE CODES
LINE 16. : FOUR CODES POUR X,Y,Z,NOCCUP FREE FORMAT
----------
NA LINES TO BE GIVEN (put always the NOCCUP code = 0)
LINE 17. : ONE CODE FOR SCALE FACTOR FREE FORMAT
----------
=====> NPAT LINES
LINE 18. : THREE CODES FOR U, V, W FREE FORMAT
----------
=====> NPAT LINES
LINE 19 : 6 CODES FOR THE CELL PARAMETERS FREE FORMAT
----------
============>>>>>>>> ATTENTION
THESE CODES APPLY TO A RECIPROCAL
METRIC TENSOR
THE CONSTRAINTS WHICH SHOULD BE APPLIED
ARE EXAMPLIFIED HERE:
EXAMPLE OF CODES FOR CUBIC : 11 11 11 0 0 0
TETRAG : 11 11 21 0 0 0
TRIGONAL OR HEXAGON : 11 11 21 0 0 11 <----------------
ORTHOR : 11 21 31 0 0 0
MONOCLI : 11 21 31 0 41 0
TRICLIN : 11 21 31 41 51 61
MONOCLINIC IS FOR beta DIFFERENT FROM 90 DEGREES
THE RHOMBOHEDRAL CASE IS ALWAYS TO BE TREATED IN TRIGONAL SETTING
_________________________________________________________________________
_________________________________________________________________________
_________________________________________________________________________
_________________________________________________________________________
OUTPUT FILES CREATED BY ARITVE:
ARCAR.IMP : results cycle after cycle
ARCAR.PAR : intermediate and final new parameters,
can be included in ARCAR.DAT
in place of the old ones
ARCAR1.PRF : ASCII file containing
observed and calculated S(2t)
for the first pattern
can be seen by DMPLOT
ARCAR2.PRF : ,, for the second pattern
OPTION OUTPUT
ARCAR.OUT : Intermediate file giving hkl
attribution to each point
ARCAR.LT4 : file containing analogous
information as the .PRF files
BIDON.DAT : Intermediate file
BIDON.IMP : Intermediate file
BIDON2.IMP : Intermediate file
BIDON.HKL : hkl and multiplicity
---------------------------------------------------------------
---------------------------------------------------------------
CPU TIME NEEDED:
For the test file calculation with ARITVE (90 refinement cycles)
on a Pentium II 266MHz, the total CPU time was 2 minutes and 14
seconds.
---------------------------------------------------------------
---------------------------------------------------------------
ARCAR.PAR output file obtained after 1 cycle
SI1
1 0.25522 0.25522 0.25572 0.04000
SI2
1 -0.00776 -0.00776 -0.00776 0.04000
O1
2 0.12533 0.12533 0.12533 0.04000
O2
2 0.66016 0.65969 0.06207 0.12000
61.67587 351787.0 5000.0 106291.8
48.46439 593292.5 10000.0 280832.2
7.15890 7.15890 7.15890 90.00000 90.00000 90.00000
That file will give the new parameters after each cycle.
In principle, you have to select the last result and insert it
in the *.dat file in place of the old parameters.
---------------------------------------------------------------
---------------------------------------------------------------
Part of the ARCAR1.PRF file obtained after 1 cycle
This file is suitable for use as data for a graphical output
by the DMPLOT program (.prf format)
Adaptation is easy since the .PRF file contains:
2theta-end 2theta-initial Step
number of phase , number of points
two unused values
All observed S(2t)
and then all calculated ones:
91.600 0.000 0.400
1 230
1 1
0. 0. 0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 2968. 3686. 4804. 6347.
7884. 8705. 8489. 7535. 6301. 5231. 4480. 3943.
3539. 3233. 3008. 2867. 2809. 2879. 3443. 4701.
......
5920. 5946. 5978. 6014. 6056. 6101. 6148. 6195.
6242. 6286. 6325. 6358. 6381. 6394. 6394. 6380.
6355. 6321. 6279. 6232. 6181. 6128. 6072. 6016.
5960. 5904. 5849. 5797. 5747. 0.
0. 0. 0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 2496. 3717. 5105. 6471.
7578. 8211. 8257. 7744. 6827. 5732. 4676. 3808.
3196. 2842. 2712. 2767. 2976. 3313. 3752. 4257.
.......
6444. 6438. 6419. 6386. 6343. 6292. 6235. 6174.
6114. 6056. 6001. 5953. 5910. 5875. 5845. 5821.
5802. 5785. 5771. 5757. 5742. 5726. 5709. 5690.
5671. 5651. 5633. 5618. 5608. 0.
---------------------------------------------------------------
---------------------------------------------------------------
B- Strategy
This part explains how to proceed. A very important point is
that your interference functions must be accurate and go to as
large Q as possible, up to the end of modulations. But the program
works with S(2t), t for theta (constant wavelength data), see
A.C. Wright, J. Non-Cryst. Solids 123 (1990) 129. If you have
only one data set (i.e. one neutron pattern, or one X-ray),
then it would be better if your material was a mono-atomic one.
