Large Area Picosecond PhotoDetectors (LAPPDs)

• LAPPDs are large-area, MCP-based imaging photodetectors that provide picosecond timing and sub-millimeter spatial resolution for high-energy physics applications.
• They employ ALD-functionalized glass capillary-array MCPs, large-area bialkali photocathodes, and custom waveform-sampling electronics to achieve high gain and cost-effective scalability.
• LAPPDs achieve gains exceeding 10^7 with timing resolutions under 65 ps, significantly improving Cherenkov light detection, neutrino event reconstruction, and other precision experiments.

Large Area Picosecond PhotoDetectors (LAPPDs) are large-area microchannel-plate photomultiplier tubes (MCP-PMTs) developed to combine picosecond-scale timing, millimeter- to sub-millimeter-scale position sensitivity, large active area, high gain, and low cost per unit area. In the LAPPD program, the central target module was an 8-inch-class, $20 \times 20$ cm planar detector using ALD-functionalized glass capillary-array MCPs, large-area bialkali photocathodes, RF-capable hermetic packaging, and custom waveform-sampling electronics (Adams et al., 2016). Subsequent work established both first-generation stripline-anode devices and second-generation capacitively coupled devices, extended the platform to cryogenic operation, and deployed LAPPDs in neutrino and Cherenkov detector environments (Shin et al., 2022).

1. Historical development and programmatic context

The Large-Area Picosecond PhotoDetector Collaboration was formed in 2009 to develop large-area photodetectors capable of time resolutions measured in pico-seconds, with accompanying sub-millimeter spatial resolution. During the next three and one-half years the collaboration developed the LAPPD design of $20 \times 20$ cm modules with gains greater than $10^7$, non-uniformity less than $15\%$, time resolution less than 50 psec for single photons, and spatial resolution of 700 $\mu$m in both lateral dimensions (Adams et al., 2016).

LAPPD was not only a detector concept but a full technology stack. The collaboration developed 8-inch-square capillary glass MCP substrates, ALD-based resistive and secondary-emissive coatings, large-area bialkali photocathodes, RF-capable hermetic packaging, a custom low-power 15-GigaSample/sec waveform-sampling 6-channel integrated circuit, and a two-level modular data acquisition system based on Field-Programmable Gate Arrays for local control, data-sparcification, and triggering (Adams et al., 2016). The collaboration ended in December 2012 with a transition from R&D to commercialization, and pilot production at Incom, Inc. was initiated in 2018 (Adams et al., 2016, Shin et al., 2022).

A notable intermediate branch was Argonne National Laboratory’s $6 \text{ cm} \times 6 \text{ cm}$ glass-body MCP-based photodetector program, developed as an R&D testbed and as a step toward larger-area devices and eventual commercialization. This smaller platform reproduced the central LAPPD design logic—ALD-coated MCPs, stripline readout, hermetic sealing, and bialkali photocathodes—while enabling rapid process iteration and cryogenic adaptation studies (Wang et al., 2016, Dharmapalan et al., 2016).

The technical history also records a significant shortcoming: a fully self-standing, hermetically sealed tile/module with a bialkali photocathode was not completed during the formal R&D period. The reported interpretation was that this was primarily a problem of scaling a delicate commercial sealing recipe from smaller devices to much larger 20 cm-class modules, rather than a fundamental flaw in the concept (Adams et al., 2016).

2. Detector architecture and operating principles

LAPPDs are planar MCP-based imaging photosensors. A photon enters through the entrance window, is converted in a semitransparent or multi-alkali photocathode, and the emitted photoelectron is multiplied through a chevron pair of MCPs before reaching the anode/readout system. In the 2022 Gen-II implementation, the device is presented as an 8" $\times$ 8" MCP-PMT built from two independently biased ALD-fabricated glass capillary array MCPs in a chevron stack, sealed inside a hermetic glass or ceramic envelope, with a UV-grade fused silica window and a Na$_2$KSb bialkali photocathode (Shin et al., 2022). Earlier baseline devices used a borosilicate glass hermetic package, a chevron pair of large-area MCPs, a Na$_2$KSb photocathode, and a microstrip anode with 28 silver strips (Lyashenko et al., 2019).

The essential distinction from conventional PMTs is that LAPPDs are imaging detectors. They measure not only photon arrival time but also hit position on the detector surface. In ANNIE-related descriptions, they are explicitly contrasted with traditional photosensors because they provide position information for each detected photon, not just a summed charge (Tiras, 2019). This distinction is central to their use in Cherenkov reconstruction, track imaging, and vertexing.

