2.1.

MXS Design

An early concept of generating x-rays by means of electrons generated by the photoelectric effect was proposed by Gendreau¹¹ (More information can be found from a later document at Ref. 12). Our design of the MXSs is based on a modified image intensifier unit (made by Photonis). It is a short vacuum tube with a photocathode on one end and a beryllium vacuum window on the other end. Light from two light-emitting diodes (LEDs) illuminates the photocathode, producing electrons, which in turn are accelerated by a 11.3-kV high voltage (HV) toward a target material sitting on top of a $300\text{-}\mu \mathrm{m}$ beryllium window. Generated x-rays exit the source through the target and the beryllium window (see Fig. 6).

Fig. 6

Schematic principle of an MXS.

X-rays will only be produced when the LED is emitting light. This allows for short duration x-ray pulses. In this way, continuous monitoring of the performance of each detector during astronomical observations is possible by selecting only those short-time periods in which the LED is on, leaving, in the remaining time, the observed astrophysical spectrum uncontaminated by the broadband spectrum emitted by the MXS.

Although the photocathode material (S20) has maximum quantum efficiency (QE) at wavelengths below 280 nm, maximum yield in terms of number of electrons, which is proportional to the number of generated x-ray photons, is at 420-nm wavelength (see Fig. 7). For a typical value of 5 mW of optical energy, this means an output of $2\times {10}^{15}\mathrm{electrons}/\mathrm{s}$. In the final flight version of the MXS, blue LEDs were used with maximum output power at 460 nm.

Fig. 7

S20 photocathode efficiency (provided by the Photonis Company). Black line shows the photocurrent (mA) related to optical power (W) as function of wavelength. The red lines show an overlay of coordinates of constant QE which represent the ratio in which electrons are generated as function of the number of incoming optical photons. Maximum QE (around 26%) is around 270 nm, but maximum cathode efficiency is at 440 nm, where QE is only 20%. This is due to the larger number of photons at longer wavelength for the same optical power.

The target consists of a layer of 25 nm of chromium on top of 150 nm of copper deposited on the beryllium window. These two types of material will generate two sets of x-ray lines: the Cu ${\mathrm{K}}_{\alpha }$ and ${\mathrm{K}}_{\beta }$ lines at 8.05 and 8.90 keV, respectively, and the Cr ${\mathrm{K}}_{\alpha }$ and ${\mathrm{K}}_{\beta }$ lines at 5.41 and 5.95 keV, respectively. The electrons will hit the chromium first. Although this order of layers will have lower efficiency generating the Cu lines, it does offer the possibility of generating the Cr lines only, when the HV is operated at a lower voltage than the nominal voltage of 11.3 kV. Figure 8 shows a measured spectrum [measured with a low-resolution silicon drift detector (SDD)], showing the chromium and copper lines plus the Bremsstrahlung continuum. For comparison, Fig. 9 shows the high-resolution spectrum measured with the SXS during the thermal vacuum test of the satellite. Due to the thick beryllium window on the Dewar gate valve, the chromium lines measured during the thermal vacuum tests are suppressed compared to what we expect for normal operations on-orbit (after the gate valve was opened). In addition, the blocking geometry of the gate valve window support structure enhances the additional Fe and Ni fluorescent lines from the MXS housing.

Fig. 8

Spectrum of the MXS obtained with low-resolution SDD. On top of the Bremsstrahlung continuum, the lines of Cr and Cu can be seen. This spectrum was measured in air, which decreases the low-energy Cr lines somewhat.

Fig. 9

Similar to Fig. 8, but now measured with the high-resolution SXS during a satellite TV test. Apart from the Cu and Cr lines, fluorescent lines of a few other elements (Ni 7.478 keV, Fe 6.403 keV) present on the MXS housing can be seen. In addition, a weak ${}^{55}\mathrm{Mn}$ 5.895-keV line from the radioactive source on the filter wheel is visible.

