---
title: Layered Water Cherenkov Tank Design
url: https://www.emergentmind.com/topics/layered-water-cherenkov-tank-design
type: topic
---

# Layered Water Cherenkov Tank Design

Layered water Cherenkov tank design denotes a class of water-Cherenkov detector architectures in which the water volume is intentionally partitioned into regions with different optical visibility or different particle-response functions, rather than used as a single undivided Cherenkov medium. In the most explicit form, a single tank is divided into two optically separated vertical layers so that the upper layer is electromagnetic-dominated and the lower layer is muon-enriched; in related veto and shielding applications, the same tank can also be partitioned into an outer passive water border and an inner reflector-defined active Cherenkov region [2508.17461] [2406.12870]. Across cosmic-ray, gamma-ray, neutrino, and rare-event detectors, the common design objective is to decouple prompt electromagnetic response, penetrating-particle tagging, shielding, and calibration within one hydraulic system, while controlling light transport, dynamic range, and long-term water quality [2209.09305] [1908.11532].

## 1. Physics drivers and detector logic

The principal motivation for a layered tank is differential sensitivity to the electromagnetic and muonic components of extensive air showers. In the layered concept studied for next-generation air-shower detectors, the top layer is intended to absorb and record most of the electromagnetic component, while the bottom layer becomes relatively enriched in signals from through-going muons [2508.17461]. The physical basis is explicit: the electromagnetic component \((\gamma, e^\pm)\) is “highly attenuated and absorbed” in the top layer, with an attenuation length of about \(\SI{40}{cm}\), whereas muons are typically through-going minimum-ionizing particles [2508.17461]. This same logic underlies the earlier “Layered Surface Detector” concept, which exploits the fact that \(e^\pm\) are absorbed in a few cm of water, \(\gamma\) deposit energy over roughly one radiation length, \(X_0 \sim 36\ \mathrm{cm}\), and muons traverse several meters of water [1405.5699].

For ground-based gamma-ray astronomy, the design objective is not only component separation but gamma/hadron discrimination. In the SWGO double-layer configuration, the upper chamber is optimized for timing and local energy measurement, while the lower chamber is dedicated to muon tagging; hadronic showers contain many more muons than gamma-ray showers, so lower-layer muon-like signals provide a direct background discriminator [2209.09305]. A related formulation appears in the SWGO single-unit design study, which states that the lower layer helps in gamma/hadron discrimination because muons are more abundant in hadronic showers and can cross the upper layer, reaching the lower layer where they are measured [2205.02148].

In ultra-high-energy cosmic-ray composition studies, the same partition enables event-by-event mass sensitivity. The LSD design was proposed to provide an independent measurement of the muonic and electromagnetic components of extensive air showers, and the paper states that, according to EPOS-LHC or QGSJetII-04, the discriminating power between iron and hydrogen primaries reaches Fisher values of \(\sim 2\) or above for energies in excess of \(10^{19}\) eV with an array layout similar to that of the Pierre Auger Observatory [1405.5699]. A closely related argument is made in the later prototype study, which states that improving the sensitivity to ultra-high-energy gamma rays and to the mass composition of ultra-high-energy cosmic rays requires the ability to measure the muonic content of air showers [2508.17461].

In underground veto systems, the motivation is different but structurally analogous. The COSINUS water tank combines a passive shield with an active Cherenkov veto; the water attenuates ambient gamma and neutron backgrounds, while the instrumented inner region tags muons and muon-induced secondaries [2406.12870]. Here the central design tension is between keeping enough passive water near the wall to suppress ambient gamma triggers and preserving sensitivity to shower-like events concentrated near boundaries [2406.12870].

## 2. Canonical geometries and regional partitioning

The most explicit layered geometry currently documented is the two-layer cylindrical air-shower tank. In the prototype tanks deployed at the Pierre Auger Observatory, the tank diameter is \(\SI{3.6}{m}\), the total height is \(\SI{1.2}{m}\), and a reflective separation layer at \(\SI{0.8}{m}\) above the bottom divides the detector into a bottom layer of \(\SI{80}{cm}\) and a top layer of \(\SI{40}{cm}\), i.e. a 1:2 top:bottom thickness ratio [2508.17461]. The original Auger-based LSD design adopts the same total height and interface location, placing the separator \(80\,\mathrm{cm}\) above the bottom so that the tank is divided into a \(40\,\mathrm{cm}\) top layer and an \(80\,\mathrm{cm}\) bottom layer [1405.5699].

