Measurement Science and Technology
Meas. Sci. Technol. 33 (2022) 115105 (12pp) https://doi.org/10.1088/1361-6501/ac8221
High-resolution MCP-TimePix3
imaging/timing detector for antimatter
physics
L Glöggler1,2,∗, R Caravita3,∗, M Auzins4, B Bergmann5, R S Brusa3,6, P Burian5,
A Camper7, F Castelli8, P Cheinet9, R Ciuryło10, D Comparat9, G Consolati11,12, M Doser1,
H Gjersdal7, Ł Graczykowski13, F Guatieri6, S Haider1, S Huck1,14, M Janik13,
G Kasprowicz13, G Khatri1,7, Ł Kłosowski10, G Kornakov13, C Malbrunot1, S Mariazzi3,6,
L Nowak1, D Nowicka13, E Oswald1, L Penasa3,6, M Piwi´
nski10, S Pospisil5,
L Povolo3,6, F Prelz12, S A Rangwala15, B Rienäcker1, O M Røhne7, H Sandaker7,
T Sowinski16, I Stekl5, D Tefelski13, M Volponi1,6, T Wolz1, C Zimmer1,7,11,17, M Zawada10
and N Zurlo18,19
1Physics Department, CERN, Geneva 23, 1211 Geneva, Switzerland
2Department of Physics, Technical University Berlin, Hardenbergstr. 36, 10623 Berlin, Germany
3TIFPA/INFN Trento, via Sommarive 14, 38123 Povo, Trento, Italy
4University of Latvia, Department of Physics, Raina Boulevard 19, LV-1586 Riga, Latvia
5Institute of Experimental and Applied Physics, Czech Technical University in Prague, Husova 240/5,
11000 Prague 1, Czech Republic
6Department of Physics, University of Trento, via Sommarive 14, 38123 Povo, Trento, Italy
7Department of Physics, University of Oslo, Sem Sælandsvei 24, 0371 Oslo, Norway
8Department of Physics, University of Milano, via Celoria 16, 20133 Milano, Italy
9Université Paris-Saclay, CNRS, Laboratoire Aimé Cotton, 91405 Orsay, France
10 Institute of Physics, Faculty of Physics, Astronomy and Informatics, Nicolaus Copernicus University in
Toru´
n, Grudziadzka 5, 87-100 Torun, Poland
11 Politecnico of Milano, Piazza Leonardo da Vinci 32, 20133 Milano, Italy
12 INFN Milano, via Celoria 16, 20133 Milano, Italy
13 Warsaw University of Technology, Faculty of Physics ul. Koszykowa 75, 00-662 Warsaw, Poland
14 Department of Physics, University of Hamburg, Jungiusstraße 9, 20355 Hamburg, Germany
15 Raman Research Institute, C. V. Raman Avenue, Sadashivanagar, Bangalore 560080, India
16 Institute of Physics, Polish Academy of Sciences, Aleja Lotnikow 32/46, PL-02668 Warsaw, Poland
17 Department of Physics, Heidelberg University, Im Neuenheimer Feld 226, 69120 Heidelberg, Germany
18 Department of Civil, Environmental, Architectural Engineering and Mathematics, University of
Brescia, via Branze 43, 25123 Brescia, Italy
19 INFN Sezione di Pavia, via Bassi 6, 27100 Pavia, Italy
Received 7 April 2022, revised 8 July 2022
Accepted for publication 18 July 2022
Published 5 August 2022
Abstract
We present a hybrid imaging/timing detector for force sensitive inertial measurements designed
for measurements on positronium, the metastable bound state of an electron and a positron, but
also suitable for applications involving other low intensity, low energy beams of neutral
∗Authors to whom any correspondence should be addressed.
Original Content from this work may be used under the
terms of the Creative Commons Attribution 4.0 licence. Any
further distribution of this work must maintain attribution to the author(s) and
the title of the work, journal citation and DOI.
1361-6501/22/115105+12$33.00 Printed in the UK 1 © 2022 The Author(s). Published by IOP Publishing Ltd
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
(antimatter)-atoms, such as antihydrogen. The performance of the prototype detector was
evaluated with a tunable low energy positron beam, resulting in a spatial resolution of ≈12 mm,
a detection efficiency of up to 40% and a time-resolution in the order of tens of ns.
Keywords: resolution, TimePix3, imagings, timings, detector, positrons, antihydrogen
(Some figures may appear in colour only in the online journal)
1. Introduction
Performing the first inertial studies with low energy neutral
antimatter atoms [1] is the main goal of the AEgIS collabor-
ation, located at the CERN Antiproton Decelerator. Of par-
ticular interest is determining the gravitational acceleration
experienced by anti-atoms in Earth’s gravitational field from
their non-relativistic motion inside a so-called Moiré deflec-
tometer [2]. The measurement concept consists in working
out the average acceleration experienced by the atoms from
their non-relativistic equation of motion in the deflectometer,
δy=aδt2, where δyis the acquired vertical displacement after
the free-fall and δtis their time of flight throughout the device.
One method is to adopt a pulsed beam of anti-atoms with a
time spread in the order of tens of ns and detect δyand δton
the same sample. If two independent detectors are employed
for the two quantities, no correlated event identification is pos-
sible between individual δyand δtmeasurements. The accel-
eration can only be determined by statistically averaging the
collected δyand δtsamples and solving their average equation
of motion. Contrarily, if an event correspondence between δy
and δtmeasurements can be reliably obtained, the possibility
of working out the acceleration for each event is attained. This
opens the possibility to measure the statistical distribution of
acceleration values and study its correlations with other exper-
imental parameters.
Pulsed inertial sensing experiments employing neutral anti-
atoms are now conceivable due to recent successes in form-
ing both antihydrogen (H, the bound state of a positron and
an antiproton) and positronium (Ps, the bound state of a
positron (e+) and an electron) with time-controlled processes.
