Le Bail extraction and continue_after_convergence

gabortel

The Le Bail extraction with continue_after_convergence seems to output the parameters from the last iteration, not from the best cycle.

To reproduce the case, take the following original example file:
C:\Topas7\test_examples\peak-intensity-extraction\zhu3lebail.inp
This works as intended.

However, if I add 'continue_after_convergence' and 'num_cycles 3' and comment out the '#define USE_PAWLEY' line, and run it, the end of the screen output looks like this:

...
390  Time   1.00  Rwp    5.104   -0.002 MC     0.00 1
391  Time   1.00  Rwp    5.102   -0.002 MC     0.00 1
392  Time   1.00  Rwp    5.100   -0.002 MC     0.00 1
393  Time   1.00  Rwp    5.097   -0.003 MC     0.00 1
394  Time   1.00  Rwp    5.097   -0.001 MC     0.00 1

Cycle:     2  T:  0.70  Ti:   1  Best/Global :     5.03848      5.03848
395  Time   1.01  Rwp   70.223   65.126 MC     0.00 0
--- 1.006 seconds ---
396  Time   1.01  Rwp   70.390    0.167 MC     0.00 0
File C:\Topas7\test_examples\peak-intensity-extraction\test.out updated
    with parameters corresponding to best Rwp

But the out file contains 'r_wp 70.3896324' that is obviously wrong and contradicts to the displayed best/global values. Also, the extracted intensities in the out file are constant 1's.
I am using tc.exe 'TOPAS-64 Version 7.16 (c) 1992-2020 Alan A. Coelho'.

Do I overlook something? Any suggestions? Thanks.

alancoelho

I just included the following:

	continue_after_convergence 
	num_cycles 3

and I get very different results with even the Rwp not being the same; ie.

0  Time   0.46  Rwp    4.869   -0.001 MC     0.00 1
 31  Time   0.47  Rwp    4.869   -0.000 MC     0.00 1

Cycle:     2  T:  0.70  Ti:   1  Best/Global :     4.86898      4.86898
 32  Time   0.47  Rwp   37.856   32.987 MC     0.00 0
--- 0.475 seconds ---
 33  Time   0.48  Rwp    4.869  -32.987 MC     0.00 0
File C:\w\test_examples\peak-intensity-extraction\zhu3lebail.out updated
    with parameters corresponding to best Rwp

Also, use Version 7.17; ie. download the file http://www.topas-academic.net/ta7.en to your main TA directory and then run TA-SETUP.BAT. If you still have probleme then sent me the INP file you are using and I will have a closer look.

gabortel

Did you also comment out the line to switch to Le Bail?

'#define USE_PAWLEY '  Comment this line out for Lebail

Using Pawley method, I get exactly the same results as you, and the .out file is OK.
The problem exists in the output file only when using Le Bail.
So, this is the complete .inp file:

continue_after_convergence 
num_cycles 3 
/*
   Pawley / Le Bail examples

   Example use and no use of
      lebail

   Note that Pawley refinement gets to lower Rwps mainly due to
   the CCG routine. Le Bail refinement has better chance of getting
   to low Rwps if the peak shapes, lattice parameters and back ground was
   approximately defined. Note that the TOPAS implementation of Le Bail
   refinment rarely diverges.
*/
'------------------------------------
'#define USE_PAWLEY '  Comment this line out for Lebail
'------------------------------------
r_wp  8.868
'------------------------------------
XDD(zhu3)
   CuKa2(0.001)
   Radius(200.5)
   LP_Factor(27)

   Full_Axial_Model(12, 20, 12, 5.2, 5.2)
   Divergence(2)
   SW(@, 1)

   ZE(@, 0)
   One_on_X(@, 1000)
   bkg  @ 0 0 0 0

   finish_X 70
   hkl_Is
      #ifdef !USE_PAWLEY
         lebail 1
      #endif
      space_group p1211
      a    @  7.676
      b    @  9.641
      c    @  7.089
      be   @ 106.22
      Strain_L(@, 0.1)
'------------------------------------

On the program versions. I did the update with that ta7.en file early in February without any problem. But there seems to be a discrepancy between displayed command line and gui versions:
tc.exe when run displays: TOPAS-64 Version 7.16 (c) 1992-2020 Alan A. Coelho
ta.exe gui window in the title bar displays: TOPAS-Academic-64 V7.17
Also the main Topas directory contains ver.txt that file contains 17.
So I think we use the same version. To double check, my tc.exe is 5,371,392 bytes and dated 2022-01-24 00:48, if that matters.

Thanks.

alancoelho

Ok, I get the same result if I used the Le Bail method.

I will put this in the to-look-at bin.

For now I suggest not to use the Le Bail with continue_after_convergence unless you can convince me that there's a need.

cheers
alan

gabortel

I can run Topas repeatedly externally, until you fix it.
Thanks.