First results about on-ground calibration of the Silicon Tracker for the AGILE satellite
ABSTRACT The AGILE scientific instrument has been calibrated with a tagged
$\gamma$-ray beam at the Beam Test Facility (BTF) of the INFN Laboratori
Nazionali di Frascati (LNF). The goal of the calibration was the measure of the
Point Spread Function (PSF) as a function of the photon energy and incident
angle and the validation of the Monte Carlo (MC) simulation of the silicon
tracker operation. The calibration setup is described and some preliminary
results are presented.
arXiv:1112.2600v1 [astro-ph.IM] 12 Dec 2011
First results about on-ground calibration of the Silicon Tracker for the AGILE satellite
P.W. Cattaneo∗,g, A. Argana, F. Boffellig, A. Bulgarellie, B. Buonomou, A.W. Chenc,d, F. D’Ammandoa,b, T. Froyslandb,d, F.
Fuschinoe, M. Gallih, F. Gianottie, A. Giulianic, F. Longof, M. Marisaldie, G. Mazzitelliu, A. Pellizzonir, M. Prestl, G. Pucellaa, L.
Quintieriu, A. Rappoldig, M. Tavania,b, M. Trifoglioe, A. Troisa, P. Valenteu, E. Vallazzaf, S. Vercellones, A. Zambrac, G.
Barbiellinif, P. Caraveoc, V. Coccoa, E. Costaa, G. De Parisa, E. Del Montea, G. Di Coccoe, I. Donnarummaa, Y. Evangelistaa, M.
Ferocia, A. Ferrarid,q, M. Fiorinic, C. Labantie, I. Lapshova, F. Lazzarottoa, P. Liparii, M. Mastropietroj, S. Mereghettic, E.
Morellie, E. Morettif, A. Morsellik, L. Pacciania, F. Perottic, G. Pianoa,b,k, P. Picozzab,k, M. Pilial, G. Porrovecchioa, M.
Rapisardam, A. Rubinia, S. Sabatinia,b, P. Soffittaa, E. Strianib,k, V. Vittorinia,b, D. Zanelloi, S. Colafrancescon, P. Giommin, C.
Pittorin, P. Santolamazzan, F. Verrecchian,1, L. Salottio
aINAF/IASF-Roma, I-00133 Roma, Italy
bDip. di Fisica, Univ. Tor Vergata, I-00133 Roma,Italy
cINAF/IASF-Milano, I-20133 Milano, Italy
dCIFS-Torino, I-10133 Torino, Italy
eINAF/IASF-Bologna, I-40129 Bologna, Italy
fINFN Trieste, I-34127 Trieste, Italy
gINFN-Pavia, I-27100 Pavia, Italy
hENEA-Bologna, I-40129 Bologna, Italy
iINFN-Roma La Sapienza, I-00185 Roma, Italy
jCNR-IMIP, Roma, Italy
kINFN Roma Tor Vergata, I-00133 Roma, Italy
lDip. di Fisica, Univ. Dell’Insubria, I-22100 Como, Italy
mENEA Frascati, I-00044 Frascati (Roma), Italy
nASI Science Data Center, I-00044 Frascati(Roma), Italy
oAgenzia Spaziale Italiana, I-00198 Roma, Italy
pOsservatorio Astronomico di Trieste, Trieste, Italy
qDip. Fisica, Universit´ a di Torino, Turin, Italy
rINAF-Osservatorio Astronomico di Cagliari, localita’ Poggio dei Pini, strada 54, I-09012 Capoterra, Italy
sINAF-IASF Palermo, Via Ugo La Malfa 153, I-90146 Palermo, Italy
tDip. Fisica Univ. di Trieste, I-34127 Trieste, Italy
uINFN Lab. Naz. di Frascati, I-00044 Frascati(Roma), Italy
The AGILE scientific instrument has been calibrated with a tagged γ-ray beam at the Beam Test Facility (BTF) of the INFN
Laboratori Nazionali di Frascati (LNF). The goal of the calibration was the measure of the Point Spread Function (PSF) as a
function of the photon energy and incident angle and the validation of the Monte Carlo (MC) simulation of the silicon tracker
operation. The calibration setup is described and some preliminary results are presented.
