Paper presented at Conference on the High Energy Radiation
Background in Space, November 3-5, 1987, Sanibel Island,
Florida, published by American Institute of Physics, AIP Proceedings 186
eds C. Restor and J. Trombka

D. E. Gruber, G. V. Jung, and J. L. Matteson
Center for Astrophysics and Space Sciences
University of California, San Diego, La Jolla, CA, 92093


  1. Introduction
  2. Instrumentation
  3. Origin and variability of the background
  4. Separation of activities
  5. References


The UCSD/MIT Hard X-Ray and Low-Energy Gamma-Ray Experiment (A4) was carried on the HEAO-1 satellite to perform a survey of the sky at energies between 10 keV and 10 MeV. Observation of radiation from cosmic sources in this energy range has been difficult because of high internal background counting rates from radioactivity induced in the material of the detectors. The activation is caused by ambient charged particles and neutrons, and leads to complex, time-variable energy-loss spectra. We report here on the results of an extensive analysis of the radioactivity induced in the NaI Medium Energy Detectors of this experiment during the 500-day HEAO-1 mission. These 7.5 cm dia by 2.5 cm thick crystals operated from 40 keV to 2 MeV. Various radioactive nuclides were identified through their observed spectra and decay times. The incident charged particle flux was monitored with three NE 102 plastic scintillators. Model activities at a variety of expected half-lives were calculated from this charged particle input. The amplitude of each of these activation terms was treated as a free parameter in standard multiple regression fits at each energy. Since the dominant behavior of each term is exponential with time, the problem is ill-conditioned. Judicious selection of data from the entire mission nevertheless permitted a reliable separation of activities. The resulting energy spectra and nuclide identifications are discussed.


In the observation of cosmic x- and gamma-ray sources at energies between 10 keV and 10 MeV the limiting factor has been the high internal background counting rate from radioactive species induced in the detectors by ambient protons and neutrons at balloon and satellite altitudes. A wide variety of species are produced, each with its own decay modes and half life, thus modeling of the internal background is a formidable task. Although such modeling has been employed for certain measurements, such as the Apollo-17 measurement of the spectrum of the diffuse component, the observation of individual sources is more usually accomplished with chopping techniques using a shutter or by comparing the source region with nearby blank sky. Since the internal background is variable on a variety of time scales, it is necessary to chop on other time scales. Generally this is done by selecting data intervals on which only slow variability is expected, then chopping as rapidly as practical.

The A4 experiment on HEAO-1 was designed at a time when detector activation and the resultant variability were not yet fully appreciated. Enough was understood so that that the dwell time on an individual source in the Medium Energy Detectors was kept short (three minutes) compared to the 45 minute background variability expected from the orbital motion in the geomagnetic field. Moreover, the instrument carried special detectors for monitoring charged particles, and an elaborate telemetry scheme returned very complete information on all detector and active shield counters on time scales of 41 seconds or less. Thus it was possible to make a rather complete analysis of the radiation environment, its variability, and the responses of the various parts of the detector system. This study formed a key role in the formation of a sky map from the A4 detectors: knowledge of the variability permitted the formulation of a scheme of data selection and accumulation which was relatively insensitive to the background changes. Furthermore, the size of residual variability in the sky map could be estimated. We report here on various aspects of this analysis. Much of this work has been reported earlier2.

Figure 1

Figure 1 The A4 instrument on HEAO-1 consisted of seven NaI/CsI phoswich scintillation detectors surrounded by massive CsI shielding. Additional organic scintillators provided for monitoring of charged particles. The instrument was designed for observing cosmic sources of x- and gamma-radiation at energies between 13 keV and 10 MeV.