Two readout generations dominate the literature:

Generation	Internal anode	External readout
Gen-I	Stripline anode for direct charge readout	Fixed by sealed package
Gen-II	Internal resistive thin-film anode	Capacitively coupled customizable PCB

In Gen-I, the anode is physically tied to the sealed vacuum package, and the readout geometry is essentially fixed after fabrication. In Gen-II, the internal resistive thin-film anode capacitively couples to an external, customizable PCB, including stripline or pad layouts, so the photodetector can be manufactured once while the signal pickup board can be adapted later to a given application (Shin et al., 2022). The paper reporting this architecture states that Gen-II does not sacrifice spatial information despite giving up the fixed stripline readout; with centroiding of neighboring pixel amplitudes, the reported position resolution is $\sim 1.30$ mm for 25 mm pitch pixels and $20 \times 20$0 mm for 6 mm pitch pixels (Shin et al., 2022).

For stripline devices, position along the strip is reconstructed from differential arrival time at the two ends. Argonne’s $20 \times 20$1 work gives

$20 \times 20$2

where $20 \times 20$3 and $20 \times 20$4 are the arrival times at the two strip ends and $20 \times 20$5 is the signal propagation speed along the strip (Wang et al., 2016). The broader LAPPD history describes the same stripline principle conceptually: $20 \times 20$6, while the average time is proportional to $20 \times 20$7 (Adams et al., 2016).

3. Materials, fabrication, and scale-up

A defining technical feature of LAPPDs is the use of ALD to impart both resistive and secondary-emission properties to glass capillary-array MCP substrates. The collaboration developed tunable resistance coatings consisting of conducting metal nanoparticles embedded in an Al$20 \times 20$8O$20 \times 20$9 matrix, as well as secondary-emission coatings such as MgO and Al$10^7$0O$10^7$1 (Adams et al., 2016). In the Argonne $10^7$2 prototypes, ALD coats both the MCPs and the glass spacers, so that the MCPs and spacers together form an internal HV divider (Wang et al., 2016).

Large-area substrate production was itself a major advance. Incom developed 8-inch-square plates with an open-area ratio exceeding $10^7$3, approximately 80 million pores each, based on $10^7$4m inner-diameter drawn glass capillaries, sliced on an $10^7$5 bias and finished into 1.2 mm-thick wafers with aspect ratio about $10^7$6 (Adams et al., 2016).

Argonne’s Small Single Tube Processing System (SmSTPS) was organized into four chambers: a vacuum load-lock, bake-and-scrub chamber, photocathode deposition chamber, and sealing chamber. The authors stated that the system was expected to produce about one detector per week with yield $10^7$7 (Wang et al., 2016). This smaller-scale system functioned as a process-development bridge between the original collaboration R&D and later commercialization.

A separate scale-up path was the “air-transfer” production method for large-area MCP-PMTs. The reported workflow was: deposit a thin Sb precursor layer on the inside of the window before assembly; assemble the tile with the MCP stack, support buttons, readout anode, and metalized sealing surfaces; clamp the window to the tile base in a dual-vacuum fixture; bake the module while it is connected to an inner UHV manifold, allowing a hermetic indium-alloy seal to form; after sealing and leak-checking, reheat the sealed module to about $10^7$8C and introduce alkali vapor to form the photocathode in situ; after photocathode synthesis and characterization, pinch off the tubulations to isolate the finished detector (Angelico et al., 2020).

In that air-transfer program, the large-perimeter window-body interface was sealed with a capillary-wick indium-alloy solder seal. Both the window and tile sealing surfaces were metalized, typically with 200 nm of NiCr followed by 200 nm of Cu, deposited without vacuum break. The window was clamped at a preset gap width, typically $10^7$9, and molten In-Ag alloy was wicked into the gap by capillary action. The In-Ag eutectic solder has a melting point of $15\%$0C, while during thermal cycling the seal forms around $15\%$1C (Angelico et al., 2020).

The same study also demonstrated in-situ photocathode synthesis in a large-area low-aspect-ratio volume, but the reported demonstration reached only about $15\%$2 QE and non-satisfactory uniformity for a commercial device. The stated causes were non-uniform window temperature during synthesis and an overly thick Sb precursor layer (Angelico et al., 2020). A separate important result was that the micro-channel plates recovered their functionality after cathode synthesis: during cesiation the MCP stack initially experienced a large drop in resistance, but after the alkali vapor was valved off the plates recovered, and the gain and noise fell back close to initial values within hours of operation (Angelico et al., 2020).