Key words: artificial satellites – gamma rays: observations – instrumentation: detectors – telescopes
1. The AGILE mission
AGILE (Astro-rivelatore Gamma a Immagini LEggero) is
a Small Scientific Mission of the Italian Space Agency (ASI)
launched on April 2007 and dedicated to high-energy astro-
physics . The AGILE satellite is designed to detect and im-
age photonsin the 18- 60keV, 30MeV - 50GeV and350keV -
100 MeV energybands with excellent spatial resolution, timing
capability, and large field of view.
AGILE is the most compact (≈ 0.25m3), light (120 kg for
the instrument, 350 kg for the whole satellite) and low power
(≈ 60W) scientific instrument ever developed for high-energy
Email address: Paolo.Cattaneo@pv.infn.it (P.W. Cattaneo )
detectors with independent detection capability. The Gamma-
Ray Imaging Detector (GRID) consists of a Si-W converter-
tracker  sensitive in the γ-ray energy range 30 MeV - 50
GeV, a shallow (1.5X0on-axis) CsI Calorimeter  and a seg-
mented AntiCoincidence system based on plastic scintillators
In additionto the GRID, a coded-maskhardX-ray imagingsys-
tem (SuperAGILE), made of a Si detector plane and a W mask,
ensures coverage in the range 18 − 60keV .
The AGILE main feature is the combination of two co-aligned
imaging detectors (SuperAGILE and GRID) sensitive in the
hard X-ray and in the γ-ray ranges with large field of view
(≈ 1.0sr and ≈ 2.5sr respectively).
Preprint submitted to Nuclear Instruments and Methods in Physics Research A December 13, 2011
Moreover the CsI MiniCalorimeter (MCAL) can operate in
stand alone ”burst mode” in the 350keV − 100MeV range to
On ground and subsequently on flight calibrations of a detector
are essential to the interpretation of its results. The purpose of
the calibration of a scientific instrument is to reproduce, under
controlled condition, the detector response in operation.
This paper describes the on-ground calibration of the silicon
tracker and some results on the instrument performances de-
rived by it.
Figure 1: A schematic view of the AGILE scientific instrument.
2. The Silicon Tracker
The core of the GRID is the Silicon Tracker (ST) that con-
verts the γ-rays and measures the trajectories of the resulting
e+/e−pairs - TheST consists of12trays withdistance be-
tween middle-planes equal to 1.9cm optimized by simulation.
The first 10 trays consist of a W converter layer 245µm thick
orthogonal to each other to provide three dimensional points
(correspondingto a total thickness 0.01(Si)+0.07(W)X0). The
last two trays have no W converter layers since the GRID trig-
ger logic requires at least three contiguous Si planes.
The detector unit is a 9.5 × 9.5cm2tile, 410µm thick with strip
pitch 121mum. Four tiles bonded together form a ’ladder’. Ev-
ery ST plane consists of four ladders.
Only every second strip is readout to limit the power consump-
tion. The non readoutstrips contributeto the resolutionthrough
the principle of capacitive charge division.
Each ladder is read-out by three TAA1 ASICs, each operat-
ing 128 channels at low noise, low power configuration (<
400µW/channel), self-triggering ability and analog readout.
The ST position resolution is below 40µm for a large range of
particle incidence angles .
2.1. The GRID simulation
The GRID as mounted on the spacecraft and as installed in
the test beam is simulated using the GEANT 3.21 package .
This package provides for a detailed simulation of the materi-
als and describes with high precision the passage of particles
through matter including the production of secondary particles.
The simulation output is formatted to be readable by the recon-
struction programs used for the analysis of in-flight data.
