The EXOSAT Data Archive at the HEASARC

L. Angelini, N. E. White

HEASARC

The European X-ray Observatory Satellite (EXOSAT) was operational from May 1983 to April 1986 and in that time made 1780 detailed observations of a wide variety of astronomical objects. This article is an overview of the EXOSAT archive available from the HEASARC. The first part gives a historical overview of the EXOSAT Observatory, describing the EXOSAT instrumentation, operational phase and post-operational data. The following sections provide a description of the different levels of data currently present in the HEASARC archive and their format. Future plans for reformatting the raw data to FITS format are briefly described.

Historical Overview

EXOSAT was launched on 1983 May 26 from the Vandenberg complex in the USA on a Thor-Delta rocket and put in a highly eccentric orbit (e ~ 0.93) with a 90.6 hr period and an inclination of 73deg.. The mission lifetime was ultimately limited by orbital decay. The spacecraft performed within specifications for almost three years. On 1986 April 9 a failure in the attitude control system caused the loss of the spacecraft. The natural decay of the orbit caused EXOSAT to re-enter on 1986 May 6.

The scientific payload consisted of two low energy imaging telescopes (LEITs or LE), sensitive over the 0.05-2 keV band, a medium energy (ME) proportional counter and a gas scintillation proportional counter (GSPC) providing a coverage over the 1-50 keV band. At the focus of each telescope was a moveable instrument bench upon which a channel multiplier array (CMA) and a positional sensitive detector (PSD) were located. A transmission grating spectrometer (TGS) could be interposed into the light path of each telescope. The high energy instruments, ME and GSPC, were collimated telescopes, with a full width half maximum field of view , FOV, of 45 arcmin. The three instruments (LE, ME and GSPC) were complimentary and designed to give a complete coverage over a wide band pass of 0.05-50 keV.

The low energy imaging telescopes gave both medium quality spectral resolution using the gratings, or broad band filter spectroscopy. The field of view was about 2deg., with a peak effective area of 10 cm² (which was reduced by a factor of ten when the gratings were utilized). The ME detector gave spectra with 20% resolution at 6 keV, with a total effective area of 1600 cm². Each half of the ME detector array could be offset to monitor the particle background. The gas scintillation proportional counter provided a factor of 2 improved spectral resolution in the 2-30 keV band, but with a factor of 10-20 less effective area than the ME. The optical axes of the four telescopes were coaligned.

An on board computer (OBC), programmable from the ground, preprocessed and compressed the data. The flexibility given by the OBC and the capability to load new OBC programs, enhanced the quality of the science obtained, by writing new data modes to address science discoveries not foreseen before launch. It also proved invaluable to work around problems with the instrumentation and the spacecraft. Figure 1 shows an exploded view of the EXOSAT satellite.

The orbit of EXOSAT was quite different from that of any previous X-ray astronomy satellite. The initial apogee was 191,000 km and the perigee 350 km (Parkes 1985). The science instruments were operated above 50,000 km, outside the earth's radiation belts. This allowed scientific operations for up to 76 hr per 90 hr orbit, without interruption. EXOSAT was visible from the ground station at Villafranca in Spain for practically the entire time that the science instruments were operated and there was no need for any onboard data storage. EXOSAT operated in a 3-axis stabilized mode. A propane cold-gas thruster system was used for both slew maneuvers and fine pointing. The attitude was controlled using one of two star trackers, three gyros, and a sun sensor and could be maintained to within ~1 arcsec. The first 8 weeks after launch were dedicated to a performance verification phase. After this, the regular guest observer program was undertaken. By the end of the operational phase, four announcements of opportunity had been made with 98 %, 94 %, 64 % and 18 % of each program completed. During these 3 years, scientists around the world used EXOSAT to study most classes of X-ray sources and made many new discoveries. Perhaps the most notable were the discovery of QPO from low mass X-ray binaries, soft excesses from AGN, the red and blue shift iron K line from SS433, many orbital periods from low mass X-ray binaries, and several new transient sources. In ~ 3 years operation the total volume of data was about ~ 160 Gbytes. The public availability of these data contributed to many notable systematic studies of individual classes. The EXOSAT archive continues to be a heavily used resource and is still yielding new results.