4. Performance envelope

The LAPPD literature reports a broad but internally consistent performance envelope spanning collaboration targets, small-form-factor prototypes, commercialized tiles, Gen-I single-photon tests, Gen-II pad-readout measurements, and beam tests. Reported values depend strongly on device generation, photocathode quality, high-voltage configuration, readout electronics, and whether the measurement is in single-photoelectron, multi-photoelectron, laser, or beam conditions (Lyashenko et al., 2019, Foster et al., 2024, Bhattacharya et al., 2023, Korpar et al., 2 Dec 2025).

Source	Configuration	Reported figures
(Wang et al., 2016)	Argonne $15\%$3 cm prototype	$15\%$4 ps single PE; $15\%$5 ps multi-PE; $15\%$6 mm position
(Lyashenko et al., 2019)	Recent stripline LAPPDs	gain up to $15\%$7; $15\%$8 ps RMS; 3.2 mm along strips; 0.76 mm across strips
(Shin et al., 2022)	Gen-II $15\%$9 capacitive readout	active area 373 cm$\mu$0; gain $\mu$1; dark $\mu$2 kHz/cm$\mu$3; timing $\mu$4 ps; QE $\mu$5 at 365 nm
(Foster et al., 2024)	Gen-I single-photon characterization	corrected TTS $\mu$6 ps; 4.4 mm along stripline; 2.0 mm across stripline
(Bhattacharya et al., 2023)	CERN PS Cherenkov beam test	about $\mu$7 ps rms single PE; $\mu$8 ps at about 400 mV
(Korpar et al., 2 Dec 2025)	Gen-II laser characterization	prompt peak $\mu$9 ps for #162; $6 \text{ cm} \times 6 \text{ cm}$0 ps for #109

Argonne’s $6 \text{ cm} \times 6 \text{ cm}$1 detectors reported quantum efficiency up to $6 \text{ cm} \times 6 \text{ cm}$2 at $6 \text{ cm} \times 6 \text{ cm}$3 nm, transit time spread of $6 \text{ cm} \times 6 \text{ cm}$4 ps for single photoelectrons and $6 \text{ cm} \times 6 \text{ cm}$5 ps for multi-photoelectrons, and measured position resolution of $6 \text{ cm} \times 6 \text{ cm}$6 mm with beam-size and electronics limitations included (Dharmapalan et al., 2016). The 2016 prototype paper explicitly described these results as achieved “without optimization,” with gain rising from about $6 \text{ cm} \times 6 \text{ cm}$7 to $6 \text{ cm} \times 6 \text{ cm}$8 as HV increased (Wang et al., 2016).

Commercially produced stripline LAPPDs later reported electron gains of up to $6 \text{ cm} \times 6 \text{ cm}$9, dark noise rates $\times$0 Hz/cm$\times$1 at a gain of $\times$2, single-photoelectron timing resolution of $\times$3 picoseconds RMS, single-photoelectron spatial resolution along and across strips of 3.2 mm and 0.8 mm RMS respectively, and high uniform bi-alkali photocathodes with mean QE values at 365 nm of $\times$4, $\times$5, $\times$6, and $\times$7 for four reported devices (Lyashenko et al., 2019). That paper explicitly stated that the timing and along-strip spatial resolutions were electronics limited.

The Gen-II $\times$8 device with capacitively coupled readout reported an active area of 373 cm$\times$9, single-photoelectron gain $_2$0, dark count rates $_2$1 kHz/cm$_2$2, single-PE timing resolution $_2$3 ps, position resolution $_2$4, and a bialkali photocathode with QE $_2$5 at 365 nm and spectral response down to $_2$6 nm (Shin et al., 2022). In the same paper, multiple consecutively sealed tiles showed average QE values above $_2$7 at 365 nm, with some tiles demonstrating $_2$8 spatial uniformity (Shin et al., 2022).

Single-photon bench characterization of a Generation I device, LAPPD 104, verified that the single-photoelectron peak is clearly isolated, that mean single-photoelectron gain is greater than $_2$9, that gains approaching $_2$0 are possible at higher voltages, and that the best corrected TTS was $_2$1 ps at 200 V photocathode voltage (Foster et al., 2024). The same study reported an average differential timing resolution of about 50 ps, corresponding to 4.4 mm spatial resolution along the stripline, and 2.0 mm perpendicular to the striplines (Foster et al., 2024).