The ideal case for a di-atomic material is that you have three
neutron interference functions with isotopic substitutions
corresponding to large scattering length variations (nevertheless,
why not adding the X-ray interference function). Of course,
the more you have data to fit, the more your model will be
credible. THE "PERFECT" FIT OF ONE INTERFERENCE FUNCTION IN
THE CASE OF A POLYATOMIC AMORPHOUS MATERIAL IS NOT SUFFICIENT
FOR CLAIMING THAT YOUR MODEL IS A POSSIBLE ONE ("PERFECT" meaning
Rp<1% may be).
B-1- The model
As the Rietveld method, applied to structure refinement from
powder diffraction data, ARITVE needs a starting structural model.
The choice of one or several models belongs to the user's
responsibility. The source of possible models is, as should be
evident, to be found in some well known structural data base
(ICSD, Cambridge, Pearson's Handbook of Crystallographic Data for
Intermetallic phases...). It is also an evidence that
the models have to be chosen among isoformula crystallized
compounds with atoms behaving similarly in crystal chemistry.
However, all amorphous materials have not a formula as simple as
"SiO2". Thus sometimes several compositions can be chosen in the
glass forming domain in order to reproduce exactly the formula of
crystallized compounds known to coexist in the diagram with the
amorphous ones. That is to say, for structure modelling of amorphous
materials, THE STRATEGY MAY BEGIN AT THE SYNTHESIS STAGE.
Suppose that possible crystalline models exist but with different
kind of atoms (ex: BeF2 instead of SiO2), then you should modify
the cell parameters of the model in a proportional way to BE
COHERENT WITH THE ATOMIC DENSITY OF YOUR AMORPHOUS COMPOUND.
B-2- An order to respect for parameter refinement:
Once the model has been chosen and all the files prepared, you cannot
start by refining all parameters. Do not expect that any case will
work as well as the test case where all parameters are refined
together and Rp decreases regularly from 8.14 and 5.35 to 3.39 and
2.90 respectively for neutron and X-ray data after 90 cycles. If
you try that directly, the calculation may "explode". Remember that
the test case is a "good" one...
First you must choose a set of U V W parameter (in order to broaden
sufficiently the reflection profiles) and perform some cycles
refining the scale factors only till stabilization of the Reliability
factor(s) Rp. If there is not any similitude at this stage between
the observed and calculated interference function, you have not A
large chance to succeed. You must have Rp<20 or even 10% at this stage.
Second: adjust U V and W manually or try to refine U and W with
V fixed to a low value (corresponding to the resolution function
of the apparatus). The alternate refinement of U, W and then the cell
parameters can be attempted (or simultaneous, but this is dangerous
because they are highly correlated).
Third, the atomic coordinates together with the scale factor
should be refined. If some atom has a very low scattering
length for all the interference functions simultaneously,
then you should not refine its coordinates.
One must note that the thermal motion effects are considered as
included in the disorder effect simulated by the line-width
variation.
The above suggestions may not work because the whole process
is very unstable... At all stages it is possible
to compromise the next step by too much a deviation of one
parameter or another. With the exception of the scale factors,
the relaxation factors have to limit heavily the variation of
the refined parameters. Easy fall down to false minima cannot
be excluded. Remember that each model will have a Rp limit.
Process slowly in introducing new refined parameters. Do not
keep new parameters that led to divergency. However, sometimes
Rp may increase before to decrease.
When significant changes on the cell parameters have been
observed, the whole first step (preparation of .hkl and .pre
files) must be done again to be consistent with the new values.
---------------------------------------------------------------
---------------------------------------------------------------
C- What can be expected from ARITVE
Rietveld application to amorphous materials should be considered
as one particular approach to their structure modelization:
essentially a way to test quickly an idea about the possibility
that there could be some similitude between the mean organization
in a given amorphous material and the strict tridimensional one
of a given crystalline structure.
This method can be used to eliminate wrong models. The method is
however insufficient to establish the validity or unicity of a
model, even if the fit is quite "good".
In all published applications (see the list of references at the
beginning of this guide), "good" fit (say Rp < 3%) could be
obtained with models that showing a few unrealistic
distances between some atom-pairs. However, the mean distances for
particular types of pairs were generally credibles. After all,
this holds also for other methods of simulating amorphous
material structures, such as molecular dynamics or model-building
followed by relaxation or Reverse Monte Carlo.
With this program I tried a lot of models for SiO2 (X-ray + neutron
data simultaneously fitted) without obtaining a fully
satisfying one... The best of the more simple ones is proposed as
a test of the program, derived from the carnegieite structure.
---------------------------------------------------------------
---------------------------------------------------------------
D- How ARITVE could be eventually improved
For those who are not satisfied by the modest performances
of ARITVE, some works are suggested:
The program efficiency would be improved by:
- Using constraints on distances during refinement, thus
ensuring the respect of external indications (from EXAFS
measurements or from well established crystal chemistry).
- Generation of the new hkl list at each cycle following
the eventual cell parameters variation.
