- Email: [email protected]
Application of Si-strip technology to X-ray diffraction instrumentation
Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Application of Si-strip technology to X-ray diffraction instrumentation a ¨ E. Gerndt a,n, W. Da˛browski b, L. Brugemann , J. Fink a, K. S´wientek b, P. Wia˛cek b a b
Bruker AXS, Karlsruhe, Germany Faculty of Physics and Applied Computer Science, AGH University of Science and Technology, Krakow, Poland
a r t i c l e in f o
a b s t r a c t
Available online 8 June 2010
We describe the successful technology transfer of High Energy Physics (HEP) silicon-strip detectors for tracking of minimum ionising particles (MIPS) to industrial X-ray diffraction instruments. In our application the detector is used to measure 1-D intensity profiles of low-energy photons. The challenges of such an application are low noise because of the relatively low energy of X-ray photons, from 5 to 22 keV, and high count rate capability. The technical implementation, with a focus on custom designed front-end electronics and optimisation of strip geometry taking into account the charge division effects, is shown and the achieved performance is summarized. The detector was launched several years ago and we report on the in-field experience. Lastly, we describe several scientific applications. & 2010 Elsevier B.V. All rights reserved.
Keywords: X-ray diffraction X-ray detectors Silicon strip detectors Spatial resolution Energy resolution Charge division Front-end electronics Low noise electronics
1. Introduction The combined technologies of silicon strip detectors and Application Specific Integrated Circuits (ASICS) have enabled 1-D position sensitive X-ray detectors to be designed to fulfil specific requirements. This combined technology is widely used in HEP experiments for building high precision tracking detectors. As a mature technology, it is a suitable one for building a detector for in-house commercial instruments used for X-ray diffraction, X-ray reflection, and X-ray scattering measurements. One-dimensional X-ray diffraction (XRD) detectors were quite popular in academics and research for a long time. They enabled special X-ray diffraction investigations also in a laboratory. One example is the kinetic study of materials under non-ambient conditions. However, due to their limited count rate capabilities, the quality of the data in terms of measured intensity and of peak position stability was considerably low. An overview of XRD detectors, the specific requirements, and relative performance can be found in Ref. [1]. With the introduction of silicon strip detectors this limitation was removed. Furthermore, due to the enormous count rate capability, robustness, and ease of use, new applications in X-ray diffraction were opened. In its first instance, the novel instrument consists of an X-ray source, primary beam modifying optics, e.g. focussing or monochromating optics, a sample, and the 1-D detector with optical components suppressing scattered radiation. The whole set-up is arranged in Bragg– Brentano geometry obeying the y–2y condition during the entire
n
Corresponding author. Tel.: +49 721 595 7080; fax: + 49 721 595 2317. E-mail address: [email protected] (E. Gerndt).
0168-9002/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2010.05.032
measurement. The 1-D detector replaces the 0-D detector in the classical set-up. This set-up allows powder diffraction measurements for qualitative and quantitative crystalline phase identification. Results for this type of X-ray diffraction, in addition to more elaborate measurements using different set-ups, are shown in Section 5 of this paper. The performance of a new technology like the silicon strip detector can easily be checked by using standard reference material (SRM) provided by the National Institute of Standard and Technology (NIST, USA). Various certified solid or loose powder samples are available for comparison measurements. An important quality criterion of a detector is its angular resolution measured as full width at half maximum (FWHM) for given reflections. Fig. 1 shows a measurement recorded on an LaB6 powder sample (SRM 660a) using the developed silicon strip detector. The FWHM of 0.0421 matches very well the value, which can be expected from the instrument set-up. There is no loss in angular resolution when compared to a point detector. This shows even though the silicon strip detector acts as 192 point detectors, recording data at the same time at slightly different diffraction angles, the data quality is equivalent to a single point detector, but measured 192 times faster. It is worth mentioning that the ideas to build a 1-D detector for powder diffraction, based on silicon strip detector technology, have been presented many years ago [2,3]. However, to our knowledge, none of these projects has materialised as a useful instrument addressing properly the requirements of the diffraction technique. Performance and measurement capability of both prototype detectors reported in the above mentioned papers were severely limited by the performance and quality of the readout electronics. While the silicon strip detector technologies, as
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
351
2.1. Detection efficiency The considered range of X-ray energy is from 5.4 up to 22 keV, while the majority of applications uses 8 keV copper X-ray tubes. The typical wafer thickness of 300 mm used in silicon strip detectors for HEP experiments appear to match well our requirement as it provides 98% absorption for 8 keV X-rays. Using a thicker detector would provide higher efficiency for X-ray energies above 8 keV; however, this would result in larger charge sharing between the neighbouring strips and compromising the single strip energy resolution due to diffusion of charge generated in the detector volume. These effects are related directly to the strip geometry and will be discussed later. Therefore, we have decided to develop an optimised design for typical applications based on 300 mm thick silicon wafers and an optional design for higher X-ray energies based on 500 mm thick wafers. Fig. 1. Measurement of the angular resolution using the standard reference material 660a (LaB6). The resolution achieved is equivalent to a point detector (0-D detector) with a 0.1 mm receiving slit (diffractometer D2 Phaser with 1.51 Soller slit1 and 0.1 mm divergence slit).
developed for high energy particle physics experiments, can be applied directly to measurement of X-rays, one needs a readout ASIC of different readout architecture and much better performance to build a useful instrument. The basic requirements for the ASIC are as follows:
counting mode with an energy window discriminator low equivalent noise charge (ENC), below 150 electrons rms, at
room temperature for a detector capacitance of 2.4 pF, allowing for discrimination between noise and signal of 1500 electrons corresponding to X-ray energy of 5.4 keV linear front-end circuitry up to an input charge of 6200 electrons (22.2 keV in Si) count rate capability significantly higher than 106 cps storage data buffers on the chip and multiplexed data readout possibility of simultaneous data taking and data readout low power dissipation of 3 mW per channel allowing operation without active cooling.
2. Silicon strip detector We have selected a conventional single-sided silicon strip detector structure with AC-coupled p + strips in high resistivity n-type silicon bulk as the sensor for our detector. Such a choice was driven, firstly, by the requirement to use a mature sensor technology and, secondly, by specific technical requirements discussed below. The strip length of 16 mm is determined by the geometry of the X-ray beam in the diffraction instruments. This allows us to maximise the use of the beam, although shorter strips would result in lower noise and better energy resolution. However, since the intention was to use the detector in preexisting instruments with a given geometry it was decided to use the full possible length. Also, the strip pitch was chosen due to the available instrumentation for which the detector was designed. Due to the commonly achieved focal spot dimensions of sealed Xray tubes, a pitch width of much less than 100 mm would no lead to better spatial resolution of the system. 1 A so called Soller slit consists of parallel plates limiting beam divergence in one direction. It is typically defined by its opening angle.
