LSMO Blog

I attempted doing a long range fit (newstd_srinner, from 1.1 to 3.8 Ang) and using it to generate standards for the iterative background removal on bilayer3 0.4-c. The results in sig^2 did not change however, and the rs files looked worse. I believe the problem was that newstd_srinner was a poor fit to the data.

I used the following process as prescribed by Bud:

1.)Longrange fit (1.1-3.8 Ang)
2.)Make rsfit redo the calculation for k=0.5-15 and r=0-7, and generate rs files (ends in '.std' in rs)
3.)Back FFT the rs files to get chi(k)_theoretical (ff.std in ks)
4.)Self-absorption distort (using sabcor) so that the chi(E) oscillations are distorted in the same manner as the original es data (un-self-absorption corrected of course) (s.std in ks)
5.)Use rmchi to remove the uncorrected chi(k)_theoretical from the pre-edged es files (.bg in es)
6.)Use sureduce to fit the resulting backgrounds and save spline fits to them (s.bgfit in es)
7.)Use rmbg to remove the spline-fit bgs from the original data and transform the results into ks (sb in ks)
8.)Use sabcor to self-absorption correct these new ks files (b in ks)
9.)FFT the ks files (b in rs)

Also, I tried the same process, but with a step 3.5.) in which I added the backFFT of the high-r portion of the original rs file (4.0-7.0 ang) to the chi(k)_theoretical, to further remove hifrequency oscillations that have nothing to do with background. However this did not work for some reason; it seemed like the phase of the hi-frequency components might have been wrong. The files end in (hifreq or hifreqff)

Took data in July (mag25) and analyzed it for 0.4-ab
Determined the following:
-The three low-T points (before the transition) are bad because the cryostat window broke just before taking these data: there was no exchange gas to cool the sample. Who knows what temperature they actually correspond to.
-There appeared to be intermittent steps in the data, which were removed using 'manip'. 'manip' at this point does not put details about the step removal in the header. This should be changed in the future, but for now the files with an appended 's' are ones that were step-removed.
-The FFT-range of 2.5-11.5 (in reduce) seems to produce good fits. The files with an 'f' appended are these files
-The 096b files have too high of a sig^2, and the 102b files are too low. There seems to be a background problem with these sets...we are dealing with this by doubling the error bars on these points.
-We attemted to deal with the inconsistencies between bilayer3 and mag25 by assuming that mag25 had resolution problems; convoluted bilayer3 with 5eV and 10eV windows to see if this fixed things. It did not. These files have a 'c' or a 'c1' appended

Haven't posted for a while since finishing on the thesis and programming. Here's where I'm at now:
-Presently attempting to determine the correct S0^2 using the following method:
Looked back at the CaMnO3 from Mag2 and reanalyzed using k=3.5-12.7. We assume that this sample has almost no static disorder at low temperature and that the S0^2 that statically offsets \sigma^2(T) such it matches the zero-offset of the appropriate debye curve (~880K) is the correct S0^2...that is to say (1) fit \sigma^2(t) to a debye model with offset to obtain the debye temperature (2) use fitcD to determine the \sigma^2 at 50K (3) do a fit of the 50K data with constraint \sigma=\sigma_debye(50K) to obtain S0^2.

-Also testing FEFF8 standards with the SCF card on. Used Atoms 3.0.1 to generate a FEFF file:
______________________________________________________

* This feff8 input file was generated by Atoms 3.0.1
* Atoms written by and copyright (c) Bruce Ravel, 1998-2001

TITLE La(2-2x)Sr(1+2x)Mn2O7, (Mitchell, J.F. et al PRB 55, 63 1997)

* Mn K edge energy = 6539.0 eV
EDGE K
S02 1.0

* pot xsph fms paths genfmt ff2chi
CONTROL 0 0 0 0 0 1
PRINT 1 0 0 0 0 0
POLARIZATION 1 0 0
# SIG2 0.0025
*** ixc=0 means to use Hedin-Lundqvist
* ixc [ Vr Vi ]
EXCHANGE 0

CRITERIA 5.0 5.0

*** Radius of small cluster for
*** self-consistency calculation
*** A sphere including 2 shells is
*** a good choice
*** l_scf = 0 for a solid, 1 for a molecule
* r_scf [ l_scf n_scf ca ]
SCF 2.0

*** Upper limit of XANES calculation.
*** This *must* be uncommented to
*** make Feff calculate full multiple
*** scattering rather than a path expansion
* kmax [ delta_k delta_e ]
# XANES 4.0

