From: Robin Shirley <R.Shirley@surrey.ac.uk> 
To: uppw@cristal.org 
Date: Mon, 10 Nov 2003 13:59:48 -0000 
Subject: UPPW-4 solution 

Dear Armel


** UPPW-4 Solution **


Again not a difficult problem, which proves to belong to the expected 
STF zeolite family and can be solved by judicious refinement starting 
from the published STF cell.  Even if it could not, all six automatic 
indexing programs that were tried solved it easily at their first attempt.


I am very confident that the resulting cell is correct, and would expect 
that, given profile data, it would not be difficult to proceed to a 
Rietveld refinement using the published STF structure as starting point.


The following C2/m cell, with standard deviations in brackets, was refined 
against 2Theta values using Celref (within Chekcell) and includes a 2Theta 
zero correction:  


  a = 13.9656 (0.0068) A
  b = 18.3743 (0.0107) A
  c =  7.4377 (0.0034) A
  beta = 99.400 (0.004) degrees     giving V = 1882.96 (0.15) A(3)
  
  2Theta zero correction (to be added) = +0.0080 (0.0067) degrees


This cell indexes all 25 observed lines using only 35 calculated lines, 
with figures of merit (including the C-lattice absences):


  M1(20) = 43.9,  M1(25) = 35.8.


Although the 2Theta zero correction is small, omitting it has a significant 
worsening effect on the standard deviations reported by Celref, in particular 
increasing the esd of beta from 0.004 to 0.12 degrees.  


No significant improvement was found by treating the correction as due to 
specimen displacement.  



** Discussion **


A quick (5 to 10 minutes) session with Crysfire showed that many indexing 
programs would solve this dataset very easily using their Crysfire defaults 
(1V and 2V refers to volume multiples of the correct C2/m cell:  
 ITO12      1V cell, "M20" = 47.2 (C-centred)
 FJZN6      1V cell, "M20" = 46.6 (C-centred) (twice)
 TREOR90    1V cell, "M20" = 20, "M25" = 17. F25 = 45 (0.0084, 67)
 KOHL (TMO) 1V cell, "M20" = 46.6 (C-centred), plus two 2V cells
 DICVOL91   1V cell, "M20" = 22.9 plus two equivalent 1V settings
 LZON       1V cell, "M20" = 46.5 (C-centred) (3 times) plus two 2V cells


With such a clear indication there did not seem any need to run further 
programs such as TAUP, LOSH, HMAP and McMaille, though I have little doubt 
that they would also have succeeded.  


The Crysfire version actually used was Crysfire 2003 (32-bit), but 
Crysfire 2002 would have produced exactly the same outcomes.


Since the resulting cell is quite close to the published STF type cell, and 
clearly belongs to the same family, the question arises as to why the 
original author did not have sufficient confidence to propose an indexing.   


Perhaps the reason that a cell was not reported is that, although UPPW-4 
is clearly at least closely related to the type-structure given in the 
zeolite database, it is not quite straightforward to proceed directly 
to its cell simply by least-squares refinement, taking the published STF 
cell as starting point.


Thus attempts to refine directly from the STF cell in the database 
(C2/m; 14.104, 18.205, 7.477; 98.989; V=1896.25) using all 25 lines, 
at least with Celref, stick in a local minimum in which 19 lines 
fit reasonably well (if not all to the expected accuracy of the data).


However, although apparently not far from the correct minimum, the 
remaining 6 lines (lines 4, 10, 12, 14, 19 & 22) remain unindexed 
and even look a little as if they might come from a second phase.


This local minimum can be escaped by starting the refinement at lower
resolution by using only the first 8 observed lines.  The STF cell then
converges with Celref to what is effectively the correct cell in just 
one cycle:


       a = 13.978,   b = 18.391,   c = 7.444 A,   beta = 99.44 deg


with the following shifts:
 
       -0.12 (-9%)   +0.19 (10%)   -0.03 (-0.4%)         -0.55°


Introducing the rest of the data and including a 2Theta zero correction
then gives the fully refined C2/m cell, as reported above.


Robin Shirley