Home‎ > ‎

Neutron Diffractometer Upgrades

A number of linear position sensitive detector array powder diffractometers built in the period from the mid-1980s through early 2000s benefit from upgrades to their detector front-end electronics, decoding electronics, software and computing control system hardware.

Upgrade of PNPD. NCSU engaged IA to upgrade the PNPD diffractometer: a neutron powder diffractometer with a 15-element linear PSD detector and a bent perfect silicon crystal Popovici monochromator.

The instrument has an interesting history. It was originally built in the mid-1990s at the University of Missouri as a replacement for the 5-element detector PSD-II at MURR. The detector decoding electronics were designed in the Instrument Development Group at MURR in a collaborative project with the Institute for Energy Technology at Kjeller Norway. Three instruments were designated to receive the (then new) electronics: the PUS diffractometer at Kjeller, PSD-III (the 15-element diffractometer now working at NCSU) and MURRSI, a 9-element instrument for residual stress measurements.

When the neutron scattering program at MURR collapsed in 1999-2000, the residual stress instrument went into storage (it is still there) and the 15-element detector, decoding electronics and detector shield went first into storage and then moved to the University of Michigan and the Ford Nuclear Reactor. After the FNR shut down in 2003, the instrument went to NCSU in 2004 where it was installed on the 1 MW PULSTAR reactor. The figure below shows PNPD shortly after the instillation was complete.

 
Figure 1. PNPD shortly after the installation was completed in 2006.

In the course of the instrument upgrade, IA rebuilt the detector, installing new detector front-end electronics and replaced the old PSD position decoding electronics with 15 PEMs. The instrument was re-configured, placing the control computer and a small rack holding the detector decoding electronics on top of the detector shield in order to shorten cable runs.

Results from the upgrade have been gratifying. Detector position resolution and linearity performance is excellent – as good as it gets with 24” long 1” dia. 3He filled neutron sensitive linear PSDs.

The two figures below show the result of a calibration measurement with the upgraded detector system.

 
Figure 2a. Detector Position Spectrum obtained by exposing the comb-mask covered detector to neutrons scattered from a plastic rod located at the specimen position. 
 
Figure 2a. Analysis of the calibration position spectrum. 

The position of the detector peaks in Figure 2a, above, is fit to Gaussians and the Gaussian peak position (in position channels) is plotted against the mask slot position in cm and marked with black triangles. A linear least-squares fit to this data gives the detector calibration – relating position channel to position in cm along the detector axis. The Gaussian width FWHM of the calibration spectrum peaks is plotted (against the RHS axis) in cm. These results must be corrected by subtracting, in quadrature, the with of the comb-mask slots (0.25 cm) which yields the true detector position resolution. The red curve is the linearity deviation – the difference between the linear least-squares fit to the slot-position data points and the data itself. This result is offset by 0.3 cm (0.3 cm = 0 cm deviation). As can be seen, the linearity deviation is quite small, typically a few mm or less. Even this deviation has little practical consequence as the rebinning algorithm,

which converts position along the detector axis to scattering angle, uses a piece wise linear fit to the calibration data to assign event positions. The results shown in the figure are typical.

The detector was re-wired and the system electronics reconfigured to put the PEMs and the other system electronics in a small electronic rack placed on the top of the detector shield. The replacement of the 1990s electronics with PEMs and their placement on top of the detector shield removes the 25' long snake of 36 RG-58 coaxial cables that carried signals from the detector to the decoding electronics shown in Fig. 1.

The figure below shows the instrument after the upgrade.

 
Figure 3. PNPD electronics mounted on the top of the detector shield

The instrument resolution also shows a marked improvement. The figure below shows the result for a measurement of Ni. Powder.

 

Figure 4. Neuton powder diffraction spectrum from a 3 mm dia. Vanadium specimen can containing annealed Ni powder. The data is composed of three 1 hour duration segments (detector positions) of 20 deg each.

The data of the figure above was analyzed by Rietveld Refinement to obtain the peak shape parameters U, V and W chat characterize the diffraction peak shape as a function of angle. A comparison of these data with the instrument results before the upgrade is shown in the Figure below.

 
Figure 5. Comparison of instrument resolution after upgrade (2017Expt00038) with results from 2008 and 2010.

The instrument U, V and W parameters were used in the Cagliotti formula to plot the inferred diffraction peak FWHM as a function of angle.

NDCS Upgrade. The resolution of a neutron powder diffractometer is dependent on many factors including the height of the sample and the span of the detector perpendicular to the scattering plane. The design of every instrument is a compromise between resolution and count rate; making the sample and detector height larger will improve the speed of data-taking at the expense of resolution.

In the case of the linear position sensitive detector array, it is possible to make corrections to the data that reduce the resolution price that must be paid for the increase in instrument throughput obtained by increasing the detector vertical height.


Figure 1. Ni powder spectrum 2017Expt00038

Debye-Scherrer line broadening is a major contributor to the width of diffraction peak lines at low scattering angles. The impact and cause of Debye-Scherrer broadening can be easily illustrated. Figure 1 shows the diffraction pattern from a 3 mm dia Ni specimen at NCSU on the 15-element linear PSD detector array of the PNPD diffractometer.

The detector peak shifts that result from the intersection of the Debye-Scherrer cone at ~43.2o with the detector plane placed at 50is shown in Figure 2a, below.

 
Figure 2a. Plot of diffraction data from low angle peak (~43o) for each of the detectors in the array and the data for the detector array sum using a Flat Rebin.
 