Figure 1: An exploded view of the EXOSAT Observatory

Instruments

The Low Energy Imaging Telescopes

The two LEIT telescopes (de Korte et al (1981)) and detectors assemblies were identical. The telescopes consisted of double nested gold coated Wolter type I grazing incidence optics, with a focal length of 1.1 meter and an outer diameter of 0.3 meters. The telescope parameters give a high energy cut-off of ~2 keV. The on-axis half energy width of the point spread function is 24 arc seconds, which degrades to 4 arc minutes 1 degree off-axis. Vignetting in the telescopes reduces the off-axis effective area to 45% its peak value 1 degree off-axis.

In the focal plane of each telescope either a channel multiplier array, CMA, or a position sensitive proportional counter detector, PSD, was interchanged. Also, a transmission grating could be inserted behind each telescope and the dispersed spectrum imaged by the CMAs. The gratings were 500 lines mm-1 in one telescope (CMA2) and 1000 line mm^-1 in the other (CMA1). The spectral resolution is 2Å and 1Å respectively for energies > 0.25 keV, and 5Å at 304Å in both telescopes.

The field of view of each CMA covers a diameter of 2deg.. A CMA has no intrinsic energy resolution in the X-ray band, but a number of different filters gave coarse spectral information (analogous to UBV photometry). The choice of filter dictates the energy response with the overall energy window covered by the filter combinations ranging from 0.05 to 2.0 keV. The CMAs were sensitive to ultraviolet photons and this caused contamination for observation of bright O and B stars, pointed or serendipitous in the field. Filters were used to determinate the degree of contamination. Observations obtained with the boron filter were free of UV contamination and the Aluminum/Parylene is relatively immune to it except for fields with the very brightest and earliest stars. Also the sum signal distribution of events in the CMA provides a crude method of differentiating between X-ray and UV sources. The most commonly used filters during the mission were 3000 Lexan , Al/P and boron.

The CMA particle background counting rate was typically 8 x 10^-6 count s^-1 pixels^-2 (a pixel is 4 arcsec) in the central region. The background counting rate depends on the strength of the solar wind. In 90% of the observations the average background counting rate is within a factor of 2 of the quiescent value. The average source detection threshold for a 10⁴ second exposure (a typical minimum observation time) within the central 12 arc minute radius region of the detector, using the 3000A lexan filter, was 2 x 10^-3 count s^-1. For observations longer than a few thousand seconds the sensitivity of the CMA is background limited. Source position can be estimated to 6 arc seconds or 8 arcsec error radius at 67% or 90% confidence level respectively within the central 12 arc min.

There were a number of notable problems with the instrumentation associated with the LEITs. Both PSDs failed early on during the performance verification phase. One of the CMA (CMA2) failed on 1983 October 28. The mechanism to insert the grating behind the other telescope (CMA1) failed in 1983 September 15, eliminating this spectroscopic capability. Only 24 sources were observed with the gratings. These spectra provided some of the highest quality spectra obtained at that time of nearby stars (e.g., Capella), white dwarfs (e.g., HZ43) and Sco X-1. The surviving CMA functioned well until the end of the mission.

The Medium Energy Proportional Counter Array

The Medium Energy (ME) instrument consisted of an array of eight proportional counter, with a total geometric area of 1600 cm² and 45 arcmin FWHM field of view, providing spectral and temporal data in the 1-50 keV energy range (Turner, Smith and Zimmermann 1981). Each proportional counter had two gas chambers separated by a 1.5 mm beryllium window with argon in the top layer and xenon in the lower. Each proportional counter used a multi-wire design with a [[Delta]]E/E of 21(E/6 keV)^-0.5% FWHM for the argon chambers and 18(E/22 keV)^-0.5% FWHM for the xenon chambers. The argon and xenon spectra were pulse height analyzed into 128 channels each, sensitive to 1-20 keV and 5-50 keV energy ranges, respectively. The ME background was very stable and dominated by particle events from the solar wind and radioactive lines caused by the decay of residual plutonium in the Beryllium windows and detector bodies. The background count rate per detector were in the argon chamber 2.4, 4.3 and 9.4 count/s in the energy range 1-6, 1-10 and 1-20 keV respectively and 40.6 and 59.1 count/s in the xenon chamber in the 10-30 and 10-50 keV energy range, respectively. Occasional background flares occurred in both the aligned and offset halves caused by enhancements in the solar wind. A major solar storm typically happened every six months, causing the background to increase by several orders of magnitude. The detectors were turned off during these events.