In the former case, a pulsed source of H atoms was recently
demonstrated, producing anti-atoms in a 250 ns-wide time
window by a charge exchange reaction with Rydberg-excited
positronium atoms, having an axial root mean square velo-
city of ≈2000 m s−1[3]. In the latter case, pulsed positronium
sources are routinely obtained by implanting pulsed e+
bunches in suitable matter converters [4]. Typical forma-
tion velocities of positronium (0.7–1.0×10−5m s−1) [5] are
much higher than the ones of H due to the small weight
of positronium. However, positronium sources feature much
shorter production time windows (8–20 ns) [6] thanks to
the higher degree of control over positron bunches reached
by employing modern plasma-based positron trapping tech-
niques [7]. In both cases, average pulse durations are much
shorter than the atoms’ average time of flights along a typ-
ical 30–100 cm baseline necessary to host a Moiré deflecto-
meter, allowing the conceptual design of the first experiments
adopting pulsed sources and this kind of inertial-sensing
devices [8,9].
An effective deflectometer/detector assembly design in
terms of efficient usage of the low flux atomic beams from
these sources consists of a two-grating deflectometer comple-
mented by a combined position- and time sensitive sensor.
Its spatial resolution needs to be high enough to resolve
the deflectometer fringe pattern, while the timing resolution
should allow for an accurate determination of the time of flight
of each atom. In the proposed experimental designs aiming
at measuring the gravitational acceleration of H, for instance,
δy≪10 mm and δt≪1µs [8]. At present, spatial resolutions
of ≈1 mm to antiproton annihilations can be reached only by
nuclear emulsion detectors [10], which however offer no tim-
ing resolution and real-time diagnostics.
One possibility to realize a combined position- and timing-
sensitive detector is to exploit annihilation processes occur-
ring when anti-atoms impact material obstacles. High energy
annihilation products (charged/neutral pions and nuclear frag-
ments for H, gamma rays for Ps) can be detected with pixelated
high energy detection systems allowing annihilation vertex
reconstruction while providing accurate time resolution. This
approach has proven successful to provide combined position-
and time detection of H atoms, due to the high average num-
ber of annihilation products (on average 2.8 charged pions
are emitted per annihilation event) and their high detection
efficiency with a modest material budget. Several collabor-
ations have developed hybrid vertexing/timing detectors for
antihydrogen. Examples are ATHENA, ALPHA and AEgIS,
whose detectors reached spatial resolutions in the order of a
few mm [11–13]. Pursuing this approach is more difficult in
the case of positronium, where either two 511 keV (para-Ps) or
three 0–511 keV (ortho-Ps) gamma rays are emitted per anni-
hilation. On the one hand, efficiently detecting high energy
gamma rays typically requires scintillation detectors in the
range of several cm to provide sufficient stopping power,
which limits its spatial resolution. On the other hand, the lim-
ited number of emitted gamma rays imposes stringent effi-
ciency limits due to the necessity of building coincidences to
reconstruct vertices. One notable detector example is J-PET
[14], which reached a 5 mm resolution and a vertexing effi-
ciency of 0.1% to ortho-Ps annihilations, having however a
diameter of ≈1 m.
An alternative approach consists of field- or photoionizing
the atoms. Low energy charged antiparticles such as positrons
and antiprotons can be detected by microchannel plates (MCP)
with phosphor screens [15] and a field ionization detector for
the study of Rydberg positronium has been developed [16].
2
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
This approach is particularly convenient when implemen-
ted within the volume of electromagnetic traps, as those used
in antimatter experiments.
The magnetic field radially confines charged particles by
cyclotron motion, allowing macroscopic separation between
dissociation and detection planes. This approach was pion-
eered recently by the AEgIS collaboration, which first demon-
strated a hybrid imaging and timing photoionization MCP-
phosphor detector operated in a 1 T magnetic field [17]. The
detector featured a dual acquisition chain: a CMOS camera
readout to acquire the time-integrated image of the phosphor
screen, and a fast electronic readout of the space-integrated
electric signal of the phosphor screen. It reached 88 µm spa-
tial and 15 ns timing resolution in the imaging plane [18]. This
proof-of-concept detector, however, also exhibited some lim-
itations hampering its adoption in a first inertial sensing exper-
iment. Primarily, the incapability of distinguishing individual
hits in the electric signal, due to the limited bandwidth and
signal-to-noise ratio of the collective readout and the intrinsic
spatial resolution limit imposed by the imaging system. Fur-
thermore, even if a better design of the readout chain would
improve the first issue, the independence of the two acquis-
ition chains would necessitate to operate the detector in the
single particle detection regime. A way to surpass these limit-
ations consists in replacing the two separate and independent
acquisition chains by a more sophisticated detection system
to simultaneously obtain the position and the timing of each
event at once.
In both antihydrogen and positronium, the embedding of
an ionization stage in the present detector layout has a central
role to enable experimentation with neutral anti-atoms. Pos-
sibilities are either an intense laser beam in front of the first
MCP layer or a metallic grid structure with a strong electric
field, or a combination of the two. Positrons released by ion-
ization would have to be guided towards the detection module
by either a strong magnetic field perpendicular to the detector
surface (as in [17]) or by a guiding electric field (as in [16]).
In this article, we present a novel hybrid position and
timing detector featuring simultaneous 12 µm imaging and
15 ns timing resolution, an up to 41% detection efficiency of
low energy antimatter-matter systems’ ionization products and
individual event identification. This detector is based on the
combination of a MCP and a TimePix3 pixel detector used
as readout anode, previously demonstrated in [19–21]. The
capabilities and advantages of combining MCP detectors with
Medipix/TimePix readout are summarized in [22]. A testing
campaign was conducted with a tunable low energy positron
beam to study its capabilities in detecting positrons in the
energy (velocity) range compatible with both, antihydrogen
and positronium photo- and field-dissociation events.
