NEW POSSIBILITIES OF PHASE, TEXTURE, AND STRESS ANALYSIS USING ENERGY DISPERSIVE DETECTORS

For angular dispersive and energy dispersive phase, texture, and stress analysis of com- pact samples (TiA1) and thin TiN layers, an electroluminescence detector and a Si(Li) semiconductor detector were applied. The parameters and application properties of the detectors are described. For angular dispersive measurements higher reflection intensities and better peak-to-background ratios can be attained, in comparison to conventional methods, by using multichannel pulse accumulation and mathematical spectrum evaluation. In energy dispersive texture and stress measurements, reflection data from several lattice planes can be obtained simultaneously. In stress measurement, variations of lattice spacings in the order of 10 -3 are reliably detectable. In principle, texture and stress information can be received from one measurement.


INTRODUCTION
For determining the texture of substances of low crystal symmetry or of thin surface layers, weak reflections must be included in general. This often leads to an insufficient accuracy and to long measuring times because of small intensities and poor peak-to-background ratios. Due to beam defocussing, in most cases the conditions cannot be improved by using a crystal monochromator behind of the sample. In addition, the reflection efficiency of, e.g. a graphite monochromator is not better than 20-25%. However, the application of detectors with high energy resolution and a sufficiently large entrance window allows a nearly lossless measurement of peak intensity at strongly reduced background. In previous investigations (Moras.and Richter, 1992), we attained a three to five-fold intensity gain by using an electroluminescence detector instead of a secondary graphite monochromator for phase analysis of ttm-TiN layers on steel.
The aim ofthe present work is to investigate the principal possibilities of the application of detectors of different energy resolution (electroluminescence and semiconductor detectors) in the angular dispersive or energy dispersive mode for phase, texture, and stress analysis. At the same time, the advantages and limitations of single channel versus multichannel pulse accumulation including mathematical spectrum evaluation are analyzed. The   X-ray photons, drift in the relatively weak field into the luminescence zone. There the field strength is adjusted to such a value, that the Xe atoms are being excited to optical luminescence, but that no impact ionization occurs. The gas amplification used in proportional counters is replaced here by a light amplification. The light photons are registered by the coupled photomultiplier (PMT). The relations are so, that the energy resolution is determined practically by the statistics of the primary ionization alone; the contributions of the light and PMTamplification are negligible. That is why the energetic resolution of the LUX detector is better by a factor 2 than that of a proportional counter, or by a factor 4 in comparison with a scintillation detector.

