Solar Physics, 197/2
337-360, 2000

J. Sylwester

Space Research Centre, Polish Academy of Sciences, 51-622 Wroclaw, ul. Kopernika 11, Poland

F. Farnik

Astronomical Institute, Academy of Sciences of the Czech Republic, 251 65 Ondrejov

O. Likin

Space Research Institute, Russian Academy of Sciences, Profsoyuznaya St. 84/32, Moscow

Z. Kordylewski, M. Siarkowski, S. Nowak, S. Plocieniak, W. Trzebiński

Space Research Centre, Polish Academy of Sciences, 51-622 Wroclaw, ul. Kopernika 11, Poland

Abstract

On 16 October 2000, INTERBALL-tail ended his usefull orbital life and has been directed to plunge into Atlantic Ocean.

1. Introduction

2. General

The RF was developed in order to facilitate high-time resolution measurements of integral solar flux in the spectral energy range between 2 keV and 240 keV and to image solar flares in the bands 2 - 4 keV and 4 - 8 keV from rotationally modulated data (X-ray Tomograph-Imager, XTI). The photograph of RF instrument is shown in Fig. 1.

The RF is equipped with five detector systems:

• Proportional gas-filled detector.

• NaI(Tl) scintillation detector with Al window.

• Three silicon detectors sensitive to charged particles (radiation belt detectors).

• Proportional detector behind the XTI.

• XTI attitude sensor (small optical refractor f = 6 mm).

Below we describe in more detail both instruments.

3. The Photometer

Table 1: Nominal energy ranges and designations for the lower energy photometer channels (formatted into standard, common block of telemetry).

    The proportional detectors consist of two cylindrical compartments with steel walls ~ 1.2 mm thick sharing the same gas. The section exposed to the solar radiation is equipped with the Be window in order to select the radiation with the energy above ~ 1 keV. The hidden (twin) section is illuminated by the radioactive calibration Fe⁵⁵ source. Corresponding calibration signal is constantly probed in order to detect ageing, leaking and other possible effects which may cause deterioration of the gas gain. Special electronic feedback system regulates accordingly the high voltage level on both anodes of the detector counteracting in this way the effects of gain variations. The energy discrimination thresholds are thus kept in place independent of the gain variations. However, towards the end of 1997, the activity of Fe⁵⁵ calibration source (7 years from fixing) decreased more than 10 times and went outside the limits of self-correction. Details of the gain time variations will be presented in a subsequent paper where the history of discrimination level positions will be given for entire duration of the mission. The average background count rate is ~ 10 cts/s for the proportional detector.
    The scintillation detector contains 8 mm thick NaI(Tl) crystal with the Al filter (0.25±0.01  g/cm² mass thickness) placed in front and a photomultiplier tube. This detector has its own calibration source (Am²⁴¹) but without feedback control.
In Figure 2, we present efficiencies of the proportional and scintillation detectors plotted against the photon energy. The incident solar photon flux is transmitted by detector window and absorbed by gas/crystal with the corresponding efficiencies. The redistribution over energies according to spectral response function takes place afterwards.
The proportional detector collects soft X-ray photons in three energy bands 2 - 3 keV, 3 - 5 keV and 5 - 8 keV (see Table 1). The detector window is placed slightly below the instrument cover opening. This geometry causes the effective aperture to vary by few percent as the solar illumination angle changes with the satellite spinning phase.

Table 2: Nominal energy ranges and designations for high-energy photometer channels recorded (in flare mode) into individual telemetry blocks. The measurements are reported every 125 msec.

Figure 2: Spectral efficiency of the proportional (left) and scintillation (right) detectors as a function of photon energy.

Another factor contributing to modulation of the signal is that the absorption depth of the photon also depends on the rotation phase. This latter effect is especially important for higher energy radiation, for which the photon absorption length in the gas is greater than the detector depth.
    The scintillation detector measures the hard X-ray flux in five energy bands (see Tables I and II). In flare mode the measurements are made every 0.125 s in the four highest energy channels. For these channels, the data are allowed to be transmitted to the telemetry provided that corresponding rate thresholds (40 cts/s) are exceeded for the four consecutive seconds. Otherwise, the data are not written in order to secure the telemetry volume. The channel h1 output is recorded differently. It is formatted together with the signal from the proportional detectors and sent to telemetry all the time (every 2 or 8 s) depending on the mode (flare or patrol). The adopted mode names may however be somewhat misleading, since the most of data have been collected in so-called flare mode. The reason for this was the selection of very low threshold for flare trigger. Actually it has been set at the level corresponding to microflare (few cts/s). Switching between patrol/flare modes is controlled onboard based on count rates recorded in s2 channel. Early in the mission, the instrument operated mostly in the flare mode, for periods up to several hours.

