LiquidO: Novel Scintillation with Stochastic Confinement

• LiquidO is a novel radiation detection concept that uses intentionally opaque scintillators and a dense lattice of wavelength-shifting fibers to confine scintillation light in localized “light balls”.
• The technique employs stochastic light confinement to preserve event topology, enabling high-resolution energy spectroscopy, vertex reconstruction, and particle tracking in MeV-scale events.
• Ongoing research has advanced LiquidO from proof-of-principle detectors to demonstrators with sub-centimetre spatial resolution and timing-based Cherenkov/scintillation separation.

Searching arXiv for LiquidO-related papers to ground the article in current literature. LiquidO is a radiation-detection concept in which an intentionally opaque scintillator confines scintillation light by repeated scattering, while a dense lattice of wavelength-shifting (WLS) fibres reads out the resulting localized “light ball” near each energy deposition. In contrast to transparent liquid-scintillator or water-Cherenkov detectors, where photons propagate ballistically to photosensors on the periphery, LiquidO uses stochastic light confinement to preserve event topology in the spatial pattern of fibre signals. Since its introduction as an opaque-detector concept for neutrino physics in 2019, the approach has progressed through proof-of-principle validation, litre-scale spectroscopy and point-like vertex reconstruction, ten-litre imaging studies, sub-millimetre muon tracking in a cubic demonstrator, and incorporation into the scatterer design of an MeV Compton telescope (Cabrera et al., 2019, Collaboration, 2024, Collaboration et al., 4 Mar 2025, Collaboration et al., 18 Jul 2025, Collaboration et al., 28 Feb 2025).

1. Conceptual basis and historical development

LiquidO was introduced as a deliberate inversion of the standard transparency requirement in liquid scintillator detectors. The foundational formulation specifies a highly scattering medium with scattering length $\lambda_s \sim 1$–$5\ \mathrm{mm}$ and absorption length $\lambda_a \sim 1$–$10\ \mathrm{m}$, combined with a dense three-dimensional grid of WLS fibres, typically with $\sim 1\ \mathrm{cm}$ spacing. In that regime, photons execute a random walk in the neighbourhood of their creation and are effectively confined to a localized sphere, or “light-ball,” around each ionization point, producing self-segmentation without dead material or opaque barriers (Cabrera et al., 2019).

The first experimental validation reported in the 2019 concept paper used a $0.25\ \mathrm{L}$ proof-of-principle detector with three Kuraray B-3 fibres and a mono-energetic $1\ \mathrm{MeV}$ electron beam. In the opaque state, the lowest fibre collected $\sim 2\times$ more light, the middle $\sim 1.4\times$ more, and the top $\sim 0.5\times$ less than in the transparent control, while the PMT signal collapsed; the reported interpretation was that light was not simply absorbed but stochastically confined around the energy deposit (Cabrera et al., 2019).

Subsequent work diversified both medium chemistry and detector architecture. A 2024 technical report characterized a $5\ \mathrm{mm}$0, 32-channel prototype based on opaque water-based liquid scintillator (oWbLS), emphasizing energy spectroscopy, optical modelling, and point-like vertex reconstruction (Collaboration, 2024). In 2025, two complementary ten-litre reports described “Mini-LiquidO,” a NoWaSH-based prototype in which the scattering length is tuned by temperature, demonstrating centimetre-scale light confinement and timing-based separation of Cherenkov and scintillation components (Collaboration et al., 4 Mar 2025, Navas-Nicolás et al., 14 Mar 2025). A further 2025 study reported event-by-event muon tracking in a $5\ \mathrm{mm}$1 cubic detector using a wax-based opaque scintillator with scattering length of approximately $5\ \mathrm{mm}$2 (Collaboration et al., 18 Jul 2025). In parallel, LiquidO was adopted as the scatterer technology in COCOA, a compact MeV Compton camera (Collaboration et al., 28 Feb 2025).

