IsoDAR Electron-Antineutrino Source

• IsoDAR electron-antineutrino source is a compact, accelerator-driven system that produces antineutrinos through 8Li decay-at-rest using a high-current H2+ cyclotron.
• The design integrates a beryllium target, lithium-based sleeve for effective neutron conversion, and optimized shielding to minimize background interference in underground detectors.
• It supports a diverse research program including sterile neutrino searches, electroweak measurements, and beyond-Standard-Model investigations through inverse beta decay and neutrino-electron scattering.

Searching arXiv for the cited IsoDAR papers to ground the article in current arXiv-indexed records. The IsoDAR electron-antineutrino source is a compact, accelerator-driven isotope-decay-at-rest source in which an intense charged beam generates neutrons in a beryllium target, those neutrons are converted into $^{8}\mathrm{Li}$ in a surrounding lithium-bearing sleeve, and the resulting $^{8}\mathrm{Li}$ undergoes $\beta^-$ decay at rest to produce an intense, isotropic $\bar\nu_e$ flux. Across the IsoDAR literature, this source is treated as a near-detector facility for precision short-baseline oscillation studies, inverse beta decay measurements, electroweak tests, and broader beyond-the-Standard-Model searches; the concept originated in KamLAND-oriented studies and later evolved into a first-of-its-kind underground facility concept at Yemilab (Adelmann et al., 2012, Abs et al., 2015, Alonso et al., 2022).

1. Historical emergence and programmatic scope

IsoDAR was introduced as an “exciting first step” within the broader high-current $\mathrm{H}_2^+$ cyclotron program associated with DAE$\delta$ALUS, but with a distinct physics goal: a localized $\bar\nu_e$ source for sterile-neutrino searches near existing liquid-scintillator detectors such as KamLAND or SNO+ (Alonso, 2012). Early cost-effectiveness studies then fixed the now-standard conceptual baseline: a compact underground cyclotron accelerating $\mathrm{H}_2^+$ to $60\ \mathrm{MeV/amu}$, delivering an effective $10\ \mathrm{mA}$ proton beam at $^{8}\mathrm{Li}$0, and producing $^{8}\mathrm{Li}$1 in a $^{8}\mathrm{Li}$2-rich converter around a beryllium target (Adelmann et al., 2012).

The KamLAND conceptual design report converted that source idea into a technical facility concept, specifying a source adjacent to the detector, a nominal five-year run, and an expected $^{8}\mathrm{Li}$3 reconstructed inverse beta decay events together with a substantial $^{8}\mathrm{Li}$4-electron scattering program (Abs et al., 2015). Subsequent studies generalized the same source architecture to JUNO, where the larger detector volume and a deuteron-enhanced source option were argued to provide a decisive test of the LSND and MiniBooNE antineutrino appearance anomalies under $^{8}\mathrm{Li}$5 invariance (Conrad et al., 2013).

Later Yemilab documents recast IsoDAR as an underground source-detector facility beside the 2.26–2.3 kt liquid scintillator detector, emphasizing short-baseline $^{8}\mathrm{Li}$6 disappearance, model-agnostic wave-pattern measurements in $^{8}\mathrm{Li}$7, precision weak-interaction measurements, and a wider BSM program. The 2022 Snowmass overview characterizes IsoDAR@Yemilab as a first-of-its-kind underground facility, while the deployment report and the preliminary design reports formalize the integration of cyclotron, MEBT, target, sleeve, and shielding in the Yemilab caverns (Alonso et al., 2022, Alonso et al., 2022, Winklehner et al., 2024, Spitz et al., 15 Aug 2025).

2. Nuclear production chain and source characteristics

The defining nuclear sequence is the conversion of beam power into $^{8}\mathrm{Li}$8, followed by $^{8}\mathrm{Li}$9 decay at rest. In the source-simulation literature this is written, for proton incidence on beryllium, as

$\beta^-$0

$\beta^-$1

$\beta^-$2

$\beta^-$3

Operationally, the beam produces neutrons in $\beta^-$4, heavy water moderates them, and capture on enriched $\beta^-$5 breeds the $\beta^-$6 that sets the $\beta^-$7 yield (Zhao et al., 2015).

A recurring misconception is that the Be target itself is the source. The detailed production studies instead show that the source is primarily a neutron-converter system: in the 2015 GEANT4 study, many neutrons are produced in Be, heavy water clearly moderates them, and $\beta^-$8 is mainly produced in the outer FLiBe sleeve because more neutrons enter that region (Zhao et al., 2015). The 2018 optimization study sharpened that picture quantitatively, finding that $\beta^-$9 of total $\bar\nu_e$0 production comes from neutron capture on $\bar\nu_e$1, while only $\bar\nu_e$2 comes from inelastic neutron interactions on Be that directly produce $\bar\nu_e$3 (Bungau et al., 2018).

