1. Introduction
Bragg scattering, which is used in conventional crystallography, gives only information about the
average crystal structure. Diffuse scattering from single crystals, on the other hand, is a prime source of
local structural information. There is now a wealth of evidence to show that the more local structure is investigated the more we are obliged to reassess our understanding of crystalline structure and behaviour [
1].
It is important to realize that the
average structure is just that: the average of many different
local structures. As a material’s properties often depend on this local structure it is important to understand of what different local structures the average is comprised. Developing, optimizing and critically assessing the tools that we have available to uncover local structure in crystals is now a vital part of crystallography [
1]. Although
X-ray diffuse scattering is of prime importance and is well documented [
2,
3], diffuse
neutron scattering can play an important role too, especially in view of the unique properties of neutrons—their ability to explore inelastic scattering events, the fact that scattering lengths do not vary systematically with atomic number and their ability to scatter from magnetic moments.
Since diffuse scattering intensities are typically several orders of magnitude lower than Bragg peak intensities it might be supposed that obtaining diffuse scattering data using neutrons is simply not a viable proposition. The neutron fluxes available at the best neutron facilities in the world are very low in comparison to the photon fluxes available for X-ray sources. For example the neutron flux on SXD [
4] at ISIS (Didcot, UK) is
neutrons per s per mm
whereas the X-ray flux at the 1-ID beamline at APS is
photons per s per mm
[
5]. Note that the X-ray flux given is the flux at the sample after monochromatisation. With the monochromator selecting energies with a resolution
this means a large fraction of the available photons are not utilized. In contrast, for experiments at a spallation neutron source, it is possible to use the whole spectrum of incident neutron wavelengths since their contribution to the scattering can be separated by time-of-flight (TOF) spectroscopy. If this is coupled with the use of extensive detector banks that cover a large fraction of the scattering solid angle a large enough enhancement factor relative to the X-ray case is obtained such that recording diffuse scattering can become quite viable for a suitably sized sample. A typical sample size recommended for SXD is
mm
while for the synchrotron a size of
mm
is more typical.
Nuclear reactors are traditional “steady state” neutron sources and have been used for diffraction experiments since about 1950. Wilson [
6] has given an account of the development and relative merits of reactor versus spallation sources. The high-flux reactor-based Institut Laue-Langevin (ILL) (Grenoble, France) is an international research centre that is still at the leading edge of neutron science and technology despite having been established in 1972. Technological limitations as well as political issues associated with building more powerful reactors have set limits on possible neutron production from reactors and further dramatic increases in flux are unlikely. Accelerator driven sources are not subject to the same technological or political constraints and these can potentially offer increasingly higher fluxes into the future. Since the experiments on SXD described below in
Section 2, which were carried out when the ISIS facility was the most intense spallation source in the world, other more powerful spallation sources have already come on stream (this includes the spallation neutron source (SNS) at Oak Ridge National Laboratory, (Oak Ridge, TN, USA) where the new CORELLI instrument described in
Section 3 is located). For diffuse scattering studies a reactor source is, moreover, not ideal since, in order to survey reciprocal space, either a monochromator has to be used or alternatively a single aperture chopper employed to allow TOF analysis as in the spallation neutron case. Either of these scenarios means that a large proportion of the neutrons generated in a reactor are discarded. Despite this, with a suitably sized sample, useful diffuse scattering data can be obtained at a reactor and an example of this is described in
Section 4.
In this paper we describe experiments on three different instruments that are available for neutron scattering experiments and that have contributed towards making diffuse neutron scattering a viable and valuable research tool. In
Section 2 and
Section 3 below we compare measurements made on the molecular material
d-benzil, C
14D
10O
2, using SXD at the ISIS spallation neutron source [
7] with more recent measurements made at the SNS spallation source using the dedicated diffuse scattering instrument CORELLI [
8]. For comparison in
Section 4 we show what results can be achieved at a reactor source as typified by measurements made on lead zinc niobate,
(PZN), using the high intensity powder diffraction instrument Wombat at the Open Pool Australian Lightwater reactor (OPAL) (Lucas Heights, Australia) [
9]. Before proceeding it is useful to recall here the relationships that exist between various quantities pertinent to the discussion. The neutron wavelength
is given by
, where
h is Planck’s constant,
m is the neutron mass and
v is its velocity. The neutron energy
is given by
. The velocity may be measured by the TOF
for the neutron to reach the detector located at a distance
, where
. For a more complete discussion see Copley and Udovic [
10].
2. Spallation Neutrons I —SXD Single Crystal Instrument at the ISIS Facility
Welberry et al. [
7] have recorded diffuse scattering from the molecular crystal
d-benzil, C
14D
10O
2, using the SXD single crystal diffractometer at the ISIS Facility of the Rutherford Appleton laboratory. The experimental details and resulting diffuse scattering data for this example are illustrated in
Figure 1 and
Figure 2.