2. Optical transport and stochastic light confinement

The optical transport model underlying LiquidO is formulated in terms of scattering and absorption mean free paths,

$5\ \mathrm{mm}$3

with the anisotropy of Mie-like scattering accommodated through the reduced scattering coefficient $5\ \mathrm{mm}$4 and reduced scattering length $5\ \mathrm{mm}$5 in the oWbLS prototype analysis (Collaboration, 2024). A more general description uses the steady-state radiative transfer equation (RTE),

$5\ \mathrm{mm}$6

where $5\ \mathrm{mm}$7 is the radiance and $5\ \mathrm{mm}$8 is the scattering phase function (Collaboration, 2024, Collaboration et al., 18 Jul 2025).

On scales much larger than the scattering length, the ten-litre studies describe photon transport with a diffusion–absorption equation,

$5\ \mathrm{mm}$9

with $\lambda_a \sim 1$0 and $\lambda_a \sim 1$1. For a point-like source in equilibrium, the reported solution is a multivariate Laplace distribution,

$\lambda_a \sim 1$2

and the confinement fraction within radius $\lambda_a \sim 1$3 is

$\lambda_a \sim 1$4

For the observed parameter range $\lambda_a \sim 1$5 and effective $\lambda_a \sim 1$6–$\lambda_a \sim 1$7, the ten-litre article reports $\lambda_a \sim 1$8 (Collaboration et al., 4 Mar 2025).

The WLS fibres are integral to the confinement mechanism rather than a downstream readout add-on. In the $\lambda_a \sim 1$9 prototype, each Kuraray Y-11 fibre was modelled with core absorption spectrum $10\ \mathrm{m}$0, emission spectrum $10\ \mathrm{m}$1, re-emission quantum yield $10\ \mathrm{m}$2, and trapping efficiency $10\ \mathrm{m}$3 for double-clad fibres. Photons entering the core are absorbed with probability $10\ \mathrm{m}$4, re-emitted at longer wavelength with probability $10\ \mathrm{m}$5, and trapped with $10\ \mathrm{m}$6 (Collaboration, 2024). The broader R&D literature treats trapping efficiencies of $10\ \mathrm{m}$7–$10\ \mathrm{m}$8 as typical for LiquidO-style fibres and expresses the detected photoelectron yield schematically as

$10\ \mathrm{m}$9

where $\sim 1\ \mathrm{cm}$0 is the scintillation yield, $\sim 1\ \mathrm{cm}$1 the geometric fibre coverage, and $\sim 1\ \mathrm{cm}$2 the survival factor against absorption (Klein et al., 2022).

3. Scintillator formulations and detector architectures

LiquidO has been realized with several distinct opaque scintillator formulations. The $\sim 1\ \mathrm{cm}$3 oWbLS detector used an emulsion of $\sim 1\ \mathrm{cm}$4 di-isopropylnaphthalene oil-based scintillator, $\sim 1\ \mathrm{cm}$5 PPO fluor, $\sim 1\ \mathrm{cm}$6 bis-MSB wavelength shifter, and $\sim 1\ \mathrm{cm}$7–$\sim 1\ \mathrm{cm}$8 water by volume; the micelles of the surfactant introduce Mie-like scattering, and the measured light yield was reported as $\sim 1\ \mathrm{cm}$9 photons/MeV (Collaboration, 2024). The ten-litre “Mini-LiquidO” prototype used NoWaSH, defined as LAB+PPO doped with $0.25\ \mathrm{L}$0 paraffin wax; at $0.25\ \mathrm{L}$1 the mixture is transparent with $0.25\ \mathrm{L}$2, while at $0.25\ \mathrm{L}$3 it crystallizes into micro-crystals yielding $0.25\ \mathrm{L}$4 (Collaboration et al., 4 Mar 2025). The muon-tracking demonstrator used a wax suspension in LAB with approximately $0.25\ \mathrm{L}$5 wax, $0.25\ \mathrm{L}$6 PPO and $0.25\ \mathrm{L}$7 POPOP, producing a scattering length of approximately $0.25\ \mathrm{L}$8 (Collaboration et al., 18 Jul 2025).