2. Detection module
The detection module studied here is schematically depicted in
figure 1. It consists of a MCP stack in chevron configuration
read out by a TimePix3 application specific integrated circuit
(ASIC) [23], used as a capture anode for the electron bursts
emerging from the MCPs. A negative potential in the kV range
is applied to the MCP front face to attract low energy positrons.
Impacting positrons with a few keV efficiently release second-
ary electrons in the first section of the MCP channels, which
are subsequently amplified by the accelerating electric field in
the MCP channels. The electron clouds emerge into the gap
between the TimePix3 and the MCP stack and are acceler-
ated by the uniform electric field between the MCP backside
and the TimePix3 surface, which is connected to ground by
the front-end readout electronics of each of its pixels’ pads.
When the electron clouds hit the surface of TimePix3, their
transverse projection typically covers multiple pixels. This is
caused by the expansion they experience during the flight in
the gap between the MCP stack and the TimePix3 chip, driven
by the Coulomb repulsion of the electrons in the charge cloud.
For this effect, and due to the TimePix3 feature of measuring
the pixel-per-pixel absolute charge deposit, sub-pixel size spa-
tial resolution of the impact position is achievable by means of
centroid reconstruction algorithms, applied to each detected
charge cloud.
In the current detector configuration, the positrons are
imaged by two MCPs (18-10, TopAG Lasertechnik GmbH) in
chevron configuration. In this configuration, the relative angle
between pores of neighbouring plates is maximized, which
reduces undesirable ion feedback by forcing the ions to hit
the channel walls instead of propagating all the way through
the channel and release another electron avalanche at the sur-
face of the MCP. An aluminum spacer between the two MCP
plates creates a gap of 300 µm. Each plate has a thickness of
0.43 mm, an active area of 18.6 mm and an open area ratio of
⩾58%. The channel diameter is 10 µm and the inter-channel
distance is ⩽12 µm. The electron gain is ⩾103for each indi-
vidual plate at a voltage of 0.8 kV between the front face and
the back face of the MCP. The readout ASIC is mounted at
a distance of ≈4 mm from the MCP stack. With TimePix3 at
ground, the voltage applied to the back face of the second MCP
plate in the chevron stack determines the acceleration of the
electron cloud emitted by the MCP stack. TimePix3 is a hybrid
pixel detector readout chip based on 130 nm CMOS tech-
nology, developed by the Medipix3 collaboration at CERN
Pixels with a size of 55 ×55 µm2are arranged in a matrix
of 256 ×256, forming an area of 14 ×14 mm2. The chip is
operated in data-driven readout mode, where individual pixels
are read out when triggered while the rest of the pixel matrix
remains active. TimePix3 allows the simultaneous measure-
ment of time-over-threshold (ToT), which is an output value
proportional to the charge collected by the pixel, and the time-
of-arrival (ToA) with a resolution of 1.56 ns (640 MHz). The
TimePix3 analog front-end consists of an input pad, a preamp-
lifier with a leakage current compensation circuit, a digital-
to-analog converter and a discriminator. The charge collected
by a pixel is integrated and preamplified. If the output voltage
of the preamplifier exceeds a threshold, a pulse is generated
by the discriminator, which starts the clock. The pulse length
corresponds to the time the preamplifier output voltage is lar-
ger than the threshold. The chip is read out with Katherine, an
ethernet embedded readout interface for TimePix3 [24]. The
chipboard is directly connected to the readout device with a
3
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
Figure 1. Left: conceptual scheme of the detector, showing the MCP stack and the TimePix3 and their electrical connections. Right:
detection module.
VHDCI (Very-High-Density-Cable-Interconnect) 68-pin con-
nector. TimePix3 dissipates 1.5 W of power. To ensure tem-
perature stability of the chip during operation, the back side of
the chipboard is in contact with a water heat exchanger via a
copper plate and copper pipes.
3. Experimental characterization
The detection module was characterized in terms of detection
efficiency, spatial resolution and time resolution. In the follow-
ing, the experimental setup and data processing procedures are
described and the results are discussed.
3.1. Positron beamline used for characterizing the detector
The detector assembly was experimentally tested with the con-
tinuous e+beam from the electrostatic positron beamline of
the AML (AntiMatter Laboratory) of the University of Trento.
The beam is tunable between 50 eV and 25 keV while keeping
the sample under study at ground potential. The beam line is
schematically depicted in figure 2and thoroughly described
in [25]. Briefly, positrons emitted by a 12 mCi Na22 source
and moderated by a tungsten foil are electrostatically trans-
ported, collimated and accelerated towards a sample region
held at ground potential. This is achieved by biasing a set
of low voltage beam shaping electrodes together with their
power supplies to high DC potential in the 50 V–25 kV range.
