source: trunk/doc/expgui1.html @ 542

Last change on this file since 542 was 505, checked in by toby, 12 years ago

# on 2001/12/19 18:55:42, toby did:
reformat to match NCNR style

  • Property rcs:author set to toby
  • Property rcs:date set to 2001/12/19 18:55:42
  • Property rcs:lines set to +20 -17
  • Property rcs:rev set to 1.6
  • Property rcs:state set to Exp
  • Property svn:keywords set to Author Date Revision Id
File size: 8.6 KB
Line 
1<html>
2<head>
3   <META NAME="Author" CONTENT="Brian H. Toby">
4   <title>EXPGUI</title>
5   <meta name="keywords" content="crystallography, Rietveld, diffraction,
6   GSAS, EXPGUI">
7</HEAD>
8<style>
9A:link {text-decoration:none}
10A:vlink {text-decoration:none}
11</style>
12
13<BODY BGCOLOR="#FFFFFF"
14      topmargin="0" leftmargin="0" marginwidth="0" marginheight="0" 
15      text="#000000" link="#0033ff" vlink="#0033ff" alink="#0033ff">
16
17<?
18   include("/var/www/include/navigation.inc");
19   include("/var/www/include/utility.inc");
20?>
21<blockquote><font face="arial, helvetica, sans-serif">
22
23<TABLE BORDER BGCOLOR="#FFFF40" ALIGN=RIGHT>
24<TR><TH><A  Href="expgui.html">EXPGUI top</A> 
25</TH><TH><A Href="expgui2.html">Next page</A>
26</TH></TR></TABLE><BR CLEAR=ALL>
27
28<center><h1>
29<HR noshade width="75%" size="2" align="center">
30EXPGUI, part 1
31<HR noshade width="75%" size="2" align="center">
32</h1></center>
33<P>
34<h3>A.1 Least Squares (LS) Controls Pane</h3>
35<DL><DL>
36<p>The LS Controls pane shows information about the
37current experiment, typically found in the EXPEDT "Least
38Squares Controls" options.
39<P>
40Note that the order that histograms appear in this
41pane is determined by the
42<a href="expguic.html#sorthist">"Sort histograms by"</a> option in
43the Options Menu.
44</DL>
45<img SRC="1.gif" align=TEXTTOP alt="EXPGUI Screen snapshot">
46</DL>
47<DL><DL>
48The entries on the upper part of this pane are overall options for the
49entire experiment.
50<P>
51<DT><B>Last History</B><DD>
52This shows the last history record written into
53the experiment file, showing the last program that modified the file
54and when it was run.
55<DT><B>Title</B><DD>
56This is a title for the refinement. users can specify any information
57they want saved in the experiment file.
58<DT><B>Number of Cycles</B><DD>
59This is the number of refinement cycles to be performed in GENLES.
60If this number is zero when GENLES is run,
61powder diffraction intensities are computed and, when requested
62(<a href="#extract">see below</a>) reflection intensities are estimated
63but parameters are not refined. Note that when a
64<a href="#lebail">LeBail extraction</a> is performed
65with the cycles set at zero, reflection
66intensities are optimized even when though no cycles of refinement are
67performed.
68<DT><B>Print Options</B><DD>
69<img SRC="1a.gif" align=right alt="EXPGUI Screen snapshot">
70This allows you to control what types of output GENLES provides. The menu of
71options is shown to the right. I recommend including the summary of shifts and
72in most cases the correlation matrix in the output.
73<DT><B>Convergence Criterion</B><DD>
74GENLES stops refinements when the sum of the squares of each
75parameter shifts divided by its standard uncertainty is less than
76this "Convergence Criterion." Since this quantity is the <I>total</I>
77sum of squares, it is reasonable to raise this value for refinements
78where large numbers of parameters will be refined.
79<DT><B>Marquardt Damping</B><DD>
80Marquardt damping increases the weighting of the diagonal elements
81in the Hessian matrix, reducing the impact of parameter correlation
82on the refinement. It increases refinement stability at the cost
83of requiring additional cycles of refinement. A Marquardt term of 1.0
84corresponds to a standard least-squares refinement with no Marquardt
85damping. The value 1.2 has been recommended to me by Lachlan Cranswick as
86a good choice.
87<P>
88</DL>
89<br clear=all>
90<DL>
91The lower section, labeled "Reflection Intensity Extraction" has options
92for each histogram that determine if reflection intensities will be estimated,
93and if so, how.
94<P>
95<DT><B>Extract Fobs</B><DD>
96When the Extract Fobs option is on, reflection intensities are computed
97using the method developed by Hugo Rietveld. In this method
98the intensity for each reflection is determined by summing the
99appropriate data points, weighed by the ratio of the computed intensity
100from that reflection to the total computed intensity at that point. This means
101that in the case of severely overlapped reflections, "observed"
102intensities are apportioned according to the relative computed reflection
103intensities. This is clearly biased since it invokes the crystallographic
104model, but is about the best that can be done. Turning this option off
105saves a very small amount of computer time.