2.2. Bias voltage and leakage current The typical full depletion voltage of 300 mm thick sensors is about 60 V. This parameter is not very critical as we want the sensor to be biased with much higher voltages, up to 250 V, in order to ensure a short charge collection time and to minimise division of charge between neighbouring strips. No active cooling is foreseen for the detector. Given that the detector is operated in the diffraction instrument at temperatures typically about 10 1C above room temperature, we have assumed a maximum leakage current of up to 1 nA per strip. This upper level of the strip leakage current is set by the requirement that the parallel current noise generated by the detector leakage current should not contribute significantly to the equivalent noise charge (ENC) of the overall system. For a given voltage noise of the preamplifier and detector capacitance this limits the peaking time of the front-end electronics to be below 300 ns. 2.3. Bias resistance Because the noise contribution of the strip leakage current cannot be made completely negligible we have used the structure with AC-coupled strips. Using DC-coupling would require an additional circuitry in the front-end electronics to compensate the DC voltage shifts in the preamplifier circuit. Such a circuitry would be an additional source of current noise. In first approximation, this circuitry would double the noise due to the detector leakage current. For the AC-coupled strip, the bias resistance becomes another critical parameter as an additional source of parallel current noise at the preamplifier input. The assumed maximum leakage current of 1 nA corresponds to an equivalent noise resistance of 50 MO. In order to avoid additional noise contribution from the bias resistor, it should be made at least an order of magnitude higher, i.e. 500 MO. Such high values are achievable only with FOXFET structures [4]. 2.4. Spatial resolution Primarily, the silicon strip detector should replace the 0-D detectors used in scanning mode so that the strip pitch should be of the same order of magnitude as the slit used with the 0-D detector. It is worth noting a different approach to the spatial resolution of silicon strip detectors used for X-ray detection compared to the common approach used for tracking relativistic charged particles in HEP experiments. In tracking detectors of charged particles one can take advantage of charge division effects between the strips and improve the spatial resolution by estimating the centre of gravity of signal amplitudes measured for
352
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
a cluster of strips. In order to make the charge sharing between several strips in the cluster more linear one can introduce intermediate strips, which are not connected to the readout electronics. They provide additional purposely introduced capacitive charge division. For relativistic charge particles spatial resolution better than 2 mm rms has been achieved [5] using silicon strip detectors with a strip pitch of 25 mm and readout electronics applying a 50 mm pitch. In order to take advantage of the charge division technique one needs readout electronics, which records signal amplitudes with a high signal-to-noise ratio. For minimum ionising particles the most probable signal generated in a typical 300 mm thick silicon strip detector is 24 000 electrons. By proper optimisation of the front-end electronics a signal-to-noise ratio of the order of 50 is achievable. This is needed to reach the ultimate spatial resolution mentioned above. The charge generated in the silicon detector by an 8 keV photon is about 10 times smaller than the charge generated by a minimum ionising particle. Achieving the required high signal-tonoise ratio would demand an extraordinary low noise of the frontend electronics. This is not achievable at room temperature. Furthermore, employing charge division effects for improvement of the spatial resolution requires recording signal amplitudes. In combination with the high count rate capability this would require massive Analogue-to-Digital Converters for processing the data in real time. Therefore, the only practical approach is to use the binary architecture of the front-end electronics, i.e. pulses with amplitudes in between a lower and an upper threshold are counted as good events. Subsequently, charge sharing effects should be minimised. Ideally, in the considered application of a strip detector to X-ray diffraction, the spatial response of the strip, i.e. the ratio of the number of incident photons to the number of counted photons, should be flat across the whole strip area.
2.5. Energy resolution Typically, the energy resolution of silicon strip detectors is defined with respect to the total charge generated in the detector volume by photons or charged particles. Since the generated charge may be divided between several strips, depending on the strip geometry, this requires summing up of the signals from the cluster of strips for each event. In the considered application, assuming readout electronics with binary architecture, no information on the pulse amplitude is preserved so that an appropriate approach is to define the energy resolution for the distribution of signals measured on a single strip. This distribution is affected by the electronic noise of the front-end electronics and by the charge division effects. Both the electronics noise and the charge division depend on the strip geometry and optimising these two parameters leads to contradictory requirements concerning the strip pitch and the strip width. In XRD applications the strip pitch is primarily defined by the required spatial resolution. In order to minimise the charge division effects one would rather make wide strips and narrow gaps between the strips. Such geometry results in a small fraction of events with significant charge sharing between two adjacent strips, which degrades the single strip energy resolution. On the other hand, narrow gaps between the strips result in large inter-strip capacitance, which affects directly the ENC. Given that the charge signals generated by X-rays are small, about 2200 electrons for 8 keV photons, the ENC is the primary parameter, which determines the single strip energy resolution. The additional point to be considered is a trade-off between the inter-strip capacitance and the strip series resistance. The strip series resistance contributes to the total equivalent voltage noise at the preamplifier input. Since we aim to achieve a low equivalent noise resistance of the front-end electronics, of the
order of 100 O, the serial resistance of aluminium readout strips, of the order of 35 OXcm, has to be taken into account. The final choice of strip geometry has been made taking into account the expected ENC of the front-end electronics to be about 100 electron rms and the charge sharing effects. The charge sharing effects have been estimated by Monte Carlo simulations [6] and measurements of the energy and spatial response performed for the prototype silicon strip detectors [7]. Based on these results we have chosen the strip pitch of 75 mm and the strip width of about 20 mm. Degradation of the single strip energy resolution due to charge sharing is of the order of 10% for the ENC level of 100 electron rms, and it depends on the X-ray energy, detector bias voltage, and also on the side (strip or ohmic contact) from which the detector is irradiated. More details can be found in Refs. [6,7]. Making the strips wider, above 20 mm, would reduce the charge sharing effects only slightly, especially for the detector irradiated from the ohmic side, while the increase of the inter-strip capacitance would have a predominant effect on electronic noise. The silicon strip detector has been designed in close collaboration with Sintef. The detectors for series production are delivered by Sintef according to the mutually agreed specification and acceptance tests.
3. Readout ASIC A simplified block diagram of the complete readout ASIC is shown in Fig. 2. The chip comprises six basic blocks: 64 analogue front-end channels, 128 18-bit counters, back-end buffers, control block, calibration circuit, and bias generators. Each channel is equipped with two pulse height discriminators providing the LOW and the HIGH discrimination level of the analogue signals. The output pulses from each discriminator are stored in an 18-bit counter. The counters are grouped in the blocks of 16 counters each and the data from each block is read out via an 8-bit bus with tristate outputs. The readout of counter content is performed in two steps: (1) parallel transfer of all bits into a buffer register and (2) serial readout of the blocks of eight channels. This scheme allows one to continue data taking while the buffer register is read out. Sequences of data collection and readout of the data from the counters are controlled by a control block. The control block provides also settings of the digital-to-analogue converters (DACs), which are used in the front-end block to control the discriminator threshold, bias currents in the analogue circuits, and the amplitudes of the internally generated test pulses. The control block receives the commands via an external serial link, decodes them, and executes by sending the control signals and data to other blocks. Up to eight chips can be addressed geographically and connected to a common bus. The chip ID is defined by three bits on three address inputs to be wire bonded to external logic levels. The ASIC has been designed and manufactured in the 0.35 mm CMOS process by Austria Microsystems. The prototype designs were manufactured on Multi-Project Wafers in the frame of service provided by EUROPRACTICE. 3.1. Front-end architecture The block diagram of the front-end channel is shown in Fig. 3. The channel consists of a charge sensitive preamplifier, a shaping amplifier, and two pulse height discriminators. The bias current in the input stage and the value of the feedback resistor in the preamplifier are controlled by internal DACs implemented on the chip. These allow for optimisation of the noise performance according to the parameters of the silicon strip detector, leakage current, and strip capacitance. The bias current in the shaper stage
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
353
Fig. 2. Block diagram of the 64-channel readout ASIC.
Fig. 3. Block diagram of the front-end channel.
and the value of the feedback resistor in the shaper are controlled by internal DACs. These allow for tuning the shaping time constant and the gain of the shaper stage. The discrimination threshold in each discriminator is controlled by a differential voltage delivered from an internal voltage generator controlled by a DAC.