The emitted antineutrinos are those of standard $\bar\nu_e$4 beta decay at rest. Several IsoDAR physics papers quote an average antineutrino energy of about $\bar\nu_e$5–$\bar\nu_e$6, with an endpoint near $\bar\nu_e$7; later Yemilab deployment and sensitivity studies describe the source as a pure $\bar\nu_e$8 flux with mean energy of order $\bar\nu_e$9 and endpoint near $\mathrm{H}_2^+$0–$\mathrm{H}_2^+$1 (Conrad et al., 2013, Conrad et al., 2013, Alonso et al., 2022, Alonso et al., 2021). The parent half-life is quoted as $\mathrm{H}_2^+$2 or $\mathrm{H}_2^+$3, which is short enough to produce a high-activity source but long enough that the source is effectively continuous on detector timescales (Alonso et al., 2022, Bungau et al., 2024).

Because each $\mathrm{H}_2^+$4 decay yields one $\mathrm{H}_2^+$5, source performance is usually parameterized by $\mathrm{H}_2^+$6 production per incident proton. This suggests that cross-comparison of IsoDAR papers should be done at the level of common assumptions—especially $\mathrm{H}_2^+$7 enrichment, duty factor, and geometry—because different papers alternately quote $\mathrm{H}_2^+$8, total $\mathrm{H}_2^+$9 over live time, or selected detector events.

3. Accelerator driver and beam-delivery architecture

The reference accelerator architecture accelerates $\delta$0 rather than protons directly. The reason given throughout the accelerator literature is the mitigation of low-energy space-charge limitations and the possibility of stripper-based conversion to two protons after extraction. In the early cyclotron design paper, IsoDAR uses the axially injected normal-conducting injector cyclotron at $\delta$1 with a continuous $\delta$2 particle-mA beam and $\delta$3 on target, while the later program documents standardize the operational description as $\delta$4 accelerated to $\delta$5 and then stripped to $\delta$6 of protons (Alonso, 2012, Alonso et al., 2022).

Three accelerator innovations recur in the mature IsoDAR design: acceleration of $\delta$7, direct axial injection via an RFQ, and the use of vortex motion as a collective beam-dynamics effect. The 2022 Snowmass overview states these points explicitly and ties them to a $\delta$8 continuous-wave proton-beam requirement at $\delta$9, with demonstration work in a $\bar\nu_e$0 test cyclotron (Alonso et al., 2022). The 2024 cyclotron-driver PDR describes the same strategy in design form: a filament-driven multicusp ion source, compensated LEBT, split-coaxial RFQ direct axial injection, optimized spiral inflector, early low-energy collimation, and a compact four-sector cyclotron purpose-built to exploit vortex motion (Winklehner et al., 2024).

The RFQ direct-injection study makes the front-end bottleneck explicit. Earlier source and spiral-inflector tests indicated that source current and inflector transmission were plausible, but longitudinal capture into the cyclotron RF bucket remained the limiting factor. The proposed four-rod RFQ, operating at $\bar\nu_e$1, accepts $\bar\nu_e$2 at $\bar\nu_e$3, accelerates it to $\bar\nu_e$4, and in simulation transmits $\bar\nu_e$5 of a $\bar\nu_e$6 beam, with about $\bar\nu_e$7–$\bar\nu_e$8 of the beam within $\bar\nu_e$9 of RF phase. The same study also identifies the unresolved post-RFQ challenge: over about $\mathrm{H}_2^+$0 between RFQ exit and spiral inflector, the beam expands enough that about $\mathrm{H}_2^+$1 is lost before the first acceleration gap unless additional focusing and rebunching elements are added (Winklehner et al., 2015).

Source-front-end R&D has therefore centered on producing high-intensity $\mathrm{H}_2^+$2 with acceptable purity and stability. The first MIST-1 commissioning paper reports stable total extracted currents of order $\mathrm{H}_2^+$3–$\mathrm{H}_2^+$4, a maximum current density of $\mathrm{H}_2^+$5, and about $\mathrm{H}_2^+$6 hours of accumulated runtime, with good stability for about $\mathrm{H}_2^+$7 hours at a time; at that stage, however, species separation and emittance were not yet measured (Winklehner et al., 2018). The later driver PDR still treats source current margin and end-to-end injection validation as major risk items, even while presenting a credible path to the nominal $\mathrm{H}_2^+$8–$\mathrm{H}_2^+$9 DC $60\ \mathrm{MeV/amu}$0 design expectation (Winklehner et al., 2024).