The UCSD/MIT Hard X-Ray and Low-Energy Gamma Ray Experiment (
Fig. 1) was a massive collection of inorganic scintillator counters with NaI detectors surrounded by anticoincidence CsI shielding in all directions except for the entrance apertures. With thickness below 10 cm in very few directions, this shielding was very effective against background charged particles and photons, but was quite permeable to neutrons, with an optical depth for scattering of the order of unity. Anticoincidence for charged particles was extended to 4 pi by covering the aperture with a thin sheet of NE 102 scintillator. The two low energy detectors, with thickness of only 3 mm, were optimized for operation below an energy of 100 keV. The four medium energy detectors (MED) had thicknesses of 2.5 cm, resulting in photon detection probabilities near unity up to several hundred keV. The single high energy detector had a thickness of 7.5 cm. The ambient charged particle flux was monitored with three 10 mm diameter spherical NE 102 scintillators surrounded by absorbers of different thicknesses to permit crude energy resolution. A complex telemetry scheme returned counting rate averages on several time scales, none longer than 41 seconds. Arrival times and energies of individual photon detections were also returned.

The HEAO-l satellite was launched into a highly circular orbit of inclination 23 and altitude 400 km. At this altitude drag from the tenuous exosphere was sufficient to cause reentry after only 580 days. Total mass of HEAO-1 was three tons.


HEAO-1 and its experiments were subjected to two dominant sources of ionizing radiation: cosmic rays and geomagnetically trapped particles. The cosmic ray flux was modulated on two time scales as the satellite moved through the geomagnetic field: twice per orbit the satellite crosses the geomagnetic equator where the shielding effect of the geomagnetic cutoff is greatest; and once per day the orbital plane lies closest to the geomagnetic equator. Geomagnetically trapped particles were encountered only at one point in the orbit, the so-called South Atlantic Anomaly (SAA), where the inner belts dip unusually low. This region was traversed six to eight times per day. Changes of a few tens of kilometers in the satellite orbit caused large changes in the particle fluxes encountered because of strong gradients of the population of trapped particles. This is immediately evident in Figure 2, where periodic changes are seen to modulate an overall strong decline which results from orbital decay. The wobbles about the general decline were resolvable into two periodicities at 28 and 43 day. These are probably identifiable with cyclic changes of two elements of the orbit: the argument of the perigee at 28 days and the right ascension of the ascending node at 48 days. The 28-day motion moves the satellite apogee in and out of the SAA region; more intense fluxes are encountered at apogee. A modulation with this period of 11.5% was measured. The 48-day orbital precession causes a variation of the local time of maximum daily SAA encounter with synodic period 43 days. Since the satellite spin axis was kept fised near the sun, the satellite orientation with respect to the local vertical at SAA also varied with this period. Proton fluxes on the inner edge of the belt are strongly anisotropic with apex to the local west. The particle monitors are not completely isotropic in response, but are partially shielded by the massive instrument and spacecraft. Therefore a modulation of the observed proton flux may be expected. In fact, a highly significant 43-day modulation of amplitude 5.6% was measured, with maximum at times when the SAA was encountered near the dawn terminator. After allowance for these two periodic terms, the monitored daily particle fluences showed no further signs of periodic structure, and evidently the considerable remaining variability was random. Sizable random variability of particle populations has been reported by Stassinopoulos3.

In Figure 2 normalized counting rates for two MED detectors are also shown. These rates are measured for a strong internal background line of I123, which results From spallation interactions of protons with the I127 of the detector. This activity decays with a half-life of 13 hours, making it effectively an instantaneous monitor of proton dosage at the detector on the time scales of interest here. Moreover, the decay is predominantly by internal capture, and the energy of 193 keV is a sum peak of a 159 keV gamma from the Te123 daughter and the The Te x-rays, thus the gammas leaking from the CsI shields, where I123 is also produced, are separately counted. The wobbles on the counting rate curves of the two detectors are obviously slower than for the particle monitor. In fact, Fourier analysis shows that the two same frequencies are present, but the relative strengths are reversed: in both detectors the 28-day modulation has an amplitude near 6% and the 43-day modulation has amplitude 15%.