An insulation transformer connected to the 220 V AC mains
provides the necessary isolation to the DC low voltage power
supplies, while a ceramic section of the vacuum chamber
effectively isolates the beam-forming region to the sample
chamber. The beam size at the target position is roughly 1–
2 mm in full-width at half maximum for all beam energies;
its alignment also does not vary by more than ≈1 mm while
spanning from 200 eV to a few keV [25]. These are very
suitable conditions for the present characterization, as the
TimePix3 chip features a 14 ×14 mm2overall size encom-
passing the entire beam spot. Fine beam steering corrections
(in both horizontal and vertical axes) could optionally be per-
formed by adding discrete magnetic correctors right before
the sample chamber, altering the e+trajectory by the little
amounts needed to either center the beam on specific posi-
tions of the detector surface, or to compensate slight posi-
tional variations during energy scans. Typically, this beamline
Figure 2. Schematic drawing of the electrostatic positron beamline
used to characterize the detection module. Components highlighted
in blue are held at a high DC negative voltage between 0 and 25 kV,
whereas the sample chamber is kept at ground potential by
insulating it from the acceleration stage with a ceramic vacuum
section.
delivers positrons with a constant rate in the 4–8 ×104e+s−1
range. This range does not stem from intensity fluctuations: it
is linked to the intrinsic degradation of the nominal moderation
efficiency from the tungsten foil moderator after exposures to
ambient pressure and due to rest gas in the vacuum. A high
temperature re-activation cycle is required to recrystallize the
tungsten foil after every exposure, and slight moderation effi-
ciency variations up to ±30% are normally observed between
different activation cycles.
3.2. Raw data processing
In a first step, the time of arrival is calculated in ns and the pixel
numbers are converted to cartesian coordinates. The event list
is sorted by increasing ToA values. To identify pixels belong-
ing to the same cluster of pixels, according to the flux of the
4
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
Figure 3. Overlapping footprints of two positron events in one time
frame.
positron beam, a coincidence time window in the order of
a few µs is chosen. All pixels hit within that time window
are interpreted as belonging to the same cluster. A minimum
cluster size in the order of about 20 pixels (depending on the
MCP gain- and acceleration voltages) is set as a threshold
to discriminate noise. Clusters containing less pixels than the
minimum cluster size are rejected. Both, the length of the coin-
cidence time window and the minimum cluster size depend on
the acceleration voltage and the gain of the MCP stack. For
each set of these parameters the coincidence window and the
minimum cluster size needs to be optimized. A compromise
needs to be found between a coincidence window long enough
to avoid clusters being sliced and interpreted as several clusters
of pixels and short enough to prevent event overlaps (figure 3)
in one timeframe. Event overlaps cannot be discriminated by
the algorithm and the center-of-mass of the combination of the
footprints is calculated. This is a systematic error in the ana-
lysis. With optimized settings, event overlaps can be reduced
to 1%–2% of the total events. The minimum cluster size needs
to be set large enough not to account for cluster fractions res-
ulting from cluster slices falsely distributed over different time
frames and to discriminate noise. However, with an increasing
minimum cluster size more events are being falsely rejected.
This process returns a list of clusters, each cluster containing
the pixels triggered, the coordinates of the pixels and their cor-
responding ToA and ToT values.
4. Results and discussion
4.1. Background event rate
An initial background measurement campaign was conduc-
ted to estimate the rate of background events observed by the
detector, not attributable to positron impacts. The measure-
ment was performed by inhibiting the transport of positrons
towards the detector by turning off the high-voltage power sup-
ply biasing the electrostatic 90◦bender in the beamline. All
other power supplies were left to their nominal setting, includ-
ing the main high voltage biasing the beam line set at 1.0 kV, to
allow transporting eventual spurious positive ions originating
from the vertical section of the transfer line, thus allowing a
more inclusive estimation of the rate of background events on
the detector.
A background count rate of 17 s−1was observed, integrat-
ing over the whole surface of the TimePix3 chip. This meas-
ured background rate, if compared to the typical detection rate
when positrons are transported in the order of ≈104s−1, is
completely negligible for the purpose of this work to charac-
terize the resolution and efficiency of the detection module.
These detected background events were most likely either pos-
itive ions originating in the transfer line and transported onto
the detector, cosmic ray events and thermally-excited electron
avalanches in the MCP stack. Shot-noise of TimePix3 was
found to be completely negligible, thanks to the high trigger
threshold that could be programmed in a way not to impact the
efficiency in detecting MCP avalanches.
4.2. Cluster size distribution
The electron clouds emerging from the channels of the MCP
stack leave a circular shaped charge footprint. The size (num-
ber of active pixels) of these electron clusters strongly depends
on the MCP gain voltage and the acceleration voltage across
the gap between the MCP stack and TimePix3. Figure 4shows
the cluster size distribution for varying beam energies, accel-
eration voltages and MCP gain voltages, respectively. Selec-
tions were applied to only take into consideration clusters lar-
ger than 20 pixels and smaller than 500 pixels, to reject cluster
fragments and event overlaps. The distributions were built
from sample sizes of 10 000 clusters.
Higher positron beam energies increase the probability of
creating secondary electrons at the surface of the first MCP,
which reflects in a slight increase in the number of large
clusters, whereas smaller and medium sized clusters seem to
be equally distributed for different beam energies. An increase
in MCP gain voltage leads to higher amplification in the chan-
nels, thus more electrons in the charge clouds and stronger
Coulomb repulsion. The charge cloud size in the plane of
the chip significantly gets larger with increasing MCP gain
voltage, which can be seen in the corresponding histogram.
The distributions also show an increase in cluster size with
higher acceleration voltages. This can be explained by the
electrons with increased kinetic energy having a higher chance
to activate a pixel, thus the former dark border of the elec-
tron cloud becomes visible. This is a systematic effect of the
detector.
4.3. Positron detection efficiency
To obtain the number of positrons hitting the detector sur-
face during a given time interval with high precision, the
photons emitted upon positron annihilation at the surface
of the detector were measured with a NaI(Tl) scintillator
(Scionix 76B76/3M) coupled to a photomultiplier tube (Can-
berra 2007P). The scintillator was calibrated with a Na22
source of known activity, mounted close to the detection
5
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
Figure 4. Cluster size distribution as a function of positron beam energy, acceleration voltage and MCP gain voltage, respectively. If not
specified, the acceleration voltage was 500 V, the MCP gain voltage 1300 V and the beam energy 0.5 keV.