Figure 2b.  Peak position for each of the detector elements given by a Flat Rebin of the diffraction data. 


Fitting the diffraction peak for each of the detectors to a Gaussian, quantifies the Debye-Scherrer broadening (Figure 2b.). Here the scattering angles for the detected neutrons are derived from thetaS = arctan(x/L) for all of the detectors. This is a Flat (for flat cone) Rebin. The elliptical signature of the intersection of the diffraction cone and the detector plane is clear.

Working backward from the neutron event position on the detector plane to the equation of the ellipse and thus the scattering angle turned out to be quite vexing and until recently only quite complicated numerical results could be obtained.

Earlier attempts to apply these Debye-Scherrer corrections to data from the NCSU instrument were not successful. Given the poor performance of the (old) PNPD electronics, which dated from the mid-1990s, and the horrible background, improvements to the rebinning algorithm did not appear to have any effect.

Since that time, the staff at NCSU has accomplished an enormous reduction in the instrument background. When the background is high and discontinuous at the segment boundaries, small changes in the rebinning algorithm have little effect on the quality of the Rietveld refinements. In short, the impact of previous efforts to implement Debye-Scherrer Cone Corrections were obscured by the generally poor quality of the Rietvelt fits.

A second contributor to instrument performance is the quality of the position encoding electronics. Shown in Figure 3. is a plot of the electronic gains and zeros for the ADCs in the IA-PEMs over the last several weeks. These gains and zeros are measured automatically, requiring only that the user lower the detector bias so that there are no signals from neutron events. The gain measurements are shown by the black line and points, the red line is the “gain target”. The zeros intercepts for the A- and B-Side ADCs are shown by the red an black lines in the lower plot. For good performance, it is only necessary that the gains be “stable” and that the zeros be close to zero. This performance is typical and makes it possible to be confident of diffraction experiments that may take several days to complete.

 
Figure 3. Electronic gains and zeros for Det00, PEM196. a, b) The electronic gain for the A- and B-side preamplifier-amplifier-ADC chains. The “Target slope is shown in red and the measurements by black triangles. c) The ADC zero-intercepts in channels for the A- and B-side ADCs.

A much simpler solution to peak broadening problem has been discovered in the course of the NCSU PNPD upgrade and implemented in the latest version of the IA Neutron Diffractometer Control System (NDCS).

When a rebin of the detector position data is performed with the new code (DSCone Rebin), the impact is immediately obvious (Compare to Figure 2a). The diffraction peaks for all of the detectors nicely overlay and the peak position derived from fitting the individual detector results to Gaussians has only minor deviations.

 
Figure 4a. Plot of diffraction data from low angle peak (~43o) for each of the detectors in the array and the data for the detector array sum using a DSConeRebin.
 

Figure 4b.  Peak position for each of the detector elements given by a DSConeRebin of the diffraction data. 

The expectation would be that all of the detector peak centers would be at the same scattering angle. As can be seen, this is not quite the case. The residual deviations are likely due to the tilt in the scattering plane (which was documented earlier at PULSTAR) but it may also result from placing the sample below the effective centerline of the beam.

The benefit of using the DSCone Rebin can also be seen in the Rietveld fits to the data. Figure 5. shows the Rietveld results using data from all 15 detectors for “Flat Rebin” and “DSCone Rebin”.

 
Figure 5. Rietvelt fits to the Ni diffraction data. a) Flat Rebin. b) DSCone Rebin. Statistical measures of the fit are shown in the figures.


Statistical summaries of the fit quality are shown in each figure. Given that this is the same data, the reduction in wR from 12% to 7.5% and the Chi2 from 9000 to 3500 is significant.

Comparing the 15 detector DSCone rebin to a 5 Det DSCone rebin, the statistics are:

  • 15Det DSCone Rebin
    wR = 7.48%, Chi2 = 3539, GOF 1.72

  • 5Det DSCone Rebin

    wR = 10.56%, Chi2 = 2369, GOF 1.41

These differences are again significant. You pay only a small price on going from 5 detectors to 15 detectors in wR and the Chi2 numbers should be compared by weighting them by the total number of counts – the 15 detector data has approximately 3 times more counts.

Currently, Rietveld fits to the data employ the standard assymetric peak broadening formula used for conventonal powder diffractometers. Implementation of a more accurate diffraction peak-shape model for the linear PSD array in the Rietveld code could make a further improvement to the Rietvelt results.

Finally, the detector resolution for 15, 9 and 5 detector elements as a function of scattering angle is shown. These curves are derived from the Cagliotti U, V and W parameters obtained from the Rietveld fits to the data for 15, 9 and 5 detectors respectively.

 
Figure 6. Diffractometer resolution FWHM calculated from the Rietveld fits comparing Flat and DSCone Rebin using different numbers of detectors. 

The FWHM curves should all have the value at “Flat Cone” (Scattering Angle = 90o) and the deviation observed here is believed to be due to a “tilt” in the scattering plane. The scattering plane is not horizontal. This effect can be observed in the analysis of the Ni diffraction peak at 89which is shown in Figure 7.

 
Figure 7. Peak position of the Ni diffraction peak at 89for each of the detector elements. 

This line should be horizontal and the tilt of the scattering plane injects a further broadening to the “sum” diffraction peak for 9, 11 and 15 detectors.

Further corrections to the rebinning algorithm might be able to correct for the diffraction plane tilt but will require additional research.