Diffuse Supernova Neutrino Background

• DSNB is a constant, nearly isotropic flux of neutrinos emitted by all past core-collapse (and failed) supernovae, serving as a relic signal of stellar death.
• It is modeled by convolving supernova rates with neutrino yields and cosmological expansion, making it a tool for probing both astrophysical events and cosmic history.
• Next-generation detectors like Super-Kamiokande-Gd, JUNO, and DUNE are set to convert DSNB observations into precision tests of neutrino properties and potential new physics.

The Diffuse Supernova Neutrino Background (DSNB) is the constant, nearly isotropic flux of neutrinos and antineutrinos emitted by all past core collapses in the observable Universe. In the MeV regime, it is the cumulative neutrino signal from core-collapse supernovae and related failed collapses over cosmic history, redshifted by cosmic expansion and mixed by flavor conversion. Although it has not yet been definitively detected, present limits are already comparable to realistic flux predictions, and the next generation of searches at Super-Kamiokande with gadolinium, Hyper-Kamiokande, JUNO, and DUNE is expected to turn the DSNB into a precision probe of core-collapse demographics, stellar formation history, cosmology, and neutrino properties (Iváñez-Ballesteros et al., 2023, Collaboration et al., 2021, Santos, 15 Jan 2025).

1. Physical basis and cosmological formulation

Core-collapse supernovae are the dominant known sources of MeV neutrinos on cosmic scales. Each collapse releases almost all of its gravitational binding energy as neutrinos, typically about $3 \times 10^{53}$ erg, and a single event produces $10^{58}$ neutrinos; averaged over the observable Universe, roughly one collapse happens every second (Ando et al., 2023, Suliga, 2022). The DSNB is therefore the time-integrated relic neutrino glow of stellar death, including both optically bright and optically invisible collapses.

A standard DSNB calculation folds the core-collapse rate with the neutrino yield per event and the cosmological expansion history. In a mass-eigenstate description, one representative form is

$$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$

with $E'_\nu = E_\nu(1+z)$, $z_{\rm max}=5$, and

$$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$

This formulation makes explicit that the observed spectrum is a convolution of source physics, population synthesis, and cosmology (Iváñez-Ballesteros et al., 2023).

The practical detection window is set by backgrounds rather than by the source spectrum alone. Reactor antineutrinos dominate at low energy, atmospheric neutrino backgrounds rise at higher energy, and the useful DSNB region is therefore typically quoted around $10$–$30$ MeV, with detector-specific refinements (Suliga, 2022, Ando et al., 2023). In that window, most of the detectable signal is generated at relatively low redshift: $z<1$ contains $\sim 87\%$ of the expected $10^{58}$0–$10^{58}$1 MeV DSNB event rate, while for the $10^{58}$2–$10^{58}$3 MeV flux only $10^{58}$4 arises from star formation above $10^{58}$5 (Lien et al., 2010, Anandagoda et al., 2020). This low-redshift weighting is what makes the DSNB simultaneously an astrophysical and cosmological observable.

2. Source modeling, black-hole formation, and dominant uncertainties

The dominant astrophysical input is the core-collapse rate, usually inferred from the star-formation-rate density and an initial mass function. The DSNB normalization is correspondingly sensitive to the adopted SFRD model. One explicit comparison found that the DSNB flux estimated using the SFRD derived from Fermi-LAT Collaboration et al. (2018) is significantly higher, by $10^{58}$6, relative to the flux estimated using the SFRD from Madau & Fragos (2017) (Anandagoda et al., 2020). Current uncertainty in the cosmic supernova rate normalization has been quoted at approximately $10^{58}$7, with synoptic optical surveys anticipated to reduce this to $10^{58}$8, dominated by systematics (Lien et al., 2010).

The treatment of failed supernovae, or black-hole-forming collapses, is equally important because these events have hotter neutrino spectra and disproportionately affect the high-energy tail. In one three-flavor DSNB decay study, the failed-supernova fraction $10^{58}$9 was varied over $$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$0, $$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$1, and $$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$2, with $$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$3 taken as fiducial; inclusion of failed supernovae affects the high-energy tail of the DSNB (Iváñez-Ballesteros et al., 2023). A later state-of-the-art calculation based on multi-dimensional, multi-second simulations and binary population synthesis found that including black-hole-forming collapses enhances the DSNB by up to $$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$4 at energies greater than $$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$5–$$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$6 MeV, while binary evolution effects change the total flux by up to a $$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$7 increase (Lunardini et al., 28 Jun 2025). This suggests that the highest-energy accessible DSNB bins are the most informative about collapse outcome.

