PICOSEC-Micromegas Detector

• The PICOSEC-Micromegas detector is a gaseous precise-timing system that uses a Cherenkov radiator and a semi-transparent photocathode to produce prompt photoelectrons.
• It shifts timing from stochastic gas ionization to controlled photoelectron emission, achieving resolutions as low as 12.5 ps for high-energy particles.
• Performance improvements through gap reduction, photocathode optimization, and refined readout design have expanded its applications in HL-LHC, 4D tracking, and large-area timing layers.

The PICOSEC-Micromegas detector is a gaseous precise-timing detector that combines a Cherenkov radiator, a semi-transparent photocathode, and a Micromegas amplification structure in order to shift the time-defining process from stochastic primary ionization in gas to prompt photoelectron emission at a well-defined surface. In this configuration, a traversing charged particle first produces ultraviolet Cherenkov photons in a radiator, those photons liberate prompt photoelectrons at the photocathode, and the electrons are then amplified in a very thin high-field drift/preamplification region followed by a Micromegas amplification gap. This architecture has enabled timing for minimum-ionizing particles in the few-$10$ ps regime, beginning with $24.0 \pm 0.3~\mathrm{ps}$ for $150~\mathrm{GeV}$ muons and $76.0 \pm 0.4~\mathrm{ps}$ for single photoelectrons, and later progressing to substantially better single-channel performance through gap reduction, parasitic-control engineering, and photocathode optimization (Bortfeldt et al., 2017).

1. Defining principle and historical emergence

PICOSEC was developed to address timing requirements in environments where extreme particle multiplicities, severe pile-up, or monitored secondary beams make sub-nanosecond timing insufficient. Its defining innovation is that it does not time direct ionization in the gas volume in the usual Micromegas sense. Instead, it times photoelectrons emitted from a photocathode illuminated by Cherenkov light. This suppresses the large fluctuations associated with the position of the first ionization cluster in ordinary gaseous detectors and is the essential reason why PICOSEC can reach picosecond-order timing. The concept was originally formulated in the context of HL-LHC pile-up mitigation and later extended toward large-area timing layers, calorimeter timing, and monitored neutrino beams (Bortfeldt et al., 2019).

A common misconception is that gaseous detectors are intrinsically limited to nanosecond timing. PICOSEC demonstrates that this limitation applies to conventional ionization-based operation, not to a Cherenkov-photocathode architecture with an ultrathin preamplification region. In this sense, PICOSEC is not merely a faster Micromegas tracker; it is a distinct detector class within MPGD-based timing detectors. The same distinction also underlies later derivative developments such as $\mu$RWELL-PICOSEC, which retain the Cherenkov-photocathode timing principle while changing the amplification structure for mechanical reasons (Gnanvo, 10 Dec 2025).

2. Architecture and operating cycle

The canonical PICOSEC detector comprises a $3~\mathrm{mm}$ $MgF_2$ Cherenkov radiator, a semi-transparent photocathode, a very thin drift/preamplification gap, a Micromegas mesh, an amplification gap, and an anode readout. In early single-channel implementations the photocathode was an $18~\mathrm{nm}$ CsI layer on a thin Cr substrate; later studies also used Aluminium, DLC, B$_4$C, Ti, and thinner CsI layers. The standard gas is the COMPASS mixture,

$$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$

typically at ambient pressure or $24.0 \pm 0.3~\mathrm{ps}$0 bar. Baseline PICOSEC geometries used a drift region of about $24.0 \pm 0.3~\mathrm{ps}$1 and an amplification gap of about $24.0 \pm 0.3~\mathrm{ps}$2–$24.0 \pm 0.3~\mathrm{ps}$3, while later reduced-gap versions used a drift gap of approximately $24.0 \pm 0.3~\mathrm{ps}$4 with an amplification region of approximately $24.0 \pm 0.3~\mathrm{ps}$5 (Manthos et al., 2022).