Table 3: Essential characteristics of the proportional detector.

Later on, following the change in the detector feedback control, the frequency of switching between the modes increased, and presently the photometer operates preferentially in patrol mode except for flares above GOES class B5 - B7.
    During the patrol mode the calibration of the scintillation detector is made approximately every 2 hours (cf. Table I). For calibration, a moving arm equipped with Am²⁴¹ radioactive source (half-decay time 432.2 years; energy 59.6 keV) is being mechanically inserted into the FOV of the detector. The calibration arm partly occults the aperture contributing to the signal in h1. In this way during the calibration the recorded signal responds in most cases to the calibration source quanta only. The other data reported at the time of calibration (s2 fluence) represent solar signal. The scintillation detector has no feedback control and therefore the ageing effects can be deduced only from analysis of time variations of the calibration level readouts. This analysis is in progress.

4. Examples of Photometer Observations

These observed modulations have few percent relative amplitudes and the period around 118 s. In order to remove the fluctuations, we have determined appropriate phase modulation profiles (for each of the channels separately) using epoch folding (details of the procedure will be described in a subsequent paper). These folded profiles have been next used in order to remove the modulation from the raw data. At present, we are in the process of systematic reformatting and regularizing the data.

At our web site (www.cbk.pan.wroc.pl/rf15-i_www/RF15-I _observations.htm) we put all the data processed so far along with short explanations. On a positive side is that the modulation of the signal due to the satellite rotation allows for distinguishing between solar X-ray emission and non- solar particle contribution (the last being usually not modulated).
    One of the important advantages of RF measurements is that there are many periods where the solar signal has been measured every 2 s continuously for up to 40 hours in the four energy bands.
    The overall data coverage of the solar X-ray photometry is ~ 85%.

    The physical analysis of data is in progress. In the example shown in Figure 4 (bottom panel) we present derived variations of coronal source plasma temperature as determined in the isothermal approximation from the ratio of fluxes recorded in the s1 and s2 bands. One may notice that the scatter in derived temperature is relatively small and that the temperatures are within realistic range also for periods where no flare in progress has been reported. Details of the physical data analysis will be published in a forthcoming paper.
    As concerns the data collected in the higher energy channels, reliable measurements have been made for hundreds of events (Siarkowski et al., 1999). Analysis of this data indicates that during strong events, pile-up effects might contribute to the flux in higher energy channels, as well as the detector dead time effects may become important (as pointed up by the Referee). The dead-time for electronics of both proportional and scintillation counters is 8ms, larger than intrinsic detector's physical dead-times. Therefore, the effects of pile-up and dead-time corrections will become important (at 10% level) for count rates above 10⁴ cts/s. A more detailed consideration of these problems will be given in the forthcoming paper.

    For the M1.2 flare shown in Fig. 6, RF recorded the fluxes in h2 and h3 energy channels. This flare has been observed also by Yohkoh HXT and BATSE. We put corresponding lightcurves in the Figure for the comparison. The overall agreement of signals in corresponding energy bands is very good (for description of the HXT construction see Kosugi et. al., 1991). Detailed cross-calibration between RF, GOES, HXT and BATSE measurements is in progress.
    The higher energy RF data are also presented and available from our web server.
While crossing van Allen radiation belts the detectors were designed to be switched-off. We noticed however that on several occasions the particle flux has been low enough not to trigger the in-belt flag. Therefore the detectors did not switch-off and recorded signal due to particles. This resulted in a characteristic pattern of recordings in all channels of RF. Such particle events can be easily recognized through their specific inverted/flat energy spectrum. On at least one occasion the non-solar hard X-ray event has been detected, originating from the SRG1900+14 burst (Sylwester et al., 1998).
    We have encountered a number of problems during data reduction and prompt analysis. For example, the count rate level from large flares (i.e. above M5-M6 GOES class) exceeds the capacity of counters (set at 32000 and 64000 hits per integration time in particular channels). This caused counter wrapping, sometimes many-fold.

5. The X-ray Tomograph-Imager

The XTI consists of two units:

• The X-ray collimator, and

• The optical solar disk sensor.

5.1. The X-ray Collimator (XRC)

    The exact location of the grid planes have been calculated according to a novel strategy. The novelty of the solution is the use of semi- numerical interactive method of placing the grids. In case of the XTI, we specified the approximate values for the collimator length, thickness of grids, desired resolution and transmission fringe separation and found the optimum collimator design as presented in Fig. 8.

