International
Tables for
Crystallography
Volume G
Definition and exchange of crystallographic data
Edited by S. R. Hall and B. McMahon

International Tables for Crystallography (2006). Vol. G, ch. 3.3, pp. 128-129

Section 3.3.9. Use of pdCIF for Rietveld refinement results

B. H. Tobya*

aNIST Center for Neutron Research, National Institute of Standards and Technology, Gaithersburg, Maryland 20899-8562, USA
Correspondence e-mail: brian.toby@nist.gov

3.3.9. Use of pdCIF for Rietveld refinement results

| top | pdf |

One of the major aims of the development of the pdCIF definitions was to be able to communicate the results of Rietveld refinements, and this is expected to be the most common use for pdCIF. To aid the development of software that prepares pdCIF output from Rietveld refinements, this section describes the blocks and loops to be found in a pdCIF, noting variations due to the type of Rietveld refinement. Programmers may also wish to look at the GSAS2CIF program, which creates CIFs for a wide range of types of diffraction data, and for multiple data sets and phases (Toby et al., 2003[link]).

It is valuable for the CIF to contain the structural model(s), the observed powder-diffraction intensities and the calculated powder-diffraction intensities so that the fit of the model to the observed diffraction pattern can be viewed graphically. It is the present author's firm belief that it is impossible to judge the quality of a Rietveld refinement by R factors or any other numerical metric, since these values describe not just how well the structural model fits the measurements, but also how well the background and peak shape are fitted as well. Very poor models can have good R factors and [\chi^2] values if there is a significant amount of non-Bragg scattering that has been well fitted. On the other hand, with high-resolution observations measured to excellent precision, even trivial imperfections in the peak shapes can result in poor agreement factors. There is no substitute for the visual examination of a plot of the observed and calculated patterns, optimally at more than one magnification level. The program pdCIFplot (Toby, 2003[link]) plots the observed and calculated powder-diffraction intensities in a pdCIF and allows the fit to be examined in more detail than can be provided by a figure showing the whole profile at once.

3.3.9.1. A single phase

| top | pdf |

When a single set of diffraction measurements is used to model a single phase, a pdCIF will usually contain only one block. There will be several important loops present.

One loop will contain atomic parameters, such as coordinates. The unit cell must also be specified.

A second loop will contain the reflection table.

A third loop will contain the observed (or processed) diffraction measurements and the simulated pattern. Other items that should be included in this loop are the least-squares weights (usually σ−2, where σ is the standard uncertainty) so that it is possible to determine the quality of the fit in individual regions. Weight values of zero can also be used to indicate that data points have been excluded from the refinement. Since background fitting is quite important in Rietveld analysis, it is also valuable to include the background values. Thus, this loop should specify:

(i) the ordinate of the Rietveld plot, using one or more of: _pd_meas_2theta_scan, _pd_meas_time_of_flight, _pd_proc_2theta_corrected, _pd_proc_d_spacing or _pd_proc_recip_len_Q; alternatively the ordinate can be specified using either _pd_meas_2theta_range_* or _pd_proc_2theta_range_*, where _* is _min, _max and _inc outside the loop.

It is recommended that all CIFs describing the results of a Rietveld refinement include either _pd_proc_d_spacing or _pd_proc_recip_len_Q.

(ii) The observed (or processed) intensity values, using the items _pd_meas_counts_total, _pd_meas_intensity_total, _pd_proc_intensity_total or _pd_proc_intensity_net.

(iii) The background, using the item _pd_proc_intensity_bkg_calc.

(iv) The least-squares weights, using the item _pd_proc_ls_weight. If these weights are not specified, then it must be presumed that all points have been used in the refinement and that the weights are the reciprocal of the intensity values (if _pd_meas_counts_total was used) or the reciprocal of the intensity standard uncertainties, if specified.

(v) The calculated pattern should appear using either _pd_calc_intensity_net or _pd_calc_intensity_total.

It is good practice always to include at least one data item from each entry in the list above.