A CERN PS beam test moved beyond laser-only measurements. For a Generation II LAPPD prototype detecting Cherenkov radiation from high-energy hadrons, the reported single-photoelectron timing resolution was about 80 ps rms, while the approach demonstrated capability to measure time resolutions as fine as 25–30 ps, with the best result $_2$2 ps at about 400 mV pulse height, interpreted as about 23 photoelectrons (Bhattacharya et al., 2023). A later comprehensive Gen-II study reported that the prompt timing peak exhibits a resolution of approximately 30 ps for one device variant and approximately 40 ps for another, while the overall timing structure is explained by photoelectron propagation and back-scattering effects from the MCP input surface (Korpar et al., 2 Dec 2025).

Timing measurements are often expressed through the relation

$_2$3

which was explicitly used in stripline-LAPPD timing studies (Lyashenko et al., 2019). A plausible implication is that direct comparison of quoted numbers across papers requires attention to laser width, digitizer calibration, electronics jitter, readout geometry, and photon statistics.

5. Reconstruction performance and scientific applications

LAPPDs were developed in large part for Cherenkov-light applications in neutrino and collider detectors. In water Cherenkov reconstruction, the key geometrical relation is

$_2$4

and the LAPPD contribution is to provide photon arrival times and hit positions with much finer precision than conventional single-pixel PMTs (Anghel, 2013).

Simulation studies for large water Cherenkov detectors showed that this precision materially affects event reconstruction. For 1.2 GeV muons in a 200 kton detector with 13% photodetector coverage, the use of 100 ps LAPPDs yielded about 3 cm vertex resolution in the direction perpendicular to the track, corresponding to an approximately factor-of-3 improvement as timing sharpened from 2 ns to 0.1 ns (Anghel, 2013). The same work emphasized that vertex reconstruction perpendicular to the track is driven mainly by timing, and that position imaging alone helps only modestly unless timing is also very fast (Anghel, 2013).

ANNIE became the first physics experiment designed to deploy LAPPDs. In its Phase II reconstruction framework, the event vertex and direction are estimated with a maximum-likelihood fit using photon hit times and the Cherenkov cone pattern. The paper defines event-level metrics

$_2$5

and reports that vertex resolution improves from about 38 cm with 128 PMTs alone to 12 cm when 5 LAPPDs are added, while angular resolution improves to about $_2$6 (Drakopoulou, 2018). In the same configuration, the muon energy resolution is 10% and the neutrino energy resolution is 14% at the 68th percentile, and the $_2$7 resolution is considerably improved across all bins (Drakopoulou, 2018).

A related ANNIE R&D paper described the deployed LAPPDs as 20 cm $_2$8 20 cm flat-panel photodetectors with a borosilicate glass window, a multi-alkali $_2$9 photocathode, and 28 readout strips per device, read out from both ends. It quoted roughly 50 ps single-photoelectron timing, gain exceeding $\sim 1.30$0, spatial resolution less than 1 cm across a strip and less than 5 mm along a strip, and a simulation result in which adding 5 LAPPDs improved position resolution from 40 cm to 20 cm at 68% CL (Tiras, 2019).

The first neutrino interactions ever observed with an LAPPD were later reported by ANNIE. In that deployment, a single Gen-I stripline LAPPD was mounted at a central position on the downstream wall of the tank, synchronized to the experiment-wide trigger and DAQ system, and operated in a fully integrated underwater package (Adams et al., 14 Aug 2025). After synchronization and event selection, the reported cut flow reached 98.7% purity with 451 estimated beam events in the spill window, establishing the first observation of beam neutrinos with an LAPPD (Adams et al., 14 Aug 2025). The paper also showed that arrival time versus distance across the detector face exhibited about a 1 ns gradient across the surface, consistent with the expected Cherenkov geometry from the MRD-reconstructed muon track (Adams et al., 14 Aug 2025).

High-rate gas Cherenkov applications form a second major branch. In Jefferson Lab Hall C tests, a Gen-I internal-stripline LAPPD detected Cherenkov signals at rates up to about $\sim 1.30$1 kHz/cm$\sim 1.30$2, and both the LAPPD and MaPMT reference system could separate single-electron and pair-production events (Peng et al., 2020). A later SoLID-oriented test of a coarse-pixelated Gen-II LAPPD, with $\sim 1.30$3 pixels of $\sim 1.30$4, reported successful operation up to 21 kHz per pixel, the ability to separate single-electron events from pair-production events while rejecting background, and Cherenkov disk images in both CO$\sim 1.30$5 and C$\sim 1.30$6F$\sim 1.30$7 (Xie et al., 2024).

The applications cited across the literature include electromagnetic calorimeter timing layers, photon-based neutrino detectors, RICH and DIRC instrumentation, high-energy collider experiments, medical imaging systems including TOF-PET, and nuclear non-proliferation systems (Shin et al., 2022, Lyashenko et al., 2019). This suggests that the LAPPD platform is best understood as a general-purpose fast imaging photosensor rather than as a device tied to one detector class.