To optimize the background subtraction each half of the ME array detector could be offset (also known as an `array swap') from the source, pointing at a source-free region of the sky to monitor the particle background. The offset half was alternated every few hours. Since the background obtained from the offset half was slightly different for the half of the detector on source, difference spectra were created to correct this effect. Background was also obtained using the slew on and/or off the source. This technique was used when the detector halves were coaligned and no array swaps made during the observation. In 1985 August 20 one of the detectors in half-1 failed. Occasionally because of small detector breakdown and/or reduction in the gain, observations were carried out with one or more detectors off for a few hours.

An important component in the operation of the ME instrument was the usage of the OBC. Depending on the objective of the observation, the OBC programs traded time resolution against spectral information. Depending on the telemetry load and the OBC programs running for the other two experiments, two or three ME programs could be run simultaneously. A typical ME observation was carried out with a primary spectral oriented program and a secondary high time resolution program. One of the major problems with the ME+OBC setup was a dominant deadtime effect mainly due to the OBC sampling.

The Gas Scintillation Proportional Counter

The gas scintillation proportional counter, GSPC, (Peacock et al (1981) on EXOSAT has a [[Delta]]E/E of 4.5(E/6 keV)^-0.5% FWHM, (a factor of 2 better than the ME experiment) with a peak effective area of ~100 cm². The energy spectra were pulse height analyzed into 256 channels. Three different electronic gain modes were used: gain 1 = 2-32 keV, gain 2 = 2-16 keV and gain 0.5 = 2-64 keV. The latter was only used on the very bright source Sco X-1. Variations in the intrinsic detector gain caused by temperature changes were removed using two background line features at 10.54 and 12.70 keV caused by fluorescence in the lead collimator and the radioactive decay of residual plutonium in the beryllium window. In addition a Xenon L feature at 4.78 keV could also be used for bright sources to calibrate the gain.

The particle background rejection used burst length discrimination, which rejects events which have exceptionally long or short duration. The background shape spectrum remains constant during the short timescale variations, but shows small changes on longer timescales. The standard OBC mode used in the GSPC gives 256 channel spectra every 8 seconds. Higher time resolution was used only for bright sources and if the telemetry needs of the other experiments was low. The GSPC worked perfectly for the entire mission.

EXOSAT Post-Operation

During the 4-year post-operational phase the EXOSAT observatory staff concentrated on producing a database of EXOSAT results and data products accessible via computer network by the whole community. During the 1987-1990 period two parallel efforts were undertaken. The first was creating basic products, lightcurves, spectra and images, for all the three experiments. The second was developing a database management system (later known as EXOSAT DBMS) not only to make those result products available, but also giving the possibility to manipulate, cross-correlate and derive results from the database. The computer resources available were a HP1000 running RTE operating system used to process the data and a VAX/VMS system dedicated to create the DBMS and provide the on-line access. The EXOSAT results database first went put on-line in April 1989. This was one of the first on-line astronomical archive systems to allow the on-line retrieval and analysis of scientific data. When the HEASARC was established in 1990, the EXOSAT system was adopted as its on-line system, and is now also in use at several other high energy astrophysics sites around the world.

Data Processing

The products, created for each experiment, were obtained through automatic processing. If the automatic analysis failed in creating reliable results, an effort was made to correct the nature of the failure. The images, lightcurves and spectra produced were in binary format designed to optimize disk space, which was a major constraint at that time. The binary table extensions to the FITS format were not well established when the EXOSAT database effort was begun (1987). Therefore, the resulting files were not portable between different operating systems, and intermediate ASCII files were used. The HEASARC has taken all the EXOSAT data products and converted them to FITS. The FITS data format was designed as general as possible to accommodate not only the EXOSAT data but also data from other missions. Guidelines for the FITS format for spectra, response matrices and lightcurves are described in the following Legacy articles: Arnaud et al. 1992, George et al. 1992 and Angelini et al. 1993, respectively. To analyze the data the XANADU software packages XSPEC (spectral fitting), XIMAGE (image analysis) and XRONOS (timing analysis) have been modified to read the data in the new format. The volume of all the products in the original binary format was about 1.7 Gbyte for a total number of files of about 40100, which expanded by a factor of 2 after transforming in FITS. The data available from HEASARC archive have been compressed using the standard unix `Z' compression to minimize network traffic. All the header information present in the original binary file has been kept in the current FITS header and stored in keywords. In some cases the original value was changed because it was not compatible with the keyword definition. To maintain the original information in all the EXOSAT FITS files, the original header information has been preserved in a COMMENT card.

The LE Products

The LE automatic analysis consisted of:

1) generating an image for each filter used during an observation;

2) searching the image for point-like sources;

3) for each detection, generating a background-subtracted lightcurve;

4) for each image, generating a background lightcurve.

