A&A 411, L71-L79 (2003)
D. Attié1 - B. Cordier1 - M. Gros1 - Ph. Laurent1 - S. Schanne1 - G. Tauzin1 - P. von Ballmoos2 - L. Bouchet2 - P. Jean2 - J. Knödlseder2 - P. Mandrou2 - Ph. Paul2 - J.-P. Roques2 - G. Skinner2 - G. Vedrenne2 - R. Georgii 3 - A. von Kienlin3 - G. Lichti3 - V. Schönfelder3 - A. Strong 3 - C. Wunderer3 - C. Shrader4 - S. Sturner4 - B. Teegarden4 - G. Weidenspointner4 - J. Kiener5 - M.-G. Porquet 5 - V. Tatischeff5 - S. Crespin6 - S. Joly6 - Y. André7 - F. Sanchez8 - P. Leleux 9
1 - DSM/DAPNIA/Service d'Astrophysique, CEA Saclay, 91191 Gif-sur-Yvette, France,
2 - Centre d'Études Spatiales des Rayonnements, 9 avenue du Colonel Roche, 31028 Toulouse, France
3 - Max-Planck-Institut für extraterrestrische Physik, Giessenbachstraße, 85748 Garching, Germany
4 - NASA/Goddard Space Flight Center, code 661, Greenbelt, MD 20771, USA
5 - Centre de Spectrométrie Nucléaire et de Spectrométrie de Masse, IN2P3-CNRS, 91405 Orsay Campus, France
6 - CEA Bruyères-le-Châtel, BP 12, 91680 Bruyères-le-Châtel, France
7 - CNES/Centre Spatial de Toulouse, 18 avenue E. Belin, 31401 Toulouse, France
8 - Universitat de Valencia, IFIC, Dr Moliner 50, 46100 Burjassot, Valencia, Spain
9 - Univ. of Louvain, chemin du Cyclotron, 1348 Louvain-La-Neuve, Belgium
Received 14 July 2003 / Accepted 26 August 2003
Three calibration campaigns of the spectrometer SPI have been performed before launch in order to determine the instrument characteristics, such as the effective detection area, the spectral resolution and the angular resolution. Absolute determination of the effective area has been obtained from simulations and measurements. At 1 MeV, the effective area is 65 cm2 for a point source on the optical axis, the spectral resolution 2.3 keV. The angular resolution is better than 2.5 and the source separation capability about 1. Some temperature dependant parameters will require permanent in-flight calibration.
Key words: instrumentation: detectors - instrumentation:
spectrographs - space vehicles: instruments -
gamma rays: observations
During the SPI ground calibration at Bruyères-Le-Châtel (BLC), low intensity radioactive sources were used at short distances (up to 8 meters) for energy resolution, camera efficiency and homogeneity measurements. In addition, specific tests were included using high intensity radioactive sources at 125 meters for imaging performance measurements and photon beams generated with a Van de Graaf accelerator were used for high energy ( MeV) calibration (Mandrou & Cordier 1997; Schanne et al. 2002). Some additional runs with standard detectors were necessary to understand the high energy lines. Different GEANT simulations which provided the Instrument Response Function (IRF) (Sturner et al. 2003) have been compared to measurements.
The spectrometer SPI is described in Jean et al. (2000) and Vedrenne et al. (2003). In the text that follows the term camera refers to the Ge detector plane enclosed by the Anti Coincidence System (ACS) and the Plastic Scintillator Anticoincidence sub-assembly (PSAC). The term telescope refers to the whole system of the camera with the mask. The description of the camera and its events types (SE and ME) is given in Vedrenne et al. (2003).
Table 1: Energies of the radioactive sources used at 8.533 meters and the associated efficiency calibration obtained at Bruyères-Le-Châtel. The efficiencies camera for the SE and ME is respectively given by and . The transparency of the mask support is and the transparency of the central pixel is . Using this transmission factors and with an air density g cm-3 the effective area of the SPI telescope is given by with the error in cm2.
The SPI imaging capability has been tested using high intensity sources (241Am, 137Cs, 60Co, 24Na), located at 125 m from the telescope, outside the experiment hall through a transparent window. For security reasons, the beam was strongly collimated with a diameter of 2.5 m at the SPI position. In order to ensure that the entire telescope was inside the beam, a scanner using a NaI detector measured the vertical and horizontal beam profiles. Before each SPI run, a standard Ge detector, whose efficiency had been thoroughly calibrated, was used for 10 min in order to obtain the real -ray flux entering SPI, thus avoiding the need for any correction for the absorption within the 125 m air column.