The collimator grid system has been co-aligned by visible light technique with diffraction effects taken into account. We measured separately the transmission in X-rays. The area averaged rms alignment accuracy has been found to be ±1 mm, and the maximum wire misalignment found is ~ 10 mm.
    As shown in Fig. 8, the front and the middle grid wired sections cover only a portion (approximately half) of the planes. The front grid has additionally an occultation strip. The width of this strip is selected to be 6.2 mm allowing thus to ensure that the solar disk will be fully eclipsed (as seen from the detector window). With the Sun crossing the bar, the solar X-rays are totally blocked permitting measurements of the background signal (cf. Fig. 8).

    In Fig. 9 we schematically draw the trajectory of a compact solar source (flare) as seen in the coordinate system of the collimator (front grid). Provided nutation is negligible, the solar source trajectory is a circular path crossing the open area, occultation bar and gridded area. Depending on the offset d of the source relative to the axis of the collimator, a number of modulation patterns are expected:

• no bar occultation for d < 0.54^°

• single partial bar occultation for 0.54^° < d < 1.55^°

• double partial bar occultation for 1.55^° < d < 2.63^°

• full modulation pattern (as in Fig. 9) for d > 3.64^°

An example of calculated characteristic pattern of the signal modulation is shown in Figure 10.

The actual position of the satellite rotation axis relative to the optical solar disk is monitored by a dedicated optical sensor unit described below.

5.2. The Optical Solar disk Sensor (SDS)

    The SDS unit consists of the optical system which projects images of the solar disk on the system of two narrow slits. The optical axis of SDS is coaligned with the axis of the X-ray collimator. The slits are adjusted to be parallel to the occulting bar. Spinning of the spacecraft causes the projected optical solar disk to circle through the double slit system presented in Fig. 11.
    Behind the slit system, the photodiode monitors the overall optical flux passing through the slit system. A dedicated on-board computer routine analyzes the slit crossing times and determines corresponding phases of rotation and the solar centre offset angle based on the history of slit crossing times from the previous rotation. The slit crossing times were envisaged to be dumped to the telemetry after every 100 satellite rotations.

    It is by misfortune that during the pre-flight preparations immediately preceding the launch from the Baikonur, the aperture of the SDS optical sensor has not been cleared. As a consequence we are missing the data from this sensor. We have fortunately foresighted for such an unfortunate situation and pre-programmed a safe-mode activity for the XTI in such circumstances.

5.3. On-board data processing

The signals incoming from the XTI are being processed in real time by the PRAM onboard computer and stored temporarily to the onboard memory before being sent to the ground.
    In order to effectively use the memory capacity, we devised an elaborated scheme of on-board signal processing, which saves the memory and allows to concentrate on flare events occuring between consecutive telemetry dumps. In particular, the signal from XTI X-ray detector is constantly monitored in order to detect the flare occurence. Flare intensity class is being estimated in real time and the decision is made on-board when and which observing sequence is to be executed. Even with such elaborated scheme, only ~ 80 compressed image modulation sequences can be stored in the on-board memory between consecutive ground telemetry dumps (say one sequence per hour on average).

The following XTI operation modes have been preprogrammed:

• low activity - only minimum and maximum X-ray rate values for a given S/C spin are stored;

• high activity - detailed modulated signal together with extreme and averaged values of the background from unmodulated signal are preserved. Interfringe zeros are discarded.

5.4. Example Observations from XTI

In Figure 12, an example of the XTI modulated signal for the 20 October 1995 flare is shown. There are hundreds of periods identified already where the imaging sequences have been executed. We will present detailed analysis of these periods in one of a subsequent publications.

Figure 12: The observed modulation of the X-ray signal in the 2 - 4 keV band of the X-ray Tomograph-Imager (XTI) for 20 October 1995 flare. In the bottom panel, initial small portion of the recorded signal is enlarged in order to present more details of the record shown above. On the horizontal axis the time elapsed from the start of this particular modulation sequence is indicated.

5.5. Deconvolution of XTI images

The solar X-ray source (flare) is scanned at various angles (directions) due to spinning motion of the satellite. The detector signal in any of the transmission fringes represents one-dimensional projection of the two-dimensional source brightness distribution. As the phase of the rotation changes views from many projection angles are recorded. A resulting set of multiple one-dimensional scans constitutes the input for the numerical image reconstruction algorithm. Below, we briefly introduce the basics of this image reconstruction algorithm.
    In Fig. 13 we present definitions used when formulating the problem of image reconstruction. In this representation f(k,l) is the X-ray brightness of the investigated source.