After each run, the image and the detections were visually inspected. The source detection algorithm used is based on a sliding cell method. The size of the search cell was such to maximize the sensitivity across the field of view. When a source was detected its sky position (RA and Dec in 1950 coordinates), count rate and error, significance of the detection, the X and Y image pixel position, dead time correction and several other parameters were calculated. The information recorded for each detected source was written as a unique entry in a database together with the associated products. The integration time of the background-subtracted light curve varies from source to source and was chosen such that the average count per second per bin was about 0.3. The automatic analysis did not include a test for UV contamination, therefore a fraction of all the sources detected are due to such contamination.

The ME Products

The ME automatic analysis was designed to cope with the complex observing configurations as well as the myriad of OBC mode used during the mission. For each observation the automatic analysis creates the following products:

1) a source lightcurve background-subtracted in the energy range between 1-8 keV with a time resolution of 30 seconds;

2) a background lightcurve in the energy range 1-8 keV with a time resolution of 30 seconds;

3) a multiband on-source background-subtracted lightcurve in the energy range 1-3.8 keV, 3.8-8 keV and 1-8 keV. The time resolution for those lightcurves varied from observation to observation and corresponds to the maximum available time resolution for the pulse height analyzed data. Typically this ranges between 1-10 seconds.

4) a background-subtracted on-source spectrum integrated over the entire observing interval. Spectra of burst sources were obtained excluding the burst event.

The number of files expected for a particular pointing observation depends on how the data have been divided (either because different ME array configurations were used throughout the observation, or because there was a change in the OBC mode). Typically, different lightcurve files were created for different array configurations as well as OBC configurations within one observation. Spectral files obtained from different array configurations were corrected for the difference spectra between the halves and then averaged together. The spectra and lightcurves were obtained only using the argon detectors sampled by the primary OBC mode (energy mode). Because of the limited resources in available disk space, the automatic analysis did not create lightcurves with high time resolution data.

The major problem in running an automatic process on the ME data was the selection of the method for the background subtraction. This mainly depended on the configuration used throughout an observation. Typically, background taken simultaneously to the source spectra gave better results, but this was not always possible.

For about 30% of the observations, a number of corrections were required, mostly to fix unsatisfactory background subtraction. The main reasons for this were: a) flaring in the background; b) detectors which appear significantly noisier than others during a particular observation; c) for those observation with no simultaneous background, the standard or the slew background was not adequate. Most of the problems were recovered by running the analysis interactively on each problematic dataset. Quality flags from 1 to 5 were assigned to give an indication of the of the overall quality of the ME products.

The GSPC Products

Although the GSPC was on during most of the mission life time, data products were obtained only for those observations where the source ME count rate was in excess of 5 count/s/half. Sources with lower count rates typically do not have sufficient signal-to-noise to justify a GSPC analysis. The products created with the automatic processing for the GSPC instrument included:

1) background subtracted lightcurves for the source in the energy range between 2-8 keV and 8-15 keV with the original time resolution (8 seconds in most cases);

2) an average background-subtracted spectra for the entire observation of the source (excluding any bursts that occurred).

The background subtraction was done using both the slew spectra (to determinate the gain) and standard spectra (adjusted to the proper gain). Again, the data were quality checked and a flag from 1 to 5 assigned to indicate the overall reliability of the products.

The TGS Products

A total of 24 sources were observed with one or both transmission grating spectrometers, from which 19 gave a useful spectrum. For each observation the following products were created:

1) a background subtracted lightcurve, per filter, centered on the zero order in a box 16 by 16;

2) multiple spectra (see later);

3) background spectra.

The spectra were extracted from the CMA images, typically using a mask of 40 to 50 pixels in width (on the Y-axis) and integrating perpendicular to the dispersion direction (along the X-axis). The spectra generated in this way consisted of a one-dimensional array of 2048 channels. A second set of spectra consisted of the positive and negative orders added together, resulting in a 1024-channel spectra. The background spectra was estimated over the same mask size in the image of a long blank-field exposure and renormalized to the source image. The source spectra obtained with the TGS are not background subtracted. Within an observation, the 2048- and 1024-channel spectra were accumulated for each filter or change in position. In addition, when it was possible, a simultaneous GSPC and/or ME spectra covering the same interval of time was accumulated.

Database Tables

The automatic processing also performed spectral and timing analysis of the products. The analysis results were included as parameters in the databases, although parameters associated with timing and/or spectral fitting should be regarded with more attention, because in same cases, the automatic processing failed.

