Peak indexing and Lattice Parameter refinement

Dr Nik Reeves-McLaren, Prof Neil C. Hyatt

Once we have identified what chemical phases are present within a sample, we may wish to
characterise these phases a little further. Can we assign Miller indices to each observed
Bragg reflection? What is the symmetry and size of the unit cell? This short primer is
intended to help people new to X-ray diffraction with such tricky subjects. Throughout this
guide, instructions for the user to choose in software tend to be given in bold.

Table of Contents
Introduction ................................................................................................................................ 2
Data Collection Strategies.......................................................................................................... 2
Indexing a powder pattern manually.......................................................................................... 3
Lattice Parameter Refinement Getting Started ....................................................................... 4
Lattice Parameter Refinement using WinXPOW ...................................................................... 7
Step 1: Creating a peak file for peak calibration.................................................................... 7
Step 2: Peak calibration ......................................................................................................... 8
Step 3: Creating a peak file for the phase of interest ........................................................... 10
Step 4: Pattern Fitting (Optional) ......................................................................................... 11
Step 5: Lattice Constant Refinement ................................................................................... 12

Introduction
Every solid crystalline compound gives its own unique X-ray diffraction pattern, consisting
of a set of Bragg peaks. The diffraction pattern for a compound can be considered
analogous to a fingerprint, or barcode, with the peak positions determined by the unit cell
symmetry and dimensions (or lattice parameters). When we collect XRD data, we can use
these fingerprints to identify not only what phases are present in our sample(s), but also
index the pattern to obtain information about the unit cell size and shape.

Data Collection Strategies

Lattice parameter refinement does not necessarily require data of very high quality. The only
important parameter is peak position, and the peaks will remain in the same place no matter
how long we collect data for. However, our ability to determine the peak position accurately
will depend to some degree on the peak shape; i.e. it will be difficult to determine the exact
position of a noisy peak, or one lost amongst the background!

For indexing and lattice parameter refinement, therefore, there are four main decisions that
should be made before carrying out experimental work:
-

Equipment: Bragg-Brentano (also known as reflection) geometry instruments (i.e. the

Philips PW1825, Siemens D500 and D5000) can be used, but specimen height
displacement errors can easily introduce significant error into refined lattice
parameters, and so use of an internal standard of known lattice parameter is essential.
For powder samples, the high resolution STOE PSD systems are excellent choices,
though an internal or external peak position standard must still be used; use of the
STOE imaging plate detectors is not recommended.

Scan range: The resolution in d-spacing increases with 2, which means that for
lattice parameter refinement we want to collect data to high angles. Avoid collecting
unnecessary angular ranges where only background noise will occur.

Scan rate: Try to collect at the fastest rate possible, and then see how good your data
are:
o For Bragg-Brentano geometry instruments, start with a scan rate of 1 per
minute. You should aim to record 2000-5000 counts on your largest peaks.

o On the STOE PSD system, the data collection time necessary will depend on
the nature of your sample. For cubic samples, 1 or 2 hours will most likely be
more than sufficient. For lower symmetry, more time will be required.
-

Step size: For lattice parameter refinement, peak resolution (i.e. being able to
distinguish two peaks that are very close together) and accurate peak position
determination are critical. Use step sizes of 0.02 2 on Bragg-Brentano instruments,
or 2-4 points to be added on the STOE PSD systems.

Again: the most common pitfall with Bragg-Brentano experiments is specimen height
displacement. Loading the sample too high, with respect to the sample holder, will cause any
Bragg peaks to be shifted to higher angles; if the sample is too low, the peaks will be shifted
to lower angles. The effect on peak positions is almost, but not quite linear with 2. It is
always worth making sure your sample preparation is as good as possible!

Indexing a powder pattern manually

Indexing is the process of determining the size and shape of the unit cell for a given phase. It
involves assigning the correct Miller indices to each Bragg reflection. The complexity of the
task depends on the symmetry of the phase it is easy for cubic, tetragonal and orthorhombic
systems, but gets significantly more complex when the angles are not 90 !

In these notes, I will deal with a cubic system:

We know that the Miller indices are integer numbers, and hence the (h2 + k2 + l2) term must
also be an integer value. Furthermore, the quotient / 4a2 must be constant for any given

diffraction pattern, since both and a are fixed values. This means that we can use ratios of
sin2 values to help us determine the correct Miller indices for reflections.
Note that for each Bravais lattice type, only certain values are permitted for (h2 + k2 + l2):.

Primitive: All values of h, k, l are allowed

Body centred: Values of h + k + l = 2n only are allowed (where n is integral)

Face centred: Values of h, k, l all odd or all even are allowed (0 counts as even)

Also, there are certain prohibited values for (h2 + k2 + l2), e.g. 7, 15, 23, which cannot be the
result of the sum of three squares. Knowing all of this can be very useful in indexing, and
means we need at least seven reflections to distinguish between primitive and body centred
cells!