3.2. Front-end performance Because of the binary readout architecture the analogue parameters of the front-end circuit, the gain of the preamplifiershaper circuit, and the discriminator offset can be extracted from multiple measurements for different discrimination thresholds,
354
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
covering the interesting range of signal amplitudes at the discriminator input. A primary result of such a discrimination threshold scan is an integral distribution of the signal amplitudes, which in the case of a mono-energetic signal and Gaussian noise is described by the complementary error function. By fitting the measured integral distribution to the complementary error function one obtains the signal amplitude and the standard deviation of noise. In the case of a more complex spectrum one can differentiate numerically the integral distribution to obtain amplitude distribution. Example of test results for all 64 channels in one ASIC obtained with the test signals generated internally in the ASIC is shown in Fig. 4. From the threshold scan curves we extract the rms value of noise at the discriminator input of each channel and from the response curves we extract the gain of the channels up to the discriminator inputs and the discriminator offsets. Having estimated noise and gain at the discriminator input we obtain the ENC for each channel. Example distributions of gain offset and ENC for one ASIC are shown in Fig. 5. The spread of discriminator offsets within one ASIC is typically 22 electron rms input equivalent and the relative spread of the gain is 0.8% rms. Thus, the uniformity of the lower discrimination threshold is primarily determined by the offset spread and is small compared to the electronic noise.
iCAL= 7
100
11
15
19
23
27[LSB]
counts
80
60
40
20
4. System architecture 4.1. Detector head and detector control unit
0 0
100
200
300 Uth [mV]
400
500
600
550 500 450 400 Uth [mV]
One should note that the parameters of the front-end circuit, gain, offset, and ENC, as well as pulse high distribution for X-rays are obtained in the test mode by scanning the discrimination threshold. These data are used to set up the working thresholds for operation of the system in the counting mode as used in the diffraction measurements. For a sensor bonded to the ASIC and working at room temperature the ENC is 115 electron rms that corresponds to the energy resolution of 985 eV FWHM. This does not take into account charge division effects in the detector. The energy resolution is due to electronic noise only and independent of the energy. Performance of the detector, exposed to a 57Co X-ray source from the strip side, is illustrated in Fig. 6. It shows the amplitude distributions for 64 channels obtained by numerical differentiation of the threshold scan curves. In order to minimise the effects of charge sharing the spectrum has been measured for a detector bias of 270 V. The source emits X-rays of energies 6.4, 7.06, and 14.41 keV. The lines 6.4 and 7.06 keV are Ka and Kb of iron, respectively. The intensity of the Kb line relative to the intensity of the Ka is only 15% and obviously the two lines cannot be distinguished given the energy resolution of about 1 keV FWHM. The low intensity peak around energy of 20 keV comes from coincidences of 6.4 and 14.41 keV photons (‘‘sum peak’’). The average energy resolution obtained is 1.12 keV FWHM for the 6.4 keV line and 1.06 keV FWHM for the 14.41 keV line. The resolution of 1.06 keV for the 14.41 keV line is only slightly worse than the resolution of 0.985 keV due to electronic noise. This observation confirms that, at sufficiently high bias voltage, the charge division effects are indeed limited to a small fraction of photons. The 6.4 keV peak is somewhat broader because of the contribution of Kb line of 7.06 keV. The continuous and almost flat background between the peaks comes from events with charge sharing between neighbouring strips. The energy resolution of the lower energy line is thus affected by the photons of higher energies absorbed in the inter-strip regions, for which the charge is shared between neighbouring strips.
350 300 250 200 150 100 1500 2000 2500 3000 3500 4000 4500 5000 5500 Qin [e-]
Fig. 4. Example test results of analogue parameters of the front-end circuits: (a) threshold scan curves for 64 channels for six different test signals and (b) response curves for 64 channels extracted from the threshold scan.
The detector head of the implemented silicon strip detector system houses a sensor module, components for mechanical alignment and X-ray optics. The X-ray optics can comprise elements reducing beam divergence or also to reduce background due to air scattered X-rays. On the sensor module, one AC-type silicon strip sensor with 192 strips and three ASICs are grouped together and connected with interface circuitries. The irradiation of the sensor module is made from the back side of the silicon strip sensor. For matching the thermal expansion of the silicon dies and the printed circuit board a ceramic material based on Al2O3 had been chosen as substrate. In comparison to the printed circuit board material like FR4, Al2O3 has a greater heat conductance for better removal of the dissipated electrical power. The detector control unit consists of the Data Acquisition Board, Main Control CPU board, very low noise power supplies for the bias supply of the silicon strip sensor (0–500 V), and the supply of the three ASICs. An FPGA device located on the Data Acquisition Board configures the three ASICs dynamically, controls the data collection, processes the measured data and forwards the results to the Main CPU board. All communication between the Data Acquisition Board and the detector head is performed via serial LVDS lines running at a clock rate of 10 MHz. The contents of the 384 counters for the LOW and HIGH
number of channels
0.15 0.1 0.05 µ = 0.10435 ± 0.00082 [mV/e-] (0.78 %) 0
10
20
30
40
50
µ = 1.6922 ± 1.6 [mV]
0 0.1
0
10
20
30
40
50
40 20
10
0
60
60
µ = 46.5703 ± 4.2 [e-] (9.0%) 10
20
30 channel
10
0 20
40
50
0 10 Offset [mV]
20
0 0
0.105 0.11 Gain [mV/e-]
20
-10
number of channels
Offset [mV]
0
-5
ENC (peak-1400 e-) [e-]
10
60
5
355
20
0
number of channels
Gain (peak-1400 e-) [mV/e-]
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
60
40 ENC [e-]
60
Fig. 5. Typical distribution of gain, offset, and ENC across 64 channels in one ASIC.
synchronization between the scanning motor drive and the data collecting FPGA. This is accomplished by using motor clock pulses for triggering. 4.2. Diffractometer system with silicon strip detector The block diagram for a silicon strip detector embedded into a diffractometer system is shown in Fig. 8. Besides the radiation detector, the main components of an XRD instrument are the X-ray tube including X-ray shutter, motorized goniometer, and sample stages. These are controlled and monitored by the diffractometer control electronics. Through the motor clock line, there is a very tight synchronisation between the spatial movement of the master scanning axis and the data collection within the detector control unit. Both diffractometer control electronics and detector control unit are connected via network with an external computer running the measurement server program and evaluation software packages. Fig. 6. Spectrum using a 57Co X-ray source. All 64 strips of a silicon strip detector are measured simultaneously.
4.3. Measurement modes
discriminators are read out in parallel via eight LVDS lines in order to achieve a minimum readout time Fig. 7. The Main Control CPU runs the LINUX-based control software that provides high level measurement commands and communicates directly with the host programs via an Ethernet interface. Some of the measurement modes need a very precise
The electronics chain allows several ways to perform measurements. Typically the detector is at a fixed radius from the sample and moving with a constant angular velocity. The count rates of the individual strips are assigned online to give 2y angle. All measurements to a given position are summed up. During the movement, the detector collects data for a short time frame, latching the counter values, and then continuing with the
356
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
Fig. 7. Block diagram of the detector head and the detector control unit.
Fig. 8. Electronics architecture of X-ray diffraction instrument with silicon strip detector.
collection of the next measurement frame. The exposure time per measurement frame is typically in the range of about 10 ms and is set by software. During this continuous scan the strips are essentially used as parallel detectors. The number of strips (192) determines approximately the gain in measurement speed when compared to a measurement with a point detector.
It is also possible to take data with the detector at a fixed position. Again, the electronics assigns angular positions to the respective strips. The strip synchronization with respect to the angular position of the 2y-arm of the goniometer is no longer needed. The angular range is given by the radius and the opening of the detector, in our case 14.4 mm. This so called snapshot mode
Counts
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
46 000 44 000 42 000 40 000 38 000 36 000 34 000 32 000 30 000 28 000 26 000 24 000 22 000 20 000 18 000 16 000 14 000 12 000 10 000 8 000 6 000 4 000 2 000 0 -2 000 -4 000 -6 000 -8 000
Quartz Hematite Gibbsite Boehmite Goethite Anatase Kaolinite (BISH) Quartz Hematite Gibbsite Boehmite Goethite Anatase Kaolinite (BISH)
10
12
14
16
18
20
22
24
26
28
357
5.64 % 8.59 % 55.34 % 14.70 % 8.77 % 1.79 % 5.17 % 5.57 % 8.95 % 57.15 % 13.44 % 8.69 % 1.82 % 4.37 %
30
32
2Th Degrees Fig. 9. Phase quantification of bauxite. Two measurements are overlaid; the lower curve was measured while setting the energy discriminator window narrow to suppress Fe fluorescence. The given quantitative composition of the sample was determined by Rietveld analysis using TOPAS software [8].2
allows time resolved measurements, for example, if the evolution of phase transitions should be observed. It is also possible to sum the counts from all strips. In this case the one dimensional information is lost, but the 0-D measurement allows more than two orders of magnitude large count rates, mostly defined by the number of strips irradiated.