6. Specialized variants, readout challenges, and operational issues

A specialized offshoot of the LAPPD program is cryogenic adaptation for noble-liquid and noble-gas detectors. Argonne reported a $\sim 1.30$8 all-glass MCP-based photodetector with a pair of ALD-coated MCPs, glass grid spacers, silk-screened anode stripline readout, bi-alkali photocathode, glass-frit bonded sidewall, and indium-sealed top window, explicitly aimed at liquid argon and cryogenic gas environments (Dharmapalan et al., 2016). The indium press seal and glass-frit bond remained intact after multiple liquid-nitrogen cycles; devices were immersed in liquid nitrogen at 77 K for up to 4 days, and 4 devices were tested with 1 failure attributed to a pre-existing hair-line crack in the glass sidewall (Dharmapalan et al., 2016). The same program reported MCP resistance-versus-temperature studies, cryogenic-tuned ALD recipes under evaluation for gain and stability, plans for a TPB-coated window, MgF$\sim 1.30$9-window detectors for direct VUV detection, and bare-MCP photodetectors in which the MCP surface itself is ALD-coated to act as a photocathode (Dharmapalan et al., 2016).

Large-area stripline readout introduced a distinct reconstruction problem: near-simultaneous arrival of multiple photons can create ambiguity in matching left- and right-propagating stripline pulses. One proposed solution was a maximum a posteriori framework with a likelihood matrix

$20 \times 20$00

combining amplitude similarity, timing consistency, and across-strip position consistency (Jocher et al., 2018). In simulated scintillator and water-Cherenkov detectors, this method achieved useful efficiency and timing/position recovery, but the reported results also showed degradation with increasing occupancy (Jocher et al., 2018).

Capacitive readout introduces its own operational complexities. Tests with Gen-II devices emphasized that the glass backplane produces substantial charge spread across neighboring pixels. A PETSYS ToFPET2 ASIC study concluded that pads smaller than about 6 mm do not collect enough signal for simple binary readout, and that below that scale one should rely on charge-center reconstruction; after TOA–TOT time-walk correction the reported timing reached about 80 ps sigma, with millimeter-scale FWHM spatial resolution from centroiding (Seljak et al., 2022). A Jefferson Lab single-photoelectron study of Incom LAPPD 38 similarly found that single-photoelectron signals were clearly detected if a tight collimation of photons impinging on the photocathode was used compared to the pixelation of the charge collection signal board (Malace et al., 2021).

Recent Gen-II operational studies have made these effects more explicit. Charge sharing and electronic cross-talk were measured on the standard $20 \times 20$01 pixelated readout board with $20 \times 20$02 pixels, and cross-talk was reported to fall below 1% beyond a 1-pixel neighbor (Stradleigh et al., 18 Jun 2026). The same paper observed resonant peaks at 180 MHz and 550 MHz when electrical pulses were injected into the readout board, suggesting that the detector/readout assembly behaves as a resonant cavity (Stradleigh et al., 18 Jun 2026). Dark-count behavior after top-MCP voltage changes was described by a tri-exponential relaxation,

$20 \times 20$03

after a spike from about 200 Hz to 28 kHz and recovery over about half a day (Stradleigh et al., 18 Jun 2026). The same study also identified delayed features at approximately 60 ns and 110 ns, interpreted in terms of backscatter, ion-related processes, and readout coupling (Stradleigh et al., 18 Jun 2026).

Analytical modeling has begun to formalize these effects. A 2025 Gen-II characterization used the Shockley–Ramo theorem,

$20 \times 20$04

to describe induced charge on segmented external electrodes, and developed models for prompt timing, MCP-input-surface back-scattering, and resistive-anode backscattering (Korpar et al., 2 Dec 2025). That work reported that smaller pads increase spatial resolution but dramatically increase charge sharing, while the dielectric constant of the back plate has only a modest effect compared with geometry (Korpar et al., 2 Dec 2025). This suggests that readout segmentation, clustering, and waveform reconstruction must be co-designed with the detector geometry rather than treated as downstream electronics choices.

Taken together, these specialized studies address a common misconception: LAPPDs are not simply “large fast PMTs.” They are coupled detector–materials–packaging–electronics systems in which photocathode synthesis, ALD-tuned MCP resistivity, hermetic sealing, anode geometry, digitizer architecture, and waveform reconstruction all directly affect measured timing, spatial resolution, dark behavior, and usability in dense optical environments (Adams et al., 2016, Shin et al., 2022, Stradleigh et al., 18 Jun 2026).