ISIJ International
Online ISSN : 1347-5460
Print ISSN : 0915-1559
ISSN-L : 0915-1559
Special Issue on "Frontier in Characterization of Materials and Processes for Steel Manufacturing"
A Probable Improvement of Wavelength Dispersive X-Ray Fluorescence Spectrometer for Steel Making
Jun Kawai
Author information
JOURNAL OPEN ACCESS FULL-TEXT HTML

2022 Volume 62 Issue 5 Pages 867-870

Details
Abstract

In steel manufacturing process, one X-ray fluorescence analyzer with 40 crystal spectrometers is sometimes used for elemental compositional monitoring to control the process. The present paper is a suggestion to improve the performance of this bulk of X-ray fluorescence spectrometers by replacing the proportional/scintillation counters by silicon drift detectors (SDD) with digital signal processors (DSP). The wavelength dispersive X-ray fluorescence spectrometer with SDDs will enable the automatic adjustment of the optimal measuring condition. The shortcomings of both SDD and proportional counters are discussed.

1. Introduction

The wavelength dispersive X-ray fluorescence (WD-XRF) spectrometers are used for steel making for controlling or monitoring the elemental compositions of iron and steels. Sometimes forty crystal spectrometers are assembled with one X-ray tube, and 40-elemental analyses are possible simultaneously as shown in Fig. 1, which is taken from Ono and Narita,1) where a flat crystal spectrometer with Soller slits2) and a scintillation counter, or a curved crystal spectrometer with pin-hole slits and a proportional counter, is found. A proportional counter is for soft X-ray counting, and a scintillation counter is for hard X-rays. The X-ray tube is an end-window water cooled type. The counting rates are usually a few tens of kcps (kilo counts per second). The deadtime ratio of a detector will increase as the increase of the counting rate, and the deadtime (τ) correction N 0 = N 1-Nτ is used to obtain the true counting rate N0, when we observe N counts in a second,3) but I will show in the present paper an example that this equation does not work well, due to the counting rate dependent amplitude of a detector. The purpose of the present paper is in order to give a proposal of improving an X-ray spectrometer of Fig. 1, by replacing the scintillation and proportional counters by Si drift detectors (Si-SDD) accompanied with a digital signal processor (DSP), in order to adjust the optimal operating condition by digital control. SDD was first reported more than 35 years ago,4) but a practical application was at the end of 1990s.5) SDD was the abbreviation of “semiconductor drift detector”5) at that time, but now many people use SDD as an abbreviation of “silicon drift detector” (major) or “surface drift detector” (minor).

Fig. 1.

A schematic illustration of WD-XRF spectrometer with 40 crystal spectrometers, modified from Ono and Narita,1) reproduced and modified with permission of the authors and ISIJ. (Online version in color.)

Murayama6) once (2005) compared the wavelength dispersive X-ray spectrometer (WDX) and energy dispersive X-ray spectrometer (EDX) for electron probe X-ray microanalyzers (EPMA), and concluded as follows:

(i) The energy resolution of EDX is inferior to that of WDX.

(ii) The sensitivity of EDX is inferior to that of WDX for especially low atomic number elements, such as oxygen, nitrogen, and carbon.

(iii) The quantitative analysis using WDX is based on the theoretical principle, and thus an open discussion is possible in order to improve the precision and accuracy of quantitative analysis. On the other hand, quantitation of EDX is not clear because the numerical algorithm is in the “black box” of EDX, in other words, the computer program is not open to users, and the algorithm is different depending on each EDX company, and sometimes program is changed without notice.

The Murayama’s opinion is quite appropriate, and thus the reliable quantitative analysis should be performed by the WDX spectrometer. The “black box” meant active analog circuits used in the Si(Li) solid state detector (SSD) before the age of 1990s and after 2000 the black box became DSP circuits in an SDD detector unit. The era between 2000 and 2010 were transient from Si(Li) detector with analog circuit in the Nuclear Instruments Module (NIM) to the SDD with digital circuit (DSP). The Si(Li) detector has the following shortcomings.

(i) Bias voltage is as high as several hundred volts (700 V to 1 kV). The increasing of bias voltage should be slow. The vacuum vessel should be evacuated down to high vacuum in order to avoid electric discharging, as well as heat insulation because liquid nitrogen is used.