5. Applications The most frequent application in X-ray diffraction is qualitative and quantitative crystalline phase identification. A reliable 1-D detector is of essential importance in case diffraction is used for quality or process control in industries like minerals and mining, cement, or pharmaceuticals. Fig. 9 shows two measurements of a bauxite sample. Bauxite is a common synonym for hydrous clay rocks of different geological origin. Bauxite is the primary source for industrial aluminium production. Alumina (Al2O3) is prepared from raw bauxite by a hydrometallurgical process. Knowing the composition and mineralogy of bauxite deposits is essential for evaluating the ‘‘processability’’ of the raw material. An instrument equipped with a robust silicon strip detector collects the data in this rough environment of a mine within minutes. The quality of the data in terms of peak position and relative peak intensities allows quantitative Rietveld refinement [8] down to the 1% level. Fe-containing minerals, like Hematite, contribute to Fe fluorescence radiation in raw materials. Even the modest energy resolution, as described above, can be used to suppress the fluorescence and thus the unwanted background is reduced, see the lower measurement in Fig. 9. Even though the peak intensity is slightly reduced, the data quality is improved by the much 2
The grey line around zero shows the deviation between data and fit.
better peak to background ratio. This leads to an improved sensitivity for minor constitutes. Not every investigation requires short measurement times. Instead, the use of 192 virtual point detectors measuring simultaneously can be used to collect data with larger statistical relevance. Texture investigations using X-ray powder diffraction commonly suffer from very long measurement times because the measurements are performed with the sample being positioned in phi (typically 0–3601) and chi (typically 0–801) with step widths of 51 each or smaller. Additionally, a couple of background measurements need to be performed at diffraction angles where no reflections occur. Using the silicon strip detector in snapshot mode or continuous scan mode offers two essential advantages over the traditionally used point detectors. Firstly, a wider diffraction angle range is measured simultaneously; thus background detection is performed in parallel, and for each individual measurement. This improves data quality significantly while reducing overall measurement time. Secondly, for each individual measurement, the entire diffraction peak is recorded. The peak position and the integral peak intensity can be fitted. The measured data are of superior quality compared to the traditional technology due to time saving considerations since the detector does not scan the diffraction angle. Fig. 10 shows two high quality pole figures determined from the same scan measurement using the silicon strip detector. Another example showing the capabilities of a fast one dimensional detector is reciprocal space mapping (RSM) on compound semiconductor material (Fig. 11). To record such RSM a time consuming series of loops of scans swinging the sample surface around a reflection is required (rocking curve). For each rocking curve the detector is positioned at a slightly different diffraction angle. Without major losses in quality, an RSM can be recorded in minutes instead hours with a 1-D silicon strip detector. If used in snapshot mode a series of different diffraction angles are present simultaneously. Also, it should be
358
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
Fig. 10. Shown are pole figures of rolled aluminium. The texture measurement of the 111 reflex (left) at 38.51 2y and the 200 reflex (right) at 44.71 2y show an almost orthorhombic symmetry. They were measured using a sealed Cu tube with the 1-D strip detector described in this paper.
Fig. 11. Reciprocal space map: The main peak at the bottom is due to the GaN (104+ ) substrate. In the upper right one can see a peak due to the thick relaxed AlN (104 +) layer. The signal marked with an ellipse stems from a thin strained InGaN (104+ ) layer.
noted that due to the high count rate capability of the detector erroneous intensity measurements are avoided when the strong reflections of the substrate fall into the detector (see the GaN 104+ reflection in Fig. 11).
Typically, in thin film applications a couple of orders of magnitude of intensities can be observed. X-ray reflectometry (XRR) uses the material property that for every material the index of refraction is smaller than unity, total external reflection
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
short range orders. For the same angular scanning range more information can be obtained. Also, unwanted fluorescence background may be reduced by switching to a different radiation wavelength. The comparison between the data recorded with silicon strips detectors with 300 and 500 mm show the expected higher quantum efficiency while maintaining the angular position resolution.
1e7 1e6 intensity [cps]
359
1e5 1e4 1000
6. Summary
100 10
0
1
2
3
4 5 2-Theta [deg]
6
7
8
Fig. 12. 500 A˚ Pt on SiO2 substrate measured using Cu-radiation (45 kV, 24 mA) ¨ and a primary Gobel Mirror.3 The strip detector was used in 0-D mode, summing the counts from all strips.
appears at angles below 11. Fig. 12 shows an XRR measurement on a single 500 A˚ thick Pt film on a quartz substrate. The oscillations are caused by interference of X-ray beam reflected from the surface and from the Pt/SiO2 interface; the distance of the interference fringes are used to determine the film thickness with an accuracy of better than 1%, without the need of calibration as is known from optical methods. The measurement covers about 7 orders of magnitude in intensity, which is outside the dynamical range of most X-ray detectors. When using a silicon strip detector turned in a way that the strips are perpendicular to the reflected beam profile, which typically is a line, one can measure this XRR curve in one scan. This is because the enormous intensity in the vicinity of the total external reflection (around 11) is distributed over all strips, simply speaking; each individual strips needs to collect 107/192 cps only. Cu radiation (8.0 keV) is by far the most commonly used energy for laboratory X-ray diffraction systems. A silicon strip sensor of 300 mm does match this energy quite well because most of the incoming photons are absorbed and detected. Transmission diffraction experiments on materials that contain heavy elements, however, benefit from harder X-ray such as Mo (17.4 keV) or Ag (22.2 keV) radiation due to their better depth penetration. Additionally, a higher energy and thus shorter wavelength extends the accessible diffraction space for the investigations of
3 ¨ A Gobel Mirror is an X-ray mirror. It consists of a parabolically bent multilayer structure with alternate layers of low-Z and high-Z materials, which monochromatises X-rays due to Bragg’s law. Typical combinations of layers would be W/B4C, W/Si.
The outcome of the project to develop a 1-D X-ray detector for XRD application using the silicon strip detector and ASIC technology is the LynxEye detector [9,10]. It was launched for series production in 2005. Careful optimisation of the silicon strip detector and ASIC design has resulted in a device that marks a significant step in the development effort to accelerate XRD measurements. The silicon strip detector technology enables XRD analysis meeting industrial requirements concerning robustness and ease of use, while providing best data. Limitations of legacy detectors were overcome. Since the new technology only delivers improvements without any disadvantages, it is soon becoming the default choice for any instrument. It is on its way to replace the scintillation counter as standard detector for laboratory instruments.
Acknowledgements We would like to thank Kurt Helming and Hugues Guerault for providing many of the excellent application data. Also, special thanks to Berit Avset for her patience and help during the sensor design phase. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
¨ L. Brugemann, E.K.E. Gerndt, Nucl. Instr. and Meth. A 531 (2004) 292. R. Turchetta, et al., IEEE Trans. Nucl. Sci. NS-41 (1994) 1063. V. Chatzisotiriou, et al., Nucl. Instr. and Meth. A 418 (1998) 173. C. Colledani, et al., Nucl. Instr. and Meth. A 372 (1996) 379. P.P. Allport, et al., Nucl. Instr. and Meth. A 310 (1991) 155. P. Wiacek, W. Dabrowski, Nucl. Instr. and Meth. A 551 (2005) 66. P. Wiacek, W. Dabrowski, Nucl. Instr. and Meth. A 580 (2007) 1355. /http://www.bruker-axs.de/topas.htmlS. Bruker AXS publication DOC-S88-EXS027 V2 (LynxEye specifications). Bruker AXS publication DOC-S88-L88_E0059 (The LynxEye detector: First choice for fast clinker analysis).