Architecturally, the earliest proposal assumed a three-dimensional fibre lattice with centimetre-scale pitch and two-end readout for longitudinal localization along each fibre (Cabrera et al., 2019). The $0.25\ \mathrm{L}$9 prototype instead implemented a compact cubic instrument: a $1\ \mathrm{MeV}$0 detector body machined from white 3D-printed polymer, with 32 Kuraray Y-11 fibres arranged in two orthogonal sets of 16, a $1\ \mathrm{MeV}$1 pitch in $1\ \mathrm{MeV}$2 and $1\ \mathrm{MeV}$3, $1\ \mathrm{MeV}$4 spacing in $1\ \mathrm{MeV}$5, read out through a Hamamatsu H12700A 64-pixel MAPMT and two CAEN V1730 digitizers (Collaboration, 2024). The ten-litre instrument used 208 double-clad Kuraray B-3 fibres in two orthogonal horizontal planes, with only 56 of the 208 fibres instrumented and only one fibre end read out, using Hamamatsu S13360-1350PE SiPMs, custom “SiC” front-end amplifiers, USB-driven “SiBB” boards, a WaveCatcher digitiser, and a reference PMT (Collaboration et al., 4 Mar 2025). The $1\ \mathrm{MeV}$6 cube employed a denser $1\ \mathrm{MeV}$7 grid at $1\ \mathrm{MeV}$8 pitch, read at both ends by SiPM arrays through PETsys TOFPET2 electronics (Collaboration et al., 18 Jul 2025). COCOA scales the concept to a hermetic $1\ \mathrm{MeV}$9 scatterer with two interleaved inclined fibre planes, 900 fibres per plane, and 1,800 channels total (Collaboration et al., 28 Feb 2025).

4. Reconstruction, topology, and timing

The simplest published LiquidO reconstruction chain for point-like events is the centre-of-mass (CoM) estimator used in the $\sim 2\times$0 oWbLS prototype:

$\sim 2\times$1

where $\sim 2\times$2 is the signal in fibre $\sim 2\times$3 and $\sim 2\times$4 its coordinate. Because edge effects and anisotropic light distributions bias the raw CoM, a mapping function from CoM to true position was derived by simulating $\sim 2\times$5 events and fitting

$\sim 2\times$6

recovering a linear response within a central fiducial region $\sim 2\times$7 corresponding to $\sim 2\times$8 (Collaboration, 2024).

More elaborate pattern-based reconstruction appears in the muon-tracking study. There, within each horizontal fibre row, the muon $\sim 2\times$9-coordinate is reconstructed as a weighted centre-of-mass from the photoelectrons collected in that row; the rowwise points are then fit to a straight line, and the one-dimensional row resolution is extracted with a leave-one-out residual method. The reported unbiased row resolution is

$\sim 1.4\times$0

where $\sim 1.4\times$1 and $\sim 1.4\times$2 are the standard deviations of residuals with and without the row included in the line fit (Collaboration et al., 18 Jul 2025).

Topological imaging is central to the LiquidO program. The ten-litre papers describe how a localized electron beam spot generates a confined light ball whose radial profile changes with scattering length, while the 2019 concept paper emphasizes that single electrons, Compton $\sim 1.4\times$3 chains, and positron annihilation signatures generate distinguishable fibre patterns (Cabrera et al., 2019, Collaboration et al., 4 Mar 2025). In the COCOA application, interaction localization is obtained by triangulation between inclined U and V fibre planes: the centroid line in one plane and the centroid line in the other define a point of closest approach, giving the reconstructed $\sim 1.4\times$4 for each Compton scatter site (Collaboration et al., 28 Feb 2025).

Timing adds a second discriminant. The ten-litre studies report clear separation of a fast Cherenkov component and a slow scintillation tail when the detector is filled with pure water or pure LAB, and conclude that statistical separation of Cherenkov vs. scintillation light with per-event timing alone is practical in slow scintillators (Collaboration et al., 4 Mar 2025). A parallel report on the same ten-litre program states that custom electronics achieve better than $\sim 1.4\times$5 time resolution per channel and defines a fast fraction $\sim 1.4\times$6 as the integral of the first $\sim 1.4\times$7 divided by the total waveform integral, with $\sim 1.4\times$8 rising from 0 below $\sim 1.4\times$9 to $\sim 0.5\times$0–$\sim 0.5\times$1 at $\sim 0.5\times$2 in LAB (Navas-Nicolás et al., 14 Mar 2025). The $\sim 0.5\times$3 technical report explicitly notes that maximum-likelihood estimation and machine-learning regressors trained on simulated light patterns can in principle handle multiple light centres and anisotropic emission (Collaboration, 2024).