The operating sequence is fixed by this geometry. A relativistic charged particle traverses the radiator and emits Cherenkov photons. These photons strike the photocathode and release photoelectrons into the drift region. The electric field in that gap is intentionally very high, so multiplication already starts there as a pre-amplification avalanche. Electrons then pass through the mesh into the Micromegas amplification gap, where a second multiplication occurs. The resulting waveform contains a fast electron peak followed by a slower ion tail; in early PICOSEC measurements the electron component had a rise time of about $24.0 \pm 0.3~\mathrm{ps}$6 ps, while later large-area ENUBET-oriented studies described a fast electron peak of about $24.0 \pm 0.3~\mathrm{ps}$7 ps and an ion tail of about $24.0 \pm 0.3~\mathrm{ps}$8 ns. Timing is extracted from the leading edge of the electron peak, commonly with a logistic or sigmoid fit and a 20% constant-fraction definition of the signal-arrival time (SAT) (Bortfeldt et al., 2019).

Later single-channel engineering work showed that detector packaging and readout parasitics are part of the timing problem rather than external details. In that redesign, a $24.0 \pm 0.3~\mathrm{ps}$9 FR4 Micromegas board, a multilayer outer board, spring-loaded signal and HV contacts, local filtering, and short quasi-coaxial signal return paths were used to reduce noise, improve stability, and preserve the leading-edge slew rate. The same study compared readout pads of 10 mm, 13 mm, and 15 mm diameter and found that pad capacitance scaled strongly with pad size, directly affecting rise time and timing (Utrobicic et al., 2024).

3. Signal formation, SAT–amplitude correlation, and timing physics

The central timing observable in PICOSEC is the SAT, defined relative to an external reference such as an MCP-PMT or a fast photodiode. In several studies the timing resolution is defined as the RMS of the SAT distribution; other analyses use a double-Gaussian parameterization and quote the corresponding fitted or combined width. A key empirical feature is that the mean SAT and the timing resolution depend strongly on the electron-peak charge: larger pulses arrive earlier and exhibit smaller jitter. For single-photoelectron data this dependence is described by the power law

$150~\mathrm{GeV}$0

with $150~\mathrm{GeV}$1 the e-peak charge and $150~\mathrm{GeV}$2 observed to be independent of drift voltage (Bortfeldt et al., 2019).

That behavior is not interpreted as a trivial electronics artifact. GARFIELD++ simulations and phenomenological modeling showed that the dominant fluctuations arise in the drift/preamplification region, not in the later Micromegas amplification stage. The microscopic variable corresponding to SAT is the average passage time through the mesh of the preamplification electrons. The largest source of jitter is the stochastic distance and time to the first ionization in the thin drift gap, together with the correlated development of the pre-amplification avalanche. Earlier first ionization produces a longer pre-amplification avalanche, a larger signal, an earlier arrival time, and better resolution. This is why increasing drift voltage improves timing more effectively than shifting gain into the amplification stage (Bortfeldt et al., 2019).

The phenomenological model further introduced a “time-gain per interaction” picture to explain several counter-intuitive microscopic observations, including the different effective drift behavior before and after the first multiplication. A consequence is that the observed time walk reflects detector avalanche physics as well as electronics. This explains why reduced drift gaps improve performance: a shorter drift region narrows the distribution of possible avalanche-development lengths and makes the pre-amplification process more prompt and less stochastic. The same picture also clarifies why segmentation or mesh transmission need not destroy timing if the early avalanche remains well controlled (Paraschou, 2020).

4. Timing performance and performance evolution

The first benchmark measurements established $150~\mathrm{GeV}$3 timing resolution for single photoelectrons and $150~\mathrm{GeV}$4 for $150~\mathrm{GeV}$5 muons in the standard single-channel detector. In later reduced-drift-gap prototypes with a drift gap of approximately $150~\mathrm{GeV}$6 and an Aluminium photocathode, the single-photoelectron resolution improved to $150~\mathrm{GeV}$7, and the same detector reached $150~\mathrm{GeV}$8 when responding to $150~\mathrm{GeV}$9 photoelectrons on average; for about $76.0 \pm 0.4~\mathrm{ps}$0 photoelectrons the timing was reported as better than $76.0 \pm 0.4~\mathrm{ps}$1. These results established the drift-gap reduction as a decisive optimization axis (Manthos et al., 2022).