2.2. Direction and Energy Reconstruction
The γ-ray direction reconstruction is obtained from the iden-
tification and the analysis of the e+/e−tracks stemming from
the conversion vertex. Each microstrip silicon plane measures
separately the X and Y hit coordinates.
The first step of the event analysis requires to find two tracks
among the possible associations of the hits detected by the ST
The second step consists in fitting the track trajectories through
the hits accounting for the presence of energy loss and multiple
scattering. These steps are performed separately for the X and
Y coordinates producing four tracks, two for each projection.
The three dimensional direction is obtained requiring a correct
association of the two projections of each track.
The track parameters are fitted by a Kalman filter smooth algo-
rithm . A special implementationof the filter  exploits the
measurementof the angularscatteringof the e±due to the inter-
actions with the material to estimate the track energies. Com-
bining the track energies the γ-ray energy is estimated.
3. The γ-ray Calibrations
3.1. Calibration goals
The goal of the calibration is to estimate the instrument re-
sponse function by exposing it to a γ-ray beam with energyand
direction known to an accuracy better than the resolving power
of the instrument.
The required accuracy of ST is driven by its use during the
AGILE mission: the systematic errors introduced by the cali-
bration shouldbe smaller than the statistic errors expectedfrom
a bright celestial source.
The detectorpropertiesto be evaluatedby the calibrationare:
the detection efficiency, the angular resolution, the energy res-
olution. In this paper we concentrate on evaluating the Point
Spread Function (PSF) as a function of the γ-ray energy and
program. This simulation will be required to complement the
calibration data in the untested parts of parameter space. In
particularthe informationabovethe maximumenergyavailable
at BTF can be obtained only through the simulation.
The calibration is designed to cover a wide range of the ge-
ometries and conditions realized in space. The ST was cali-
brated at the INFN LNF in the period 2-20 November 2005,
thanks to the collaboration between the AGILE Team and
3.2. Calibration strategy
To meet the calibration accuracy requirements, we have de-
termined the number of γ-rays required for the calibration of
AGILE, taking into account the photon fluence of a character-
istic γ-rays reference source as Vela.
With a cover-up efficiency of 50% and an effective area of
≈ 500cm2, the number of counts estimated is about 104for E
> 100 MeV, after two months of observation. The requirement
on the number of calibration photons detected by the GRID is
4 × 105for Eγ> 30MeV, 4 × 104for Eγ> 100MeV.
3.3. Calibration set up
3.3.1. The Beam Test Facility
For the ST calibration we used the Beam Test Facility (BTF)
in the Frascati DAΦNE collider complex, which includes a
LINAC at high e−/e+currents, an accumulator of e−/e+and
two accumulation rings at 510 MeV.
ring to be subsequently injected in the principal ring. When the
beam is not transferredin the accumulator,it can be transported
fer line: the BTF line. The BTF provides a collimated beam of
e−/e+in the energy range 20-800 MeV with a pulse rate of 50
Hz. Thepulse durationcan varyfrom1 to 10 ns andthe number
of particles for bunch can range from 1 to 105.
We operated with energy beam of 463 MeV and a pulse dura-
tion of 2 ns.
γ-rays were produced by Bremsstrahlung in a thin Silicon
target; subsequently a magnet bent away the e−while the γ-
rays could impinge on the GRID.
The target is constituted by two pairs of silicon microstrip sin-
gle sided detectors of 8.75 × 8.75cm2and 0.41mm thick, in-
cluding 384 strips with 228µm pitch. The target measures the
passage of the e−and cause the emission of Bremsstrahlung
impact of reduced
BTF e beam
Figure 2: A schematic of the γ-ray line: the target, the bending magnet and the
3.3.3. Tagging system
Our Team developed and installed in the BTF area a Photon
Tagging System (PTS) for the detection of the particles inter-
acting with the target. The e−are tagged using microstrip Si
detectors located on the internal walls of the bending dipole
magnet (see Fig.2). Depending on the energy loss in the tar-
get, the e−impinge on different strips. The correlation of the
measurements of the e−by the target Si planes and by the PTS
tags the photon; the position on the PTS measures the photon
The PTS operates in self-trigger mode, i.e. it is readout inde-
pendently from the GRID. This point has a great relevance for
the following analysis.