5. Reported performance and validation

LiquidO performance has been reported at several scales, and the published numbers are strongly configuration-dependent.

System	Measurement	Reported value
$\sim 0.5\times$4 oWbLS prototype	Mean position reconstruction error	$\sim 0.5\times$5 at $\sim 0.5\times$6-equivalent; $\sim 0.5\times$7 at $\sim 0.5\times$8-equivalent
$\sim 0.5\times$9 oWbLS prototype	Energy resolution	$5\ \mathrm{mm}$00 at $5\ \mathrm{mm}$01; $5\ \mathrm{mm}$02 at $5\ \mathrm{mm}$03
Ten-litre Mini-LiquidO	Light confinement	$5\ \mathrm{mm}$04 within $5\ \mathrm{mm}$05, $5\ \mathrm{mm}$06 within $5\ \mathrm{mm}$07, $5\ \mathrm{mm}$08 within $5\ \mathrm{mm}$09
Ten-litre Mini-LiquidO	Total light yield	$5\ \mathrm{mm}$10 opaque at $5\ \mathrm{mm}$11; $5\ \mathrm{mm}$12 transparent at $5\ \mathrm{mm}$13
$5\ \mathrm{mm}$14 cube	Average row position resolution	$5\ \mathrm{mm}$15 opaque; $5\ \mathrm{mm}$16 transparent
COCOA scatterer simulation	Spatial resolution (FWHM)	$5\ \mathrm{mm}$17 for $5\ \mathrm{mm}$18 deposits

In the $5\ \mathrm{mm}$19 detector, laser pulses injected at four positions and at two intensities corresponding to $5\ \mathrm{mm}$20 and $5\ \mathrm{mm}$21 yielded mean reconstruction errors of $5\ \mathrm{mm}$22 and $5\ \mathrm{mm}$23, respectively, with the article stating that these values are uniform across the fiducial region. The same study defined energy resolution as $5\ \mathrm{mm}$24 and measured $5\ \mathrm{mm}$25 at $5\ \mathrm{mm}$26 and $5\ \mathrm{mm}$27 at $5\ \mathrm{mm}$28; these values include broadening from laser pulse stability $5\ \mathrm{mm}$29 but not the intrinsic scintillator resolution $5\ \mathrm{mm}$30 at $5\ \mathrm{mm}$31 (Collaboration, 2024).

Model validation has also been reported in detail. For the $5\ \mathrm{mm}$32 prototype, a $5\ \mathrm{mm}$33 pulsed-laser scan was used to fit six parameters—fibre tip position $5\ \mathrm{mm}$34, reduced scattering length $5\ \mathrm{mm}$35, absorption length $5\ \mathrm{mm}$36, and wall reflectivity $5\ \mathrm{mm}$37—by minimizing $5\ \mathrm{mm}$38 between simulated and measured channel signals with a genetic algorithm in TMVA. The optimized values were $5\ \mathrm{mm}$39, $5\ \mathrm{mm}$40, and $5\ \mathrm{mm}$41, with mean $5\ \mathrm{mm}$42 and $5\ \mathrm{mm}$43. Using those tuned parameters, Geant4 reproduced the $5\ \mathrm{mm}$44-ray spectra of $5\ \mathrm{mm}$45Co, $5\ \mathrm{mm}$46Cs, and $5\ \mathrm{mm}$47Na with $5\ \mathrm{mm}$48–$5\ \mathrm{mm}$49 (Collaboration, 2024).