Fixed coordinate system (k,l) is used. During the scan, the value of the measured signal in each position (time bin represented as column m perpendicular to the scan direction) represents a result of convolution of the collimator transmission function g and the source intensity integrated along this perpendicular direction. Thus, the flux measured in bin number j can be expressed as:

For the purpose of image reconstruction we have assumed the collimator transmission function to consist of a number of triangular profiles (windows) with FWHM = b = 11 arcsec. The consecutive transmission windows are separated by the angular distance p  = 288 arcsec. In the assumed coordinate system, the transmission g in bin j is:

where d is the bin width and n = 0, 1, 2, ... is the index of particular transmission window. In case of scan taken at the position angle j_i, the value of m in Eq. (2) is k cos j_i + l sin j_i, so the signal measured in bin j can be expressed as:

For compact sources with spatial extension smaller than the separation between transmission fringes, only a single transmission FOV contributes to the signal. By rotating the collimator relative to the source, its brightness distribution f_kl is recorded in a set of intensity vectors I_ij (corresponding to individual scans). This transformation is handled by a transmission tensor g_ijkl.

The purpose of image restoration process is to solve the inverse problem i.e. to determine f(k,l) from measured I_ij. We have developed an elaborated maximum likelihood iterative procedure to solve this inverse problem. Within the algorithm, we assume a uniform brightness distribution as an initial approximation (f_kl = const). The subsequent (improved, n+1) approximation for fⁿ⁺¹_kl distribution is obtained by multiplying the preceding result by the appropriate correction matrix. This correction matrix represents the ratio of the observed (O) and calculated (C) flux in each bin, weighted by the collimator transmission function:

For instance, if the observed flux at bin j is greater the calculated one, we increase the brightness in each of the pixels (k,l) which contribute to this bin (in proportion to the corresponding individual weights of the collimator transmission function).
    The quality of the iterative convergence is judged in terms of c² statistics. The process of image reconstruction is terminated when the value of reduced c² approaches unity.
    The inversion technique used here is a straightforward extension of one- and two-dimensional maximum likelihood iterative inversion methods commonly used for solar X-ray data analysis (Sylwester, Schrijver and Mewe, 1980;  Svestka et al., 1983, Sylwester and Sylwester, 1998, 1999). It is analogous of the computer tomography image reconstruction procedure used in medical applications, where scans from many directions are used to generate two-dimensional picture of (slices) of the body internal structure.
    The method used is inherently positively constrained since it uses multiplicative corrections based on observed to calculated (weighted) flux ratios. Within the present method it is possible to vary the pixel size in the reconstructed image, independently of the size or the number of modulation scans used for image reconstruction. At present, we perform deconvolution using 10 arcsec ×10 arcsec pixel size in the deconvolved image. We will look for the solutions with finer mesh in the future. In Fig. 14 we present an example of the reconstructed image for the 20 October 1995 flare observed around 06:14 UT. The lightcurves of this flare are shown in Fig. 4. The flare took place on the disk (S09 W55).

The limiting spatial resolution on the reconstructed image depends on many factors such as the signal/noise ratio of the modulated signal, the amplitude of nutation of the spacecraft rotation axis and the time variations of brightness within the flaring structure during the modulation period (we assume the latter not to change during this time). There are hundreds of identified modulation sequencies available for the image reconstruction. We will perform their analysis in the near future.

6. Conclusions

7. Acknowledgements

References

van Beek H.F.: 1976, Space Science Instrumentation, 2, 197.

Kosugi T., Mikishima K., Murakami T., Sakao T., Dotani T., Inda M., Kai K., Masuda S., Nakajima H., Ogawara Y., Sawa M., and Shibasaki K.: 1991, Solar Phys., 136, 17.

Lin R.P. at al.: 1998, 'High-Energy Solar Spectroscopic Imager (HESSI) Small Explorer Mission for the Next (2000) Solar Maximum', in C. M. Korendyke (ed.) 'Missions to the Sun II', Proc. SPIE Vol. 3442, p. 2

Oda M. Muranaka N., Matsuoka M., Miyamoto S., and Ogawara Y.: 1976, Space Science Instrumentation, 2, 141.

Siarkowski M., Sylwester J., Gburek S., and Kordylewski Z.: 1999, 'Review of RF15-I X-ray Photometer Observations', in A. Wilson (ed.) Magnetic Fields and Solar Processes, ESA, SP-448, p. 877.

Svestka Z., Schrijver J., Somov B.R., Dennis B.E., Woodgate B.E., Furst E., Hirth W., Klein L., and Raoult A.: 1983, Solar Phys., 85, 313.

Sylwester J., Kordylewski Z., Siarkowski S., Gburek S., Farnik F., and Likin O.:1998, Acta Astronomica, 48, 819.

Sylwester J., Schrijver J., Mewe R.: 1980, Solar Phys., 67, 285.

Sylwester J. and Sylwester B.:1998, Acta Astronomica, 48, 519.

Sylwester J. and Sylwester B.:1999, Acta Astronomica, 49, 189.

Tsuneta S.: 1984, Ann Tokyo Astron. Obs., 20, 1.