Figure 1a shows the diffraction geometry for a single stationary crystal setting and for a single detector bank whose centre is positioned at the scattering angle 2
=
. The figure is drawn to show how the neutrons are scattered in directions emanating from the centres of the Ewald spheres towards the detector bank for the two cases corresponding to the minimum and maximum wavelength present in the incident beam. The radii of the two spheres shown were chosen for clarity. In practice the value of
is around
smaller than
.
Figure 1b shows the arrangement of detector banks on SXD. There are 11 detectors each of which has
pixels with each pixel recording a complete TOF spectrum. The blue shaded area in
Figure 1a shows that region of the reciprocal lattice that lies within the Ewald spheres corresponding to the maximum (
) and minimum (
) wavelengths and covers all points that are scattering simultaneously for this crystal setting. The green shaded area covers all reciprocal points that give rise to scattering that is directed towards the
detector bank (detector 5 in this case).
The combination of the two-dimensional nature of the detectors and the TOF spectrum allows a large 3D region of reciprocal space to be recorded at any stationary position of the crystal sample. To get a complete mapping of reciprocal space a number of exposures are required with the crystal rotated by an angle
about the vertical axis (normal to the plane of the paper in
Figure 1b). Since the equatorial detectors subtend an angle of
of
(i.e.,
of
) an increment of
in
is required to ensure successive exposures give an unbroken mapping of reciprocal space. In practice
is generally incremented by
to allow some overlap. In the time available three complete such exposures were made and a fourth partially completed (approximately half the time of the others). The length of each full exposure (typically ∼18 h) was determined by monitoring the number of microamps of the beam current. The data presented in
Figure 2 were selected from a complete 3D set of data that was collected in just over three days in total. The benzil crystal used to record the data shown in
Figure 2 measured
mm.
Figure 2 shows reciprocal lattice plots of diffuse scattering in the
section of benzil at 150 K recorded on the SXD instrument.
Figure 2a shows data from a single detector (detector 1 in
Figure 1b) for a single crystal orientation setting.
Figure 2b shows the coverage obtained using the 6 equatorial detectors (Nos.
) for a single crystal orientation setting. For comparison
Figure 2c shows the coverage given by a single detector but using 3 crystal settings. These 3 settings corresponded to successive rotations of the sample by ∼
. Since a
d-benzil crystal has
m symmetry these three settings are sufficient to provide complete coverage of the asymmetric unit of data.
Figure 2d shows a complete reciprocal section of data in which 3 crystal orientations have been used together with four detector banks and with
m symmetry averaging applied. This gives better signal to noise and a smoother scattering pattern. For
Figure 2a–d the data extends out to a maximum Q = 4
of 15 Å
.
Figure 2e,f highlight the data at lower values of Q. In both
Figure 2e,f three crystal orientations and symmetry averaging were used with a single detector. For
Figure 2e detector 1 was used while for
Figure 2f detector 3 was used.
For a given Q the high-angle and low-angle detectors measure the data with longer and shorter wavelengths and this impacts on the way the inelastic features are sampled.
Thus there is a marked contrast between data shown in
Figure 2f,e. In
Figure 2f there are prominent diffuse lines forming a hexagon shape near the centre of the pattern (and other diffuse lines at higher
Q) while these have largely been dispersed in
Figure 2e. The diffuse lines arise from a low-energy phonon mode in benzil corresponding to a shearing motion of the molecular rows and columns. Since the energy of this mode is ∼1 meV, which is small compared to the energies of the neutrons detected in detector bank 3, the scattering here is almost elastic and the pattern looks very similar to what is observed with X-rays. For
Figure 2e the neutron energies are much lower and these scatter quite inelastically with the phonon mode so that the intensity in the diffuse line is dispersed. For further details see reference [
7]. Since the energy (or wavelength) of each neutron recorded is known from its TOF the scattering pattern can be explored further by, for example, binning only those neutrons within a certain energy range, hence tuning the range of energy transfers over which
is integrated.
3. Spallation Neutrons II—CORELLI Cross Correlation Spectrometer at SNS
While SXD uses the spallation pulse itself to allow time-of-flight resolution of the energy of each scattered neutron the CORELLI instrument at SNS additionally uses a statistical chopper together with a cross correlation analysis technique to allow discrimination of elastic and inelastic scattering events. Use of cross correlation techniques in diffuse scattering studies were pioneered many years ago [
11]. The operation of this cross correlation method as used in CORELLI has been described by a number of authors [
8,
12,
13,
14]. We here give a brief description based on these sources.