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Application of Si-strip technology to X-ray diffraction instrumentation a ¨ E. Gerndt a,n, W. Da˛browski b, L. Brugemann , J. Fink a, K. S´wientek b, P. Wia˛cek b a b
Bruker AXS, Karlsruhe, Germany Faculty of Physics and Applied Computer Science, AGH University of Science and Technology, Krakow, Poland
a r t i c l e in f o
a b s t r a c t
Available online 8 June 2010
We describe the successful technology transfer of High Energy Physics (HEP) silicon-strip detectors for tracking of minimum ionising particles (MIPS) to industrial X-ray diffraction instruments. In our application the detector is used to measure 1-D intensity profiles of low-energy photons. The challenges of such an application are low noise because of the relatively low energy of X-ray photons, from 5 to 22 keV, and high count rate capability. The technical implementation, with a focus on custom designed front-end electronics and optimisation of strip geometry taking into account the charge division effects, is shown and the achieved performance is summarized. The detector was launched several years ago and we report on the in-field experience. Lastly, we describe several scientific applications. & 2010 Elsevier B.V. All rights reserved.
Keywords: X-ray diffraction X-ray detectors Silicon strip detectors Spatial resolution Energy resolution Charge division Front-end electronics Low noise electronics
1. Introduction The combined technologies of silicon strip detectors and Application Specific Integrated Circuits (ASICS) have enabled 1-D position sensitive X-ray detectors to be designed to fulfil specific requirements. This combined technology is widely used in HEP experiments for building high precision tracking detectors. As a mature technology, it is a suitable one for building a detector for in-house commercial instruments used for X-ray diffraction, X-ray reflection, and X-ray scattering measurements. One-dimensional X-ray diffraction (XRD) detectors were quite popular in academics and research for a long time. They enabled special X-ray diffraction investigations also in a laboratory. One example is the kinetic study of materials under non-ambient conditions. However, due to their limited count rate capabilities, the quality of the data in terms of measured intensity and of peak position stability was considerably low. An overview of XRD detectors, the specific requirements, and relative performance can be found in Ref. [1]. With the introduction of silicon strip detectors this limitation was removed. Furthermore, due to the enormous count rate capability, robustness, and ease of use, new applications in X-ray diffraction were opened. In its first instance, the novel instrument consists of an X-ray source, primary beam modifying optics, e.g. focussing or monochromating optics, a sample, and the 1-D detector with optical components suppressing scattered radiation. The whole set-up is arranged in Bragg– Brentano geometry obeying the y–2y condition during the entire
n
Corresponding author. Tel.: +49 721 595 7080; fax: + 49 721 595 2317. E-mail address: [email protected] (E. Gerndt).
0168-9002/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2010.05.032
measurement. The 1-D detector replaces the 0-D detector in the classical set-up. This set-up allows powder diffraction measurements for qualitative and quantitative crystalline phase identification. Results for this type of X-ray diffraction, in addition to more elaborate measurements using different set-ups, are shown in Section 5 of this paper. The performance of a new technology like the silicon strip detector can easily be checked by using standard reference material (SRM) provided by the National Institute of Standard and Technology (NIST, USA). Various certified solid or loose powder samples are available for comparison measurements. An important quality criterion of a detector is its angular resolution measured as full width at half maximum (FWHM) for given reflections. Fig. 1 shows a measurement recorded on an LaB6 powder sample (SRM 660a) using the developed silicon strip detector. The FWHM of 0.0421 matches very well the value, which can be expected from the instrument set-up. There is no loss in angular resolution when compared to a point detector. This shows even though the silicon strip detector acts as 192 point detectors, recording data at the same time at slightly different diffraction angles, the data quality is equivalent to a single point detector, but measured 192 times faster. It is worth mentioning that the ideas to build a 1-D detector for powder diffraction, based on silicon strip detector technology, have been presented many years ago [2,3]. However, to our knowledge, none of these projects has materialised as a useful instrument addressing properly the requirements of the diffraction technique. Performance and measurement capability of both prototype detectors reported in the above mentioned papers were severely limited by the performance and quality of the readout electronics. While the silicon strip detector technologies, as
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
351
2.1. Detection efficiency The considered range of X-ray energy is from 5.4 up to 22 keV, while the majority of applications uses 8 keV copper X-ray tubes. The typical wafer thickness of 300 mm used in silicon strip detectors for HEP experiments appear to match well our requirement as it provides 98% absorption for 8 keV X-rays. Using a thicker detector would provide higher efficiency for X-ray energies above 8 keV; however, this would result in larger charge sharing between the neighbouring strips and compromising the single strip energy resolution due to diffusion of charge generated in the detector volume. These effects are related directly to the strip geometry and will be discussed later. Therefore, we have decided to develop an optimised design for typical applications based on 300 mm thick silicon wafers and an optional design for higher X-ray energies based on 500 mm thick wafers. Fig. 1. Measurement of the angular resolution using the standard reference material 660a (LaB6). The resolution achieved is equivalent to a point detector (0-D detector) with a 0.1 mm receiving slit (diffractometer D2 Phaser with 1.51 Soller slit1 and 0.1 mm divergence slit).
developed for high energy particle physics experiments, can be applied directly to measurement of X-rays, one needs a readout ASIC of different readout architecture and much better performance to build a useful instrument. The basic requirements for the ASIC are as follows:
counting mode with an energy window discriminator low equivalent noise charge (ENC), below 150 electrons rms, at
room temperature for a detector capacitance of 2.4 pF, allowing for discrimination between noise and signal of 1500 electrons corresponding to X-ray energy of 5.4 keV linear front-end circuitry up to an input charge of 6200 electrons (22.2 keV in Si) count rate capability significantly higher than 106 cps storage data buffers on the chip and multiplexed data readout possibility of simultaneous data taking and data readout low power dissipation of 3 mW per channel allowing operation without active cooling.
2. Silicon strip detector We have selected a conventional single-sided silicon strip detector structure with AC-coupled p + strips in high resistivity n-type silicon bulk as the sensor for our detector. Such a choice was driven, firstly, by the requirement to use a mature sensor technology and, secondly, by specific technical requirements discussed below. The strip length of 16 mm is determined by the geometry of the X-ray beam in the diffraction instruments. This allows us to maximise the use of the beam, although shorter strips would result in lower noise and better energy resolution. However, since the intention was to use the detector in preexisting instruments with a given geometry it was decided to use the full possible length. Also, the strip pitch was chosen due to the available instrumentation for which the detector was designed. Due to the commonly achieved focal spot dimensions of sealed Xray tubes, a pitch width of much less than 100 mm would no lead to better spatial resolution of the system. 1 A so called Soller slit consists of parallel plates limiting beam divergence in one direction. It is typically defined by its opening angle.