Two types of MXS sources exist. Each type is present as a redundant pair, which gives a total of four sources. One type is the so-called direct source type, which illuminates the SXS with the x-ray spectrum generated by the Cr/Cu anode as shown in Fig. 6. These sources are mounted with a 45-deg angle (see Fig. 5) with respect to the telescope optical axis, in order to move the MXS exit window as close to the telescope beam as possible. The other type is an indirect source, which has low-energy fluorescent targets mounted in front of the MXS exit window. These fluorescent targets allow the low-energy lines of aluminum (1.49 keV) and magnesium (1.25 keV) to be incident on the SXS. The MXS vacuum windows absorb these low-energy photons, hence they must be generated (in vacuum) outside of the MXS. Due to the low fluorescent yield, the “indirect” source strength will be much lower than the “direct” sources. For the same settings (LED current and duty cycle), the fluorescent sources are estimated to be about a factor in between $100\times$ to $1000\times$ weaker than the direct sources. For this reason, these fluorescent sources will not be pulsed, such as the direct sources, but operated (occasionally) in a continuous mode. This multitude of available calibration lines from the low (fluorescent) to the higher (direct source) energies allows calibration of nonlinearities in the SXS pixel response. Information on the detailed shapes of the lines present in the MXS spectrum can be found elsewhere.¹³

Figure 10 provides an idea of the actual size of an MXS source. Figures 11 and 12 show a crosscut of the actual direct and fluorescent sources design.

Fig. 10

A bare MXS source, without source holder and LEDs, showing the size, next to a 1€ coin, for comparison. The top of the source shows the beryllium exit window.

Fig. 11

Drawing of a crosscut through the MXS source mounted in its holder (diameter 5.1 cm, length 3.1 cm), showing the LEDs on the right, illuminating the photocathode. The beryllium window, which contains the x-ray target, is to the left.

Fig. 12

Drawing of a crosscut through the fluorescent MXS source, with the Al-Mg fluorescent target mounted in front of the MXS.

2.2.

MXS Performance

In order to monitor the (slowly) varying SXS pixel gains during observations, it was estimated that a flux of $0.3\mathrm{counts}/\mathrm{s}/\mathrm{SXS}$ pixel from the MXS calibration lines would be required. This flux translates into $5\times {10}^6\text{photons}/\mathrm{s}$ at the MXS source, given a source–detector distance of about 85 cm and pixel size of 0.82 mm. The $0.3\text{counts}/\mathrm{s}$ estimate was based on a required energy scale stability of around 0.5 eV and the assumption that typical time scales of SXS gain variation would be in the order of 10 min. Terms in the detector noise budget were based on tracking gain changes in 10-min integrations, using the required limits on the temperature coefficients of the electronics. Also, given the orbital period of 90 min, the relevant tracking time scale was around 10 min. Based on ground operations, temperature changes within the Dewar were expected to be much slower than the 10-min time scale and, thus, could be well tracked in 10-min integrations.

The sources are expected to show degradation over time, due to damage of the photocathode by ions generated by the electron impact on the target, and accelerated from the x-ray target back to the photocathode. Hence, source degradation with time is a function of the generated x-ray flux.

The direct sources are intended to be operated in a pulsed mode. The source intensity can be varied by changing the LED current. In addition, the time-averaged flux depends on the duty cycle of the LED pulses.

Several lifetime tests were performed to demonstrate that the MXS flux requirements could be met during the design life of the satellite mission with nominal operations of the MXS sources. In a first test,⁹ a source was initially set to operate at $100\times$ the required flight values. This was accomplished by running the source at a 100% duty cycle (continuous operations). LED current was kept constant and the flux was allowed to decrease during the tested time period. With this setup, the source is expected to degrade a factor of 100 faster than during the nominal flight operations, which use 1% duty cycle. This was confirmed when comparing the observed degradation at the 100% duty factor with a long measurement at nominal (1% duty cycle) operations. The degradation at 1% was 100 times slower than at 100% duty cycle.

It was found that the flux half-life of a direct source during such operations will be 2.9 years. Renormalizing the flux at the end of the test to the initial flux levels by increasing LED current, it was found that given the observed half-life of the source, the required LED current at the nominal end of mission (5 years) would be well below the maximum allowed LED current of 40 mA.

A second lifetime test was conducted, controlling the MXS to operate at a constant flux level. The source was operated at 50% duty cycle at a time-averaged flux, which was $50\times$ the nominal flux required in flight. The MXS flux during the LED pulse was identical to the intended flux in flight during pulses, but pulse duration during the test was $50\times$ longer. This meant 1 week of the test was equivalent to about 1 year in orbit. The test was executed for a total of 1300 h, equivalent to about 7.4 years in orbit. Figure 13 shows the count rate during the test, together with the LED temperature. LED performance depends on temperature. The spiky character of the count rate curve is due to the periodic (automatic) updates of the LED current, to keep the count rate at an approximate constant level. (Each spike corresponds to steps of integer numbers in the DAC controlling the LED current. For this reason, the spikes decrease with time since the relative size of the integer steps decrease with increasing number, which corresponds to current). The count rate as shown in Fig. 13 is the one measured by the SDD detector in the test setup and corresponds to a source intensity providing $50\times$ the $0.3\mathrm{counts}/\mathrm{s}/\mathrm{pixel}$ in the SXS. Figure 14 shows the LED current during the test. The LED current increases due to the degradation of the source. At the very end of the test (equivalent of 7.4 years of the mission), the LED current is around 2.0 mA, well below the maximum allowed LED current of 40 mA. This test shows that the MXS can easily accommodate the nominal required flux during the 5 years of design mission lifetime. Figure 14 also shows the temperature dependence of the LED performance. A drop in ambient temperature corresponds to a discontinuity in the LED current increase.