4. Target, sleeve, and yield optimization

The source assembly in the early GEANT4 studies consists of a central $60\ \mathrm{MeV/amu}$1 target, inner and outer FLiBe sleeves, a heavy-water region between Be and the inner sleeve, a graphite reflector, and thick concrete shielding. FLiBe is described there as a molten salt of $60\ \mathrm{MeV/amu}$2, with lithium enriched to $60\ \mathrm{MeV/amu}$3 (Zhao et al., 2015). That configuration established the canonical IsoDAR logic—Be for primary neutron production, heavy water for moderation and cooling, graphite for reflection, and enriched lithium for $60\ \mathrm{MeV/amu}$4 breeding—but it did not settle the converter design.

The 2018 optimization paper reworked the sleeve design in detail and argued that the nominal FLiBe sleeve is not optimal. In FLiBe, about $60\ \mathrm{MeV/amu}$5 of neutron interactions in the sleeve occur on fluorine, whereas F does not directly produce $60\ \mathrm{MeV/amu}$6 and has sub-unity neutron multiplication, with $60\ \mathrm{MeV/amu}$7 per neutron interaction. By contrast, Be has neutron multiplication above unity—about $60\ \mathrm{MeV/amu}$8 in the FLiBe discussion and about $60\ \mathrm{MeV/amu}$9 in the Li-Be optimized case—so removing fluorine and replacing the sleeve with a homogeneous Li-Be mixture increases the useful neutron economy (Bungau et al., 2018).

That study finds that the optimum sleeve composition is approximately $10\ \mathrm{mA}$0 Be by mass, implemented practically by small Be spheres in enriched lithium. For the modeled target+sleeve system, the optimized Li-Be sleeve reaches $10\ \mathrm{mA}$1 for a $10\ \mathrm{mA}$2 target and $10\ \mathrm{mA}$3 for the thermally safer $10\ \mathrm{mA}$4 target, compared with $10\ \mathrm{mA}$5 and $10\ \mathrm{mA}$6, respectively, for the earlier FLiBe design (Bungau et al., 2018). The same paper identifies a practical sleeve scale of about $10\ \mathrm{mA}$7 length and $10\ \mathrm{mA}$8 radius as favorable once the cost of enriched $10\ \mathrm{mA}$9 is included (Bungau et al., 2018).

A complementary line of optimization asked whether one should mix the moderator and converter rather than separating them. In “Getting the Most Neutrinos out of IsoDAR,” FLUKA and GEANT4 studies of LiOD, LiOD$^{8}\mathrm{Li}$00D$^{8}\mathrm{Li}$01O, heavy-water LiOD solution, and related configurations concluded that mixing the moderator and the $^{8}\mathrm{Li}$02 converter can increase the $^{8}\mathrm{Li}$03 yield by as much as 50%, especially for $^{8}\mathrm{Li}$04 masses of order $^{8}\mathrm{Li}$05 ton or less (Ciuffoli et al., 2016). The same note also found that a target–converter gap is particularly useful for $^{8}\mathrm{Li}$06 tungsten-target cases because it reduces neutron bounce-back losses that can otherwise remove up to $^{8}\mathrm{Li}$07 of neutrons, and that liquid-nitrogen cooling can improve yield by up to about $^{8}\mathrm{Li}$08 for very small converters (Ciuffoli et al., 2016). These were not adopted as the reference IsoDAR geometry, but they established important design principles for neutron moderation and capture.

Beam-species choices at the target were also explored directly. The 2015 GEANT4 proton/deuteron study fitted neutron and $^{8}\mathrm{Li}$09 yields as

$^{8}\mathrm{Li}$10

$^{8}\mathrm{Li}$11

with $^{8}\mathrm{Li}$12 values close to $^{8}\mathrm{Li}$13, and concluded that in the low-energy region around $^{8}\mathrm{Li}$14 the $^{8}\mathrm{Li}$15 production rate for deuterons is about three times that for protons, with the advantage decreasing toward about $^{8}\mathrm{Li}$16 at higher energies below $^{8}\mathrm{Li}$17 (Zhao et al., 2015). A separate KamLAND/JUNO study then used a deuteron variant for JUNO and quoted a factor of 2.7 increase in antineutrino production rate relative to the proton option, together with a reduction in source size (Conrad et al., 2013). Within the main IsoDAR program, however, the reference accelerator baseline remained the $^{8}\mathrm{Li}$18, $^{8}\mathrm{Li}$19 proton-equivalent source delivered by an $^{8}\mathrm{Li}$20 cyclotron.