Table 1. Detection efficiency in % as a function of MCP gain voltage and acceleration voltage at a positron beam energy of 0.5 keV. The
errors are statistical.
Acceleration
voltage
MCP gain voltage
1200 V 1250 V 1300 V 1350 V 1400 V
100 V <1<1 3.9 ±0.8
200 V <1 1.9 ±0.4 9.8 ±1.4 23.7 ±1.9
300 V <1 8.5 ±1.3 20.1 ±1.6 32.6 ±1.6
400 V <1 2.4 ±0.5 12.7 ±1.3 26.2 ±1.7 37.4 ±1.6
500 V <1 4.9 ±0.7 17.6 ±1.8 31.5 ±1.6 40.4 ±0.8
600 V <1 7.1 ±1.1 20.1 ±1.6 33.0 ±1.7 41.1 ±2.1
module and facing the surface of the first MCP. The calibration
procedure yielded the photopeak detection efficiency DEγat
511 keV of the scintillator.
Varying the parameters beam energy, acceleration voltage
and MCP gain voltage, for each set of parameters data was
taken simultaneously with the MCP-TimePix3 detection mod-
ule and the scintillator. For each measurement, the number of
counts Nγin the 511 keV gamma peak was extracted from
the spectra. With a factor of 2 accounting for the two pro-
duced photons per positron annihilation event, the number of
positrons delivered by the AML positron beamline in each
time interval is:
Ne+=Nγ
2·DEγ
.(1)
The positron detection efficiency of the detection mod-
ule DEe+was calculated as the ratio of the rate of clusters
registered by the detector Nclusters and the rate of positrons
delivered Ne+:
DEe+=Nclusters
Ne+
.(2)
The detection efficiency is limited by the open-area-ratio
(OAR) of the MCPs and the probability of a positron to pro-
duce a secondary electron at the MCP front face. The probab-
ility of the algorithm to correctly interpret a cluster of pixels
as an event also plays an important role. Table 1shows the
detection efficiency in % for different MCP gain voltages and
acceleration voltages at a positron beam energy of 0.5 keV.
Figure 5shows that the detection efficiency strongly
increases with both, MCP gain voltage and acceleration
voltage, respectively. Higher MCP gain voltages lead to more
amplification and larger electron clouds emerging from the
MCP stack. These are more likely to trigger the minimum
6
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
Figure 5. Detection efficiency in % as a function of MCP gain voltage and acceleration voltage, respectively, at positron beam energies of
0.25 keV, 1 keV and 2 keV. For better distinguishability of the individual data points, an offset of 8 V on the voltage axis has been introduced
artificially between the data sets belonging to different beam energies. The MCP gain (acceleration) voltage was kept at 1300 V (500 V)
during the scan of the acceleration (MCP gain) voltage.
number of pixels necessary for a cluster to be interpreted as
a positron event by the algorithm than smaller electron clouds.
The charge cloud effective diameter scales as the square root
of the acceleration voltage [26]. This relation also reflects in
the detection efficiency, as can be observed in figure 5(left).
Slightly increased detection efficiencies are expected with
higher beam energies, which can be attributed to the probab-
ility of secondary electron creation being higher with larger
beam energies.
4.4. Spatial resolution
To obtain sub-pixel spatial resolution, a centroiding algorithm
was applied to calculate the center-of-mass of each cluster. An
image of a sharp knife edge was used to determine the spatial
resolution of the detection module.
4.4.1. Centroiding algorithm. The center-of-mass is extrac-
ted from each cluster identified in the first step of the ana-
lysis procedure described in section 3.2. Therefore, for each
pixel the ToT value, proportional to the deposited energy, is
integrated over the entire time interval of the cluster. With the
pixel numbers converted into cartesian coordinates, the ToT
values are summed for each pixel along the X- and Y-axis,
respectively. This yields two histograms proportional to the
charge distribution across the chip. The mean values of these
histograms are taken as the center-of-mass coordinates of the
cluster. The process is illustrated in figure 6.
4.4.2. Edge spread function. The spatial resolution of the
detection module was determined with the edge spread func-
tion (ESF) method. The ESF is the response of a system to a
high contrast edge. To obtain an image of a sharp edge, half
of the MCP surface was covered with a fine metal blade. The
metal blade was positioned parallel to the MCP surface, with
a 300 µm aluminum spacer between the blade and the MCP to
avoid damaging the MCP. With increasing distance between
the MCP and the knife edge, the spatial resolution is expected
to get worse, however this has not been tested experimentally.
The blade was kept at the same potential as the front face of
the first MCP plate.
The image of the edge was reconstructed, applying the
centroiding algorithm described above to all clusters. The
intensity along the edge within a subset of the image is then
integrated and fitted with an error function:
y=a+b
2[erf(x0−x
σ√2)+1].(3)
This process is illustrated in figure 7. The resolution is
extracted from the parameters of the fit by multiplying σwith
the pixel size of 55 µm.
4.4.3. Results. Displayed in table 2are the results as a func-
tion of MCP gain voltage and acceleration voltage between
the MCP back face and TimePix3, taken at a positron beam
energy of 0.5 keV. The errors are statistical. Figure 8shows the
spatial resolution as a function of the acceleration voltage and
MCP gain voltage, respectively. Higher acceleration voltages
lead to better spatial resolutions. The minimum is reached at
around 12 µm, corresponding to the inter channel distance of
the MCPs. The MCP gain voltage seems to have no significant
influence on the spatial resolution.
5. Time resolution
The timing resolution of the detector was estimated by study-
ing the distribution of the active pixels in time per detected
electron avalanche.