The energy calibration and efficiency measurements were performed with the mask removed so that the whole camera was illuminated uniformly from a distance of about 8 m. A preliminary 6-day monitoring campaign in the calibration hall with a Ge standard detector demonstrated the absence of significant background variations.
For the low energy measurements, eleven radioactive sources were used in the range from 60 keV to 1.8 MeV. Sources emitting single -ray lines or well separated lines were preferentially selected. All source characteristics are listed in Table 1. In these data acquisitions, the sources were placed at 8.533 m from the Ge detector plane. At this distance, we can consider that each Ge detector is illuminated under the same solid angle.
For the high energy range calibration, a high intensity proton beam was directed onto a water-cooled thick 13C target (100 g/cm2), with SPI at 45 from the beam axis. Two resonances of the 13C(p,)14N nuclear reaction at and 1747 keV produce photons up to 9 MeV with sufficient intensities. Relative line intensities at an angle of 45 depend on the angular dependence of the -ray emission. This effect has been measured taking into account all absorption processes (Gros et al. 2003). Since the intensity of the proton beam on the target and the photon yield are not well known, absolute efficiencies are not directly calculable. Thus we used a two step process. The efficiencies obtained from accelerator spectra were normalized to the 1638 keV line efficiency (Table 2 for 550 keV). The absolute efficiency at this energy was calculated from the interpolation of the low-intensity source efficiencies, the absolute efficiencies could then be derived for the other accelerator lines.
A calibration phase with the INTEGRAL satellite completely integrated was performed at the ESA center of Noordwijk (ESTEC). During the measurements INTEGRAL was operated vertically and irradiated by the sources previously used for the 8 m distance measurements at BLC permitting a comparison with the BLC calibration results corrected for the mask.
The natural radioactivity background spectrum is subtracted from
the source spectrum. The photopeaks in the resulting spectrum are
The spectral resolution and the energy-channel relation have been
measured during thermal tests for each Ge detector (Paul 2002).
They are temperature dependent. The mean resolution can be fit by
a quadratic function of E:
The energy-channel relations are almost linear for all detectors. The variation of the peak position with temperature was at (measured in the temperature range 93 K-140 K). The relations obtained in the calibration process are useless for in-flight data. The energy-channel relations will change following the temperature differences.
|Figure 1: Spectral resolution ( ) for the flight model camera at 90 K. For each energy, the resolutions for the 19 individual detectors are plotted. The fit function is given by Eq. (2).|
|Open with DEXTER|
To obtain the full-energy peak efficiency of the SPI telescope we corrected the efficiency of the camera alone for the absorption of photons by the mask. These results are compared to simulations.
For a Ge detector i, and a -ray line at energy E
produced by a source, the full-energy peak efficiency
is defined by the ratio
In the case of a source, the absolute flux
at the detector plane is given by
For the accelerator case, the intensities
lines in the accelerator spectrum, corrected for all absorption
effects, are relative to the 1638 keV line intensity
(Table 2). The efficiency is:
|Figure 2: Full-energy peak efficiency for sources on the SPI optical axis ( ), without mask. An average for detectors 1 to 18 is shown; detector 0 is plotted separately (see Sect. 5.2). The measurements are compared to Monte-Carlo simulations.|
|Open with DEXTER|
For detectors 1 to 18, efficiencies are comparable. The value is representative of the efficiency of a single detector. Efficiencies and are compared to the GEANT simulation average efficiency . is 10% higher than (Fig. 2).
Using 14 of the 18 energies of Table 1, we compared for each detector i the efficiency homogeneity functions to the corresponding mass homogeneity functions where . In Fig. 3 we display the homogeneity functions and their energy dependence for sources for a range of energies E on the optical axis ( ) and for .
If , the efficiency of detector 0 seems to be 10 to 20% less than , and the deviation increases when the energy decreases. This behaviour is the signature of absorption, which in the case is due to an Hostaform plastic device inserted in the center of the plastic anticoincidence scintillator (PSAC) for alignment purposes. Detectors 2 and 3 are also affected. It was subsequently found that during these calibration runs the sources were actually slightly off-axis, causing the alignment device to partially shadow detectors 2 and 3. This explains the slight attenuation observed for them (Fig. 3, left panel).