The ten-litre program measured spatial confinement directly with $5\ \mathrm{mm}$50 electrons. In opaque NoWaSH at $5\ \mathrm{mm}$51, $5\ \mathrm{mm}$52 of the detected light is reported within a $5\ \mathrm{mm}$53 radius sphere, $5\ \mathrm{mm}$54 within $5\ \mathrm{mm}$55, and no appreciable light beyond $5\ \mathrm{mm}$56; the same work reports an opaque-mode yield of $5\ \mathrm{mm}$57 under a readout with 25% fibre coverage and one-end readout, then enumerates cumulative improvement factors that take this value to $5\ \mathrm{mm}$58 for extrapolated designs (Collaboration et al., 4 Mar 2025). A related ten-litre report describes a cylindrical vessel with 208 fibres, 56 instrumented channels, better than $5\ \mathrm{mm}$59 time resolution per channel, total yield of $5\ \mathrm{mm}$60, preliminary energy resolution of $5\ \mathrm{mm}$61–$5\ \mathrm{mm}$62 at $5\ \mathrm{mm}$63, and sub-cm interaction localization (Navas-Nicolás et al., 14 Mar 2025). This suggests that reported light-yield figures are not directly interchangeable across publications without accounting for fibre coverage, one-end versus two-end readout, coupling efficiency, photosensor PDE, and scintillator formulation.

The $5\ \mathrm{mm}$64 muon-tracking demonstrator extends LiquidO from point-like imaging to continuous track reconstruction. Event displays show that light spreads over $5\ \mathrm{mm}$65 fibres in the transparent medium and remains confined to $5\ \mathrm{mm}$66 fibres per light ball in the opaque medium, yielding a clear “light cylinder.” Quantitatively, the number of SiPM channels hit per muon falls from $5\ \mathrm{mm}$67 in the transparent case to $5\ \mathrm{mm}$68 in the opaque case, while the total light yield rises from $5\ \mathrm{mm}$69 to $5\ \mathrm{mm}$70; the central-row average position resolution is reported as $5\ \mathrm{mm}$71 in opaque NoWaSH (Collaboration et al., 18 Jul 2025).

Published simulation studies set an upper-performance context for optimized designs. The 2019 concept paper reports $5\ \mathrm{mm}$72 for $5\ \mathrm{mm}$73 fibre pitch and $5\ \mathrm{mm}$74, energy resolution $5\ \mathrm{mm}$75, and a $5\ \mathrm{mm}$76 $5\ \mathrm{mm}$77-as-$5\ \mathrm{mm}$78 misidentification probability below $5\ \mathrm{mm}$79 at $5\ \mathrm{mm}$80 $5\ \mathrm{mm}$81 efficiency for $5\ \mathrm{mm}$82, improving to $5\ \mathrm{mm}$83 if $5\ \mathrm{mm}$84 is reduced to $5\ \mathrm{mm}$85 (Cabrera et al., 2019). The Snowmass white paper gives closely related small-module estimates of $5\ \mathrm{mm}$86, $5\ \mathrm{mm}$87 at $5\ \mathrm{mm}$88, and vertex resolution of $5\ \mathrm{mm}$89 in prototype conditions (Klein et al., 2022). In COCOA, a full Geant4+DETECT simulation of a $5\ \mathrm{mm}$90 LiquidO scatterer yields $5\ \mathrm{mm}$91, $5\ \mathrm{mm}$92, $5\ \mathrm{mm}$93 FWHM, an energy-resolution parameterization of $5\ \mathrm{mm}$94, and an angular resolution measure of $5\ \mathrm{mm}$95 (Collaboration et al., 28 Feb 2025).

6. Physics applications, comparative position, and open technical questions

The principal comparative claim for LiquidO is that opacity provides virtual segmentation without inactive mass. Conventional transparent liquid scintillator is described in the ten-litre overview as having vertex resolution of order $5\ \mathrm{mm}$96 and no intrinsic imaging, whereas LiquidO is described as providing effective self-segmentation at millimetre scale, real-time imaging of MeV-scale events, topological light-ball patterns for single electrons, Compton $5\ \mathrm{mm}$97 chains, and $5\ \mathrm{mm}$98 annihilation signatures, together with statistical Cherenkov separation (Collaboration et al., 4 Mar 2025). The 2019 neutrino paper adds that the concept naturally tolerates heavy dopant loading beyond the transparency constraints of conventional detectors, citing examples such as $5\ \mathrm{mm}$99Li or Gd for neutron tagging and $\lambda_a \sim 1$00In for flavour tagging, with delayed-coincidence topologies that are explicitly described in fibre images (Cabrera et al., 2019).