2.2. Bias voltage and leakage current The typical full depletion voltage of 300 mm thick sensors is about 60 V. This parameter is not very critical as we want the sensor to be biased with much higher voltages, up to 250 V, in order to ensure a short charge collection time and to minimise division of charge between neighbouring strips. No active cooling is foreseen for the detector. Given that the detector is operated in the diffraction instrument at temperatures typically about 10 1C above room temperature, we have assumed a maximum leakage current of up to 1 nA per strip. This upper level of the strip leakage current is set by the requirement that the parallel current noise generated by the detector leakage current should not contribute significantly to the equivalent noise charge (ENC) of the overall system. For a given voltage noise of the preamplifier and detector capacitance this limits the peaking time of the front-end electronics to be below 300 ns. 2.3. Bias resistance Because the noise contribution of the strip leakage current cannot be made completely negligible we have used the structure with AC-coupled strips. Using DC-coupling would require an additional circuitry in the front-end electronics to compensate the DC voltage shifts in the preamplifier circuit. Such a circuitry would be an additional source of current noise. In first approximation, this circuitry would double the noise due to the detector leakage current. For the AC-coupled strip, the bias resistance becomes another critical parameter as an additional source of parallel current noise at the preamplifier input. The assumed maximum leakage current of 1 nA corresponds to an equivalent noise resistance of 50 MO. In order to avoid additional noise contribution from the bias resistor, it should be made at least an order of magnitude higher, i.e. 500 MO. Such high values are achievable only with FOXFET structures [4]. 2.4. Spatial resolution Primarily, the silicon strip detector should replace the 0-D detectors used in scanning mode so that the strip pitch should be of the same order of magnitude as the slit used with the 0-D detector. It is worth noting a different approach to the spatial resolution of silicon strip detectors used for X-ray detection compared to the common approach used for tracking relativistic charged particles in HEP experiments. In tracking detectors of charged particles one can take advantage of charge division effects between the strips and improve the spatial resolution by estimating the centre of gravity of signal amplitudes measured for
352
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
a cluster of strips. In order to make the charge sharing between several strips in the cluster more linear one can introduce intermediate strips, which are not connected to the readout electronics. They provide additional purposely introduced capacitive charge division. For relativistic charge particles spatial resolution better than 2 mm rms has been achieved [5] using silicon strip detectors with a strip pitch of 25 mm and readout electronics applying a 50 mm pitch. In order to take advantage of the charge division technique one needs readout electronics, which records signal amplitudes with a high signal-to-noise ratio. For minimum ionising particles the most probable signal generated in a typical 300 mm thick silicon strip detector is 24 000 electrons. By proper optimisation of the front-end electronics a signal-to-noise ratio of the order of 50 is achievable. This is needed to reach the ultimate spatial resolution mentioned above. The charge generated in the silicon detector by an 8 keV photon is about 10 times smaller than the charge generated by a minimum ionising particle. Achieving the required high signal-tonoise ratio would demand an extraordinary low noise of the frontend electronics. This is not achievable at room temperature. Furthermore, employing charge division effects for improvement of the spatial resolution requires recording signal amplitudes. In combination with the high count rate capability this would require massive Analogue-to-Digital Converters for processing the data in real time. Therefore, the only practical approach is to use the binary architecture of the front-end electronics, i.e. pulses with amplitudes in between a lower and an upper threshold are counted as good events. Subsequently, charge sharing effects should be minimised. Ideally, in the considered application of a strip detector to X-ray diffraction, the spatial response of the strip, i.e. the ratio of the number of incident photons to the number of counted photons, should be flat across the whole strip area.
2.5. Energy resolution Typically, the energy resolution of silicon strip detectors is defined with respect to the total charge generated in the detector volume by photons or charged particles. Since the generated charge may be divided between several strips, depending on the strip geometry, this requires summing up of the signals from the cluster of strips for each event. In the considered application, assuming readout electronics with binary architecture, no information on the pulse amplitude is preserved so that an appropriate approach is to define the energy resolution for the distribution of signals measured on a single strip. This distribution is affected by the electronic noise of the front-end electronics and by the charge division effects. Both the electronics noise and the charge division depend on the strip geometry and optimising these two parameters leads to contradictory requirements concerning the strip pitch and the strip width. In XRD applications the strip pitch is primarily defined by the required spatial resolution. In order to minimise the charge division effects one would rather make wide strips and narrow gaps between the strips. Such geometry results in a small fraction of events with significant charge sharing between two adjacent strips, which degrades the single strip energy resolution. On the other hand, narrow gaps between the strips result in large inter-strip capacitance, which affects directly the ENC. Given that the charge signals generated by X-rays are small, about 2200 electrons for 8 keV photons, the ENC is the primary parameter, which determines the single strip energy resolution. The additional point to be considered is a trade-off between the inter-strip capacitance and the strip series resistance. The strip series resistance contributes to the total equivalent voltage noise at the preamplifier input. Since we aim to achieve a low equivalent noise resistance of the front-end electronics, of the
order of 100 O, the serial resistance of aluminium readout strips, of the order of 35 OXcm, has to be taken into account. The final choice of strip geometry has been made taking into account the expected ENC of the front-end electronics to be about 100 electron rms and the charge sharing effects. The charge sharing effects have been estimated by Monte Carlo simulations [6] and measurements of the energy and spatial response performed for the prototype silicon strip detectors [7]. Based on these results we have chosen the strip pitch of 75 mm and the strip width of about 20 mm. Degradation of the single strip energy resolution due to charge sharing is of the order of 10% for the ENC level of 100 electron rms, and it depends on the X-ray energy, detector bias voltage, and also on the side (strip or ohmic contact) from which the detector is irradiated. More details can be found in Refs. [6,7]. Making the strips wider, above 20 mm, would reduce the charge sharing effects only slightly, especially for the detector irradiated from the ohmic side, while the increase of the inter-strip capacitance would have a predominant effect on electronic noise. The silicon strip detector has been designed in close collaboration with Sintef. The detectors for series production are delivered by Sintef according to the mutually agreed specification and acceptance tests.
3. Readout ASIC A simplified block diagram of the complete readout ASIC is shown in Fig. 2. The chip comprises six basic blocks: 64 analogue front-end channels, 128 18-bit counters, back-end buffers, control block, calibration circuit, and bias generators. Each channel is equipped with two pulse height discriminators providing the LOW and the HIGH discrimination level of the analogue signals. The output pulses from each discriminator are stored in an 18-bit counter. The counters are grouped in the blocks of 16 counters each and the data from each block is read out via an 8-bit bus with tristate outputs. The readout of counter content is performed in two steps: (1) parallel transfer of all bits into a buffer register and (2) serial readout of the blocks of eight channels. This scheme allows one to continue data taking while the buffer register is read out. Sequences of data collection and readout of the data from the counters are controlled by a control block. The control block provides also settings of the digital-to-analogue converters (DACs), which are used in the front-end block to control the discriminator threshold, bias currents in the analogue circuits, and the amplitudes of the internally generated test pulses. The control block receives the commands via an external serial link, decodes them, and executes by sending the control signals and data to other blocks. Up to eight chips can be addressed geographically and connected to a common bus. The chip ID is defined by three bits on three address inputs to be wire bonded to external logic levels. The ASIC has been designed and manufactured in the 0.35 mm CMOS process by Austria Microsystems. The prototype designs were manufactured on Multi-Project Wafers in the frame of service provided by EUROPRACTICE. 3.1. Front-end architecture The block diagram of the front-end channel is shown in Fig. 3. The channel consists of a charge sensitive preamplifier, a shaping amplifier, and two pulse height discriminators. The bias current in the input stage and the value of the feedback resistor in the preamplifier are controlled by internal DACs implemented on the chip. These allow for optimisation of the noise performance according to the parameters of the silicon strip detector, leakage current, and strip capacitance. The bias current in the shaper stage
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
353
Fig. 2. Block diagram of the 64-channel readout ASIC.
Fig. 3. Block diagram of the front-end channel.
and the value of the feedback resistor in the shaper are controlled by internal DACs. These allow for tuning the shaping time constant and the gain of the shaper stage. The discrimination threshold in each discriminator is controlled by a differential voltage delivered from an internal voltage generator controlled by a DAC.