The chopper (see
Figure 3a) comprises 127 transparent and 128 opaque apertures (sectors) positioned in a random sequence around a disc. The chopper is positioned
m downstream from the target so the initial spallation pulses have become quite spread out by the time they reach the chopper, with the high energy neutrons arriving first and lower energies later. Each transparent aperture in the chopper then acts as a secondary pulse source of virtually monochromatic neutrons of an energy
and intensity
dependent on this initial TOF
and a phase offset due to its sequence around the disc. For the cross-correlation method described below it is necessary to sample the total TOF spectrum for all possible offsets with the same probability and this is easily and almost instantaneously obtained by rotating the chopper asynchronously with the source. The current maximum chopper rotation speed is rather less than 300 Hz (18,000 rpm). Since a speed of exactly 300 Hz would result in each chopper aperture producing a secondary pulse exactly 5 times per primary pulse cycle (60 Hz), the rotation speed must be reduced somewhat to achieve the desired asynchrony.
The transmitted neutrons then go forward to scatter from the sample which is located a short distance (
m) downstream from the chopper. This is shown schematically in
Figure 3b. The neutrons that are able to pass through the chopper are scattered from the sample with probability
and arrive at a detector pixel at a time
t after the initial pulse. The intensity
recorded in a single pixel of the detector as a function of the total TOF is seen to be the sum of all such scattering events and may be expressed as in Equation (
1) (see reference [
12]). The time at which the neutron beam transits the first sector of the chopper is
.
Figure 3b also shows that at the sample neutrons may be scattered either elastically (green line) or inelastically (red and blue lines) as their energy is increased or decreased (e.g., by the creation or annihilation of phonons). Successive apertures act identically but with neutrons of a different energy and intensity
where
is a time shift dependent on the sector number. The intensity recorded at the detector for a given total TOF
t is then the sum of over all possible scattering events
, modulated by the chopper modulation function
:
Here, is the scattering function for incident neutrons with phase offset and total time-of-flight, t. is the chopper modulation function, denotes the start of the modulation sequence with respect to the source pulse, is the neutron flux for the corresponding incident energy, and is the background at the detector that is uncorrelated with the chopper.
Rozenkrantz and Osborn [
8] have shown that if the cross-correlation
is formed between the detector intensity
and the chopper modulation sequence
this yields the scattering function
, where
and
t are the incident and scattered times of flight respectively. This relies first of all on the data being recorded for a sufficient time that all possible phase offsets occur with equal probability and secondly that the pseudorandom modulation sequence,
, is characterized by a delta-function autocorrelation,
Here is the fraction of of the chopper elements that are open (%).
The second term in this equation is proportional to the total scattering (integrated over all energy transfers) and this must be subtracted to obtain the desired signal (or . This is not too much of a problem if the desired signal is large but for a small signal (e.g., weak inelastic excitation) the result of subtracting [small signal + average] − [average] will have large error bars. The third term (background) can be measured independently and subtracted with no difficulty.
It should be noted that
in Equation (
1) is the intensity recorded in a single pixel of the detector at a given total TOF,
t. If the chopper phase is ignored altogether and, simply, a histogram is made of the intensity measured at the detector as a function of
t only, what is obtained is the energy-integrated intensity (provided that the chopper is running properly asynchronously to the source). This then is comparable to the TOF data measured with any standard diffractometer (such as SXD described in
Section 2). To obtain the elastic only scattering the cross-correlation has to be applied to the measured neutron intensity over the chopper phase as in Equation (
3).
The result of these two modes of data treatment are shown in
Figure 4 for the example of benzil. Data recorded at two different temperatures are shown.
Figure 4a,b were recorded at 300 K while
Figure 4c,d were at 100 K. Panels (a,c) were recorded without cross-correlation applied and thus comprise the energy-integrated intensity (elastic + inelastic) while panels (b,d) are with the cross-correlation applied and show the elastic component only. Since the benzil diffuse scattering is thermal diffuse scattering (TDS) the purely elastic scattering shows only the Bragg peaks. Note that (b,d) are printed on the same intensity scale as (a,c) respectively. The resolution in energy afforded by the cross-correlation analysis is quoted [
15] as 8% of the incident energy.
The benzil crystal used to record the data shown in
Figure 2 measured
mm. The volume of this crystal was therefore approximately 1/70 of the volume of that used in the SXD experiment shown in
Figure 2. The time to record the data (
2 days) was comparable in the two cases.
4. Measurement of Diffuse Scattering on a Reactor Source
Whitfield et al. [
9] have recorded diffuse scattering data for a number of ceramic materials using the Wombat high intensity powder diffractometer at the OPAL nuclear reactor at ANSTO (Australian Nuclear Science and Technology Organisation, Lucas Heights, Australia). Although Wombat (see
Figure 5b) was not designed specifically for measuring diffuse scattering data the instrument scientists/designers were well aware of this possible application from the start of the design process. It has attributes that make it quite suitable for recording diffuse scattering, namely a relatively high neutron flux and a large curved area detector. The area detector subtends
of
in the horizontal plane and comprises 968 pixels in the
direction, with each pixel corresponding to
deg in
. There are 128 pixels in the vertical direction giving some, though not extensive, out-of-plane coverage. Diffuse scattering data are collected by incrementally rotating the sample about a vertical axis,
, in steps of
. The arrangement is ideal for collecting data in or close to the zero-level section of an aligned sample but by judicious choice of a number of different sample orientations and by using the sample symmetry more extensive coverage of reciprocal space may be achieved.