*** Radius of cluster for Full Multiple
*** Scattering calculation
*** l_fms = 0 for a solid, 1 for a molecule
* r_fms l_fms
#FMS 4.25472 0

*** Energy grid over which to calculate
*** DOS functions
* emin emax eimag
LDOS -30 20 0.1

*** for EXAFS: RMAX 7.0 and uncomment
*** the EXAFS card
RPATH 0.1
EXAFS 20
RMAX 7.0

POTENTIALS
* ipot Z element l_scmt l_fms stoichiometry
0 25 Mn 2 2 0.001
1 57 La 3 3 6
2 25 Mn 2 2 4
3 8 O 1 1 14

and so on...

This makes a difference in fitting; particularly at low-r it matches better with the data. Better fits overall

I have discovered that the three high points in mag22 are due to a step in the data. I wrote a program called 'manip' to shift, rescale, flatten, and subtract data files relative to a reference file. This allowed me to look at the difference data to see where the step laid:




The program was also tailored to fit a constant to the step and shift the data to attmempt to remove the step. This was done on files 185, 203, 204, 205, 206, and 207. They were reglitched and rereduced, and here are the results (from newstd_rcstr.42):




The data almost fit in with Bilayer now, but it seems that a few points may still need deglitching.

I've refit mag22 and bilayer 3 over the k=4.0-11.5 range, and adjusted the s02's so that the plots line up. The rogue point seems to be a problem with the beam changing (the lab book notes that just before those scans, the glitch profile changed, which means that somehow the beam-angle changed or something like that). Here are the parameters of my fits:

mag22:
KRANGE 4.0 11.5 0.3 1
RRANGE 1.1 1.7

MAXITER 8000

E0SAME
CONSTRAINT E1 E1 E1 0 0 -1.47
CONSTRAINT A1 A1 A1 0 0 0.720
CONSTRAINT A2 A1 A1 1 0 0
CONSTRAINT S2 S1 S1 1.00 0 0
#LIMITS 10 0.00 10 10 10 10
output: newstd_rcstr.31
bilayer3:
KRANGE 4.0 11.5 0.3 1
RRANGE 1.1 1.7

ALLPATHSOUT
MAXITER 8000

E0SAME
CONSTRAINT E1 E1 E1 0 0 1.88
CONSTRAINT A1 A1 A1 0 0 0.790
CONSTRAINT A2 A1 A1 1 0 0
CONSTRAINT S2 S1 S1 1.00 0 0
#LIMITS 0.1 0.00 .5 10 0 0
output: newstd_rcstr.53


The following plots were generated using pproc, with the following inparam.dat file in mag22:
BASEFILENAME = newstd_rcstr
EQUIVFILE = mag22.equiv
NUMOFSTANDARDS = 2
POSPLOTADDENDUM = "diffdata.dat" using 1:2, "diffdata.dat" using 1:3, "tpe-bil3_newstd_rcstr.53_col1.dat" w e, "tpe-bil3_newstd_rcstr.53_col2.dat" w e
SIGPLOTADDENDUM = "ts2e-bil3_newstd_rcstr.53_col1.dat" w e

I'm stumped with this anomalous point...I believe I can see the portion of the curve in k-space that causes the increased amplitude in r-space, but I'm not so sure if its a glitch; if it is a glitch, then its a very broad glitch. From what I see, a negative 'glitch' drew the lower part of an oscillation down and the corresponding positive glitch drew the adjacent upper part of the oscillation upward. To fix this anomaly, I've judged that I would have to replace about 10 points on the lower curve and about 10 points on the upper curve. It feels to me like this would cross the line of data modification to change this many points. Should this point be thrown out?

For now I'm going to move on to the AB plane.

Now I should go back and look at mag22 0.4-ab and add it in to bilayer3/bilayer1 to see if there's a transition there too. That should be enough for the thesis.

I have reanalyzed mag22, and with the exception of one point (220K), the data now fits in with bilayer3. I used s0^2=0.8 for bil3 and s0^2=0.73 for mag22. I fit using the single peak fit, and subtracted the sigma^2 due to the long-short bond distribution from the data using the previously mentioned method.
Also, I changed my k-range to k=4.0 to 11.5, and allowed y2 a larger range to search for eMin. This made a huge difference; it appears that the background subtraction isn't great for low-k, but above 4 seems to be OK.

Here is a plot!