RESULTS AND DISCUSSION
Some experiments on angular dispersive and energy dispersive diffractometry were carried out on TiN layers. Figure 2 shows the energy spectrum of a TiN layer on steel measured with a Si(Li) detector at 2 20 (background position). On the left side the fluorescence peaks from Ti and Fe are seen, additionally the electronically generated zero peak for calibration purposes. The Cu peaks are from diffuse scattering of copper radiation (they are no Bragg peaks). On the right side there are Laue peaks caused by the reflection ofwhite radiation at TiN lattice planes. These peaks shift to lower energies when 2t increases.
In Fig. 3 the spectrum of a TiN(Fe) layer on a Si substrate is shown, measured with a LUX detector at low incidence angle and at the (220) reflection position (2 62). The energy resolution is a bit worse here so that there is some overlapping between Cu and Fe peaks.
On the basis of these spectra the different measuring modes for diffraetometry with energy dispersive detectors can be illustrated. occurs, that means in many cases when using a semiconductor detector but only relatively seldom with the electroluminescence detector.
(b) Angular dispersive diffractometry with a multichannel analyzer (MCA). In this mode, at each angular position the whole pulse height spectrum is being recorded and a certain range around the CuK peak is evaluated by peak profile analysis in order to correct for peak overlapping and background. In the ideal case, this procedure results in almost 100% monochromatic Ks intensity.
Results using the LUX detector for this method are shown below. (c) Energy dispersive diffractometry with an MCA. The energy spectrum is recorded at suitable angular positions and the Laue peaks are evaluated for peak position and integral intensity, if necessary, including overlap correction. The simultaneously measured Bragg and fluorescence peaks can be used for calibration purposes, the latter ones can give additional information on chemical composition or the thickness of metallic layers. For this mode, the Si(Li) detector is more suitable due to its higher energy resolution. The typically recordable energy range is from 10 to 40 keV which corresponds to a d-value range, e.g., from 0.1 to 0.4 nm if a 2 angle of 20 is used.
These measurement and evaluation methods can be used in connection with different diffractometer techniques, such as scans for phase analysis, for texture, or for stress analysis.
In a previous investigation (Bermig et al., 1993) texture analysis of a Ti-A1 sample containing the tetragonal "-and hexagonal c2-phases was carried out. For this analysis pole figures of some weak reflections had to be measured. Figure 4(a) shows the pulse height spectra of the sample measured with a scintillation (curve b) and an electroluminescence detector (curve a). It is evident that, due to the strong Ti fluorescence, with the scintillation counter no acceptable peak to background ratio can be achieved. On the other hand, using an electroluminescence detector in single channel mode, the Cu and Ti peaks could be completely separated and a well evaluable diagram was attained ( Fig. 4(b)). This analysis could be carried out successfully only by using such an energy resolving detector.
Using the LUX detector, the conventional and the new detection modes were compared for Ti-Al phase analysis. In Fig. 5  diffraction diagram is shown measured (a) with a secondary graphite monochromator + scintillation counter, (b) using a LUX detector with single channel analyzer, (c)with LUX detector + multichannel analyzer including mathematical curve fit. The use of a graphite monochromator succeeds in a very low background but leads to weak reflection intensities. With the LUX detector in SCA mode, high reflection intensities but also a higher background are attained. The application of the LUX detector in MCA mode results in high reflection intensities at a low background level almost as with graphite monochromator, corresponding a high peak-to-background ratio and good detection sensitivity. It has to be remarked, that for the last measurement a relatively simple evaluation of the energy spectrum was made, i.e. a curve fit including only the CuK and K peaks and a linear background function.
The evaluation algorithm in a much more complicated case of the energy spectrum was tested, carrying out an angular dispersive diffraction scan with MCA accumulation at the above-mentioned TiN(Fe) layer on Si substrate. See Figs. 3 and 6 for the energy spectrum, the first one at the (220) maximum, the second one at a background position, where the CuK peak is missing, at constant (low) incidence angle w. There are six overlapping peaks. The Ti peaks should not really affect on the CuK peak, but they were included here for testing. In some preliminary tests it was found that Gauss functions are suitable to fit the peaks, but that some narrow, physically founded, bounds and constraints for the peak parameters are necessary during the fitting procedure in order to ensure the convergence of the algorithm. Figure 7 shows a part of the diffraction diagram. The points of the lower curve represent the integral over the CuK peak function corrected for background. It must be pointed out, that exactly the same algorithm, without parameter changes, was applied to all Bragg angles.
That means an automatic fitting procedure for the whole diagram is possible.
For comparison, in the upper curve of Fig. 7, the measured values for SCA mode are shown, corresponding to a "window" from ch 300 to ch 344 (see Fig. 3). It is evident, that by the MCA mode, including spectrum evaluation, the background is reduced to 25% whereas the net intensity remains at approximately 90%. The peak-to-background ratio is improved by a factor of 2.5. Compared to a measurement with a secondary graphite monochromator it can be estimated, that an improvement of the intensity-to-background ratio by a factor of 5-8 should be attainable. (Remark: Because it was only a test investigation, the angular steps in Fig. 7 were chosen relatively large (0.5), too large for good angular resolution.) As an example for energy dispersive diffractometry using the Si(Li) detector we have investigated the possibilities and sensivity of texture and stress determination in a TiN layer on steel. For the energy spectrum of this sample look at Fig. 2  angle 20 , there is no overlapping with fluorescence or CuK Bragg peaks. The spectrum was accumulated for five tilting angles X between 0 and 50 at o=0 , and the six Laue peaks were fitted by Gauss functions in a single procedure. An overlapping correction was necessary only for the TIN(111) peak with the peak at ch 423. In Fig. 8 the dependence of the integral intensity of the six peaks on the angle X is shown, this corresponds, in principle, to sections across the pole figures for each peak at o 0. It must be remarked here, that the measuring conditions here were not optimum for texture analysis (too large steps) because they were chosen for stress analysis (see below). The measured spectra for the various angles were evaluated also concerning peak shift which is to be expected if the sample has internal stress. In Fig. 9 the dependence of peak position and d-value on the x-angle is shown for the (220) reflection. The following conclusions can be drawn from the results.
(a) The observed peak shifts are small. This requires a high precision of the determination of the peak position, that means a sufficiently good counting statistics and an accurate peak approximation. (b) Changes of the lattice constant d in the order of 10 -3 are reliably detectable.
It should be noted, that the used Si(Li) detector with an energy resolution of < 250 eV was not optimal for stress investigations. With a detector of better resolution (e.g. _< 140eV) a higher precision for d value determination should be attainable.
In summary, it must be emphasized, that from the measured and evaluated energy dispersive spectra, the texture and the stress information ofthe (in this example) six peaks could be obtained simultaneously. In practise, however, this requires a compromise concerning the measurement conditions (point or line focus ofthe tube; slits or collimators).

CONCLUSIONS
The investigations concerning employment of energy dispersive detectors in the X-ray diffraction phase, texture, and stress analysis lead to the following remarkable results.
In the angular dispersive phase and texture analysis, by the application of energy dispersive detectors in connection with multichannel pulse accumulation and mathematical spectrum evaluation, substantially higher peak intensities and better peakto-background ratios can be attained in comparison to conventional methods.
For angular dispersive analysis with an electroluminescence detector, especially with the newly developed type LUX, good results are obtained. For energy dispersive texture and stress analysis the Si(Li) semiconductor detector is more suitable due to its better energy resolution.
In energy dispersive texture and stress analysis the parameters of several diffraction peaks can be recorded simultaneously. In stress measurement, variations of the lattice spacing in the order of 10 -3 are reliably detectable. Making a compromise in optimum measuring conditions, texture and stress information can be obtained simultaneously from the same measurement.