McMaille – Sous le Capot(Under the Bonnet)

A. Le BailUniversité du Maine

Laboratoire des Oxydes et Fluorures

CNRS – UMR 6010FRANCE

alb@cristal.orghttp://cristal.org/

Content

Overview of functionality of the McMaille (pronounce Mac My)(what does it do?)

Features that differentiate the program (why/when should I use it?)

Summary of implementation (language, platform, gui/cmd line - how/where can I use/get it?)

Description of 2-3 favorite algorithm/implementation tricks (what is under the bonnet?)

No way to escape theindexing

bottleneck in the SDPD

maze

(figure extracted from the IUCr monograph 13)

SDPD =Structure

Determination by Powder

Diffractometry

Indexing powder diffraction patterns.

Using not only peak positions, also intensities.

Applying the Monte Carlo method in order to generate randomly cell parameters tested against an idealized powder profile.

What does it do ?

The columnar peak shapes used by McMaille v3

The width in the automated mode is calculated as :0.3 x / 1.54056

it depends on the user in manual mode

« Simplicity » of the Monte Carlo algorithm in McMaille

Features that differentiate the program (why/when should I use it?)

But… slow… Up to 1-15 mn in « black box mode ».

Up to 1-15 hours in manual mode.(depending if monoclinic and triclinic are examined)

To be used if TREOR, DICVOL (etc) fail, or if you have time or if you love it…

Good for providing near-complete lists of « solutions » ordered by

R factor (FoM)Cell volume

Most frequently found cellsLargest number of peak indexed

Summary of implementation

Language : old good FORTRAN 77

Compilers : Compaq Visual Fortran, Intel Visual Fortran, G77

Platforms : Windows (XP, etc), Linux

will benefit of dual-core and future multi-core processors(once the code will be paralellized : project by using the Intel

Visual Fortran compiler)

Gui/CMD line : CMD line

Typical « DOS box » with CMD line :

You just give the filenameand can stop by typing

K (capital letter),saving all results.

Where can I get it ?

Program download (GNU Public Licence) :

http://www.cristal.org/McMaille/

Everything is there(executable, Fortran code, manual, examples)

How can I use it ?

The first recommended approach with McMaille is to use the quite simple automated « black box » mode.

Peak positions and intensities can be extracted by using the WinPlotr program which is able to build the McMaille entry

data file directly for the automated mode.

If the peak positions are extracted by WinPLOTR, an entry file is directly prepared for McMaille for an

automated run.

Y2O3! Wavelength, zeropoint and NGRID (NGRID=3: black box mode)1.540560 0.000 3! List of 2theta positions, intensity (min.=20) 20.50481 1411.71155 29.15784 11198.95700 33.79123 2632.66553 35.91046 531.02966 37.92135 135.21042 39.84880 606.23480 41.70092 118.49068 43.48877 881.15845 46.88918 271.40564 48.52835 4213.47021 50.12016 230.08562 51.68559 66.62399 53.20726 506.43854 54.69715 117.27492 56.16987 404.76215 57.61465 2603.39868 59.03540 563.80334 60.43375 471.21359 61.81146 194.23712 Etc

Typical file created by

WinPLOTRfor

McMaille

Description of 2-3 favorite algorithm/implementation tricks

(what is under the bonnet?)

Simplicity of use in black box mode

(contains an internal expert system presenting ordered lists of possibilities according to several criteria)

Is a kind of Swiss knife for indexing :

You may look deeply at selected possibilities by using the alternative grid search mode (if ambiguity on one parameter, etc).

You can try multi-pattern indexing.

Final plot produced by McMaille, displayed by WinPLOTR :

Other software compatible with McMaille outputs : CHEKCELL, CRYSFIRE

with Gaussian peak shapes and width / 2.

Compatibility with Chekcell

Relative insensitivity to impurity

The user decides by two control parameters in manual mode:

N’ : number of unindexed lines (= 3 in automated mode).

R2 : consider only proposals with R < R2. Fixing it at 15% means that cell proposals explaining at least 85% of the peaks total intensity will be listed.

An impurity should not concern more than 15% of the total intensity, right ? But the number of (small) peaks belonging to the

impurity can be high…

Indexing a 2-phases powder pattern with McMaille

Preliminary conclusions about two-phases indexing with McMaille

Provided at least 30 lines are examined with 13-17 lines belonging to each phases, and 40-60% of the total intensity

distributed to each phase, then :

McMaille appears to be able to produce solutions in reasonable times (<1 hour) for combinations of two phases either cubic or

hexagonal or tetragonal or orthorhombic.

Monoclinic and triclinic not examined (too long).

Example of 2-phases indexing (20 minutes) :One mixture of a tetragonal with an orthorhombic phaseMcMaille combines the cell proposals by couples and detects the best

combinations indexing the largest number of peaks :

Last publication from Robin Shirley et al. :

« Renewed interest in powder diffraction data indexing », J. Bergmann, A. Le Bail, R. Shirley and V. Zlokazov, Z. Kristallogr. 219 (2004) 783-790.

From the indexing benchmarks :Do not limit yourself to only one indexing software…

Conclusions about McMaille

- Promising (?) method (please verify)…

- Already quite efficient if you have time and a fast computer.

- Needs some skills in manual mode, but nothing to do in « black box »  mode (except finding the zeropoint).

- Improve it if you have some ideas (GNU Public Licence).

- Completely free access.

- Use cautiously the 2-phases mode…