During the 18 days of calibration, about 2,105tagged γ-rays
were produced, of which ≈ 40% interacted with the GRID.
3.4. Instrument Ground Support Equipment (GSE)
We developed and installed specific equipment required to
coordinate and, whenever possible, automate the instrument
management and the data gathering and analysis as required
by the calibrationprocedures. The MechanicalGroundSupport
Equipment (MGSE)  hosts the payload and allows the pre-
cise motorized translations and manual rotations of the detec-
tor volume in front of the beam. In near real time, the Science
Console(SC)  archivesall the instrumentdataandperforms
the quick look to check the instrument behaviour. It is also in
charge of producing the energy histogram of the PTS data to
verify the actual statistic of the PTS measurement and decide
the measurement duration.
3.5. Trade-off on the number of e−/bunch
The GRID performance should be evaluated in a ’single-
photon’ regime without simultaneous multi-photon interac-
tions. Multiple photon events are not representative of astro-
physical conditions and may introduce a significant bias in the
The best configuration was with 1 e−/bunch, but considering
the time available for calibration and to obtain a higher effi-
ciency we adopted 3 e−/bunch.
The overall system including the beam terminal section, the
target, the bending magnet, the PTS and the GRID are simu-
lated in detail using GEANT 3.21 package .
That allows a direct comparison between the resolutions mea-
suredin simulatedandreal dataprovidinga checkofthe quality
of the MC simulations.
A significant improvements of the comparison between data
and MC were obtained by overlapping a uniform flux of low
energy γ-rays to the Bremsstrahlung γ-ray. These γ-ray rep-
resents a background that cannot be precisely and is tuned to
match the experimental data.
4. Data Analysis
4.1. Data Samples
The γ-ray beam was directed to the ST at fixed θ and φ with
respect to the detector planes. The beam spot on the detector is
small, ≈ 2 − 3mm as measured in the target. Yet, ideally the
photon beam should illuminate uniformly the ST.
An approximated uniform illumination is obtained by translat-
ing the GRID with respect to the beam on a run by run basis.
The beam impinge in four to eight different positions per orien-
tation, called spills.
Data were collected for different combinations of θ
0◦,30◦,50◦and φ = 0◦,45◦,135◦,225◦,270◦,315◦.
Runs for different spills and same orientation are grouped to-
gether. Also runs for different φ and same θ are grouped to-
gether after having verified that they are compatible.
4.2. The GRID trigger
The GRID trigger for AGILE operation is described in .
The relevant point for this calibration is the following: the
GRID is self-triggering, that is no external signal nor phase
locking with the accelerator is present.
During the calibrationthe in flight triggerconfigurationwas ac-
tive except for the AntiCoincidence veto that was turned off.
This choice was imposed by the high rate of background in-
duced hits in the experimental hall that was reducing the live
time to an unacceptable level. In this configuration the trig-
gered events were contaminated by charged particles crossing
the AntiCoincidence panels.
Mean y 136.3
RMS y RMS y
0 50 100150 200250300350400 450
Mean y 136.3
Figure 3: Relation between ST and PTS energy
4.3. Event reconstruction: the filter
The eventreconstructionprogram,called filter, selects events
with a γ-ray convertingin a e+/e−pair. The events are expected
to have two tracks from a vertex within the ST. The kinematic
of the event is reconstructed applying a Kalman filter. The en-
ergy of the tracks are estimated by the multiple scattering in the
In addition the filter returns a flag assigning an estimation for
the event being a γ-ray or background.