A later single-channel redesign aimed specifically at minimizing parasitics and preserving signal integrity produced another major step. At a common operating point of $76.0 \pm 0.4~\mathrm{ps}$2 cathode and $76.0 \pm 0.4~\mathrm{ps}$3 anode, the detector with a 10 mm diameter readout pad achieved $76.0 \pm 0.4~\mathrm{ps}$4 within a central $76.0 \pm 0.4~\mathrm{ps}$5 mm region, compared with $76.0 \pm 0.4~\mathrm{ps}$6 and $76.0 \pm 0.4~\mathrm{ps}$7 for the 13 mm and 15 mm versions. After further optimization, the same 10 mm design reached a record PICOSEC-Micromegas time resolution with a CsI photocathode of

$76.0 \pm 0.4~\mathrm{ps}$8

for tracks within the central 4 mm diameter region at $76.0 \pm 0.4~\mathrm{ps}$9 cathode and $\mu$0 anode (Utrobicic et al., 2024).

Photocathode optimization pushed the single-pad concept further. A systematic 2026 photocathode study found that a 5 nm CsI photocathode on a 2.4 nm Ti interlayer delivered

$\mu$1

with $\mu$2 extracted photoelectrons per MIP and 99.9% detection efficiency for fully contained events. That result was described as the most precise time resolution achieved by PICOSEC-Micromegas to date. The same study also quantified a robust-material frontier in the $\mu$3 ps range, showing that the detector concept remained viable even when optimized for durability rather than absolute timing (Lisowska et al., 20 Apr 2026).

5. Segmentation, large-area detectors, and mechanical tolerances

The transition from a single active pad to large-area segmented detectors introduced a new systems problem: preserving drift-gap uniformity across many channels. The first 19-pad multi-pad prototype, with hexagonal pads of 1 cm diameter, demonstrated that precise timing survives segmentation. After correction, the all-pad single-pad timing for tracks near pad centers was

$\mu$4

while a three-pad corner-sharing region yielded

$\mu$5

The dominant degradation was traced not to segmentation itself but to drift-gap thickness non-uniformity caused by PCB deformation. This study established that impact-position-based flatness corrections could restore uniform timing, and it identified better than $\mu$6 drift-thickness control as a practical design target (Aune et al., 2020).

That lesson directly motivated the first large-area 100-channel, $\mu$7 module. A rigid hybrid ceramic/FR4 Micromegas board was designed to maintain planarity better than $\mu$8 over the full area. With a $\mu$9 drift gap, the detector reproduced below-$3~\mathrm{mm}$0 ps timing in the central $3~\mathrm{mm}$1 regions of more than ten pads. A thinner $3~\mathrm{mm}$2 version, combined with newly developed RF pulse amplifiers, reached an average SAT RMS of $3~\mathrm{mm}$3 ps, with reported pad values between $3~\mathrm{mm}$4 ps and $3~\mathrm{mm}$5 ps (Utrobicic et al., 2023).

Large-area development then bifurcated into application-driven resistive modules. A $3~\mathrm{mm}$6 resistive PICOSEC Micromegas with 100 channels and a $3~\mathrm{mm}$7 DLC resistive layer achieved $3~\mathrm{mm}$8 ps for an individual pad and about $3~\mathrm{mm}$9 ps mean over the measured channels with SAMPIC, showing that a resistive Micromegas could preserve timing while improving discharge robustness (Lisowska et al., 2023). A broader large-area integration program later reported around $MgF_2$0 ps for individual pads in MIPs for $MgF_2$1 and $MgF_2$2 prototypes, including a four-crystal $MgF_2$3 assembly and dedicated RF-AM + WDM readout (Meng et al., 9 Jan 2025). In the ENUBET program, a 96-pad resistive demonstrator with $MgF_2$4 active area and $MgF_2$5 pads achieved 43 ps timing with good SAT uniformity, and explicitly quantified planarity within $MgF_2$6 as necessary to maintain good timing over the full surface (Kallitsopoulou et al., 5 Dec 2025).

6. Photocathodes, robustness, and the performance–durability trade-off

Photocathode technology is the principal materials bottleneck of PICOSEC. CsI remains the benchmark because its high ultraviolet quantum efficiency yields the highest number of extracted photoelectrons and therefore the best timing. Earlier PICOSEC work associated standard CsI implementations with about 10 photoelectrons per MIP or $MgF_2$7 in some later single-pad configurations. However, CsI is highly hygroscopic and susceptible to ion backflow, sparks, discharges, and humid handling. In one comparative study, CsI lost 80% of its QE after $MgF_2$8 accumulated charge, and in another it dropped to about 30% of its initial normalized QE after roughly $MgF_2$9 of ion bombardment (Lisowska et al., 2024).

