COMPTEL - An Orbiting Compton
Telescope for Gamma-Ray
Astrophysics
Mark McConnell
and
James Ryan
Space Science Center
University of New Hampshire
Durham, NH
COMPTEL Collaboration:
Max Planck Institute (MPE)
University of New Hampshire
SRON - Utrecht
ESA - Space Science Dept.
Launch and Deployment
Space Shuttle Atlantis - April 5, 1991
COMPTEL Description
For each event which scatters from D1
to D2 without triggering the veto domes,
we record:
-
event location (x,y) in D1
energy deposit in D1
pulse-shape in D1
time-of-flight (ToF) from D1 to D2
event location (x,y) in D2
energy deposit in D2
absolute time (1/8 msec)
Event locations provide a direction
vector for the scattered photon.
Energy deposits provide an estimate of
the photon scatter angle.
COMPTEL D1 Detectors
Each of the 7 D1 detectors consists of a cylindrical volume of NE213A liquid
scintillator, 28 cm in diameter by 8.5 cm deep. The interaction location of
each event can be determine to about 1-2 cm.
Individual D1 Cell
Complete Assembly of 7 D1 Cells
COMPTEL D2 Detectors
Each of the 14 D2 cells consists of a cyclindrical volume of NaI scintillator,
28 cm in diameter and 7.5 cm deep. Event interactions can be located with
a spatial resolution of ~1-2 cm.
Individual D2 Cell
Nearly Complete Assembly of 14 D2 Cells
COMPTEL Simulations
Photon simulations employ CERN
GEANT Monte Carlo package.
Used to better define the response
of COMPTEL (via PSF generation).
The model used in the simulations
is shown here.
COMPTEL Mass Model
D1
Assembly
D2
Assembly
Complete Mass Model
COMPTEL Effective Area
The effective area of COMPTEL (as determined by Monte Carlo
simulations) for several different point-source angles. Nominal
in-flight data selections reduce these values by ~20% – 40%.
30
θ = 0°
2
Effective Area (cm )
40
10°
20°
20
30°
40°
10
50°
0
1
10
Photon Energy (MeV)
Relative Efficiency
COMPTEL Energy Resolution
1
10
Total Energy Deposit, Etot (MeV)
Simulated energy-loss spectra for
various incident photon energies
Photopeak resolution as a function
of energy for normal incidence photons.
COMPTEL Angular Resolution
ARM distributions for various energies
ARM ≡ ϕ − ϕ geo
-10
-5
0
5
10
Angular Resolution Measure, ARM (deg)
Angular Resolution
vs. energy
Event Circle Imaging
Simplest approach to generating an image is simply to overlay
the event circles from each individual event.
Here are shown the summation of ~100 event circles from the
gamma-ray burst of 3-May-1991.
Works well only for those cases (such as this one) in which
the signal-to-noise is high.
The COMPTEL 3-d Dataspace
For a specified energy range, we can define
a 3-dimensional distribution of events.
ϕ
—
The 3-dimensional dataspace is defined by:
1) the photon scatter direction (χ,ψ)
2) the photon scatter angle (ϕ)
χ
ψ
( χ ′, ψ ′)
Idealized PSF
in the 3-d Dataspace
The scatter direction comes from event
locations in D1 and D2.
The scatter angle comes from energy
deposits in D1 and D2:

cos ϕ = 1− mec 2  1 −
ED1 ( E


1

+ ED2 ) 
D1

The Point-Spread-Function (PSF)
Simulated PSF for a mono-energetic (2.2 MeV) point source.
3-d view (of 1/4 of full PSF), with
point source located at the origin.
Individual slices of the 3-d cone.
Imaging Reconstruction
Techniques
In all cases, image reconstruction involves some process which correlates
the PSF with the measured 3-d dataspace. Must be done once per (total)
energy interval.
Two methods are normally employed:
1) maximum entropy – used primarily to define the gross structure.
Seeks the flatest image which is consistent with the measured data.
2) maximum likelihood – used to perform a more quantitative analysis.
For each point within the FoV, it determines the likelihood of a model
containing a point source plus background versus a model containing
only background.
Both methods require some independent estimate of the distribution of
background events within the 3-d dataspace.
Background Modeling
Several modeling methods have been developed. These include:
1) High-Latitude Background – Uses observations at high galactic latitudes
(where there are few sources) to estimate a background for low galactic
latitudes (where there are many point sources plus a diffuse component).
2) Dataspace Smoothing (SRCLIX) – Performs a smoothing within the
measured 3-d dataspace which removes the high-frequency (source)
components and maintains the low-frequency (background) components.
3) Synthesis Using Adjacent Energy Bands (BGDLNE) – Uses independent
estimate of the (χ,ψ) distribution from adjacent energy bands in
conjunction with the φ distribution from the energy band of interest.
Sensitive to line emission only (e.g., 26Al at 1.8 MeV).
4) Physical Background Modeling – Seeks to determine the source of all
background components. Uses Monte Carlo modeling to generate
background estimate. (Still under development.)
SRCLIX Background Model
A 3-d smoothing process
defines the estimated
background level for each
point in the dataspace.
Since the smoothing is
also applied to the source
signal, this effect must
also be accounted for via
a modified PSF.
Sources of Background
Event Types A, B:
Internally-generated Photons
Ex- thermal neutron capture (2.2 MeV),
40K (1.46 MeV)
Event Type C:
Two photons which are spatially and
temporally correlated.
“cascade events”
Ex - 27Al (p,3pn) 24Na
Event Type D:
Two photons which are spatially and
temporally uncorrelated.
“random coincidences”
Event Type E:
Two photons which are spatially
uncorrelated, but temporally correlated.
Contributions to ToF Spectrum
Time-of-Flight [nsec]
2
4
6
8
10
5000
Event Types
All γs
Down-Scatter γs
Local-Multiple γs
Accidental γs
Counts per Channel
4000
3000
Typical ToF spectrum
showing the various
components.
Can be characterized as
a Gaussian superimposed
on a smooth continuum.
The signal is contained
in the Gaussian component
(the down-scattered
component).
2000
1000
0
110
120
130
Time-of-Flight [channels]
140
Background events are
contained in the contunuum
conponent.
Observed Background Features
Etot spectrum showing 1.46 MeV
line (40K) and 2.22 MeV line
(thermal neutron capture in D1).
Etot spectrum showing 4.4 MeV
line resulting(?) from excitation of
12C (in D1) by energetic neutrons.
E2 spectra showing the 1.37 MeV
and 2.75 MeV lines from decay
of 24Na.
Background Rate vs. Particle Flux
Down-Scatter-Event Rate [1/s]x10
3
Event-Rate versus the Veto-Rate
Particle flux can be estimated using:
1) observed veto rates
2) magnetic rigidity
40
30
20
In either case, extrapolation to zero
particle flux removes prompt background
component.
10
0
500
1000
1500
2000
2500
Veto-Rate [1/2.048 s]
Both techniques yield consistent results.
Down-Scatter-Event Rate [1/s]
Event-Rate versus the Cutoff-Rigidity
This approach has been successfully
employed in the study of the cosmic
diffuse background.
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0.00
4
6
8
10
12
14
16
Geomagnetic Vertical Cutoff-Rigidity [GV]
Remaining counts are due to long-lived
background and true source events.
Long-Lived Background
Long-lived radioactive isotopes contribute to the background below ~4 MeV.
These species can be identified by their event signatures.
Simulations are then used to generate the full instrument response to each
identified isotope.
Activity of each isotope is determined directly from the data by fitting the
observed photopeaks.
The dominant long-lived
component is 24Na,
whose signature is
shown here.
Cosmic Diffuse Emission
Difficult measurement to make - requires an absolute flux measurement.
The Cosmic Diffuse Gamma-Ray Spectrum
100
APOLLO
2
E .Flux (keV cm
-2
s
-1
-1
sr )
Latest COMPTEL results yield results which are inconsistent with earlier data.
These results suggest that earlier data suffered from incomplete
treatment of the background.
HEAO-1
10
1
COMPTEL
EGRET
0.1
0
10
1
10
2
10
3
10
4
10
Energy (keV)
5
10
10
6
7
10
Neutron Imaging
Image of Solar Neutrons
15-June-1991 Solar Flare
Neutrons produce a signature similar
to that of photons.
In this case the scattered neutron
energy (Es) comes directly from the
ToF measurement and the scatter
angle must be calculated assuming
elastic neutron scattering:
tan 2 ϕ =
E
D1
Es
Similar techniques can therefore be
employed to ‘image’ neutrons, such
as those produced by solar flares.
Full-Sky Map : 1.8 MeV
This map represents the emission from the decay of radioactive 26Al
distributed throughout the galaxy.
This maximum entropy map was generated using a background
synthesized from adjacent energy bands (BGDLNE methd).
Full-Sky Map : 3 – 10 MeV
Maximum entropy map for the full sky in the 3-10 MeV energy band.
Background generated from smoothing of the observed dataspace.
The data used to generate this image comes from phases 1-3 of the
CGRO mission (April, 1991 – September, 1994).
Gamma-Ray Bursts
COMPTEL can image GRBs
which occur within the FoV.
Burst locations determined
with an accuracy of ±1-2°
(depends on burst intensity).
The Rapid Response Network
distrubutes burst locations
(as derived from COMPTEL
images) within ~15 minutes.
The Galactic Anticenter Region
A composite of the galactic
anticenter region showing
COMPTEL imaging results
for several different types
of sources.
Summary
• COMPTEL represents the first in-orbit Compton telescope.
• First all-sky survey in the 1-30 MeV energy range.
• Event spatial resolutions (1-2 cm) determined by relative PMT pulsehights.
• Energy resolution limited by scintillator technology.
• No tracking of secondary electron.
• Analysis generally performed in a 3-d dataspace.
• Various techniques employed for generating images.
• Background modelling crucial to successful imaging.
• Several background modeling procedures have been developed.
• Work on a physical background modeling scheme is progressing.