3.2. Front-end performance Because of the binary readout architecture the analogue parameters of the front-end circuit, the gain of the preamplifiershaper circuit, and the discriminator offset can be extracted from multiple measurements for different discrimination thresholds,
354
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
covering the interesting range of signal amplitudes at the discriminator input. A primary result of such a discrimination threshold scan is an integral distribution of the signal amplitudes, which in the case of a mono-energetic signal and Gaussian noise is described by the complementary error function. By fitting the measured integral distribution to the complementary error function one obtains the signal amplitude and the standard deviation of noise. In the case of a more complex spectrum one can differentiate numerically the integral distribution to obtain amplitude distribution. Example of test results for all 64 channels in one ASIC obtained with the test signals generated internally in the ASIC is shown in Fig. 4. From the threshold scan curves we extract the rms value of noise at the discriminator input of each channel and from the response curves we extract the gain of the channels up to the discriminator inputs and the discriminator offsets. Having estimated noise and gain at the discriminator input we obtain the ENC for each channel. Example distributions of gain offset and ENC for one ASIC are shown in Fig. 5. The spread of discriminator offsets within one ASIC is typically 22 electron rms input equivalent and the relative spread of the gain is 0.8% rms. Thus, the uniformity of the lower discrimination threshold is primarily determined by the offset spread and is small compared to the electronic noise.
iCAL= 7
100
11
15
19
23
27[LSB]
counts
80
60
40
20
4. System architecture 4.1. Detector head and detector control unit
0 0
100
200
300 Uth [mV]
400
500
600
550 500 450 400 Uth [mV]
One should note that the parameters of the front-end circuit, gain, offset, and ENC, as well as pulse high distribution for X-rays are obtained in the test mode by scanning the discrimination threshold. These data are used to set up the working thresholds for operation of the system in the counting mode as used in the diffraction measurements. For a sensor bonded to the ASIC and working at room temperature the ENC is 115 electron rms that corresponds to the energy resolution of 985 eV FWHM. This does not take into account charge division effects in the detector. The energy resolution is due to electronic noise only and independent of the energy. Performance of the detector, exposed to a 57Co X-ray source from the strip side, is illustrated in Fig. 6. It shows the amplitude distributions for 64 channels obtained by numerical differentiation of the threshold scan curves. In order to minimise the effects of charge sharing the spectrum has been measured for a detector bias of 270 V. The source emits X-rays of energies 6.4, 7.06, and 14.41 keV. The lines 6.4 and 7.06 keV are Ka and Kb of iron, respectively. The intensity of the Kb line relative to the intensity of the Ka is only 15% and obviously the two lines cannot be distinguished given the energy resolution of about 1 keV FWHM. The low intensity peak around energy of 20 keV comes from coincidences of 6.4 and 14.41 keV photons (‘‘sum peak’’). The average energy resolution obtained is 1.12 keV FWHM for the 6.4 keV line and 1.06 keV FWHM for the 14.41 keV line. The resolution of 1.06 keV for the 14.41 keV line is only slightly worse than the resolution of 0.985 keV due to electronic noise. This observation confirms that, at sufficiently high bias voltage, the charge division effects are indeed limited to a small fraction of photons. The 6.4 keV peak is somewhat broader because of the contribution of Kb line of 7.06 keV. The continuous and almost flat background between the peaks comes from events with charge sharing between neighbouring strips. The energy resolution of the lower energy line is thus affected by the photons of higher energies absorbed in the inter-strip regions, for which the charge is shared between neighbouring strips.
350 300 250 200 150 100 1500 2000 2500 3000 3500 4000 4500 5000 5500 Qin [e-]
Fig. 4. Example test results of analogue parameters of the front-end circuits: (a) threshold scan curves for 64 channels for six different test signals and (b) response curves for 64 channels extracted from the threshold scan.
The detector head of the implemented silicon strip detector system houses a sensor module, components for mechanical alignment and X-ray optics. The X-ray optics can comprise elements reducing beam divergence or also to reduce background due to air scattered X-rays. On the sensor module, one AC-type silicon strip sensor with 192 strips and three ASICs are grouped together and connected with interface circuitries. The irradiation of the sensor module is made from the back side of the silicon strip sensor. For matching the thermal expansion of the silicon dies and the printed circuit board a ceramic material based on Al2O3 had been chosen as substrate. In comparison to the printed circuit board material like FR4, Al2O3 has a greater heat conductance for better removal of the dissipated electrical power. The detector control unit consists of the Data Acquisition Board, Main Control CPU board, very low noise power supplies for the bias supply of the silicon strip sensor (0–500 V), and the supply of the three ASICs. An FPGA device located on the Data Acquisition Board configures the three ASICs dynamically, controls the data collection, processes the measured data and forwards the results to the Main CPU board. All communication between the Data Acquisition Board and the detector head is performed via serial LVDS lines running at a clock rate of 10 MHz. The contents of the 384 counters for the LOW and HIGH
number of channels
0.15 0.1 0.05 µ = 0.10435 ± 0.00082 [mV/e-] (0.78 %) 0
10
20
30
40
50
µ = 1.6922 ± 1.6 [mV]
0 0.1
0
10
20
30
40
50
40 20
10
0
60
60
µ = 46.5703 ± 4.2 [e-] (9.0%) 10
20
30 channel
10
0 20
40
50
0 10 Offset [mV]
20
0 0
0.105 0.11 Gain [mV/e-]
20
-10
number of channels
Offset [mV]
0
-5
ENC (peak-1400 e-) [e-]
10
60
5
355
20
0
number of channels
Gain (peak-1400 e-) [mV/e-]
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
60
40 ENC [e-]
60
Fig. 5. Typical distribution of gain, offset, and ENC across 64 channels in one ASIC.
synchronization between the scanning motor drive and the data collecting FPGA. This is accomplished by using motor clock pulses for triggering. 4.2. Diffractometer system with silicon strip detector The block diagram for a silicon strip detector embedded into a diffractometer system is shown in Fig. 8. Besides the radiation detector, the main components of an XRD instrument are the X-ray tube including X-ray shutter, motorized goniometer, and sample stages. These are controlled and monitored by the diffractometer control electronics. Through the motor clock line, there is a very tight synchronisation between the spatial movement of the master scanning axis and the data collection within the detector control unit. Both diffractometer control electronics and detector control unit are connected via network with an external computer running the measurement server program and evaluation software packages. Fig. 6. Spectrum using a 57Co X-ray source. All 64 strips of a silicon strip detector are measured simultaneously.
4.3. Measurement modes
discriminators are read out in parallel via eight LVDS lines in order to achieve a minimum readout time Fig. 7. The Main Control CPU runs the LINUX-based control software that provides high level measurement commands and communicates directly with the host programs via an Ethernet interface. Some of the measurement modes need a very precise
The electronics chain allows several ways to perform measurements. Typically the detector is at a fixed radius from the sample and moving with a constant angular velocity. The count rates of the individual strips are assigned online to give 2y angle. All measurements to a given position are summed up. During the movement, the detector collects data for a short time frame, latching the counter values, and then continuing with the
356
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
Fig. 7. Block diagram of the detector head and the detector control unit.
Fig. 8. Electronics architecture of X-ray diffraction instrument with silicon strip detector.
collection of the next measurement frame. The exposure time per measurement frame is typically in the range of about 10 ms and is set by software. During this continuous scan the strips are essentially used as parallel detectors. The number of strips (192) determines approximately the gain in measurement speed when compared to a measurement with a point detector.
It is also possible to take data with the detector at a fixed position. Again, the electronics assigns angular positions to the respective strips. The strip synchronization with respect to the angular position of the 2y-arm of the goniometer is no longer needed. The angular range is given by the radius and the opening of the detector, in our case 14.4 mm. This so called snapshot mode
Counts
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
46 000 44 000 42 000 40 000 38 000 36 000 34 000 32 000 30 000 28 000 26 000 24 000 22 000 20 000 18 000 16 000 14 000 12 000 10 000 8 000 6 000 4 000 2 000 0 -2 000 -4 000 -6 000 -8 000
Quartz Hematite Gibbsite Boehmite Goethite Anatase Kaolinite (BISH) Quartz Hematite Gibbsite Boehmite Goethite Anatase Kaolinite (BISH)
10
12
14
16
18
20
22
24
26
28
357
5.64 % 8.59 % 55.34 % 14.70 % 8.77 % 1.79 % 5.17 % 5.57 % 8.95 % 57.15 % 13.44 % 8.69 % 1.82 % 4.37 %
30
32
2Th Degrees Fig. 9. Phase quantification of bauxite. Two measurements are overlaid; the lower curve was measured while setting the energy discriminator window narrow to suppress Fe fluorescence. The given quantitative composition of the sample was determined by Rietveld analysis using TOPAS software [8].2
allows time resolved measurements, for example, if the evolution of phase transitions should be observed. It is also possible to sum the counts from all strips. In this case the one dimensional information is lost, but the 0-D measurement allows more than two orders of magnitude large count rates, mostly defined by the number of strips irradiated.