Table 2: Strongest -ray lines of the 550 keV resonance of the 13C(p,)14N reaction. is the relative intensity emitted by the target under 45, is the relative intensity at SPI position after absorption in the target backing and holder, the cooling water, 8 m of air and the window separating SPI from the experimental hall (normalized to the 1637.9 keV line). The 8076 keV line is Doppler shifted from its nominal 8060 keV, as all lines coming from the 8062 keV level. For the others values see Table 1.
|Figure 3: For each detector, a plot of the efficiency versus energy is shown (in each plot, the line energy increases from left to right according to the 14 energies used). The crosses represent the mass of the crystal. When the detector is not shadowed, the efficiency depends mainly on its mass. Left panel: at , source on axis, the detectors 0, 2 and 3 are partially shadowed by the PSAC-alignment device. Right panel: when the source is at , detectors 16 to 18 are shadowed by the ACS. The alignment device is projected outside the detection plane.|
|Open with DEXTER|
If , the alignment device is projected on one or more other detector(s). During measurements at , it is projected outside the camera, detector 0 shows a normal efficiency. The very low efficiency of detectors 16 to 18 is due to the shadow of the ACS (Fig. 3, right panel).
In the case of ME, the incoming photon cannot be associated with a specific detector. So, the whole camera must be considered. We constructed a spectrum for each calibration source and multiplicity m using events from the whole camera. We then fit the lines in the background corrected spectra. The ME efficiency for a multiplicity m, , is defined as for the SE in Eq. (3). We find that above 4 MeV, the total ME efficiency is greater than the SE efficiency. The contribution of different multiplicities is displayed in Fig. 4.
During the acquisition of the data used in the previous sections, the mask was removed to let all detectors be illuminated by the sources. Deriving the telescope efficiency from the camera and detector efficiencies need to be corrected for the absorption of photons by the open and closed elements of the mask. This correction is evaluated for on-axis sources at infinite distance, thus the rays are considered to arrive parallel on the mask.
|Figure 4: Full-energy peak efficiency for SE, ME and all events (SE+ME) and for ME with multiplicity m = 2, 3 and >3, for the camera without mask. The effective area is the geometric area multiplied by the efficiency.|
|Open with DEXTER|
The transparency of the 63 open mask pixels have been measured individually with different radioactive sources from 17 keV to 1.8 MeV (Sánchez et al. 2003). Using these data, a mathematical model was fit to reproduce the mask absorption for the open pixels and especially for the central pixel, affected also by the alignment device. Note that above 2 MeV the absorption values have been extrapolated.
The efficiencies computed from on-axis 8-meter source data analyses for the detectors 0 and 3 were corrected for absorption due to the mask support (Fig. 5). For detector 3, these efficiencies are in good agreement with the efficiencies obtained with the mask installed at ESTEC and at BLC using the 125-meter source data. Thus, for an unshadowed detector, the correction method can be applied. For detector 0, when the source is on axis, the presence of the alignment devices in the PSAC and the mask complicates the efficiency calculation. Below 1 MeV, we adjusted the correction (Sánchez et al. 2003) to fit the BLC and ESTEC measurements.
be the mask support transmission for the
the mask element
transmission multiplied by the mask support transmission for the
shadowed ones. For SE, Mi(E) is the transmission of the
element i of the mask at the energy E for the detector i(
). For SE we correct
the efficiency for each detector. For ME we correct the global
camera efficiency obtained in Sect. 5.3 by the global
mask absorption. The total effective area
of the telescope is then:
|Figure 5: The on-axis, full-energy peak efficiencies of detector 0 and 3 for SPI with its mask. The data are: 8 m BLC data without mask corrected for the mask absorption (), 125 m BLC data with mask, () and 8 m ESTEC data with mask (). For detector 0, the disagreement between the curves shows the difficulty of alignment device modelisation.|
|Open with DEXTER|
For imaging, Eq. (6) is not fully valid for ME. In this case, the number of the detector where the photon had its first interaction can be known only with a probability p < 1. The (1-p) fraction is attributed to other pixels of the mask (closed or open), and so the real ME efficiency is always less than (ME camera efficiency) (mask absorption corrections).