A second, deeper double-layer geometry was developed for SWGO. The preferred reference design is a cylindrical tank of \(3.8\,\mathrm{m}\) diameter with an upper chamber depth of \(2.5\,\mathrm{m}\) and a lower chamber depth of \(0.5\,\mathrm{m}\), each chamber instrumented with one 8-inch PMT [2209.09305]. The paper also studied alternatives with upper depths ranging roughly from \(1.7\) m to \(2.9\) m and lower depths around \(0.7\) m, but identified \(2.5\) m upper plus \(0.5\) m lower as the preferred reference geometry because the upper chamber must both contain electromagnetic cascades and shield the lower chamber from punch-through [2209.09305]. A related SWGO design survey considered double-layer circular, hexagonal, and square tanks with upper-layer height determined by Cherenkov-angle geometry plus \(1\) m of additional water for gamma conversion, and lower-layer heights of \(0.5\), \(0.75\), and \(1.0\) m [2205.02148].

Layering is not always realized as two active vertical chambers. COSINUS introduces an “optically invisible region,” also called the dead layer or optical dead layer, in which a peripheral water region remains physically present for passive shielding while being excluded from the active optical volume [2406.12870]. The dead layer is created by mounting the reflector and PMTs inward from the steel wall, so the water between reflector and tank wall is outside the PMTs’ effective optical view [2406.12870]. The study examined dead-layer thicknesses of \(0, 10, 20, 30, 40, 70,\) and \(100\) cm and recommends a dead layer on all sides of the tank—top, bottom, and wall—of \(30\)–\(40\) cm in the final design [2406.12870].

A broader functional generalization appears in lake-deployed detectors. The lake concept is explicitly based on a double-layer water Cherenkov detector in which the lower chamber is focused on muon tagging and the upper chamber provides the timing-rich electromagnetic response used for air-shower detection and angular reconstruction [2504.10043]. In the reference simulated geometry, the detector diameter is \(3.6\,\mathrm{m}\) and the upper-to-lower chamber height ratio is \(2.5:0.5\,\mathrm{m}\), giving a total active depth of \(3.0\,\mathrm{m}\) [2504.10043]. In addition, lake water above and around the detector acts as passive top overburden and lateral shielding, so the total system functions as a multi-zone detector rather than merely a two-compartment tank [2504.10043].

By contrast, some large water Cherenkov systems often associated with “layered” design are not vertically layered in this sense. The LHAASO WCDA consists of three ponds with an effective water depth of \(4.5\,\mathrm{m}\), horizontally segmented into \(5 \times 5\,\mathrm{m}^2\) cells by black curtains to prevent penetration of light from neighboring cells; the paper is explicit that this is optical segmentation of repeated modular water cells, not stacked active media or internal horizontal partitions [2009.12737]. That distinction addresses a recurring misconception: horizontal optical segmentation and vertical layering solve different detector problems.

## 3. Optical boundaries, reflectivity, and sensor placement

Optical separation is the defining condition of a layered tank. In the two-layer air-shower prototypes, the separator is repeatedly described as a reflective separation layer and as an optically separated barrier; the whole decomposition method depends on distinct layer responses [2508.17461]. In the LSD, the separator is a horizontal reflective layer forming two independent light-tight volumes, and in the prototype it was implemented using a Tyvek laminate inserted in one of the Auger liners [1405.5699].

Wall reflectivity is one of the main performance knobs. In the SWGO double-layer study, “black” is modeled as polypropylene-like reflectivity about \(10\%\) at \(450\,\mathrm{nm}\), while “white” is Tyvek-like diffuse reflectivity about \(92\%\) at \(450\,\mathrm{nm}\) [2209.09305]. For a \(3.8\,\mathrm{m}\) diameter, \(2.5\,\mathrm{m}\) deep upper chamber with an 8-inch PMT, the average number of photoelectrons per \(20\,\mathrm{MeV}\) gamma-ray energy is reported as \(2.9\) pe/20 MeV for an all-white chamber, \(1.9\) for white with black top, \(1.6\) for white with black bottom, \(1.4\) for white with black bases, and \(1.0\) for an all-black chamber [2209.09305]. The same study reports that making one or both of the top and bottom surfaces black in an otherwise white upper chamber reduces the late tail of the photon arrival-time distribution by about \(30\)–\(40\%\) at \(80\,\mathrm{MeV}\) [2209.09305]. This establishes the standard tradeoff: reflective walls increase light yield and low-energy efficiency, while dark walls preserve prompt timing.