To index our pattern, we first collate a list of peak positions, which are then converted
manually to sin2 values. We then divide these sin2 values by allowed (h2 + k2 + l2) for
primitive, body centred or face centred unit cells. We then inspect the sin2 / (h2 + k2 + l2)
values are they all constant? If not, repeat for the next unit cell type. Once we find
constant values we can then assign Miller (hkl) indices for each reflection and determine the
cell constant, a.

Consider the following example, where a number of Bragg peaks were recorded from an
unknown material using Cu K radiation.

The resolution in d-spacing increase, while systematic errors decrease, with diffraction angle.
Therefore, to calculate the unit cell constant we should use high angle reflections in the
above example, we would choose the ninth reflection:

From: West, A.R., Basic Solid State Chemistry, John Wiley & Sons Ltd, 1996.

Lattice Parameter Refinement Getting Started

The WinXPOW software can also be used to index and refine lattice parameters from your
data, but this should be approached with caution. The software does not understand your
sample, and does not have any intelligence. It may give you an answer, but that does not
mean it is correct that is for the user to judge! What you will need is a starting model:
-

a space group, or at least the symmetry and centring (e.g. if you were looking at a
spinel, you should know that the space group for a spinel is likely Fd-3m, or at the
very least know that you are looking at something face-centred cubic), and also

a good estimate of the lattice parameters.

This is made easier if the researcher has something to compare against. The most useful and
powerful source of assistance available in the Department of Materials Science &
Engineering is a database made by the International Center for Diffraction Data (ICDD)
called the Powder Diffraction File (PDF). The software we use to access this database is
made by STOE, and is called WinXPOW. The PDF contains a mixture of experimentallydetermined and calculated X-ray diffraction patterns. In the 2004 edition licensed by
Engineering Materials, there are a total of 134,000 patterns belonging to inorganic and
organic compounds. It is likely that one or more of these reference patterns belong either to
your material or to something analogous to it i.e. with the same crystal structure, and a
similar unit cell. This may give you the information you need.

Lattice Parameter Refinement using WinXPOW

Step 1: Creating a peak file for peak calibration
The first thing we need to do is to create a peak file a list of all the peak positions for a
given phase for our internal or external peak position standard. The Graphics (Raw
Data/Graphics) program can be used to view STOE *.raw files and create peak (*.pks) files
for indexing/refining unit cells.

To open a STOE *.raw file, click File/Open. Then, we must find the standard peaks in your
diffraction pattern; clicking Options/Find Peaks will bring up a child window where the
Peak Search Parameters may be entered:
Estimated Halfwidth characteristic of both the sample and the instrument the data
were collected upon. Values of ~ 0.1 (for the Stoe PSD) and ~ 0.15 (for all our other
instruments) are reasonable.
Significance Level - sets a lower limit to the peak heights, defined as the ratio of the
maximum peak intensity to background. Default value of 2.5 is fine.
Intensity Limit - All peaks having an absolute height less than the input value will be
discarded. Entering low values significantly increases the chance that the software
will find artificial peaks in the random noise of the background!
Remove Alpha2 Peaks - Checking this box will scan the peak list for pairs of
Alpha1/Alpha2 peaks and discard the Alpha2 reflection.
Remove Spikes - Peaks of a halfwidth less than 25% of the estimated halfwidth
defined above will be discarded as spikes, or electrical noise.
Clicking OK will perform the peak search. If you are using an external standard, this is
straightforward all the peaks will be from the reference material. Check the peaks have all
been selected appropriately, and save the peakfile.

If you are using an internal standard, the diffraction pattern will contain peaks from both the
standard material and from your specimen. At this stage, it is vital that we only include those
peaks from the reference material; we can manually add or edit peaks as well - double-click
the left mouse button to add a peak, right-click to remove a peak, or hold the left-mouse
button to drag a peak in to a chosen position. Check the peaks have all been selected
appropriately, and save the peakfile.

Step 2: Peak calibration

The Calibrate Peaks program allows the user to calculate correction terms for the 2 scale of
data collected with an internal (or external) standard reference material for the purposes of

this guide it will be assumed that you are using silicon. The program calculates a correction
term which may be used to correct raw data files in the Raw Data Handling program.

To perform a peak calibration on your data set, you will need a copy of the si-icdd.pks file, a
peak file for the Si standard powder. You will also need a Si peak file, created from data
collected on the same day as your data (in the case of the Stoe PSD) or from a Si internal
standard (i.e. Si powder added to your sample). This latter course of action is recommended
if you wish to refine lattice parameters from data collected on any instrument other than the
Stoe PSD.

