[Ifeffit] Basic questions

Scott Calvin SCalvin at slc.edu
Mon Apr 6 15:25:37 CDT 2009

Hi Andy,

Just a few notes I'd like to add to Carlo's helpful answer:

On Apr 6, 2009, at 1:27 PM, Carlo Segre wrote:

> Hi Andy:
> On Mon, 6 Apr 2009, Andy Connelly wrote:
>> 2. If I want to refine the degeneracy of a path can I write the
>> expression "amp*N_1" in the SO2 box? (where the N box is set to 1)  
>> and
>> then refine "amp" and/or "N_1"? where N_1 equivalent to the  
>> degeneracy
>> and amp is effectively SO2. Clearly they will be highly correlated,
>> but does a formula like this cause IFEFFIT any problems?
> This expression is OK, but you can only refine one or the other.   
> Refining
> both has no meaning.  You might as well just have a single parameter  
> then.

Unless, of course, you manage to get more than one shell of data. Even  
for glasses, that is often possible.

Another possibility is to collect data on a related crystalline  
material, and corefine the crystalline material, requiring a single  
S02 for each. (The crystal, presumably would have a known value of N.)

>> 3. If, for initial fits, I am only interested in the first shell
>> should I set R-range max to a lower value to isolate that shell and
>> then increase it as I look at shells at longer distances?
> Yes, it is a good idea to start with a limited range and then move up.
> One additional positive aspect of this is that you get a better idea  
> of
> the number of degrees of freedom you have for that shell.  If you make
> Rmax too big and you oly fit the first part of the range, you will
> artificially have too many degrees of freedom.

Yes, and don't forget that it is not sufficient to choose an rmax just  
below the paths you are omitting. This is for two reasons: paths have  
some width in R-space (as you can see from Artemis if you plot the  
path) and the center of each path is shifted down from Reff. To be on  
the safe side, I usually try to stay a half angstrom or so below the  
first important path that I am not including in my fit. For example,  
if the nearest neighbors are at 2.0 angstroms, and the next nearest  
neighbors at 2.4, but I'm not including the next nearest-neighbor,  
then I might start with rmax at 1.9 angstroms. That sounds strange,  
because it sounds like I'm not even including the nearest neighbor,  
but it too will be shifted down and have width, so I actually am  
including most of it.


--Scott Calvin
Sarah Lawrence College

More information about the Ifeffit mailing list