We have compared the measured effective areas of the SPI telescope to those found in the SPI Imaging Response Files (IRFs), see Fig. 6. Here we have limited our comparison to the on-axis full-energy peak effective areas. The IRFs used in the ISDC data analysis pipeline have been simulated using a GEANT-based software package (Sturner et al. 2003). Note that the version of the IRFs released in November 2002 have subsequently been corrected at low energies using calibration data. The total effective area is about 125 cm2 at 100 keV and 65 cm2 at 1 MeV.
|Figure 6: Full-energy peak effective area of SPI telescope using all events (SE+ME). This curve is representative of the in flight effective area for an on-axis source. We compare it to the response matrix before the launch (release 1, Nov. 2002) and after the launch (release 3, July 2003).|
|Open with DEXTER|
The long-distance source tests were designed to verify the capabilities of the entire instrument in the imaging mode, the response matrix derived from Monte Carlo simulations, and the performance of the instrument/software/response matrix combination. For 125 m, the beam divergence was about 4' and the angular size of the sources 1'.
The angular resolution and Point Spread Function (PSF) of an instrument such as SPI is a function not only of the characteristics of the instrument but of the sequence of pointings (the "dithering pattern") used and the image reconstruction technique adopted. Here we take the PSF to be the response in an image obtained by correlation mapping (Skinner & Connell 2003) to a point source observed according to a specific dither pattern. We then take as the angular resolution the full width at half maximum ( ) of this response. Figure 7 shows some results.
|Figure 7: The PSF determined from measurements of 4 different sources at 125 m. Each curve is the mean of cross-sections through the response in two orthogonal directions; a background level has been subtracted from each and the curves normalized to the same peak height. The mean FWHM is 2.55. (241Am: 60 keV; 137Cs: 662 keV; 60Co: 1332 keV; 24Na: 2754 keV).|
|Open with DEXTER|
Given that the hexagonal mask elements are 60 mm "across flats" and that the mask-to-detector distance is 1710 mm, the expected angular resolution is 2.0. But because of the finite detector spatial resolution, the will be larger. Although the detector pitch is equal to the mask element size, the detectors are somewhat smaller (56 mm). This gives an expected of about 2.5. The measured values are consistent with predictions. The does not vary significantly with energy for the sources used (59-2754 keV).
The source location accuracy depends on the signal-to-noise ratio, , of the measurement as well as the angular resolution of the instrument. Some analysis results of the analysis software Spiros (Connell et al. 1999; Skinner & Connell 2003) are shown in Fig. 8. The signal-to-noise ratio of the BLC data is extremely high. It is important to verify the performances with values of more representative of flight values. To do this, random subsets of events were taken. These subset were also diluted by adding increasing amounts of Poisson-distributed noise. We found that the position accuracy does not increase when the signal-to-noise ratio exceeded 50-100. This suggests that there are systematic effects, which limit the accuracy to >2.5'. This could be due to uncertainties in the telescope stand alignment, which are about that level. Assuming that the reference axis was displaced from the source direction by a fixed 2.5', the residual errors suggest that the intrinsic limit of the instrument may be about five times lower (filled symbols).
|Figure 8: The error in the determination of a narrow line source position from analysis of calibration data. Squares: subsets of data, circles: mean of 10-20 trials with samples of random noise added. Open symbols: deviations from the nominal position, filled symbols: deviations from an assumed displaced source position. The line corresponds to 2.5.|
|Open with DEXTER|
|Figure 9: Maximum entropy image in the 1173 keV 60Co line shows the capability of the instrument to separate sources closer than the angular resolution, where the signal-to-noise ratio is good. The sources are separated by 1.|
|Open with DEXTER|
Even if the angular resolution is about 2.5, one can distinguish sources separated by less than this angle if data have a good signal-to-noise ratio. The BLC calibration runs were restricted to single sources, but it is possible to combine the data from runs at different source angles to emulate data for double sources. Figure 9 shows Spiskymax (Strong 2003) image for sources separated by 1 using the 1173 keV line of 60Co. The source are clearly separated, thus showing that SPI, in the very high signal-to-noise regime of BLC, is able to resolve sources at least as closely spaced as 1. With the lower signal-to-noise flight conditions the same performance will not always be achieved.