A set of several thousand individual clusters with the
same acceleration and gain conditions was collected, and each
cluster’s starting time was defined by the time of the first pixel
firing (i.e. the minimum ToA value in the cluster). The distribu-
tion of active pixels per time bin was constructed by summing
over the number of active pixels in the [ToA, ToA+ToT] inter-
val in steps of 1.56 ns. The distribution was normalized by the
number of clusters in the sample.
Figure 9shows the average number of active pixels in an
individual typical electron avalanche. Three main structures
7
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
Figure 6. Reduction process from one cluster to the coordinates of its centroids: the charge deposited on the chip is integrated for all pixels
along the X- and Y-axis respectively. The mean value of the resulting distributions is calculated and used as Xand Ycoordinate of the
center-of-mass of the cluster.
Figure 7. On the left a subset of an image of a sharp edge is shown. This image is composed of the center-of-mass coordinates of the
clusters. Integrated along the X-axis, the resulting intensity profile is fitted with an error function. Multiplication of the fit parameter σwith
the pixel side length of 55 µm yields the spatial resolution of the detector.
are identified: a first primary pulse lasting up to 24 ns after
the cluster starting time, a secondary pulse starting at about
24 ns and lasting until 62 ns, and a tertiary pulse after 62 ns.
Each of these structures is related to a different spatial dis-
tribution of pixels firing (shown in figure 9, insets): the first
pulse involves mainly pixels corresponding to the center of the
cluster, whereas the second and the third pulse pixels are dis-
tributed in a ring-shaped corona around the pixels of the first
pulse.
These observations can be explained by ion feedback in
the MCP stack and the dead time of the TimePix3 pixels.
MCP detectors can give rise to afterpulses due to positive
ion feedback propagating backwards in the channels and pro-
ducing delayed, secondary avalanches [27]. These secondary
avalanches are spatially aligned with respect to the primary
avalanche, despite being separated in time by the round-trip of
the ion avalanche in the MCP stack, typically between a few ns
to few 100 ns after the first signal. This round-trip time, on the
8
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
Table 2. Spatial resolution in µm as a function of MCP gain voltage and acceleration voltage at a positron beam energy of 0.5 keV. The
errors are statistical.
Acceleration
voltage
MCP gain voltage
1200 V 1250 V 1300 V 1350 V 1400 V
100 V 20.2 ±1.6 20.3 ±1.8 16.8 ±1.5
200 V 13.4 ±1.6 14.1 ±0.7 14.6 ±0.5 16.8 ±0.7
300 V 15.3 ±1.1 14.3 ±0.7 12.8 ±0.4 14.7 ±0.5 14.3 ±0.9
400 V 13.3 ±0.9 12.0 ±1.4 14.3 ±0.9 12.6 ±0.4 16.2 ±0.6
500 V 12.4 ±1.3 12.6 ±0.5 11.3 ±0.9 13.5 ±0.9 13.9 ±0.5
600 V 11.8 ±0.7 10.5 ±1.1 11.9 ±0.4 11.1 ±0.3 12.5 ±0.6
Figure 8. Spatial resolution in µm as a function of acceleration voltage and MCP gain voltage, respectively, at a positron beam energy of
0.5 keV. In the plot on the left (right), for each acceleration voltage (MCP gain voltage) data point, the resolution was averaged over all
corresponding MCP gain voltages (acceleration voltages). The MCP gain (acceleration) voltage was kept at 1300 V (500 V) during the scan
of the acceleration (MCP gain) voltage.
other hand, is shorter than the pixel dead time of the TimePix3
chip of 475 ns [23]. This implies that the core of the second-
ary avalanche cannot be detected by pixels that have fired in
the primary avalanche, which are still recovering. Only the
outer corona of the secondary avalanche is effectively detec-
ted, as a consequence of the overall charge accumulation from
the primary avalanche and the secondary avalanche. The same
principle applies to further avalanches.
Several ways are possible to estimate the average spread
of the time distribution and work out an estimate of the time
resolution of this detector in determining the time of impact
of isolated events. Conservatively, one can consider the whole
time distribution in figure 9and estimate the resolution from
its whole standard deviation, leading to a time resolution in the
order of ≈40 ns. Alternatively, one can apply a cut at 25 ns and
isolate only the pixels from the first pulse, while rejecting those
that were most likely triggered by an ion-feedback avalanche.
This second approach pinpoints that a more accurate estimate
of the time resolution of isolated events is 12 –15 ns.
The observed timing spread of the first avalanche is likely
to originate from the energy spread of the electrons as they
get amplified within the channels and space charge effects
during their accelerated flight in the gap between the MCP
stack and the TimePix3 chip. It has to be noted that the
time-walk effect of the TimePix3 pixels is also likely to
contribute to a broadening of the distribution, as different
pixels in the cluster reach the triggering threshold at differ-
ent times. Methods to compensate for such time-walk effects
with TimePix3 are known in literature [28], yet their study
Figure 9. Evolution of the number of active pixels during the
registration of one electron cloud (corresponding to one individual
positron event).
in the present case, as well as an accurate simulation of the
electron effects in the channels, goes beyond the scope of this
experimental first investigation and will be subject of future
studies.
The time resolution was extracted from the first 25 ns in
the pulse by calculating the standard-deviation of the distri-
bution, averaging over several thousands of such pulses (with
the same MCP gain voltage, acceleration voltage and beam
energy). The pulse was normalized by the number of clusters
of the sample. For samples of different MCP voltage settings
9
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
Figure 10. Time resolution as a function of acceleration voltage and MCP gain voltage, respectively, for different beam energies. For better
distinguishability of the individual data points, an offset of 8 V on the voltage axis has been introduced artificially between the data sets
belonging to different beam energies. The MCP gain (acceleration) voltage was kept at 1300 V (500 V) during the scan of the acceleration
(MCP gain) voltage.
and beam energies, the amplitude of the pulses varies slightly
with the average cluster size. The errors are statistical.