To perform the peak calibration:

1. Run Raw Data/Peak Calibration
2. Click File/Open. A child window will appear; enter the location of the si-icdd.pks
file in the upper box, and your Si file in the lower box.
3. From the menubar, select Calibrate/Match Peaks. In the child window that appears,
there are two options:
a. Intensity limit - a cutoff-value below which the program will ignore all peaks
from both the calibration and the observed peak files. Set to 0 %.
b. 2-Theta window: the 2theta range around a calibration peak which the
program will search for matching observed peaks. Set this to 0.05.
You will need at least three peaks, but more is better.
4. You may need to connect a measured peak to a calibration peak; to do this, left-click
the (blue) calibration peak, then attach the measured (grey) peak by left-clicking on it.
5. Once all your measured peaks are attached to calibration peaks, click
Calibrate/Calculate Polynomials. The results are shown in the lower part of the
main window. The 2theta offsets for all peak pairs are shown as black squares and the
calculated correction polynomials as curves of different colours.
6. Click Calibrate/Calibrate Peakfile.

A child window will appear asking which

polynomial should be applied; choose the one with order > 0 with the lowest chi^2
value, and check that this fits the observed data well. Click OK and exit from the
Calibrate program without saving changes.
7. Go to Raw Data/Raw Data Handling and Open the file you wish to calibrate. Click
Operations/Angle Correction.
ticked, and click OK.

Make sure the use stored polynomial box is

8. Save your corrected data.

The next thing we need to do is to create a peak file a list of all the peak positions for a
given phase. The Graphics (Raw Data/Graphics) program can be used to view STOE *.raw
files and create peak (*.pks) files for indexing/refining unit cells.

Step 3: Creating a peak file for the phase of interest

To open a *.raw file, click File/Open.

Then, we must find the peaks in your diffraction pattern; clicking Options/Find Peaks will
bring up a child window where the Peak Search Parameters may be entered:
Estimated Halfwidth characteristic of both the sample and the instrument the data
were collected upon. Values of ~ 0.1 (for the Stoe PSD) and ~ 0.15 (for all our other
instruments) are reasonable.
Significance Level - sets a lower limit to the peak heights, defined as the ratio of the
maximum peak intensity to background. Default value of 2.5 is fine.
Intensity Limit - All peaks having an absolute height less than the input value will be
discarded. Entering low values significantly increases the chance that the software
will find artificial peaks in the random noise of the background!
Remove Alpha2 Peaks - Checking this box will scan the peak list for pairs of
Alpha1/Alpha2 peaks and discard the Alpha2 reflection.
Remove Spikes - Peaks of a halfwidth less than 25% of the estimated halfwidth
defined above will be discarded as spikes, or electrical noise.
Clicking OK will perform the peak search. It is important to make sure that if we want to
refine the lattice parameters for one given phase, then at this point we should ensure that only
the peaks from that phase are selected. We can manually add or edit peaks as well - doubleclick the left mouse button to add a peak, right-click to remove a peak, or hold the left-mouse
button to drag a peak in to a chosen position.

At this stage, you may well choose to omit peaks if you are not confident about the position
that they have been placed in.

Step 4: Pattern Fitting (Optional)

The Pattern Fitting routine is used to determine profile parameters and integrated intensities
from observed powder patterns. A variety of constraints makes it possible to achieve
meaningful results even for complicated patterns.

This routine does not require any

information on the structure of the phase.

1. Open up the *.raw and *.pks file for the pattern you wish to fit.
2. Click the Parameter menu option. You will notice three choices:
a. Overall background and asymmetry functions; unless the background is
strange, leave these to the program.
b. Peaks Lists the 2 values, absolute and integrated intensities, and halfwidths
of all the peaks.
c. Groups Parameters that may be fixed or varied during refinement.
3. Choose the Group option.

Start by refining the halfwidths only.

The initial

parameters should be W = 0.01, V = 0, U = 0. Tick Halfwidth as function, and

unfix W make sure that initially all the other parameters are fixed. Click Replace
to enforce any changes you have made, and then OK.
4. You are now ready to start the refinement: click Operations/Nonlinear Regression.
Once the refinement has converged, go back to Groups and allow U and V to vary.
Remember to click Replace after youve made any changes in the Groups window!
5. Once convergence is achieved, allow the profile function to vary. There are two
Pearson and Pseudovoigt try both to see which fits your data best. The default
parameters for the profile function are 0.5 for Pseudovoigt and 2 for Pearson.
6. Inspect the pattern for missed peaks, extra peaks, poorly fitted peaks and edit
accordingly by clicking Operations/Edit Peaks.
7. Inspect the peak file for negative intensities, halfwidths that are very large (may be
two peaks?) or very small (may not be a peak!)
8. Some peaks at low 2 values may be asymmetric. This can be modelled using the
asymmetry function Click Parameter/Overall Parameter and click Use

asymmetry and untick fixed. The starting value should be greater than 0.03.
Note: This step will not be necessary if you dont have any peak asymmetry!
9. Once you are happy with the fit, save the fitted peak file as <filename>.pft.