Figure 5c shows data for the
section of lead zinc niobate (PZN) collected at 160 K and
Figure 5d shows data collected from the same sample at 500 K. Neutrons of wavelength 1.54 Å were used and this allowed data to be recorded between
and 7.5 Å
. For each run 600 exposures of 60 s were recorded at
intervals, making a total collection time of 10 h. The sample was contained in a cryofurnace but a collimator was used to eliminate scattering from this and only allow neutrons emitted radially from the sample at the centre of the furnace to reach the detector. The data shown in
Figure 5c compare well with data shown in
Figure 5a that were recorded earlier on SXD (see reference [
16]). These SXD data were recorded in
20 h. The sample sizes in the two different experiments were comparable (
versus
). The neutron flux at the sample on Wombat is given as
10
neutrons per s per mm
[
9]. In both cases the space group symmetry of the crystal was used to average symmetry-related data and obtain optimum signal to noise.
One anomalous feature of the Wombat data are the radial streaks that extend along prominent lattice directions such as
,
and
. The authors attribute these to “incoherent scattering from components in the monochromator and mounting” [
9] p. 1425. They appear very similar to features observed on some X-ray patterns recorded using Mo sealed-tube microsources monochromated with doubly reflecting multilayer mirrors [
17,
18]. In either case, being radially oriented, these streaks cannot be interpreted as disorder or thermal diffuse scattering but must be attributed to a lack of monochromaticity in the incident beam. That these features are less prominent in the high temperature data is perhaps indicative of some increase in mosaic spread of the sample.
The PZN example illustrated in
Figure 5 is a good example of the special place that neutron diffraction can have in the study of disorder and local order using diffuse scattering. With X-rays the scattering patterns are dominated by the heavy cations and the effect of the oxygen anions is small while for neutrons the oxygen anions have a much stronger contribution. The patterns in
Figure 5 show a number of features displaying strong asymmetry—for example the strong diffuse peaks surrounding a number of Bragg positions such as
,
,
, etc. and the zig-zag arrangement of diffuse lines just above
. Corresponding X-ray patterns show much less asymmetry. Such asymmetry is indicative of strong
size-effect distortions that must occur on a local scale in the structure.
5. Conclusions
There is no doubt that the last 10–15 years has seen a substantial improvement in our ability to collect neutron diffuse scattering data from single crystals. Though the patterns for benzil obtained on SXD in 2003 [
7] stand up well to comparison with the recent data recorded on CORELLI it must be noted that the crystal volume for the latter experiment was only about 1/70 of that of the former experiment. This increase in performance may be attributed partly to the increased power of the SNS source relative to ISIS but also to the moderator, the instrument design and higher efficiency detectors used. From the beginning CORELLI was designed and optimized from the source all the way to the sample [
19].
The choice of the benzil example for the CORELLI experiments was useful for making the comparisons with the earlier SXD results. However, because the diffuse scattering from benzil is entirely of thermal origin, it was not an ideal molecular crystal to use for demonstrating the energy discrimination capabilities of CORELLI since the only part of the scattering that is truly elastic is the Bragg peaks. All molecular crystals display substantial amounts of TDS at ambient temperatures but there are many examples in which there is additional occupancy disorder (see for example references [
20,
21]). CORELLI would allow the study of both the static and dynamic disorder in such materials. It was intended that the CORELLI experiments would be performed alongside experiments on the TOPAZ single crystal diffractometer at SNS [
22]. Unfortunately a major mechanical break-down prevented this. TOPAZ is a diffraction instrument similar in concept to SXD with numerous detector banks covering a large solid angle.
The example of PZN described in
Section 4 usefully demonstrates that diffuse neutron scattering can be measured on a reactor source if a suitably sized sample is available. This sample size needs to be substantially larger than would be required for CORELLI but comparable to what was used in earlier experiments on SXD. The data obtained for PZN was of comparable quality to data obtained on SXD in 2005. The Wombat instrument at ANSTO is primarily a high intensity powder diffractometer and is not purpose built for diffuse scattering but its use of a large area detector that subtends an angle of
in the horizontal plane means that useful diffuse scattering data can be collected. The reciprocal space coverage is by no means complete, however, since out-of-plane coverage is limited. Full 3D data can be obtained using a number of crystal orientations or by making use of the crystal symmetry for high-symmetry specimens.