All the products and analysis results were made available from the EXOSAT X-ray Observatory as databases using the EXOSAT DBMS. Below is the list of the databases created:

* EXO_LOG contains the complete list of all EXOSAT observations, configuration and observing modes, and principal investigators.

* EXO_PUBS contains information about EXOSAT publications in refereed journals.

* CMAIMAGE contains references about all the images created for each observation using both low energy telescopes (CMA1 and CMA2). Associated products: images.

* LE contains the detections found within 6 arcmin from the pointing position obtained with both low energy telescopes (CMA1 and CMA2). Associated products: images, lightcurves.

* CMA contains the complete list of detections over the full field of view obtained with both low energy telescopes (CMA1 and CMA2). Associated products: images, lightcurves.

* GS contains the results obtained from the GSPC only for sources with an ME count rate of at least 5 count/s/half. Associated products: spectra, multiband lightcurves.

* ME contains the results obtained from the ME argon detectors. Associated products: spectra and multiband lightcurves

* TGS contains the results obtained from the TGS instruments. Associated products: multi spectra (L and R or L+R order), lightcurves

Each database contains several parameters specific to the instrument and associated products, and others obtained by cross-correlating the different databases with stellar catalogs.

The Raw Data

Data obtained from the four instruments were originally distributed to observers as Final Observation Tapes, FOT, in a form of one or more 1600 BPI tapes per instrument. A FOT contains all the scientific, housekeeping and calibration data for a single observation and can be used to make a more detailed analysis. There were a total of 8340 1600 BPI tapes , for a volume of about 160 Gbyte of data.

The complete archive of tapes was kept at the ESA European Space Operations Center, ESOC, in Germany. Requests for copies by the community were sent to ESOC which sent back a copy to the observer. This process typically took 4-6 weeks. In 1992 ESIS/ESRIN/ESA (Italy) took the responsibility to copy all the FOTs into a more permanent optical jukebox archive and distribute the tapes to the community. An agreement made between ESA and HEASARC allowed HEASARC to obtain a copy of all the FOT data, in return ESA obtained a copy of the data reformatted to FITS format by the HEASARC. From ESRIN the data arrived at HEASARC via EXABYTE magnetic tape, each of them containing about 1.8 Gbyte of data (usually compressed using the `zoo' compression scheme). The incoming data was loaded into the HEASARC rewriteable jukebox.

HEASARC archive

The HEASARC archive currently contains two different sets of the data files:

1) the data product files in FITS format and

2) the original raw data from the Final Observation Tape, FOT.

The Data Product File Contents

Lightcurves

The FITS EXOSAT lightcurves in the HEASARC archive have two different layouts which reproduce exactly the content of the original products. The ME and GS experiment data are in a bintable containing two columns only: RATE and ERROR. The RATE is the background subtracted count rate within the integration time, corrected for collimator efficiency and dead time. Gaps are padded correctly. The time associated with each bin can be calculated as TIMEZERO + ((n-1) * TIMEDEL), where TIMEZERO and TIMEDEL are header keywords which represent the zero time in mission elapsed time and the integration time, and n is the bin number. The LE and TGS lightcurve data are in a bintable containing four columns TIME, RATE, ERROR, and FRACEXP. The TIME column contains the residual time from an offset stored in the TIMEZERO and the FRACEXP is the fractional exposure for each bin. The RATE is the background subtracted count rate within the integration time, corrected for vignetting and dead time.

Spectra

The FITS EXOSAT spectral files have similar formats for the GS, ME and TGS experiments. The ME file, in addition to the spectral data stored in the first extension, also has a second extension containing information on the specific detector used during the observation. During the re-formatting stage for each ME and GS spectrum, the corresponding response matrices were generated and formatted into FITS format.

Images

The FITS EXOSAT images are stored using the FITS primary array. The images are 2048 x 2048, and are made using the linearized detector coordinates. The projection is `tangent plane' and the images do not have the y-axis direction aligned with celestial north. The FITS header contains two coordinates systems. The first is the so called sky-coordinates system which maps each pixel in the array into a RA and Dec in the sky. This system is a FITS standard and uses the keywords CRPIXn, CDELTn, CROTAn, CRVALn, and CTYPEn where values of n are 1 and 2. The secondary system, which is not a FITS standard convention, describes the detector characteristics. This system uses the keywords DLMINn, DLMAXn (the minimum and maximum values for the X and Y axes), DLABELn (the type of coordinates), DDELTn (the increment), DUNITn (unit for DDELTn), DRPIXn (value of CRPIXn in this secondary frame defined by DLMINn, DLMAXn), and DOPTICn (X and Y coordinates of the optical axis in the secondary frame).