Figure 2

Figure 2 Daily dosage of trapped protons in the SAA region and counting rate of two detectors in a strong internal background line at 191 keV, arbitrarily normalized. The overall decline is due to orbital decay. The wobbles result from 28-day cyclic changes in the eccentricity, and a 43-day precession of the orbital plane.
It is also obvious that the wobbles in the two MED curves of Figure 2 are anticorrelated. Analysis shows that the Detector 2 maxima occur when the SAA passages are encountered at the sunset terminator, and the Detector 4 maxima at the sunrise terminator. The explanation is simple, involving only the orientation of the spacecraft and anisotropies in the particle fluxes and detector shielding. Each MED has a direction of minimum shielding (plan view, Figure 1) corresponding to a particle threshold of roughly 80 MeV. Because the spacecraft spin axis is fixed near the sun, this direction of minimum shielding is in the solar direction for detector 4 and antisolar for detector 2; At the sunrise terminator the westward peak of particle flux is pointed at the thin spot of detector 4's shielding, and likewise for detector 2 at the sunset terminator.

In Figure 2 it is seen that the MED count rates drop only by a factor of two in the course of the mission, whereas the particle flux drops by an order of magnitude. An extra, relatively constant activation component is therefore present, which is attributable to cosmic rays. It is clear that at the beginning of the mission activation from trapped particles and cosmic rays occurs in roughly equal amounts.

Figure 3

Figure 3 MED background spectra at two different times of day, showing activation from resulting from exposure to SAA trapped protons and its decay. Except below 100 keV the detector background is dominated by this internal radioactivity.
Variability of internal background was, of course, not confined to longer time scales, but was quite pronounced in the course of a day. The SAA region was traversed in six to eight successive satellite passes out of the daily fifteen. Cosmic ray fluxes also varied with time as different geomagnetic latitudes were traversed, but the dynamic range of this variability was probably not more than a factor of two. The net daily variability of MED internal background is indicated in Figure 3. The daily variability of internal background is shown in more detail in Figure 4, which shows the response of the HED detector and one shield segment to trapped particle doses and to cosmic rays, whose flux is parametrized by the McIlwain L-parameter.

Figure 4

Figure 4 A typical daily history of fluxes of activating particle and the responses in selected instrument registers. Fluxes of cosmic rays and equilibrium neutrons are proportional to McIlwain L. The inner shield ULD is sensitive to the incident particles, while the HED events counter shows a delayed response dominated by the decay of I128.


Given the nearly complete data on particle input from the particle monitors, it was decided to attempt a straightforward modeling of the detector response, at least for the species which were recognized as major contributo the background. For selected half lives, then, calculations were made of model functions representing the instantaneous activity at half lives taui resulting from particle dosage D(t):
[Equation 3]
A distinction must be preserved between SAA and cosmic ray dosages, because of differing spectra and average energies. For convenience, the McIlwain L-parameter was used to represent the cosmic-ray dosage; the low counting rate of the 1 cm2 particle monitors introduced unnecessary statistical noise. Only four half-lives were employed: 25 min (I128 ), 2.1 hr (Xe123), 13.3 hr (I123), and 4.2 days (I124) A prompt response ( taui < 25 min) function was also added for the cosmic ray (L) input. A prompt response term was not needed for SAA dosage because the instrument was turned off at these times. The behavior of these functions during a typical day of operation is shown in
Figure 5. Only the prompt and 25 min terms were retained for the cosmic ray input; longer half-life terms tended to average out the 45 min variations of L. Computer limitations necessitated the elimination of one of the remaining seven terms (including constant). This was achieved by joining the 0.4 hr functions for SAA and cosmic-ray input, using a normalization parameter, determined to be of the order of unity, which represents the relative efficiency of activation by the two particle populations. Although selected because of certain dominant activities, this set of half lives rather evenly sampled the range of time scales amenable to analysis; the half lives of 0.4, 2.1, 13 and 100 hrs stand in the ratios 5:6:7.

The analysis then was reduced to a very standard problem of multiple linear regression. For selected data intervals one accumulated the usual weighted sums of observed counting rates, R(ti, E), their products with the activity models, Aj(ti)R(ti, E), and the cross-products Aj(ti) Ak(ti). The matrix equations:
Equation 1
are then solved for Sk(E), the activation spectra at each half life, plus time-average spectrum So(E). The detector background spectrum is then:
Equation 2

As a practical matter, the various sums were accumulated with the spectral data during the preparation of the sky map for each detector. Great circle scans divided into ninety sky bins were accumulated on each four-day interval of the mission. For each 4-day sky strip a single background spectral file was generated by adding all of the blank sky data, using as selection criterion a distance of 20° or more from the galactic plane, Cen A, NGC 4151 and 3C273. Since the problem is linear, 4-day spectral files could be added in any combination before solution.