106<a name="extract">
107</a><DT><B>Intensity Extraction Methods</B><DD>
108There are two approaches to reflection intensity determination. In the
109conventional <B>Rietveld</B> approach, if the "Extract Fobs" flag is on,
110reflection intensities are determined
111as part of the Rietveld refinement, reflection R-factors are
112computed, and the reflection intensities
113are saved on disk file for use in Fourier or other computations.
114<P>
115In the extraction method developed by Armel <B>LeBail</B>,
116reflection intensities are "optimized" by treating the setting the F<sub>calc</sub> value
117for each reflection to the F<sub>obs</sub> value extracted
118during the previous cycle.
119By iterating, the F<sub>calc</sub> values slowly converge to a
120set of reflection
121intensities that yields a best fit to the pattern.
122The F<sub>obs</sub> values are determined every time GENLES is run,
123or a least squares
124refinement cycle is run. This it is possible to improve the LeBail fit, by
125running GENLES with the "Number of Cycles" set to zero.
126<P>
127Note that due to reflection
128overlap, there are usually many different ways to apportion intensities with 
129fits of comparable quality, depending on what starting values are used for
130F<sub>obs</sub>. Any time POWPREF is run, the reflection list
131is regenerated and the first time that GENLES is run, the
132starting F<sub>calc</sub> values are set one of two ways:
133<P><DL>
134<a name="lebail">
135<DT><B>F(calc) weighted</B><DD>
136In a "F(calc) weighted" LeBail extraction the initial F<sub>calc</sub> values are computed
137from the crystal structure model. If the model is fairly close to being
138correct, it will likely apportion intensity for overlapping reflections in
139a manner that is fairly close to correct. Thus, the F<sub>calc</sub> values obtained
140from a "F(calc) weighted" LeBail extraction are about as good as can be
141done for the case where the structure is pretty close to correct.
142<DT><B>Equally weighted</B><DD>
143On the other hand, if one has no good structural model, but would like to
144use LeBail extraction as a way to obtain F<sub>obs</sub> values for use in structure
145solution, for example, by direct methods, then it is best to assume that all
146reflections are equally likely to contain intensity. In the "Equally weighted"
147mode, all reflections are given an identical F<sub>obs</sub> starting value. Thus, if
148two reflections are completely overlapped, in this methods, they will
149be assigned equal F<sub>obs</sub> values through the LeBail fit.
150</DL>
151<P>
152It is possible to refine unit cell, background, profile and other
153non-structural parameters at the same time as a LeBail extraction is
154performed. I often do this, for two reasons. One is that the final LeBail
155R<sub>wp</sub> and Chi<sup>2</sup> provides a better measure of the
156best possible fit
157than the statistical values, particularly if the material has non-ideal
158peak shapes or other factors that cannot modeled. The second reason is the
159LeBail fit provides excellent starting values for the unit cell, background
160and profile parameters, so these terms need not be refined again
161until all structural terms have been fit well.
162<P>
163These LeBail refinements, alas, are prone to diverge
164if the the extracted intensities are changing rapidly and
165other parameters, such as unit cell parameters are compensating. It is
166a good practice to run GENLES several times with
167the "Number of Cycles" set to zero each time POWPREF is run -- to allow the
168reflection intensities to converge before refining parameters.
169<P>
170Note that extraction different methods can be used for different phases
171in a histogram.
172It can be convenient to use LeBail extraction for an impurity phase,
173in the case where the impurity has preferred orientation or has a known
174unit cell, but an unknown structure. I have also used LeBail fits to
175obtain precise lattice constants via profile fits of materials where the
176exact structure is not known, so a Rietveld refinement cannot be performed.
177<DT><B>LeBail Damping</B><DD>
178The shifts to the reflection intensities can damped. This is useful when
179refining lattice constants or other terms that might otherwise cause the
180reflection intensities to shift dramatically and in turn cause the refinement
181to diverge.
182</DL></DL>
183<hr>
184<TABLE BORDER BGCOLOR="#FFFF40" ALIGN=RIGHT>
185<TR><TH><A  Href="expgui.html">EXPGUI top</A> 
186</TH><TH><A Href="expgui2.html">Next page</A>
187</TH></TR></TABLE>
188
189<a href="http://www.ncnr.nist.gov/staff/toby/">Brian Toby</a> (<a href="mailto:brian.toby@nist.gov">Brian.Toby@NIST.GOV</a>)
190<br>
191$Revision: 505 $ $Date: 2009-12-04 23:07:17 +0000 (Fri, 04 Dec 2009) $
192</blockquote>
193</body>
194</html>
Note: See TracBrowser for help on using the repository browser.