5. Applications The most frequent application in X-ray diffraction is qualitative and quantitative crystalline phase identification. A reliable 1-D detector is of essential importance in case diffraction is used for quality or process control in industries like minerals and mining, cement, or pharmaceuticals. Fig. 9 shows two measurements of a bauxite sample. Bauxite is a common synonym for hydrous clay rocks of different geological origin. Bauxite is the primary source for industrial aluminium production. Alumina (Al2O3) is prepared from raw bauxite by a hydrometallurgical process. Knowing the composition and mineralogy of bauxite deposits is essential for evaluating the ‘‘processability’’ of the raw material. An instrument equipped with a robust silicon strip detector collects the data in this rough environment of a mine within minutes. The quality of the data in terms of peak position and relative peak intensities allows quantitative Rietveld refinement [8] down to the 1% level. Fe-containing minerals, like Hematite, contribute to Fe fluorescence radiation in raw materials. Even the modest energy resolution, as described above, can be used to suppress the fluorescence and thus the unwanted background is reduced, see the lower measurement in Fig. 9. Even though the peak intensity is slightly reduced, the data quality is improved by the much 2
The grey line around zero shows the deviation between data and fit.
better peak to background ratio. This leads to an improved sensitivity for minor constitutes. Not every investigation requires short measurement times. Instead, the use of 192 virtual point detectors measuring simultaneously can be used to collect data with larger statistical relevance. Texture investigations using X-ray powder diffraction commonly suffer from very long measurement times because the measurements are performed with the sample being positioned in phi (typically 0–3601) and chi (typically 0–801) with step widths of 51 each or smaller. Additionally, a couple of background measurements need to be performed at diffraction angles where no reflections occur. Using the silicon strip detector in snapshot mode or continuous scan mode offers two essential advantages over the traditionally used point detectors. Firstly, a wider diffraction angle range is measured simultaneously; thus background detection is performed in parallel, and for each individual measurement. This improves data quality significantly while reducing overall measurement time. Secondly, for each individual measurement, the entire diffraction peak is recorded. The peak position and the integral peak intensity can be fitted. The measured data are of superior quality compared to the traditional technology due to time saving considerations since the detector does not scan the diffraction angle. Fig. 10 shows two high quality pole figures determined from the same scan measurement using the silicon strip detector. Another example showing the capabilities of a fast one dimensional detector is reciprocal space mapping (RSM) on compound semiconductor material (Fig. 11). To record such RSM a time consuming series of loops of scans swinging the sample surface around a reflection is required (rocking curve). For each rocking curve the detector is positioned at a slightly different diffraction angle. Without major losses in quality, an RSM can be recorded in minutes instead hours with a 1-D silicon strip detector. If used in snapshot mode a series of different diffraction angles are present simultaneously. Also, it should be
358
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
Fig. 10. Shown are pole figures of rolled aluminium. The texture measurement of the 111 reflex (left) at 38.51 2y and the 200 reflex (right) at 44.71 2y show an almost orthorhombic symmetry. They were measured using a sealed Cu tube with the 1-D strip detector described in this paper.
Fig. 11. Reciprocal space map: The main peak at the bottom is due to the GaN (104+ ) substrate. In the upper right one can see a peak due to the thick relaxed AlN (104 +) layer. The signal marked with an ellipse stems from a thin strained InGaN (104+ ) layer.
noted that due to the high count rate capability of the detector erroneous intensity measurements are avoided when the strong reflections of the substrate fall into the detector (see the GaN 104+ reflection in Fig. 11).
Typically, in thin film applications a couple of orders of magnitude of intensities can be observed. X-ray reflectometry (XRR) uses the material property that for every material the index of refraction is smaller than unity, total external reflection
E. Gerndt et al. / Nuclear Instruments and Methods in Physics Research A 624 (2010) 350–359
short range orders. For the same angular scanning range more information can be obtained. Also, unwanted fluorescence background may be reduced by switching to a different radiation wavelength. The comparison between the data recorded with silicon strips detectors with 300 and 500 mm show the expected higher quantum efficiency while maintaining the angular position resolution.
1e7 1e6 intensity [cps]
359
1e5 1e4 1000
6. Summary
100 10
0
1
2
3
4 5 2-Theta [deg]
6
7
8
Fig. 12. 500 A˚ Pt on SiO2 substrate measured using Cu-radiation (45 kV, 24 mA) ¨ and a primary Gobel Mirror.3 The strip detector was used in 0-D mode, summing the counts from all strips.
appears at angles below 11. Fig. 12 shows an XRR measurement on a single 500 A˚ thick Pt film on a quartz substrate. The oscillations are caused by interference of X-ray beam reflected from the surface and from the Pt/SiO2 interface; the distance of the interference fringes are used to determine the film thickness with an accuracy of better than 1%, without the need of calibration as is known from optical methods. The measurement covers about 7 orders of magnitude in intensity, which is outside the dynamical range of most X-ray detectors. When using a silicon strip detector turned in a way that the strips are perpendicular to the reflected beam profile, which typically is a line, one can measure this XRR curve in one scan. This is because the enormous intensity in the vicinity of the total external reflection (around 11) is distributed over all strips, simply speaking; each individual strips needs to collect 107/192 cps only. Cu radiation (8.0 keV) is by far the most commonly used energy for laboratory X-ray diffraction systems. A silicon strip sensor of 300 mm does match this energy quite well because most of the incoming photons are absorbed and detected. Transmission diffraction experiments on materials that contain heavy elements, however, benefit from harder X-ray such as Mo (17.4 keV) or Ag (22.2 keV) radiation due to their better depth penetration. Additionally, a higher energy and thus shorter wavelength extends the accessible diffraction space for the investigations of
3 ¨ A Gobel Mirror is an X-ray mirror. It consists of a parabolically bent multilayer structure with alternate layers of low-Z and high-Z materials, which monochromatises X-rays due to Bragg’s law. Typical combinations of layers would be W/B4C, W/Si.
The outcome of the project to develop a 1-D X-ray detector for XRD application using the silicon strip detector and ASIC technology is the LynxEye detector [9,10]. It was launched for series production in 2005. Careful optimisation of the silicon strip detector and ASIC design has resulted in a device that marks a significant step in the development effort to accelerate XRD measurements. The silicon strip detector technology enables XRD analysis meeting industrial requirements concerning robustness and ease of use, while providing best data. Limitations of legacy detectors were overcome. Since the new technology only delivers improvements without any disadvantages, it is soon becoming the default choice for any instrument. It is on its way to replace the scintillation counter as standard detector for laboratory instruments.
Acknowledgements We would like to thank Kurt Helming and Hugues Guerault for providing many of the excellent application data. Also, special thanks to Berit Avset for her patience and help during the sensor design phase. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
¨ L. Brugemann, E.K.E. Gerndt, Nucl. Instr. and Meth. A 531 (2004) 292. R. Turchetta, et al., IEEE Trans. Nucl. Sci. NS-41 (1994) 1063. V. Chatzisotiriou, et al., Nucl. Instr. and Meth. A 418 (1998) 173. C. Colledani, et al., Nucl. Instr. and Meth. A 372 (1996) 379. P.P. Allport, et al., Nucl. Instr. and Meth. A 310 (1991) 155. P. Wiacek, W. Dabrowski, Nucl. Instr. and Meth. A 551 (2005) 66. P. Wiacek, W. Dabrowski, Nucl. Instr. and Meth. A 580 (2007) 1355. /http://www.bruker-axs.de/topas.htmlS. Bruker AXS publication DOC-S88-EXS027 V2 (LynxEye specifications). Bruker AXS publication DOC-S88-L88_E0059 (The LynxEye detector: First choice for fast clinker analysis).