164 / 89
0.024 1.069 1.069
164 / 89
Figure 4: Eγmeasured from the GRID fitted with p0+ p1E−p2for θ = 0◦and
φ = 0◦
There are four flags tagged as G (gamma), L (limbo), P (par-
ticle), S (single). The events flagged as G satisfies very strict
requirements to be a converted γ-ray.
Those flagged as limbo are possible but not certain γ-ray.
Those flagged as particle are estimated to be particle crossing
the ST (e.g. cosmic muons).
Thoseflaggedas singleare estimated to besingle particles from
a vertex within the ST.
The Point Spread Function (PSF) of the ST ideally should be
studiedwitha sampleofG eventshavingaminimalbackground
contamination. However, when the PSF is studied for each ori-
entation and versus the energy, sufficient statistic is required.
The unselected triggered events at θ = 0◦are more than 3106
events. The flag fractions are: G(2.3%), L(48.3%) P(38.6%),
S(10.8%). The striking feature is the low fraction of G events.
This is a feature of the high backgroundenvironmentpresent in
the BTF. Inthis regardthein flightenvironmentis muchcleaner
and the fraction of G events is much higher.
4.4. Event selection for data: the PTS
The PTS can be used for two different but related pur-
poses: 1) as an off-line trigger to identify the emission of
Bremsstrahlung γ-ray in the target and 2) as a device to mea-
sure the γ-ray energy regardless from the ST.
The PTS and GRID events are paired off-line exploiting the
event times measured in both devices up to 1µs precision.
The PTS is requiredto have a very clean signal to reduce multi-
γ-ray events and various background sources. That implies a
low efficiency as off-line trigger. In the configuration θ = 0◦
the tagged events are only 23596. The fractions of events in the
four flags are: G(5.2%), L(44.5%) P(30.3%), S(20.0%). There
is a significant increase in the fraction of G events that never-
theless remain a small fraction. The same pattern is present for
the other orientations.
The other task of the PTS is the measurement of Eγ. This is ob-
tained calibrating with the MC the relation between Eγand the
position of interaction of the e−on the PTS. A close relation
between Eγmeasured by the PTS and by the ST is expected.
Fig.3 shows the profile plot of ST energy versus PTS energy.
The correlationis significant, but the spread is large and the lin-
earity is poor. The GRID energy resolution cannot be evaluated
precisely with this method.
The quality of the GRID energy measurement can be estimated
from Fig.4 where the Eγspectrummeasured by the GRID is fit-
ted with the function p0+
spectrum. The limited distortion indicates that the GRID en-
ergy resolution and the energy dependence of the efficiency do
not alter significantly the Bremsstrahlung spectrum.
Ep2as expected for a Bremsstrahlung
4.5. Event selection for data: the phase approach
A drawback of the PTS approach is the low efficiency ≈ 1%.
If high quality reconstruction (flag G) is required, the number
of events available for the PSF determination in any given con-
figuration may become very small.
An alternativeapproachconsists in exploitingthe BTF bunched
periodicity at 50Hz. The intra spill period is TBTF= 20ms.
That implies that subsequent beam related γ-ray are spaced in
time of multiples of TBTF, in phase with the beam period. The
event time on the GRID is measured with a resolution of 1µs
that defines the precision of the selection.
In Fig.5 the distribution of time differences between consecu-
tive events is displayed, showing high peaks in correspondence
Fig.6 presents the event phase versus the event time. Events
are in phase if the time difference between consecutive events
is an integer multiple of TBTFwithin 100µs. These events are
marked in lighter color.
Another prominent feature of Fig.6 can be interpreted as fol-
lows: there are time intervals (approximately0-500s and 1800-
2700 s) showing no accumulation of events in phase. That is
a sign of beam off time when the GRID measures only beam
unrelated background. Outside these intervals there is an accu-
mulation of events in phase with decreasing numbers when the
number of TBTFincreases. Restricting to the beam on intervals,
the fraction of events in phase is ≈ 25%.