(ii) The Si(Li) detector spectra changed during several hours, and thus energy calibration is needed, say, every 12 hours. Additionally, the spectra changed its profile for long term (months) history of its usage. The same model has different spectrum for the same specimen. Some examples are shown by Papp et al.7)

However the above shortcomings of Si(Li) SSD have been improved by SDD, because of the low bias voltage, of the very small vacuum can package, and of Peltier cooling down to −25°C. However if a very low noise measurement is needed, still liquid nitrogen cooled SSD is better, and for high energy X-ray detection, SSD with thick Si layer or Ge is better than SDD.

Those who want to improve the quantitative XRF algorithm, they want to check the X-ray pulse signal by an analog circuit and analog oscilloscope. Now the X-ray pulse is observable by a digital oscilloscope as shown in Fig. 2, which is a photo of computer display of a Si-PIN detector with DSP to mimic a digital oscilloscope. The “black box” problem of DSP of an EDX spectrometer is not existing, because we can use the DSP as a digital oscilloscope. “A DSP is a kind of digital oscilloscope”, which was repeatedly stated by late Tibor Papp, who passed away in December 2020 because of COVID-19. Papp presented at Denver X-ray Conference (DXC) in 2009, that the response function of silicon detectors were important for appropriate quality assurance.8) Nakaye and Kawai9) published a paper of “recording” the X-ray fluorescence spectra by using an audio input of a notebook computer as a DSP. Then improved by using digital music amplifier.10) Kawai et al.11) gave a workshop in the Denver X-ray Conference, how to use the DSP in order to measure the reliable X-ray spectra. These three publications from the present author’s group were owing to the statement of Papp.

Fig. 2.

Digital oscilloscope usage of SDD-DSP (Courtesy of Dayun Liu). (Online version in color.)

2. Proportional Counter

A proportional counter has also a complicated characteristics as follows. The center wire accumulates carbon contamination due to the decomposition of CH4 in PR (proportional) gas (Ar 90% and methane 10%). As the increase of the thickness of the carbon layer, the inhomogeneous electric field will asymmetrically broaden the line shape of the energy spectra as shown in Fig. 3.

Fig. 3.

Pulse height distribution of a proportional counter for monochromatic Ti Kα1 X-ray depending on the counting rate (top 68 cps, bottom 998 cps.), taken from Kawai.12)

(i) The high voltage e.g. 1750 V is usually applied to the center wire, and the voltage rising rate when the system is on, should be low in order to achieve the stable operation, which is quite similar to old SSD.

(ii) Most serious shortcoming of the proportional counter is the intensity dependent pulse height distribution.

The second point is not widely known, and I would like to explain in detail in the following paragraphs.

Figure 3 shows the pulse height distributions of a proportional counter,12) where the incident X-ray was titanium Kα1 (4.5 keV),13) which was monochromated by a double crystal spectrometer.14) The pulse height distributions in Fig. 3 were, in other words, EDX spectra of Ti Kα measured by the proportional counter as an energy dispersive X-ray spectrometer with an energy resolution of a few keV. The energy resolution of the double crystal monochrometer was less than 1 eV. The two pulse height distribution spectra shown in Fig. 3 were measured at the peak top and foothill of Ti Kα1. Therefore, the energy difference between the peak and foothill of Ti Kα1 was within 1 eV, which was negligible compared to the energy resolution of the proportional counter (a few keV). However, the energy was shifted as is found from Fig. 3 between the peak and foothill. The reason of this energy shift is the counting rate dependent gas amplification. When the X-ray intensity is as strong as 998 cps, the gas amplification is significantly reduced compared with the weak intensity as 68 cps. We must note here that 998 cps is not a high counting rate, it is rather low counting rate, and the 68 cps is very low counting rate. The energy spectra of a proportional counter will change its energy spectra depending on the counting rate. An escape peak, which is due to the argon Kα escape, is found in Fig. 3. It is better to include the escape peak for total titanium Kα, but when the low characteristic lines overlap the escape peak, and the inclusion of the escape peak is not always the optimal condition when we use the proportional counter. However when an SDD is used in place of a proportional counter, the energy resolution of an SDD becomes better than that of a proportional counter, and it is easier to select the optimal condition related to the inclusion of the escape peak automatically.

When we measure the X-ray fluorescence intensity by a proportional counter, the characteristic X-ray intensity is measured by setting a window of pulse height distribution, by limiting the upper and lower limit of voltages. However if the upper limits were decided by a low concentration specimen, then when a higher concentration sample is measured, the pulse height distribution will soak out the window, due to the counting rate dependent gas amplifying phenomena. The counting window should be changed as the change of the atomic number, and the window should be changed as the 2θ scan, as long as we use a proportional counter. Some companies change the applied voltage of the proportional counter when 2θ is scanned. Some companies change the gain of the main amplifier when 2θ is scanned, in order to adjust the pulse height distribution within the fixed window. Although most of the X-ray companies know of the presence of counting rate dependent gas amplification, but how to avoid the nonlinear response of the proportional counter is not open to us, users, this is also a “black box”.