The SWGO size-and-shape study reaches the same conclusion in a more general form. Using Tyvek in the upper layer yields about \(3\times\) more photoelectrons than the non-reflective polypropylene case for 1 GeV particles, across circular, hexagonal, and square geometries, but reflective walls worsen first-photon timing, reaching \(\sim 18\) ns for 10 MeV particles in the most delayed case [2205.02148]. A plausible implication is that upper-layer optics must be chosen jointly with the intended timing reconstruction method, not as a purely photostatistical optimization.

PMT placement is equally geometry-dependent. In the Auger-based layered prototypes, the upper layer retains the original arrangement of three 9-inch PMTs viewing from above, while the lower layer is read out by one additional PMT dedicated to the lower optical volume [2508.17461]. In the LSD concept, the same arrangement is described as three 9-inch PMTs in the upper segment plus one 9-inch PMT in a central cylinder for the lower segment [1405.5699]. In the SWGO reference double-layer tank, the upper-chamber PMT is one 8-inch PMT placed centrally at the bottom of the upper chamber and facing upward, while the lower-chamber PMT is one 8-inch PMT placed centrally at the top of the lower chamber and facing downward [2209.09305]. The same study reports that letting the lower PMT base protrude about \(10\,\mathrm{cm}\) into the upper chamber increases the mean lower-chamber muon light yield by a factor of about \(1.6\) relative to leaving the full PMT body and base inside the lower chamber [2209.09305].

In veto configurations, low photocathode coverage can be compensated by reflectivity and multiplicity logic. The JUNO-TAO veto prototype uses 16 three-inch PMTs in a \(1\,\mathrm{m}^3\) Tyvek-lined tank, arranged as four viewing up on the bottom, four viewing side on the corner, four viewing inside at middle height, and four viewing down from the top [2503.13249]. For the full TAO veto, the water tank thickness is \(1.2\,\mathrm{m}\), the PMT count is 300 three-inch tubes, and the PMT coverage is approximately \(1\%\) [2503.13249]. The lake-based double-layer design pushes the optical split further: the upper chamber uses low reflectivity for prompt timing, while the lower chamber uses Tyvek 1082D with simulated reflectivity \(0.92\) at \(410\,\mathrm{nm}\) to maximize muon-tagging efficiency [2504.10043].

## 4. Signal decomposition, calibration, and trigger logic

The mathematical core of layered-tank design is linear signal decomposition. In the prototype layered detectors, the layer signals are written as
\[
\begin{split}
& S_{\textrm{top}} = a \cdot S_{\gamma,e^\pm} + b \cdot S_{\mu} \\
& S_{\textrm{bottom}} = (1-a) \cdot S_{\gamma,e^\pm} + (1-b) \cdot S_{\mu}
\end{split}
\]
with coefficients that depend mainly on geometry; the paper quotes earlier simulation estimates of
\[
a \approx 0.6, \qquad b \approx 0.4
\]
and states that these coefficients are largely independent of primary species, energy, direction up to \(60^\circ\), and hadronic model [2508.17461]. The LSD presents the same response-matrix idea in matrix form and, for \(a=0.6\) and \(b=0.4\), gives the explicit inversion
\[
S_{EM}=3S_{top}-2S_{bot}, \qquad S_\mu=3S_{bot}-2S_{top}
\]
[1405.5699].

The separator height in the LSD is itself derived from this linear model. With total tank height \(H\), upper-layer thickness \(h\), and water radiation length \(X_0\), the fractions of electromagnetic and muon signal seen in the top layer are approximated as
\[
a = 1-e^{-h/X_0} \qquad \text{and} \qquad b = \frac{h}{H},
\]
with response-matrix determinant
\[
\mathcal{D}=a-b
\]
maximized for
\[
h = X_0 \ln(H/X_0).
\]
For \(H=120\,\mathrm{cm}\) and \(X_0=36\,\mathrm{cm}\), the optimum is \(h \approx 43\,\mathrm{cm}\), which supports the adopted \(40\,\mathrm{cm}\) top layer and \(80\,\mathrm{cm}\) bottom layer in the Auger-sized tank [1405.5699].