The proposed application space is correspondingly broad. In neutrino physics, published studies connect LiquidO to low-energy solar $\lambda_a \sim 1$01 spectroscopy, reactor and geo-$\lambda_a \sim 1$02 studies down to $\lambda_a \sim 1$03, diffuse supernova background searches, supernova burst flavour tagging, and decay-at-rest beam measurements (Cabrera et al., 2019, Navas-Nicolás et al., 14 Mar 2025). A dedicated geoneutrino study argues that LiquidO makes potassium geoneutrino detection feasible by combining alternative inverse-$\lambda_a \sim 1$04 targets with the positron topology of a central $\lambda_a \sim 1$05 blob plus two displaced $\lambda_a \sim 1$06 annihilation-$\lambda_a \sim 1$07 clusters, and with delayed coincidence in suitable isotopes such as $\lambda_a \sim 1$08Cu (Consortium et al., 2023). Outside neutrino detection, the COCOA concept uses LiquidO as a low-$\lambda_a \sim 1$09 Compton scatterer for the historically underexplored MeV band (Collaboration et al., 28 Feb 2025), while the ten-litre and muon-tracking papers explicitly point to applications in particle imaging and precise tracking (Collaboration et al., 4 Mar 2025, Collaboration et al., 18 Jul 2025). Medical imaging and dosimetry are also identified in the literature, including compact hybrid scintillation–Cherenkov imaging and positron-related imaging R&D (Navas-Nicolás et al., 14 Mar 2025, Consortium et al., 2023).

The main technical challenges are likewise well defined. The Snowmass white paper lists the need for stable, high-light-yield cocktails with $\lambda_a \sim 1$10–$\lambda_a \sim 1$11 but $\lambda_a \sim 1$12, WLS fibres with long attenuation length and high trapping efficiency, low-noise high-PDE SiPM arrays, large-scale mechanical support for fibre lattices, precision calibration of $\lambda_a \sim 1$13, $\lambda_a \sim 1$14, and re-emission spectra, and reconstruction methods capable of handling $\lambda_a \sim 1$15–$\lambda_a \sim 1$16 channels at full scale (Klein et al., 2022). The $\lambda_a \sim 1$17 report adds more immediate R&D directions: optimize fibre pitch, replace the MAPMT with high-QE SiPMs, develop direct optical-property measurement apparatus, and explore advanced reconstruction and particle identification by pulse-shape discrimination (Collaboration, 2024). A modelling study focused on WLS-fibre light collection recommends target reduced scattering lengths of $\lambda_a \sim 1$18–$\lambda_a \sim 1$19, bulk absorption lengths $\lambda_a \sim 1$20, and fibre pitch of $\lambda_a \sim 1$21–$\lambda_a \sim 1$22, while emphasizing that the diffusion approximation breaks down near boundaries or for $\lambda_a \sim 1$23 (Wilhelm et al., 2023).

Taken together, the literature presents LiquidO as a detector class defined less by a single scintillator chemistry than by an optical regime: long absorption length, deliberately short scattering length, and dense internal WLS-fibre readout. Within that regime, the experimentally established capabilities range from sub-centimetre point-like localization in litre-scale devices to $\lambda_a \sim 1$24 per-row muon tracking in a cubic demonstrator, while the broader programme remains oriented toward MeV-scale topology reconstruction, timing-assisted Cherenkov/scintillation discrimination, heavy-dopant neutrino targets, and scalable imaging detectors in particle physics and adjacent domains (Collaboration, 2024, Collaboration et al., 18 Jul 2025, Collaboration et al., 4 Mar 2025).