The File Naming Convention for the Data Products

The product file names are composites of 6-character strings, a letter followed by a 5-digit number, generated automatically by the computer, based on the start time of the observation. The first character, which identifies the product type has a different convention for the 3 experiments, but within an experiment different products associated with the same observation share the 5 digit number. The LE images have `a' as the first character, `b' is used for the background lightcurve, and from `c' on, alphabetically, are the lightcurves for the different sources detected in the field of view. The ME 30-second lightcurves have `d' as first character, and the multi-band use `a' (lower energy),'b' (higher energy), and 'c' (total band). The background lightcurves have `r' as a suffix and the spectra use `s.' GS lightcurves have the first character `d' or `c' depending on the energy range and `s' for the spectra.

The Raw Data

The EXOSAT raw data were made available from the HEASARC in May 1994. Data can be immediately retrieved from the anonymous FTP account at heasarc.gsfc.nasa.gov. All the data have been compressed using the `zoo' compression program, as originally provided by ESA (for these data, this algorithm typically gives a factor 2 smaller files than the standard unix compress). The zoo software is available from the legacy anonymous FTP under the directory /software/zoo.

The database EXOFOT contains 6614 entries, each of them corresponding to a FOT. It does not contain the coordinate information. To search for data within this database, it is recommended to first use the EXOSAT observation log database (EXOLOG) to retrieve the experiment, target name, start and stop time. With this information, search in the EXOFOT database using the parameters TIME and/or TIME END with a time window which includes the time value found in the EXOLOG database for the parameters FOT TIME and FOT END. Currently we are merging the EXOLOG and EXOFOT database such that the filename of FOT will be a new parameter associated to an entry in the EXOLOG database. The matching is quite tedious because it is based only on the start and stop time information stored in the log which can be different from the FOT start time and stop time.

The data format is the original binary format, the description of which can be found in the rather large FOT-handbook. This document was distributed by ESA, with the data to the original PI. It does not exist in electronic form. On request, a copy can be obtained by contacting us via our Feedback form. Rather than rewrite software to access these data it is recommended to use the EXOSAT Interactive Analysis, IA, System. This can be obtained from the anonymous ftp site at ESTEC/SSD (astro.estec.esa.nl). The IA system is fully described in the companion article by Parmar et al. in this Legacy issue.

Future Plans

The HEASARC plans to reformat the EXOSAT raw data to FITS format over the coming 2 years. The effort will make use of the IA system to write out the data in FITS format for each of the modes used, at the highest time resolution available. These FITS data will then be moved to the anonymous FTP area. This effort began in July 1995 with the LE data, writing FITS files in the same event format as is used by the imaging instruments on ROSAT, ASCA and Einstein. The ME standard mode data will be reformatted next, followed by the GSPC. The most commonly used modes will be done first, followed by the more complex, rarely used modes. As each stage is completed, regular updates will appear in future issues of Legacy.

Acknowledgments

Every member of the EXOSAT Observatory team gave a substantial contribution to the realization and success of the EXOSAT database system. Particular thanks are due to P. Barr, G. Giommi, P. Giommi, M. Gottwald, F. Haberl, L. Osborne, J. Osborne, A. Parmar, L. Stella, G. Tagliaferri, G. Thorner, and H. van der Woerd. The HEASARC staff are also thanked for their support. In particular, I. George and B. O'Neel for helping in the reformatting effort of the products, and Pat Tyler and Song Yom for creating and maintaining the databases and the database management system.

References

Arnaud K.A., George I.M, and Tennant A, 1992 Legacy 2, 65.

Angelini L., Pence W. and Tennant A. 1993 Legacy 3, 32.

George I.M, Arnaud K.A., Pence W. and Ruamsuwan L. 1992 Legacy 2, 51.

De Korte, P.A.J. et al, 1981, Sp. Sci. Rev., 30, 495.

Parmar, A, E, et al 1995, in this Legacy issue.

Peacock, A., et al. 1981, Sp. Sci. Rev., 30, 525.

Turner M.J.L., Smith, A., and Zimmermann, H.U., 1981, Sp. Sci. Rev., 30, 513.

White N. E. and Peacock A., Memorie della Società Astronomica Italiana, 1988 Vol. 59,7.

Proceed to the next article

Return to the previous article

Select another article