Calibration diverges sharply between upper and lower layers. In the fielded two-layer prototypes, one Vertical Equivalent Muon is defined as the charge deposited by a muon crossing the center of the tank vertically, and the bottom layer shows a clear atmospheric-muon peak around 135 FADC counts, making the calibration “clean and precise” [2508.17461]. The top layer is harder: its shorter track length and stronger electromagnetic contamination cause the charge deposits from electromagnetic particles and muons to overlap substantially, so the paper proposes a coincidence selection between the bottom PMT and a top PMT to make the top-layer muon peak significantly more prominent [2508.17461]. The paper identifies top-layer muon-peak extraction as the main calibration challenge specific to the layered design [2508.17461].

Trigger logic is similarly layer-dependent. In the COSINUS veto, a PMT is considered triggered if its event-integrated PE count reaches threshold, and the veto is then based on an \(L\)-fold coincidence among PMTs [2406.12870]. The probability of at least one accidental trigger in a time window \(\Delta T\) is written as
\[
p(\Delta T)=1-e^{-R_B\Delta T},
\]
and, for independent dark counts in \(N\) PMTs, the \(L\)-fold probability is
\[
P_{d,L}^{(N)} = C_{N,L}\left[1-e^{-\Delta T R_d}\right]^L e^{-(N-L)\Delta T R_d},
\qquad
C_{N,L}\equiv \frac{N!}{L!(N-L)!}.
\]
The study concludes that multiplicity is a more efficient suppressor of accidental rate than increasing dead-layer thickness [2406.12870].

The JUNO-TAO prototype implements that principle in software. It runs in triggerless mode, then forms events within coincidence windows and applies the practical muon-selection condition: fired PMTs \(\ge 3\), each fired channel charge \(\ge 0.3\) p.e., and total event charge \(\ge 10\) p.e. [2503.13249]. The preferred coincidence window after filling is \([-50,+50]\,\mathrm{ns}\), and for 16 PMTs at 5 kHz dark rate each, the random coincidence rate for a 100 ns window is \(\sim 1.83\,\mathrm{Hz}\) [2503.13249].

## 5. Performance, optimization, and observatory-level consequences

Layered designs are evaluated by different metrics in different experimental contexts. In cosmic-ray composition, the LSD reports reconstructed muon-size resolution at 1000 m of \(20\%\) for proton and \(16\%\) for iron at 10 EeV, improving to about \(10\%\) for both at 70 EeV [1405.5699]. The same study states that timing alone gives an \(X_{\max}\) resolution around \(50\ \mathrm{g/cm^2}\), potentially \(30\ \mathrm{g/cm^2}\) with time-shape information, and that proton–iron Fisher separation is \(\sim 2\) or above above \(10^{19}\,\mathrm{eV}\) [1405.5699]. The later prototype paper does not rederive those numbers but recalls that earlier results showed muon-number relative resolutions as good as 10–15% in a single detector unit [2508.17461].

In SWGO-oriented gamma-ray arrays, the primary figure of merit is gamma/hadron separation combined with threshold and angular resolution. The preferred \(3.8\,\mathrm{m}\) diameter, \(2.5\,\mathrm{m}\) upper, \(0.5\,\mathrm{m}\) lower configuration yields background rejection power of about \(10^3\) at reconstructed energies of a few TeV while keeping high gamma efficiency, and entirely white double-layered WCDs achieve background efficiency of \(3\times10^{-4}\) with good gamma efficiency [2209.09305]. The same study reports several arc-minutes angular resolution at 10 TeV in its idealized reconstruction setup, with only modest differences between wall-reflectivity choices once the timing PDFs are modeled correctly [2209.09305]. This directly counters the common assumption that reflective optics necessarily spoil direction reconstruction.

Single-tank optimization studies clarify the geometry trade space. In the shape-and-size scan for SWGO, performance worsens as tank width increases unless PMT coverage scales accordingly; Circular-DLT and Hexagonal-DLT have similar performance and both outperform Square-DLT in PE yield, detection efficiency, and timing [2205.02148]. The paper also states that in the lower layer, 10 and 100 MeV electrons and gamma rays are rarely detected, confirming the intended role of the lower layer as a muon-sensitive region [2205.02148]. The prototype-depth study likewise finds that decreasing diameter from \(3.6\) m to \(3.0\) m increases photoelectron yield, while increasing top-layer thickness should improve electromagnetic absorption and therefore electromagnetic/muon separation [2508.17461].