- Etc...
---------------------------------------------------------------
---------------------------------------------------------------
E- Final recommendations
- You must not limit yourself to the use of this program, even
if the results seem satisfying. Try other methods and programs
as the RMC of McGreevy, molecular dynamics.
- Any consequence of using this program and consequence of result
interpretation should not be considered as depending of the
program author responsibility.
- This program has not been tested extensively, problems may
occur...
-----------------------------------------------------------------
-----------------------------------------------------------------
F- BIBLIOGRAPHY - REFERENCES
References for the original Rietveld method are:
H.M. RIETVELD ACTA CRYST. 22,151-152 (1967)
H.M. RIETVELD J. APPLIED CRYST. 2, 65-71 (1969)
ARITVE was built from the multipattern Rietveld program of:
M.W THOMAS & P.J. BENDALL ACTA CRYST. A34, 5351 (1978)
The original text (in French) describing the principle of
the ARITVE program and some applications may be found in:
A. LE BAIL, THESE DE DOCTORAT D'ETAT, LE MANS (1985)
The method together with an application was first introduced
at the Thirteenth International Congress of Crystallography,
9-18 Aout 1984, Hambourg:
A. LE BAIL & C. JACOBONI, ACTA CRYST. A40, Suppl. C477 (1984)
Applications and some discussions on the method can be found in:
A. LE BAIL, C.JACOBONI & R. DE PAPE, J.DE PHYSIQUE, COLL. C8,
46, 163-168 (1985)
A. LE BAIL, C. JACOBONI & R. DE PAPE, MATER. SCI. FORUM 6,
441-448 (1985)
M. LEBLANC, G. FEREY, J.M. GRENECHE, A. LE BAIL, F. VARRET,
R. DE PAPE & J. PANNETIER, J.DE PHYSIQUE, COLL. C8, 46,
175-179 (1985)
A. LE BAIL, B. BOULARD & C. JACOBONI, MATER. SCI. FORUM,
19-20, 127-136 (1987)
M. MARET, P. CHIEUX, J.M. DUBOIS & A. PASTUREL, J. PHYS.:
CONDENS. MATTER 3, 2801-2817 (1991)
The glassy SiO2 modelling is in :
"Modelling the silica glass structure by the Rietveld method,"
A. Le Bail, J. Non-Cryst. Solids, in press (1995).
The parallelism with modelling crystallite size/microstrain
in Rietveld analysis is described in:
A. LE BAIL, NIST SPECIAL PUBLICATION 846, 142-153 (1992)
The method has been cited in some review articles:
A. C. WRIGHT, J. NON-CRYST. SOLIDS 106, 1-16 (1988)
A. C. WRIGHT, J. NON-CRYST. SOLIDS 123, 129-148 (1990)
and may be others...
In these articles, the method was classified in the same
group as the reverse Monte Carlo method(s), may be improperly.
In the first version, a strict tan(theta) dependence constraint
was applied to the FWHM in order to follow exclusively a
microstrain effect. However, for d(hkl) > dmin, the FWHM was forced
to be equal to that calculated for dmin (with dmin near of 1.9
Angstroem). Otherwise the line profiles became Dirac peaks when
tan(theta)--> 0. This was a way to take account of the instrumental
resolution function and eventually a size effect.
The present version gives similar results with a Cagliotti-type
FWHM variation law expected to take account of all possible
effects. This law is the classical one used in Rietveld-type
refinement. A demonstration that this law is able to model
microstrain and size effect together with the instrumental
resolution may be found in:
R.A. YOUNG & P. DESAI, Arkiwum Nauki o Materialach, 10,
71-90 (1989).
-----------------------------------------------------------------
-----------------------------------------------------------------
In case of use, references that may be cited are :
A. LE BAIL, ARITVE User Guide, Universite du Maine,
France (2000).
or
"Modelling the silica glass structure by the Rietveld method,"
A. Le Bail, J. Non-Cryst. Solids, in press (1995).
---------------------------------------------------------------
---------------------------------------------------------------
G- LIST OF CAUSES OF MAJOR PROBLEMS:
--------------------------------
-Your data go at too large angle. It would be better if you limit
them to nearly 90 2-theta. You could work on your whole dataset
by a regeneration at a shorter wavelength using interpolation. The
problem may be that to simulate your data at large angle, the
addition of the contributing reflexions occuring up to 180
degrees 2-theta are unsufficient.
-U (and may be V or W) is too large and the last reflections are
contributing yet to many of the last points of your pattern(s).
-Non-respect of maximal limits of the program:
3 patterns
60000 (hkl) per pattern
20000 overlapping (hkl) at a diffracting angle
1200 points per pattern
70 refined parameters
4 different atom-type in your sample
.......
-Lack of experience in refining crystalline structures. It is not
recommended to study amorphous without a good knowledge of
crystalline compounds and of techniques to determine and refine
structures from single crystal or powder diffraction data.
-You have not read this guide...
---------------------------------------------------------------
---------------------------------------------------------------
GOOD LUCK