Robust alternatives therefore became a major research direction. DLC emerged as a particularly durable transmissive photocathode. A dedicated study found an optimized DLC thickness of approximately 3 nm, with $18~\mathrm{nm}$0, 97% detection efficiency for $18~\mathrm{nm}$1 muons, and about 42 ps time resolution at $18~\mathrm{nm}$2 preamplification voltage and $18~\mathrm{nm}$3 amplification voltage. The same work reported less than 20% degradation after around $18~\mathrm{nm}$4 ion bombardment, in strong contrast to CsI (Wang et al., 2024). In parallel studies of robust photocathodes, 1.5 nm DLC yielded $18~\mathrm{nm}$5 and 96.8% efficiency in a $18~\mathrm{nm}$6 mm device, improving to $18~\mathrm{nm}$7 and 99.4% efficiency when a 5 mm $18~\mathrm{nm}$8 radiator and 15 mm active area were used to increase Cherenkov light yield (Lisowska et al., 2024).

B$18~\mathrm{nm}$9C and Ti define an additional robust branch. Earlier B$_4$0C results reached $_4$1 with 95.4% efficiency, and the same material showed better aging than CsI after $_4$2 (Lisowska et al., 2024). The 2026 comparative photocathode program improved B$_4$3C to

$_4$4

with $_4$5 and 99.1% efficiency, while 2.4 nm Ti reached

$_4$6

with $_4$7 and 98.8% efficiency (Lisowska et al., 20 Apr 2026). The trade-off is therefore explicit rather than controversial: CsI maximizes timing, whereas DLC, B$_4$8C, and Ti improve robustness and lifetime at the cost of lower photoelectron yield and a timing penalty of order $_4$9–$$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$0 ps.

7. Signal processing, applications, and related developments

PICOSEC timing was originally established with full offline waveform analysis, but several later studies focused on online-capable reconstruction. In a laser data set corresponding to an average of $$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$1 photoelectrons, the benchmark full waveform method—logistic leading-edge fit plus $$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$2 CFD and time-walk correction—achieved

$$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$3

A constant-threshold method corrected with multi Charge over Threshold reached

$$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$4

and an ANN timing estimator reached

$$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$5

showing that reduced-information or ML-based timing can match the standard offline precision in an optimized detector regime (Kallitsopoulou et al., 17 Sep 2025). This line of work is directly relevant to scalable readout, since full waveform storage at $$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$6 is impractical for high-channel-count systems (Manthos et al., 2022).

The application space has broadened accordingly. For ENUBET, PICOSEC-Micromegas has been evaluated both for timing individual particles, mainly muons, and for tagging electromagnetic showers. A 7-pad resistive detector reached

$$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$7

for $$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$8 muons with about 11 photoelectrons per muon track, and in an electromagnetic-shower configuration with a 5 cm iron absorber the central pad achieved

$$\mathrm{Ne}~(80\%) + \mathrm{C_2H_6}~(10\%) + \mathrm{CF_4}~(10\%),$$9

with overall detector response to showers around 30 ps on average (Kallitsopoulou et al., 2024). More generally, the detector has been proposed for HL-LHC-type 4D tracking, timing layers in electromagnetic calorimetry, hadron-dump muon stations, and T$24.0 \pm 0.3~\mathrm{ps}$00 tagging layers.

A related but distinct development is $24.0 \pm 0.3~\mathrm{ps}$01RWELL-PICOSEC, which preserves the Cherenkov-photocathode timing concept while replacing the Micromegas amplification structure with a resistive micro-well. That program is motivated by the mechanical complexity of stretched meshes, support pillars, and planarity requirements in large-area PICOSEC-Micromegas. Preliminary single-channel $24.0 \pm 0.3~\mathrm{ps}$02RWELL-PICOSEC results showed better than 24 ps, approaching but not surpassing the established PICOSEC-Micromegas benchmark (Gnanvo, 10 Dec 2025). This suggests that the Cherenkov-photocathode timing principle is portable across MPGD topologies, while the original PICOSEC-Micromegas remains the reference implementation for the highest demonstrated timing precision.