Solution, however, was not immediate. Two difficulties were present: statistical precision on shorter data stretches was rather limited for solution into six separate spectra; and the five activity functions were far from independent basis functions. Fortunately, data from the entire mission was sufficient to overpower the first difficulty. Overcoming the second difficulty required some careful manipulations. As a glance at Figure 5 will show, the 13 hour and 100 hour functions are very similar when allowance is made for arbitrary offsets and multiplicative factors Sk. The 13 hour and 2 hour functions are also similarly coupled. The degree of similarity can be quantified in a cross-correlation factor obtained from the error matrix following solution. The procedure employed was to find a subset of the data for which the cross-correlation factor between the 13 hour and 100 hour terms was as small as possible. A reduction from -0.98 to -0.70 was eventually obtained, which was taken as an indicator of a sufficiently stable and reliable solution. The best-fit solution was then used to subtract the 100 hour dependence from the data. Thus the most pathological term in the problem was isolated and removed. The procedure was then re-applied using the remaining terms to find the 2 hour activity spectrum and subtract it. The four terms still included were now sufficiently independent so that a good direct solution was obtained. The fitting was performed unconstrained, and a good indicator of the correctness of the fits was given by the absence of large, unphysical, negative excursions of the spectra, and also by their approach to zero at high energies. A final but very important verification comes from the expected spectral shapes. For example, the 25 min spectrum showed a very convincing representation of the I128 beta spectrum with end point at 2.12 MeV.

Figure 5

Figure 5 Model activity functions at several half-lives for SAA dosages received in a typical day of operation. The curves represent populations of radioactive species induced in the detectors by trapped particles in the SAA region. The "McIlwain L" curve represents prompt response to cosmic ray dosages. The longer half-life terms fail to decay fully in a day.
By this method reliable activation spectra, Figure 6 for example, were obtained. The non- independence of the basis functions adds an important benefit: each function collects, in the fit, all the activity at a wide range of related half-lives. If 13 hour and 100 hour activity can be distinguished only with difficulty, then 13 hour and, say, 25 hour activity must appear almost identical. Thus the solution extends to a full range of variability between a few minutes and many days, provided that the initial assumptions of SAA and cosmic-ray activation are correct. It is evident, though, from Figure 2 that the particle monitors gave an incorrect measure, by up to 30%, of the SAA dosage at the detectors, at least for time scales near one month. This difficulty could be overcome, however, by selection of 4-day spectral files only at times of peak activation.

Figure 6

Figure 6 Activation spectrum at 13 hour nominal half life. Line and continuum deexcitation spectra are present from nuclides with half lives between roughly 5 hours and 30 hours. The strong feature at 193 keV results primarily from decay of I123 with a 159 keV gamma and K x-rays following electron capture.
The activation spectra thus obtained were examined for spectral features, usually line features but occasionally beta continua, and identifications of radioactive nuclides attempted. The results are in Table 1. Many observed lines are broader than the instrumental resolution, indicating blending. Most of the identifications are already familiar from earlier work4,5, but some new identifications are proposed. Many of the lines result from electron capture and therefore include the K-level energy daughter. A few lines are attributed to inelastic scattering of neutrons. Some observed lines remain unidentified. This work was supported by NASA grant NAG-8-499.


  1. Trombka, J. I., Dyer, C. S., Evans, L. G., Bielefeld, M. G., Seltzer, S. M., and Metzger, A. E., Ap. J., 212, 925 (1977).
  2. Jung, G. V., unpublished dissertation, UCSD (1986).
  3. Stassinopoulos, E. G., these Proceedings (1988).
  4. Dyer, C. S., and Morphill, G. E., Astrophys. Space Sci., 14, 243, (1971).
  5. Fishman, G. J., unpublished NASA report NASA CR-150237 (1977).

HEAO A4 WWW page (at CASS, UCSD)
High Energy Astrophysics at CASS
HEAO 1 mission information (at the HEASARC, GSFC)