Once again, remember that more detailed help is available from the help files in the Pattern
Fitting program.

Step 5: Lattice Constant Refinement

WinXPOWs Index & Refine module comprises three indexing routines and a lattice parameter
refinement option. The implemented indexing algorithms will give a list of possible unit
cells, and are based on:

Werner's TREOR program (see e.g. P.-E.Werner, L.Eriksson, M.Westdahl, J. Appl.

Cryst. 18 (1985) 367)

Visser's ITO routine (see e.g. J.W.Visser, J. Appl. Cryst. 2 (1969) 89)

Lour's DICVOL (see e.g. A.Boultif, D.Lour, J. Appl. Cryst. 24 (1991) 987).

The Index & Refine software can be accessed from the main WinX POW console window
under Cell. This program can be used to analyse peak files in *.pks or*.pth format.
The Index menu options
Under the Index heading, there are four entries. Before running any of the algorithms,
however, you first need to define the refinement parameters.
1. Click on Index/Parameters. A child window will appear, with three tabs one for
each of the available algorithms. For the algorithm you intend to use, fill in all the
boxes you can a detailed list of what each item means can be found in the help files
accessible from the Index program.
2. To run your algorithm of choice (try each one to see which provides the best possible
solutions), click Index/Run <algorithm name>.

This will start the calculation

process this may take some time! You can then try refining your sample based on
these solutions.

When Visser's and Lour's routines have finished, they open a child window displaying their
output.

Werner's routine writes possible solutions into the main window and after

completing the search - the user may double-click on any of them to open a child window
showing the indexing of the data with the currently selected cell (highlighted in red in the
main window). Werner's algorithm may be interrupted anytime by clicking the toolbar button
with the red traffic light.

Lattice Constant Refinement

You may already have a possible solution, from literature, the Powder Diffraction File, or one
of the algorithms above. The software allows you to refine the lattice constants for your
observed data using the Refine menu option. To work through such a refinement:
1. Open your peakfile (<filename>.pks) in the Index & Refine program.
2. Click Refine. A child window will appear, offering you various data entry options,
all of which should be self-explanatory. Fill these boxes in:
a. Symmetry - Enter the space group, if you know it, or the crystal system (e.g.
cubic) and Bravais lattice (e.g. Primitive) if you dont.
b. Lattice Parameters Enter in the dimensions (in ) and angles of the cell
upon which you wish to refine your data.
c. Settings
i. 2Theta window - If the absolute difference between an observed peak
position and a calculated peak position is less than the 2Theta window
(in degrees), the observed peak will be considered indexed. The
program will try to reduce the window as much as is possible without
observed peaks becoming unindexed. A value of 0.1 is reasonable.
ii. 2Theta zero - allows the input of a constant 2theta-error introduced by
poor 2Theta-calibration of the diffractometer. The check box Refine
2Theta zero' is used to enable or disable refinement of this value
together with the cell parameters (not advised!).
iii. Use peak file indices - if checked, starting values for refinement are not
taken from the Lattice parameter' edit fields but are generated from all
peaks in the peak file which are indexed. Note: it is possible to enter

hkl values into a peak file using Notepad just place the values
directly under the H, K and L headings.
3. Press OK to perform the refinement.
4. The result of the attempted refinement will now be displayed. There are basically two
possible outcomes success (to some level) or failure.
a. Failure You will see a window much like the one below. Your solution is
not the answer, or the lattice parameters may be too far from the correct
values; try working them out from individual reflections first. Your starting
model must be very close to the final answer if you are to successfully refine
the lattice parameters!

b. Success? You will be prompted with a list of text, detailing the refined
values for your lattice constants. You should now plug these lattice constants
back into the Refine window, and try refining again. Keep doing this until
the values no longer change. You will also be told how many peaks are fitted
by the model; if all of them are fitted, then well done! If there are reflections
not accepted by the model, consider possible reasons for this is the model
correct? Could they be from secondary phases? Is the 2 window too small?
If you know that any unidentified peaks should be there, try indexing these
manually in Notepad as described above.
For a good lattice parameter refinement, the Final 2 window, i.e. the biggest value from
the obs-calc column, should be smaller than 0.03. This means that all of the observed
reflections fall within 0.03 of a peak whose position has been calculated using the refined
lattice parameters.

Once you have reached a satisfactory answer, you should save your refined peak file
(File/Save As). Saving the file also automatically saves a *.lrf file, with the same filename
as your *.pks file, with all the lattice refinement data. This *.lrf file can be viewed in
Notepad, and contains all of the results of the lattice parameter refinement.