A threshold calibration was performed using two radioactive
sources (203Hg: 279 keV; 137Cs: 662 keV) placed close to
each of the 91 BGO crystals. The redundant cross-connections
between pairs of crystals, PMTs (photomultiplier tube) and
electronics give a broad threshold function, which can be
|Figure 10: For PMT1(solid), = 100 keV and , for PMT2(open), 120 keV and . These curves illustrate the large differences between ACS crystals and PMTs combinations.|
|Open with DEXTER|
|Figure 11: ACS permeability as a function of the angle from the axis (BLC, 137Cs source)|
|Open with DEXTER|
The light yield of BGO crystals varies with the temperature. This has an significant effect on two main characteristics of the ACS:
Table 3: Background environment lines observed by SPI in absence of sources.
These two ACS effects will modify the in-flight sensitivity of the telescope: variations of temperature on the orbit will affectthe Compton background under the lines and the dead time. Current in-flight variations are about 10 K. They are permanently monitored so that these effects could be further quantified.
We produced a catalogue of all the -ray lines detected by SPI on the ground (Table 3) by summing over all of the background data. Some never before observed very-high energy lines were detected. These lines are also visible in the keV spectra, but not in the keV ones. A careful study of BLC environment excluded any kind of non-natural emission during our background tests. A complementary test of s using an isolated standard Ge detector did not produce these lines. So their observation in the SPI background spectra must be related to the presence of low energy or thermal neutrons and the Ge detector array structure of the camera.
During background measurements, the neutron flux comes mainly from the spallation of SPI materials by cosmic rays. Thermal neutron capture for Ge isotopes has been studied in details by Islam et al. 1991. They give the neutron separation energies: Sn (71Ge ) = 7415.9 keV, Sn(74Ge ) = 10 196.3 keV, Sn(75Ge ) = 6505.2 keV. Lines are detected in SPI at these energies, but the most striking feature is that the 10 196 keV line corresponds to a forbidden transition between the Sn level and the ground state.
A simulation of some of the numerous possible cascades in the de-excitation of 74Ge nuclei showed that such a line can only be observed due to the summation of different transitions of the cascade by at least two detectors close to one another, i.e. in coincidence. On the other hand, the line corresponding to Sn(73Ge ) = 6782.9 keV is not observed, but a 6717 line is. This line corresponds also to the summation of transitions, but only down to the 67 keV level of the isomeric Ge which decays independently with a lifetime of 0.499 s.
The determination of line origins is necessary to understand the SPI camera behaviour in space, where neutron and spallation induced lines will affect the observations. The relative intensities obtained in the ground calibration can then be used when subtracting the background from astrophysical data.
SPI will be able to detect nuclear astrophysics lines and continuum. The total effective area of SPI was found to be >100 cm2 in the energy range 40 keV to 300 keV with a maximum of 136 cm2 at 125 keV. At 1 MeV the resolution of the instrument is 2.5 keV with an effective area of 65 cm2, (at 511 keV: 90 cm2; at 1.8 MeV: 52 cm2).
The angular resolution of SPI was found to be roughly 2.5 and it has been shown that SPI will be able to resolve sources with a separation of 1 and probably less.
Using ground calibration data and GEANT simulations, we derived an absolute calibration of the SPI effective area. We noticed the temperature influence on ACS effects and energy restitution. For these two points, we must underline that permanent in-flight calibration is required. Neutron or spallation induced background lines will be used as tracers to extract the background component of some lines of astrophysical interest.
The authors would like to thank: C. Amoros, E. André, A. Bauchet, M. Civitani, P. Clauss, I. Deloncle, O. Grosjean, F. Hannachi, B. Horeau, C. Larigauderie, J.-M. Lavigne, A. Lefèvre, O. Limousin, M. Mur, J. Paul, N. de Séréville, J.-P. Thibault who took part in the shifts at BLC, J.-P. Laurent (CISBIO) for the very tight logistics of high intensity 24Na sources, the BLC Van de Graff team, M. A. Clair (CNES), our project manager, R. Carli (ESA) ESTEC calibration manager, D. Chambellan, B. Rattoni (DIMRI) for their essential contribution to source preparation, P. Guichon and S. Leray (CEA-Saclay), P. Bouisset, R. Gurrarian and E. Barker (IRSN - Orsay) for fruitful discussion about nuclear physics.