Diffraction pattern simulation

rowlesmr

I want to simulate some diffraction patterns for given weight fractions, so I'm using Known_Weight_Percent. I've got several structures with this macro, with the last structure having a scale factor that can refine.

The diffraction data I'm putting in consists of just 1s over a range of 5-90 °2T, and I'm only interested in the model output. The only refineable parameter is the scale factor of one structure.

With one set of known weight percents, and allowing the last scale factor to refine, Topas does one iteration, and gives me back the weight percents that I asked it to give. The refined scale factor doesn't move. I get a maximum peak intensity depending on the scale factor I choose, but I choose it to be ~10000.

If I use the same weight percent values and fix the last scale factor I get wrong values.

If I use another set of weight fractions, and allow the last scale factor to refine, then Topas does about 50 iterations, and refines the scale factor such that the peak intensity is about 1.

What is going on? Why is the behaviour so different with the second set of weight percents?

.

Input file:


#define IT_WORKS


#ifdef IT_WORKS
	macro NaAlH4_wt  { 57.2  }
	macro NaH_wt     {  0.21 }
	macro Na3AlH6_wt { 27.70 } 
	'Al should be ~14.9
#else	 'IT_DOESNT_WORK
	macro NaAlH4_wt  	{  1.71 }
	macro NaH_wt     	{ 29.22 }
	macro Na3AlH6_wt	{ 23.52 } 
	'Al should be ~45.55
#endif


xdd test.xy
	x_calculation_step = Yobs_dx_at(Xo); convolution_step 4
	bkg 1
	LP_Factor(0)
	CuKa1(0.0001)
	str 
		phase_name "Al"
		scale @  0.0035`
		CS_L(, 100)
		MVW( 107.926, 66.430, 14.911`)
		space_group Fm-3m
		Cubic( 4.050)
		site Al1 num_posns  4 x  0 y  0 z  0 occ Al  1 beq  0.6		
		
	str 
		phase_name "NaAlH4"
		Known_Weight_Percent(NaAlH4_wt)
		MVW( 216.012, 286.252, 57.179`)
		CS_L(, 100)
		space_group I41/a
		Tetragonal( 5.022, 11.35)                                        
		site Al1 num_posns  4 x  0 		y  0 		z  0 			occ Al+3  1 beq  1.78
		site Na1 num_posns  4 x  0 		y  0 		z  0.5 		occ Na+1  1 beq  3.20
		site H1 num_posns  16 x  0.228 y  0.117 z  0.838 	occ H-1   1 beq  2.45
	
	str 
		phase_name "NaH"
		Known_Weight_Percent(NaH_wt)
		MVW( 95.991, 116.930, 0.210`)
		CS_L(, 100)
		space_group Fm-3m
		Cubic( 4.89)
		site Na1 num_posns  4 x  0 		y  0 		z  0 		occ Na+1  1 beq  1
		site H1  num_posns  4 x  0.5 	y  0.5 	z  0.5 	occ H-1   1 beq  1	

	str 
		phase_name "Na3AlH6"
		Known_Weight_Percent(Na3AlH6_wt)
		MVW( 203.997, 232.309, 27.700`)
		CS_L(, 100)
		space_group P21/n
		be  89.86
		a  5.412
		b  5.533
		c  7.758
		site Al1 num_posns  2 x  0 			y  0 			z  0 			occ Al+3  1 beq  1.16
		site Na1 num_posns  2 x  0 			y  0 			z  0.5 		occ Na+1  1 beq  2.49
		site Na2 num_posns  4 x  -0.006 	y  0.461 	z  0.252 	occ Na+1  1 beq  2.49
		site H1  num_posns  4 x  0.091 	y  0.041 	z  0.2153 	occ H-1   1 beq  3.1
		site H2  num_posns  4 x  0.234 	y  0.328 	z  0.544 	occ H-1   1 beq  3.1
		site H3  num_posns  4 x  0.165 	y  0.2671 	z  0.944 	occ H-1   1 beq  3.1

rowlesmr

If I replace the Known_Weight_Percent macro with the procedure outlined in http://topas.dur.ac.uk/topaswiki/doku.php?id=refining_setting_weight_percents_directly for Version 4.1, then it works when I fix the scale factor for the phase that I don't set a weight fraction for.

alancoelho

Known_Weight_Percent looks like:

macro Known_Weight_Percent(& w)
{
scale = (w/(100-w)) Get(sum_smvs_minus_this) / (Get(cell_mass) Get(cell_volume));
}

It changes 'scale' and is a function of all 'scale's except the one it is defined in. Thus Known_Weight_Percent should only be defined for one phase. In the INP file it is defined in three out of the four phases.

rowlesmr

Thanks Alan.