5. Shielding, underground integration, and Yemilab deployment

Shielding is not peripheral to the IsoDAR source; it is a constitutive part of the source concept because the source is intended to operate very close to a delicate underground detector. The KamLAND shielding study treats this problem explicitly for a $^{8}\mathrm{Li}$21 proton beam on Be surrounded by highly enriched lithium. There the target-side source term is characterized by approximately $^{8}\mathrm{Li}$22 neutrons/POT escaping the neutrino-producing reflector, and the underground design challenge is to suppress both rock activation and detector backgrounds (Bungau et al., 2019).

The shielding solution developed for KamLAND is strongly asymmetric. Steel is used to reduce high-energy neutrons in energy; boron-rich concrete and hydrogenous material then absorb moderated neutrons. A baseline compact shield of $^{8}\mathrm{Li}$23 steel plus $^{8}\mathrm{Li}$24 boron-rich concrete achieved very low fluxes on a surrounding $^{8}\mathrm{Li}$25 sphere, but activation compliance ultimately drove the reference source shielding to $^{8}\mathrm{Li}$26 steel plus $^{8}\mathrm{Li}$27 boron concrete, with an additional $^{8}\mathrm{Li}$28 steel block toward KamLAND, making $^{8}\mathrm{Li}$29 total shielding on the detector side (Bungau et al., 2019). With that design, the detector-region rate above $^{8}\mathrm{Li}$30 was reduced to about $^{8}\mathrm{Li}$31 neutrons/POT, corresponding to only about $^{8}\mathrm{Li}$32 neutrons over the full 5-year run, and the analogous gamma background was also reduced to about $^{8}\mathrm{Li}$33 gammas over 5 years (Bungau et al., 2019).

The Yemilab deployment reports preserve the same asymmetrical philosophy but adapt it to a different site and a different cavern layout. The 2022 deployment report places the cyclotron, MEBT, and target in dedicated underground rooms, with the target room intersecting the LSC hall near the detector mid-plane and a source-detector center-to-center baseline of $^{8}\mathrm{Li}$34 (Alonso et al., 2022). The 2024 source-and-shielding study then models the underground source with a nested-shell Be target, Li-Be sleeve, local Fe and boron-loaded concrete shielding, and a beam orientation away from the detector. In that study the local shield is $^{8}\mathrm{Li}$35 iron plus $^{8}\mathrm{Li}$36 boron-loaded concrete, while a staged transport calculation through an additional $^{8}\mathrm{Li}$37 iron block yields a residual neutron flux of order $^{8}\mathrm{Li}$38 neutrons/proton/MeV, below the level required to keep source-induced neutrons in the detector below natural backgrounds (Bungau et al., 2024).

The same Yemilab study also shows that rock activation is unusually mild at that site because the surrounding rock is largely limestone with very low sodium content. Using a $^{8}\mathrm{Li}$39-thick rock block above the shield, it finds that even a reduced shield of $^{8}\mathrm{Li}$40 Fe plus $^{8}\mathrm{Li}$41 boron-loaded concrete gives hotspot activity $^{8}\mathrm{Li}$42, far below the Korean limit of $^{8}\mathrm{Li}$43 (Bungau et al., 2024). The 2025 Yemilab PDR adopts an even more conservative baseline of approximately $^{8}\mathrm{Li}$44 steel plus $^{8}\mathrm{Li}$45 borated concrete around the source and derives a detector-side iron-wall requirement of roughly $^{8}\mathrm{Li}$46–$^{8}\mathrm{Li}$47, corresponding to a volume of $^{8}\mathrm{Li}$48 and a mass of about $^{8}\mathrm{Li}$49 metric tons (Spitz et al., 15 Aug 2025).

The underground orientation of the beamline is itself part of the shielding strategy. The 2025 PDR turns the beam by $^{8}\mathrm{Li}$50 through the final MEBT so that the target is struck while the proton beam is heading away from the detector; the report judges the roughly one-meter increase in standoff preferable to another meter of steel shielding because backward-emitted neutrons are both fewer and softer than forward-going neutrons (Spitz et al., 15 Aug 2025). This design choice illustrates the central systems point of IsoDAR: target geometry, beam transport, shielding, activation, maintainability, and detector backgrounds are inseparable.