Fig. 13

Count rate plot during the test. The green line is the LED temperature, corresponding to the room temperature during the test. The room temperature was changed on purpose (the sharp drop to 20 C) to see the effect of a temperature change.

Fig. 14

LED current during the constant flux measurements. The sharp drop in room temperature (green line) also causes a drop in the LED current, due to the increased LED efficiency.

In addition, the relation between source degradation and peak flux during LED pulses was investigated. The time-averaged flux is the result of the pulse duty cycle and the flux within the pulse. A lower duty cycle can give the same time-averaged flux when the LED current is higher. Figure 15 shows the increase in LED current when the flux is kept constant, but the duty cycle is changed. For the low duty cycle (high LED pulse current), the increase is $261\pm 30\mathrm{nA}/\mathrm{h}$, whereas for the high duty cycle (low LED pulse current), the increase is $97\pm 4\mathrm{nA}/\mathrm{h}$. It appears that the increase in the current (and thus the degradation of the source) is lower when the flux at the pulse (proportional to the LED pulse current) is lower. Hence, the source will degrade faster when the flux during the pulse is set to higher levels, even if the time-averaged flux remains constant.

Fig. 15

Linear fits to the LED current increase due to aging of the MXS around the time when the LED current in the pulse is dropped by a factor of two (around the time stamp of 1650 h in the test) and duty cycle is doubled. The average x-ray flux remains the same in this way. It shows that at half the LED current (and double duty cycle), aging goes less than twice slower for the same x-ray flux.

Finally, the MXS response over time was investigated. This was done using nonflight electronics in a special setup. Figure 16 shows the MXS flux as function of time within the MXS pulse cycle. A 1-ms LED pulse was executed within a 40-ms cycle time. As can be seen, the original flux in the pulse of $5.2\times {10}^4\mathrm{events}/\mathrm{bin}$ drops within one-time resolution bin (0.2 ms) to about $400\mathrm{events}/\mathrm{bin}$ (0.7% of the initial flux) and then exponentially decreases to the background level with an e-folding time of 1.7 ms. Although the photocathode material (S20) is supposed to react very fast, at $\mu \mathrm{s}$ time scales, to changing light levels (the same material is also used in photomultiplier tubes), it may be that positive ions released by the impact of electrons on the anode continue to (re)activate the cathode or that the abundant x-ray photons interact with the cathode, slowing down the expected fast response.

Fig. 16

MXS pulse shape (logarithmic scale) for a 1-ms pulse within a 40-ms pulse cycle, using nonflight electronics. The “pulse tail” starts at 0.7% of the original pulse flux, exponentially decaying to a background level (not associated with the MXS). The red line is the fitted exponential curve with 1.7-ms e-folding time.

The above measured x-ray response is the response of the MXS itself. Unfortunately, the flight electronics generating the LED pulse for the filter wheel had an e-folding time of about 5 ms. Thus, the MXS pulse on the filter wheel had a slower decay than an isolated MXS by itself (see Fig. 17).

Fig. 17

MXS pulse shape (logarithmic scale) for a 5-ms pulse, from the Hitomi thermal-vacuum tests, using the flight electronics. As can be seen, the decay of the pulse tail is much slower and the tail starts at higher levels, compared to the MXS using nonflight electronics (Fig. 16).

2.3.

MXS High-Voltage Control

Each (main/redundant) MXS pair of direct and fluorescent sources is powered by an HV unit. These units can be set to produce 500, 4000, 7000, or 11300 V. Primarily, these settings were designed to allow the voltage to be slowly ramped up to the default operating voltage of 11.3 kV. The lower 500-V setting was devised to be used during the satellite’s South Atlantic Anomaly passages. The 7-kV setting was defined to allow the 5.41- and 5.95-keV Cr lines to become active only, without going to the highest voltage setting. Finally, the maximum voltage was, in part, limited by the operating boundaries specifications of the HV cables.

2.4.