RMS 3.252e+04RMS 3.252e+04
Figure 5: Time difference between consecutive events in a run
Events in phase can be selected regardless the presence of
tagging to enhance the available statistic and the PSF can be
estimated by these samples.
If this approachis correctall taggedevents are expectedto be in
phase. That is the case confirming the validity of the approach.
Event Time (us)
Time Diff (us)
Figure 6: Time difference between consecutive events versus time for a run.
Events in phase are marked in lighter color
Figure 7: PSF (68%) versus energy for MC.
Figure 8: PSF (68%) versus energy for data.
Figure 9: Gaussian PSF versus energy for MC.
Figure 10: Gaussian PSF versus energy for data.
5. Comparison of PSF for MC and real data in phase
The PSF was evaluated in two different ways:
• a Gaussian fit plus a polynomial background, identifying
the PSF with the Gaussian σ
• the PSF is identified with the angular spread including
68% of the events
The result of the 68% estimation for the PSF for various θ ver-
sus Eγis shown in Fig.7 for MC and in Fig.8 for real data.
The result of the Gaussian estimation for the PSF for the same
configurations is shown in Fig.9 for MC and in Fig.10 for real
The 68% PSF is significantly larger than the Gaussian PSF as
expected in presence of background.
The data and MC Gaussian PSF are compatible with each oth-
ers within the statistical errors.
For the 68% PSF, the data show somehow larger values espe-
cially at low Eγ. That is likely due to the low energy γ-ray
the Gaussian PSF should reflect more directly the quality of the
GRID simulation, rather than the beam simulation.
An interpretationof these results is that the compatibility of the
Gaussian PSF for data and MC represents a validation of the
GRID simulation within the experimental requirements.
This paper presents some preliminary results of the calibra-
tion of the AGILE ST at the BTF of the LNF in 2005.
The setup is described in detail as well as the calibration re-
quirements. We discussed the problems encounteredin exploit-
ing the PTS originally designed and a novel approach devised
to circumvent those problems: the phase analysis.
We concentrated on the measurements of the PSF presenting
two possible definitions: the Gaussian and the 68% PSF.
The calibration results are compared with the MC simulations
for a broad set of variables, showing good consistency with
some poorer agreement for 68% PSF mainly at low energies.
These results give confidence on the use of the MC simulation
in the untested part of the γ-ray parameters (e.g. higher Eγ)
especially in flight conditions, i.e. without low energy back-
ground, and in the measurement of detector parameters, like
absolute efficiency and energy resolution, that are difficult to
measure without exploiting the PTS information.
We want to remember the memory of our coworker Dr. Ful-
vio Mauri who greatly contributed to all aspects of the calibra-
tion of AGILE and left us prematurely.
 M. Tavani et al. Nucl. Instr. and Meth. A, 588:52, 2008.
 M. Prest et al. Nucl. Instr. and Meth. A, 501:280, 2003.
 C. Labanti et al. Nucl. Instr. and Meth. A, 598:470–479, 2009.
 F. Perotti et al. Nucl. Instr. and Meth. A, 556:228, 2006.
 M. Feroci et al. Nucl. Instr. and Meth. A, 581:728, 2007.
 G. Barbiellini et al. Nucl. Instr. and Meth. A, 490:146, 2002.
 R. Brun et al. Geant detector description and simulation tool, 1993.
 R. Fr¨ uwirth. Nucl. Instr. and Meth. A, 262:444, 1987.
 A. Giuliani et al. Nucl. Instr. and Meth. A, 568:692–699, 2006.
 F. Gianotti et al. In Martin J. L. Turner and Kathryn A. Flanagan, editors,
Space Telescopes and Instrumentation 2008: Ultraviolet to Gamma Ray,
volume 7011, page 70113D. SPIE, 2008.
 A. Bulgarelli et al. In Martin J. L. Turner and Kathryn A. Flanagan, edi-
tors, Space Telescopes and Instrumentation 2008: Ultraviolet to Gamma
Ray, volume 7011, page 70113C. SPIE, 2008.