The time resolution was studied as a function of the accel-
eration voltage between the MCP stack and TimePix3, as a
function of the MCP gain, and for different energies of the
positron beam (shown in figure 10). A worsening of the time
resolution was observed for higher MCP gain- and accelera-
tion voltages, linked to the increase of the average cluster size.
No significant dependence upon the positron beam energy was
observed.
Finally, we want to emphasize that this assessment aimed
at determining the precision of this detector in timing isol-
ated events, giving no information about its absolute accur-
acy. In fact, no measurement could be performed to measure
the travelling time from the positron impact at the surface
of the first MCP plate to the TimePix3 chip. This would
require an external timing signal finely calibrated with the
positron impact on the detector plane, unavailable on a con-
tinuous positron beamline relying on stochastic nuclear disin-
tegrations. An experiment like such would require a ns-pulsed
positron source, such as that of AEgIS [6].
6. Conclusion and outlook
In this work, we have demonstrated a high resolution ima-
ging detector for low energy positrons featuring single event
tagging and simultaneous position and timing readout. This
detector is based on the combination of a MCP and a TimePix3
integrated circuit, providing for each pixel the precise ToA
and total charge deposit given as the ToT. This hybrid detector
combines the high detection efficiency to incoming low energy
charged particles of MCPs with the high timing and spatial
resolution of the TimePix3 readout anode. A characteriza-
tion campaign conducted on a low energy positron beamline
showed that spatial and timing resolutions of δy≈12 µm and
δt≈12–15 ns can be simultaneously achieved in combination
with a 41% total detection efficiency, i.e. 71% probability that
a positron generates a secondary electron avalanche, if one
accounts for the 58% open area ratio. The availability of this
tool opens several new possibilities in the antimatter research
field. One area of impact are experiments dealing with accurate
determination of the impact of positrons. The enhanced spatial
resolution of this method compared to known MCP-phosphor
screen designs for positron physics, typically above 50 µm
[17,29], make it appealing, for instance, for positron interfero-
metry [30]. Another area are inertial studies with antihydrogen
or positronium, where accurate event-by-event determination
of position/timing of positron impacts is of high relevance.
Indeed, most proposed inertial schemes to date are based on
a free-falling region and detectors sampling the atoms impact
positions and time of flights, such as in [8] or [9]. Both anti-
hydrogen and positronium can either be field [31] or photo-
ionized [17], leading to low energy detectable positrons. The
detector presented here would allow working out gfrom grav-
itational displacement laws of the type y=kgt2on an event-
by-event basis20, allowing the distribution of gvalues to be
empirically determined and the different contributions from
δyand δtto the total error budget δg2=(g
yδy)2
+(2g
tδt)2
quantified on the same sample.
Similar considerations apply to inertial sensing exper-
iments with positronium atoms, in its either 23S [9] or
Rydberg [32] long-lived states. Little experimentation has
been performed so far with the external degrees of freedom
of positronium [4]. Measuring optical forces (both dipolar and
reactive), forces due to DC magnetic and electric gradients
and ultimately gravitational forces with positronium are still
open items of research. A detector as the one presented here,
deployed on an existing long-lived positronium source with
a Moiré deflectometer and a suitable ionization stage, would
enable the detection of an acceleration of ≈10 000 m s−2in a
week of data taking, according to [9]. This would allow for
more accurate charge neutrality tests of positronium, as well
as measurements of its electric and magnetic dipole moment.
Several ways are being considered to enhance the spa-
tial resolution, efficiency and detection surface of the cur-
rent design. Minimizing the interdistance between the MCP
20 Here kis a geometrical constant a-priori set by the experimental
scheme chosen, e.g. k=0.5 for free-fall in vacuum and k=1 for a moiré
deflectometer.
10
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
stack and TimePix3 will lead to more localized charge dis-
tributions at the detection plane. This will reduce the recon-
structed charge quantization noise by making a better use
of the 10 bit dynamics in the measurement of the time-over-
threshold of active pixels, leading to a more accurate event
centroiding reconstruction. The adoption of the TimePix3 chip
variant employing through-silicon vias and bump-bonded con-
nections would remove the ≈1 mm clearance limitation of the
wire bonds connectors of the current version. Similarly would
do the adoption of the new bump-bonded TimePix4 chip ver-
sion as it will become available, featuring also a four-fold
detection surface. Ultimately, the resolution will be limited
by the MCP interchannel distance; MCPs down to 3 µm chan-
nels are commercially available. Furthermore, adopting MCPs
with higher open area ratios than the current 58% would pro-
portionally increase the maximum efficiency (90% OAR are
commercially available).
Data availability statement
The data that support the findings of this study are available
upon reasonable request from the authors.
Acknowledgments
This work was sponsored by the Wolfgang Gentner Pro-
gramme of the German Federal Ministry of Education and
Research (Grant No. 05E18CHA), by the European’s Union
Horizon 2020 research and innovation program under the
Marie Sklodowska Curie Grant Agreement No. 754496,
FELLINI and by Warsaw University of Technology within
the Excellence Initiative: Research University (IDUB) pro-
gramme and the IDUB-POB-FWEiTE-1 project grant. We
thank the Medipix collaboration for kindly providing us with a
TimePix3 chip and the readout system. In particular we thank
Michael Campbell and Jerome Alozy for their support and
fruitful discussions. We thank Angela Gligorova for an intro-
duction to TimePix3 and Marco Bettonte for technical support
at the positron beamline. We thank Anton Tremsin and John
Vallerga for their feedback and advice on how to improve the
detection module.