Apart from the information contained in these loops, information from almost all sections of the pdCIF dictionary can be valuable. Such items include data items that define how the diffraction measurements were made, how the sample was prepared and characterized, how the refinement was performed, and least-squares parameters and R factors. A template and an example pdCIF showing the combined use of pdCIF and core data items form part of the Acta Crystallographica instructions for authors at http://journals.iucr.org/services/cif/powder.html .

3.3.9.2. Multiple phases

| top | pdf |

When more than one phase is present, multiple CIF blocks are needed. The resulting CIF will contain much the same information as would be found in a single-phase pdCIF, as described in Section 3.3.9.1[link]. However, there will be a separate block for each phase containing information specific to that phase, such as the unit cell and the loop containing the atomic parameters.

The CIF will usually (see Section 3.3.7[link]) contain one additional block with the observed and calculated pattern and a reflection table, as well as the other data items that define how the diffraction measurements were made, how the refinement was performed etc. While reflection tables for each phase can be placed in each phase block, it is better to include a single reflection table in the block that contains the diffraction data. This block will also contain a phase table that uses the block pointer _pd_block_diffractogram_id to link to the phase blocks. The phase blocks can also be linked to the data block using the block pointer _pd_phase_block_id. For most Rietveld refinements, each phase is allowed to have different profile parameters, so _pd_proc_ls_profile_function should also be included in the phase-table loop.

3.3.9.3. One phase, multiple sets of measurements

| top | pdf |

It is fairly common to use more than one diffraction data set to determine a model for a single phase. Some examples include: combined refinement using both neutron and X-ray powder diffraction; use of multiple X-ray wavelengths to make use of anomalous dispersion; and the use of single-crystal X-ray and powder neutron diffraction data in a single refinement. For these cases, there will be a CIF block for each data set. Each of these blocks will contain a reflection table and a loop with the observed and calculated diffraction intensities, as described in Section 3.3.9.1[link].

As explained in Section 3.3.7[link], the resulting structural parameters could be placed in a block with one of the sets of diffraction data. However, it is better practice to create one additional block for these parameters, as it then becomes clear that the result is from a combined refinement. This is indicated by linking the phase block and the data-set blocks using a loop of _pd_block_diffractogram_id values in the phase block. The data-set blocks can also have a link to the phase information using the block pointer _pd_phase_block_id.

3.3.9.4. Multiple sets of measurements and phases

| top | pdf |

Multiple data sets may be used for mixtures as well as single phases. This is becoming increasingly common as more complex materials are studied using powder diffraction. The treatment of this case follows logically from that of Sections 3.3.9.2[link] and 3.3.9.3[link]. If there are M diffraction data sets and P phases, there will be P blocks containing the crystallographic parameters for each phase. There will be M blocks with the observed and calculated diffraction intensities, as well as reflection tables. Depending on the Rietveld software, there may be M × P sets of some parameters, for example phase fractions and profile descriptions. These parameters may be placed in the phase table loop within the data-set block(s).

Ideally, the same specimen, or at least the same sample, will be used for all measurements. Sometimes, however, different samples are used for combined refinements to extend the number of observations, despite the possibility that the samples might have slightly different structures or compositions. If there are S samples, there will be an additional S blocks that record the sample and specimen preparation and characterization information. Thus, in this case there will be a total of M + P + S blocks.

As before, the phase blocks will use the block pointers _pd_block_diffractogram_id to link to the data-set blocks. Likewise, the data-set blocks will have phase tables with _pd_phase_block_id values that link to the phase blocks. The sample blocks can use both _pd_block_diffractogram_id values and _pd_phase_block_id values to link to the the diffraction data and the analysis results. This is shown in the CIF in Example 3.3.7.1[link]. The program GSAS2CIF (Toby et al., 2003[link]) can create CIFs for multiple sets of measurements and phases.

References

Toby, B. H. (2003). CIF applications. XII. Inspecting Rietveld fits from pdCIF: pdCIFplot. J. Appl. Cryst. 36, 1285–1287.
Toby, B. H., Von Dreele, R. B. & Larson, A. C. (2003). CIF applications. XIV. Reporting of Rietveld results using pdCIF: GSAS2CIF. J. Appl. Cryst. 36, 1290–1294.








































to end of page
to top of page