MXS-Intended Operations

The MXS calibration sources were planned to be switched on during the very contact pass when communications to the Hitomi satellite could not be established due to its problems. All operations stopped at that time. (The MXSs had only started on the 500-V HV setting when operations stopped.) Thus, no data on the actual flight performance were obtained.

For the MXS-intended operations, in order to minimize dead time, the duty cycle would be set to the lowest practical level, limited by the instantaneous flux that the MXS can safely produce during a pulse. The initial plan for normal flight operations was to start operating the direct sources at 1% duty cycle during most of the observing time, at a flux of $0.3\mathrm{counts}/\mathrm{s}/\mathrm{pixel}$, depending on the SXS gain stability and background levels. In order to prevent energy-resolution degradation due to crosstalk in adjacent detector read-out channels, pulse width ideally should be set short enough that the expectation value for photon detection on three pixels in the same pulse is much less than one. Thus, in principle, optimization tends to be short periods but is limited by the time constant of the MXS response.

The fluorescent sources would be switched on occasionally, possibly once a day, to measure the nonlinear effects in the SXS response. Of course, an early learning period was foreseen in which the actual MXS operations would be tuned in order to allow for optimal dynamic calibration of the SXS.

4.1.

Beryllium Filter

A spectral region of great interest for the SXS spectrometer is around the Fe-K lines (6.4 keV and above). High count rates decrease the resolution of the instrument. At high count rates, a larger fraction of detected x-ray photons (events) end up as low-resolution grade events (see Fig. 20). Hence for bright sources, the energy resolution in the Fe-K region can be degraded by excess flux at low energies. To mitigate this effect, the beryllium x-ray filter has been designed. Simulations were performed to estimate the effect of a beryllium filter on bright sources. It was found⁸ that a typical black hole candidate transient (BHCT) type source represents the best case, in which a beryllium filter can improve the quality of the observations. Figure 21 shows the expected SXS effective areas when employing different Be filter thicknesses. Figure 22 shows the effect of the same filters on the number of observed high- and medium-resolution events for a BHCT model source. The number of good resolution events around the Fe-K band increase for all beryllium filter thicknesses. Although the most positive effect is gained when employing the thickest filter, a filter with a beryllium thickness of $25\mu \mathrm{m}$ was selected for flight in order to preserve all types of science in case of a filter-wheel malfunction.

Fig. 20

Ratio of detected counts with respect to input photons for the different event grades¹⁵ as function of flux. High grade events have the highest energy resolution.

Fig. 21

SXS effective area for a (potential) set of different beryllium x-ray filters. In addition, the effect of 25% and 10% neutral density filters is shown.

Fig. 22

Expected high- plus med-res (H+M) event counts for the beryllium filters applied to a model BHCT spectrum. The solid black line (labeled “source”) are the expected number of events when all incident photons would be recorded as H+M events. Dashed line (“open”) is the actual number of recorded (H+M) events when no filter is applied. Colored lines are the number of H+M events for the different beryllium filters.

Two flight-model (FM) beryllium filters were calibrated at the Berlin Synchrotron (BESSY II) in terms of transmission. Unfortunately, due to a workmanship error, these filters were not able to withstand the launch vibration levels. Instead, a qualification model (QM) beryllium filter, for which it was proven that it could withstand the vibration levels, was mounted in the FM filter wheel. Since the batch of the beryllium used for the FM and QM models was identical, the calibration data for the beryllium spectral transmission were valid. The absolute thickness of the filter would then have been calibrated in flight. It was found that the only relevant contaminant in the beryllium foil was a small amount of iron (Fe). Figure 23 shows the tiny drop in filter transmission around the Fe-K edge.

Fig. 23

Calibration of the transmission of a $25\text{-}\mu \mathrm{m}$ beryllium filter around the Fe-K edge. A small drop of about 0.25% is observed. Measurement uncertainty is around a 0.0003 level in transmission.

4.2.

Neutral Density Filter

The second x-ray filter is the neutral density filter. This consists of a molybdenum sheet with holes (see Fig. 24). The sheet has a thickness of 0.25 mm, and the hole diameter is 1.1 mm. The total opening of all holes defines the transmission of the filter, which was computed at 24.5%. Molybdenum was selected as filter material because the major fluorescent lines of this material are outside of the SXS energy band. The cosmic particle background radiation could otherwise generate additional background lines when the neutral density filter would be in the beam. This filter was intended to be used for very bright sources, to limit the flux equally across the energy band, in order to prevent the energy resolution to degrade due to high flux.

Fig. 24

Picture of the neutral density filter, mounted in the filter wheel disk. It consists of a molybdenum sheet with holes.