ORCID iDs
L Glöggler https://orcid.org/0000-0003-0194-8680
Ł Kłosowski https://orcid.org/0000-0002-5463-5381
L Penasa https://orcid.org/0000-0003-3117-5826
M Piwi´
nski https://orcid.org/0000-0001-5847-2578
L Povolo https://orcid.org/0000-0003-1451-1947
M Volponi https://orcid.org/0000-0002-5048-8708
M Zawada https://orcid.org/0000-0002-2826-5129
References
[1] Testera G et al 2015 The AEgIS experiment Hyperfine
Interact. 233 13–20
[2] Aghion S et al 2014 A Moiré deflectometer for antimatter Nat.
Commun. 54538
[3] Amsler C et al 2021 Pulsed production of antihydrogen
Commun. Phys. 419
[4] Cassidy D B 2018 Experimental progress in positronium laser
physics Eur. Phys. J. D72 53
[5] Amsler C et al 2019 Velocity-selected production of 23S
metastable positronium Phys. Rev. A99 033405
[6] Aghion S et al 2015 Positron bunching and electrostatic
transport system for the production and emission of dense
positronium clouds in vacuum Nucl. Instrum. Methods
Phys. Res. B362 86–92
[7] Fajans J and Surko C 2020 Plasma and trap-based techniques
for science with antimatter Phys. Plasmas 27 030601
[8] Doser M et al 2012 Exploring the WEP with a pulsed cold
beam of antihydrogen Class. Quantum Grav. 29 184009
[9] Mariazzi S, Caravita R, Doser M, Nebbia G and Brusa R S
2020 Toward inertial sensing with a 23S positronium beam
Eur. Phys. J. D74 79
[10] Kimura M et al 2013 Development of nuclear emulsions with
1µm spatial resolution for the AEgIS experiment Nucl.
Instrum. Methods Phys. Res. A732 325–9
[11] Fujiwara M C et al 2004 Three-dimensional annihilation
imaging of trapped antiprotons Phys. Rev. Lett.
92 065005
[12] Andresen G B et al 2012 Antihydrogen annihilation
reconstruction with the ALPHA silicon detector Nucl.
Instrum. Methods Phys. Res. A684 73–81
[13] Storey J et al 2013 Particle tracking at 4 K: the fast
annihilation cryogenic tracking (FACT) detector for the
AEgIS antimatter gravity experiment Nucl. Instrum.
Methods Phys. Res. A732 437–41
[14] Czerwinski E et al 2017 Commissioning of the J-PET detector
for studies of decays of positronium atoms Acta Phys. Pol.
B48 1961
[15] Andresen G B et al 2009 Antiproton, positron and electron
imaging with a microchannel plate/phosphor detector Rev.
Sci. Instrum. 80 123701
[16] Jones A C L et al 2016 Large-area field-ionization detector
for the study of Rydberg atoms Rev. Sci. Instrum.
87 113307
[17] Amsler C et al 2019 A ≈100 µm-resolution position-sensitive
detector for slow positronium Nucl. Instrum. Methods Phys.
Res. B457 44–48
[18] Caravita R et al 2020 Hybrid imaging and timing Ps laser
excitation diagnostics for pulsed antihydrogen production
Acta Phys. Pol. A137 96–100
[19] Vallerga J , Tremsin A, DeFazio J, Michel T, Alozy J, Tick T
and Campbell M 2014 Optical MCP image tube with a quad
Timepix readout: initial performance characterization
J. Instrum. 9C05055
[20] Vallerga J , McPhate J, Tremsin A and Siegmund O 2008 High
resolution UV, alpha and neutron imaging with the Timepix
CMOS readout Nucl. Instrum. Methods Phys. Res. A
591 151–4
[21] Vallerga J , Raffanti R, Tremsin A, McPhate J and Siegmund O
2011 MCP detector read out with a quad Timepix at
kilohertz frame rates J. Instrum. 6C01049
[22] Tremsin A S and Vallerga J 2020 Unique capabilities and
applications of microchannel plate (MCP) detectors with
Medipix/Timepix readout Radiat. Meas. 130 106228
[23] Poikela T et al (TimePix3 collaboration) 2014 A 65K channel
hybrid pixel readout chip with simultaneous ToA/ToT and
sparse readout J. Instrum. 9C05013
[24] Burian P et al 2020 Ethernet embedded readout interface for
Timepix3 J. Instrum. 15 C01037
[25] Zecca A, Bettonte M, Paridaens J, Karwash G P and Brusa R S
1998 A new electrostatic positron beam for surface studies
Meas. Sci. Technol. 9409
11
Meas. Sci. Technol. 33 (2022) 115105 L Glöggler et al
[26] Vallerga J, McPhate J, Tremsin A S and Siegmund O
2008 Optically sensitive MCP image tube
with a Medipix2 ASIC readout Proc. SPIE
7021 702115
[27] Gatti E, Oba K and Rehak P 1983 Study of the electric field
inside microchannel plate multipliers IEEE Trans. Nucl.
Sci. 30 1
[28] Turecek D , Jakubek J and Soukup P 2016 USB 3.0 readout
and time-walk correction method for TimePix3 detector
J. Instrum. 11 C12065
[29] Vinelli G, Ferragut R, Giammarchi M, Maero G, Romé M and
Toso V 2020 Real-time monitoring of a positron beam
using a microchannel plate in single-particle mode
J. Instrum. 15 11030
[30] Sala S, Ariga A , Ereditato A, Ferragut R, Giammarchi M,
Leone M, Pistillo C and Scampoli P 2019 First
demonstration of antimatter wave interferometry Sci. Adv.
57610
[31] Storry C H et al (ATRAP collaboration) 2004 First
laser-controlled antihydrogen production Phys. Rev. Lett.
93 263401
[32] Cassidy D B and Hogan S D 2014 Atom control and gravity
measurements using Rydberg positronium Int. J. Mod.
Phys.: Conf. Ser. 30 1460259
12