Veto systems exhibit a different optimization. COSINUS predicts a final configuration with veto efficiency \(99.63 \pm 0.16\%\) for muon events and \(44.4 \pm 5.6\%\) for shower events, reducing the cosmogenic neutron background to \(0.11 \pm 0.02\ \mathrm{cts\cdot kg^{-1}\cdot year^{-1}}\), corresponding to less than one background event in the region of interest for the whole COSINUS-\(1\pi\) exposure of 1000 kg·days [2406.12870]. The same work shows that each extra PMT required in coincidence lowers the ambient-gamma trigger rate by about an order of magnitude, whereas every additional 10 cm of dead layer lowers it by about a factor of two, with diminishing returns beyond 40 cm [2406.12870]. This is one of the clearest general design lessons available for passive/active layered tanks.

At the array scale, layered stations alter deployment strategy. The GCOS study uses the minor axis of the shower-footprint ellipse at 90% single-station trigger probability to infer a spacing of about \(2.2\,\mathrm{km}\) for full efficiency above \(10\,\mathrm{EeV}\), and notes that with spacing below \(2.5\,\mathrm{km}\), more than 15,000 detectors are required to cover \(\SI{60\,000}{km^2}\) [2508.17461]. This suggests that station simplicity, durability, and calibration stability become as important as per-station discrimination power.

## 6. Engineering implementations, operational constraints, and unresolved issues

The engineering realization of layered tanks is highly diverse. The two-layer Auger prototypes have been operating since 2014 for more than 10 years; among five prototypes, two remained in continuous operation, with only a recalibration intervention in 2016 for Clairon Jr., an electronics fix in 2021 for Guapa Guerrera, and both upgraded to new Auger electronics in April 2023 [2508.17461]. The paper repeatedly stresses durability, low maintenance, and suitability for huge arrays, but also notes that top-layer calibration remains unresolved in field data [2508.17461].

Floating and submerged designs introduce additional layers of mechanical and environmental complexity. The lake concept uses light-tight, water-tight floating bladders, polyethylene floater and stretcher rings, and a central dual-PMT assembly suspended from an access hatch; the bladder material must provide water tightness, light tightness, UV tolerance, mechanical durability, and low contamination, with multi-layer polyethylene liners identified as the best candidates and Tyvek 1082D as the working reflective inner surface for the lower chamber [2504.10043]. Prototype tests suggest an optimum fill level of about 90%, leaving slack in the walls to tolerate wave-induced deformation [2504.10043]. This implies that in compliant layered systems, internal partitions should avoid rigid structures that fight bladder motion unless exceptionally compliant.

Submerged membrane-contained neutrino tanks provide a different form of layering. CHIPS uses surrounding water as shielding and hydrostatic support, with the detector implemented as a vertical cylinder enclosed in XR-5 geomembrane liner, separated from external pit water, and supported by stainless-steel end-cap frames, buoyancy tubes, and Dyneema cables [2401.11728]. The tank does not implement a formal inner detector/outer veto annulus in the deployed prototype, but the architecture is a layered stack of pit water, support floats, steel frame, geomembrane hull, purified detector water, and instrumented PMT planes [2401.11728]. A plausible implication is that hydrostatic pressure balancing can substitute for a rigid pressure vessel when the external environment is itself water.

Water chemistry becomes a first-order design variable once chemical layering is introduced. EGADS, although not a structurally layered detector, is the principal demonstration that \(0.2\%\) gadolinium sulfate by mass can be stably maintained in a 200-ton stainless-steel water Cherenkov tank with no detectable Gd loss after 2.5 years and 650 full water-volume passes through the purification system [1908.11532]. The project also shows that dissolved Gd sulfate quickly becomes homogeneously distributed and exhibits no stratification [1908.11532]. That result is especially important for layered design because it implies that persistent doped and undoped zones are unlikely to survive ordinary circulation without physical separation.

Several limitations recur across the literature. The fielded layered air-shower papers do not yet provide a complete industrialized solution for top-layer calibration, inter-layer optical cross-talk control, or final optimization of layer depths for future observatories [2508.17461]. The SWGO design studies provide strong comparative trends but do not fix one universally optimal geometry for all array layouts and cost models [2205.02148] [2209.09305]. LHAASO provides essential contrast by showing that not every segmented water Cherenkov architecture is a layered tank; its WCDA is horizontally segmented into black-curtain-isolated cells with 4.5 m depth and bottom-centered PMTs, but it does not describe internal vertical layers, wall materials, roof, or optical zoning by height [2009.12737]. That absence underscores a final point of terminology: “layered” should be reserved for designs in which vertical or functionally distinct water regions are intrinsic to detector response, not merely for any segmented water Cherenkov system.

Source: https://www.emergentmind.com/topics/layered-water-cherenkov-tank-design