3. SDD

SDD has at least two problems in order to replace the proportional counters in WD-XRF spectrometers. One is the thermal problem, and the other is problem associated with unknown black-box algorithm.

The thermal problem is described as follows. If we measure a soldering iron by an SDD, we can measure spectra by changing the temperature of the specimen from around 200 to 400°C. When Tanaka et al.15) compared the X-ray spectra by changing the temperature of the soldering iron at a fixed geometry, separated enough (say 10 cm) the distance between SDD detector and the soldering iron, the X-ray spectral intensity changed. In steel making analysis, the specimen temperature of 200°C is not rare. If we insert a cooking aluminum foil (15 μm thick), the intensity becomes slightly weak for the 3d-transition-metal element X-rays, but the temperature effect became a negligible level.

The second problem is associated with the numerical processing inside the DSP. If we measure an ED-XRF spectrum using an EDX detector, a sum peak11,16) appears. The sum peak has twice the energy of a single photon when two photons come to the detector simultaneously; the detector produces one pulse, but its height is twice the single photon. For iron and steel analysis, Fe Kα (6.4 keV) is usually the strongest peak, and we observe the sum peak at 12.8 keV, which overlaps the Pb Lβ (12.6 keV). Therefore, it is sometimes very difficult to analyze Pb, a trump element, in iron and steel analysis by an ED-XRF spectrometer. Arsenic, also a trump element, Kα (10.5 keV) overlaps Pb Lα.

However, the two X-ray photons come not always simultaneously, but the two photons come to the detector with a short time difference. In such a case the sum peak tails to the low energy side. This is often observed in the analog EDX spectra. However, when the same spectrum is recorded by the DSP, the tails of the sum peak are not observable.11) The sum peaks become sharp symmetric peaks without tails when recorded by the DSP. This is due to a numerical data processing in the DSP. It is true that the narrower sum peak without tail is better from the view point of avoiding the overlap of peaks, but such kind of black box was not welcome for the iron and steel analysis. This kind of problem should be solved by collaborations among steel companies, the X-ray company, and an SDD company.

If we check the commercially available SDD detectors, some SDD has both analog and USB outputs, but other SDD has only digital output. We must use two-way output SDD in order to check both the analog and digital spectra.

The size the specimen of WD-XRF analysis is usually large. The diameter of the specimen is usually larger than one inch, and sometimes rotated during measurement. This is because the wider area is needed for high precision in quantitative analysis.17) Therefore the X-ray beam size at the detector is larger than the window size of usual SDD detectors used, and we must check the window size effect when we use the flat crystal and Soller slits spectrometer.

4. Concluding Remarks

I have suggested and proposed a method to improve the multi crystal WD-XRF spectrometer for process control. The goal of the present suggestion is an automatic digital control of the parameters of the WD-XRF spectrometer, which enables us an automatic adjustment of the optimal measuring condition, in order to improve the precision of the WD-XRF analysis. This is probable based on the combination of (i) higher energy resolution of SDD compared with a proportional counter, (ii) low bias voltage, and (iii) digital control.

The progress of SDD is now quite rapid. The window of the SDD was only beryllium and polymer a few years ago, but windowless and graphene has been added. The windowless is good for light element detection, but not suitable for hot specimens. Polymer window is not appropriate for hot (IR radiation) and visible light emitted objects. Usually the XRF spectra are displayed after 2θ to energy axis conversion, but we must include the intensity factor when we convert the WDX spectra into EDX spectra, which was found less than 10 years ago.18)

Acknowledgements

Thanks are due to Liu Dayun (master student) for Fig. 2, and Ryohei Tanaka (Assistant Professor) for temperature dependent X-ray data. The present research was partially supported by The Iron and Steel Institute of Japan (ISIJ) through the Research Group of Non-Destructive/On-Site Analysis. The use of Fig. 1 after modified has been permitted by Akihiro Ono representing the authors and ISIJ as publishing institute.

References
 
© 2022 The Iron and Steel Institute of Japan.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs license.
https://creativecommons.org/licenses/by-nc-nd/4.0/
feedback
Top