6. Experimental uses, sensitivities, and interpretive issues

The flagship application of the IsoDAR source is short-baseline $^{8}\mathrm{Li}$51 disappearance measured through inverse beta decay,

$^{8}\mathrm{Li}$52

In the KamLAND and JUNO studies, the relevant survival probability is written in the effective two-flavor form

$^{8}\mathrm{Li}$53

with $^{8}\mathrm{Li}$54 in meters and $^{8}\mathrm{Li}$55 in MeV (Conrad et al., 2013). The crucial idea is not only rate sensitivity but oscillation-wave imaging: because the source is compact, the spectrum is hard, and large liquid-scintillator detectors reconstruct both vertex and energy, IsoDAR can measure event distributions directly in $^{8}\mathrm{Li}$56 rather than infer disappearance from an integrated normalization.

Historically, this was already sufficient in the KamLAND/JUNO analysis to imply a decisive test of the LSND and MiniBooNE antineutrino appearance anomalies under $^{8}\mathrm{Li}$57 invariance. For IsoDAR@KamLAND, a five-year run was argued to cover the reactor-anomaly disappearance region up to about $^{8}\mathrm{Li}$58, while for IsoDAR@JUNO a five-year run with the deuteron-enhanced source yielded an estimated $^{8}\mathrm{Li}$59 reconstructed IBD events and $^{8}\mathrm{Li}$60 coverage of the entire global short-baseline antineutrino appearance region (Conrad et al., 2013).

At Yemilab, the scale is between the original KamLAND and the speculative JUNO deployment but closer to the latter in statistics. The 2021 Yemilab physics study assumes a $^{8}\mathrm{Li}$61 detector, a $^{8}\mathrm{Li}$62 center-to-center baseline, a source yield of $^{8}\mathrm{Li}$63 $^{8}\mathrm{Li}$64/proton, and a total of $^{8}\mathrm{Li}$65 antineutrinos in 4 years livetime, leading to $^{8}\mathrm{Li}$66 detected IBD events and about $^{8}\mathrm{Li}$67–$^{8}\mathrm{Li}$68 elastic-scattering events (Alonso et al., 2021). The same paper extends the source’s use to wavepacket decoherence, sterile-neutrino decay scenarios, and searches for a light $^{8}\mathrm{Li}$69 boson produced in the target and decaying to $^{8}\mathrm{Li}$70 (Alonso et al., 2021). The Snowmass overview casts this more generally as a model-agnostic oscillation program probing $^{8}\mathrm{Li}$71 to $^{8}\mathrm{Li}$72 with a single-isotope, high-rate $^{8}\mathrm{Li}$73 flux (Alonso et al., 2022).

A second major use is precision $^{8}\mathrm{Li}$74-electron scattering,

$^{8}\mathrm{Li}$75

which is especially clean because the source is flavor-pure and the $^{8}\mathrm{Li}$76 spectrum lies above much of the low-energy radioactivity background. The original KamLAND analysis projected $^{8}\mathrm{Li}$77 ES events in $^{8}\mathrm{Li}$78–$^{8}\mathrm{Li}$79 visible energy and a $^{8}\mathrm{Li}$80 measurement of $^{8}\mathrm{Li}$81 (Conrad et al., 2013). The later Yemilab study, using the larger detector, projected a precision of $^{8}\mathrm{Li}$82 or $^{8}\mathrm{Li}$83, improving to $^{8}\mathrm{Li}$84 or $^{8}\mathrm{Li}$85 with directional reconstruction (Alonso et al., 2021).

The broader BSM reach follows from the same source properties: very large IBD statistics, a single-isotope spectrum, and a nearby large detector. The Yemilab overview and Snowmass documents explicitly identify sensitivity to non-standard neutrino interactions through $^{8}\mathrm{Li}$86-electron scattering, wavepacket effects, multiple sterile states, sterile-neutrino decay, new bosons created at the target and decaying to neutrino pairs, and axion-like particles (Alonso et al., 2021, Alonso et al., 2022). At the same time, the source’s large neutron and photon production has motivated non-neutrino ideas such as “neutrons-shining-through-walls” and target-based dark-sector production in the KamLAND technical report (Abs et al., 2015).

The resulting picture is that the IsoDAR electron-antineutrino source is best understood not as a single target technology but as an integrated accelerator–target–converter–shield platform. Its distinctive scientific value comes from the conjunction of a $^{8}\mathrm{Li}$87 decay-at-rest source, unusually high beam current, compact underground siting, and detector-proximate operation. The literature also shows that its design is not static: FLiBe and Li-Be sleeves, proton and deuteron incidence studies, KamLAND and Yemilab deployment constraints, and multiple shielding baselines all belong to the source’s technical history rather than to a single frozen configuration (Zhao et al., 2015, Bungau et al., 2018, Spitz et al., 15 Aug 2025).