Is there any way to set the wt% of multiple phases, as in the 4.1 example, but using v5 keywords? http://topas.dur.ac.uk/topaswiki/doku.php?id=refining_setting_weight_percents_directly#version_41

(I've got up to 7 phases to simulate)

alancoelho

Hi Matthew

The way it is written up at http://topas.dur.ac.uk/topaswiki/doku.php?id=refining_setting_weight_percents_directly for V4.1 is the best I can think of. You would need of course to extend it to 7 phases.

cheers
alan

rowlesmr

Thanks Alan.

rowlesmr

I put together a quick and dirty macro to help out:

http://topas.dur.ac.uk/topaswiki/doku.php?id=refine_weight_percentage

alancoelho

Looks great Matthew

The #m_if should work. Using #prm and #if definitely works and a slight reworking would look like:

macro wt_dets(n_,p_)
{
prm m##n_ = Get(cell_mass) Get(cell_volume);
prm x##n_ = w##n_ / (m##n_ (100 - w##n_));

#prm p = p_;
#prm n = n_;

#if p == 2; 'there are 2 phases in total
#if n == 1;
'do nothing, as s1 should already be defined elsewhere
#elseif n == 2;
prm s##n_ = x2*s1*m1;
#endif
#endif
etc...
}

Note #prm's can be redefined. Also the double quotation for comments should be replaced with a single quotation.

If you need any help with it then let me know.

cheers
alan

johnsoevans

Matthew,

This looks super useful. Macro of the year award!

John

rowlesmr

Thanks y'all!

I'll see if I can have another look at it next week.

rowlesmr

Thanks y'all!

I'll see if I can have another look at it next week.

mfisch

Hi Matthew

It is possible that I missed your exact goal, but there may be a much simpler way for multiple-phase pattern simulation.

Just get rid of the Rietveld normalization to 100 wt% by using the Raven and O'Connor external standard approach. This will also allow you to get a result close to a real measurement (without including effects as fluorescence, etc.).

(1) Determine K and the MAC from another sample/standard
(2) You know wt.-% (set by you), Z, M, V of each phase and the MAC of the final mixture (given by wt.-% set by you)
-> Calculate the required scale factors of each phase.

If you begin the inp file with

yobs_eqn = 0;
min 5
max 60
del 0.02

there is even no need for real data - and in the end not a single parameter to refine.

rowlesmr

Hi Martin

I just want to simulate a series of diffraction patterns given some wt% inputs.

That does seem to work, but the difficulty in setting it up is that you need to calculate the mixture LAC (from the individual phase LACs, not from the mixture LAC macro) and I don't know how to automatically to add up a series of prms without explicitly writing it down.

BTW, you don't need data for mine either.

edit: Just figured it out: existing_prm is a good thing! You just need to make sure that you put all of the macros in the correct order, or Topas doesn't like it.

Now I just need to figure out why my mixture LAC is a different value to Topas' mixture LAC.

mfisch

rowlesmr wrote That does seem to work, but the difficulty in setting it up is that you need to calculate the mixture LAC (from the individual phase LACs, not from the mixture LAC macro) and I don't know how to automatically to add up a series of prms without explicitly writing it down.

I did that in Excel.

rowlesmr wrote Now I just need to figure out why my mixture LAC is a different value to Topas' mixture LAC.

Are the Topas values larger than if you use those from the International Tables?
I don't remember exactly - it's some years ago - but I think there was a bug in LAC and/or MAC calculation in Topas 4.
Something was wrong with the interpolation. Maybe Alan remembers...

alancoelho

Hi Matthew, Martin

Yes, V4 did a linear interpolation on the data in the MAC directory. V5 uses additional data (some more data points) plus it uses a spline. Here are some results for Mixture_LAC_1_on_cm:

V4
Mixture_LAC_1_on_cm( 588.21315`)
mixture_density_g_on_cm3 5.86705118`
prm = Get(mixture_density_g_on_cm3); : 5.86705`

V4 with data from the MAC directory of V5
Mixture_LAC_1_on_cm( 605.99018`)
mixture_density_g_on_cm3 5.86705118`
prm = Get(mixture_density_g_on_cm3); : 5.86705`

V5
Mixture_LAC_1_on_cm( 609.01249`)
mixture_density_g_on_cm3 5.86696826`
prm = Get(mixture_density_g_on_cm3); : 5.86697`

Thus more accurate values can be obtained by V4 by using the MAC directory data from V5. I can e-mail the V5 MAC data to anyone that wants it.

cheers
alan