Modern DSNB modeling is moving away from single-template spectra toward population-level mappings between pre-collapse stellar structure and neutrino emission. One recent framework uses the Carbon-Oxygen core mass and the compactness parameter to connect binary-affected stellar populations to neutrino spectra from both neutron-star and black-hole-forming collapses (Lunardini et al., 28 Jun 2025). A plausible implication is that future DSNB inference will be driven increasingly by forward population synthesis rather than by a small set of benchmark supernova spectra.

Alternative collapse phenomenology has also been explored. In the Standard Collapse and Double Collapse models motivated by the double neutrino burst of SN 1987A, the Double Collapse scenario predicts a significantly higher electron-neutrino flux and a harder spectrum, especially above $$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$8 MeV (Libanov et al., 2022). This remains a specialized scenario, but it illustrates how strongly the DSNB can depend on the assumed ensemble of collapse channels.

3. Flavor conversion, mass ordering, and standard flux expectations

The DSNB is produced in flavor eigenstates but propagates as an incoherent mixture shaped by flavor conversion in the supernova envelope and by vacuum propagation. Mikheyev-Smirnov-Wolfenstein effects are the standard robust ingredient, while collective oscillation effects remain uncertain in DSNB applications because of time and energy averaging (Ando et al., 2023). In an adiabatic MSW scenario with normal ordering, one convenient representation is

$$\phi_{\nu_i}(E_{\nu}) = c \int_0^{z_\mathrm{max}} \frac{dz}{H(z)} \int_{8~\rm M_\odot}^{125~\rm M_\odot} dM~R_{\rm SN}(z, M)~Y_{\nu_i}(E'_\nu, M),$$9

$E'_\nu = E_\nu(1+z)$0

so the detected $E'_\nu = E_\nu(1+z)$1 component is dominated by originally heavy-lepton neutrinos, whereas the $E'_\nu = E_\nu(1+z)$2 component remains a mixed channel (Tabrizi et al., 2020).

Mass ordering is a central control parameter for DSNB interpretation. The importance of identifying the neutrino mass ordering is explicit in DSNB decay studies, where ordering determines whether decay enhances or suppresses the detectable flux (Iváñez-Ballesteros et al., 2023). More generally, simultaneous measurements of $E'_\nu = E_\nu(1+z)$3 in water Cherenkov or scintillator detectors and $E'_\nu = E_\nu(1+z)$4 in liquid argon provide the flavor complementarity needed to test oscillation scenarios and disentangle source physics from propagation effects (Ando et al., 2023, Tabrizi et al., 2020).

A representative no-decay benchmark gives the following integrated fluxes: $E'_\nu = E_\nu(1+z)$5

$E'_\nu = E_\nu(1+z)$6

with normal-ordering values listed first and inverted-ordering values in parentheses (Iváñez-Ballesteros et al., 2023). These numbers place the standard DSNB directly adjacent to present detector sensitivities.

4. Detection channels, backgrounds, and experimental status

The principal DSNB detection channel is inverse beta decay,

$E'_\nu = E_\nu(1+z)$7

which dominates searches in water Cherenkov and liquid-scintillator detectors (Ando et al., 2023, Suliga, 2022). The defining strategy is prompt-delayed coincidence, with neutron tagging dramatically reducing backgrounds. In Super-Kamiokande with gadolinium, neutron-capture efficiency exceeds $E'_\nu = E_\nu(1+z)$8, and the first Gd results showed background-limited sensitivity already equivalent to prior pure-water exposure with only $E'_\nu = E_\nu(1+z)$9 the statistics (Ando et al., 2023). DUNE provides complementary $z_{\rm max}=5$0 sensitivity through charged-current interactions on $z_{\rm max}=5$1Ar, while coherent elastic neutrino-nucleus scattering in dark-matter detectors has been discussed as an all-flavor channel (Ando et al., 2023, Suliga, 2022).

The dominant backgrounds are reactor antineutrinos at low energy, atmospheric charged-current and neutral-current events at higher energy, spallation products, invisible muons, accidental coincidences, and fast neutrons (Ando et al., 2023, Collaboration et al., 2021, Collaboration et al., 2022). Atmospheric neutrino induced neutral-current events are the most critical background for JUNO after cuts, and the experiment’s projected reach depends strongly on pulse-shape discrimination and triple-coincidence rejection (Collaboration et al., 2022). In water Cherenkov detectors, neutron tagging and Cherenkov-angle-based classification are central to separating inverse beta decay from atmospheric backgrounds (Collaboration et al., 2021).

Experimentally, the strongest published $z_{\rm max}=5$2 limit remains the Super-Kamiokande bound of about $z_{\rm max}=5$3 at $z_{\rm max}=5$4 C.L. for $z_{\rm max}=5$5 MeV from the combined SK I–IV analysis (Collaboration et al., 2021). For $z_{\rm max}=5$6, SNO set $z_{\rm max}=5$7 in the $z_{\rm max}=5$8–$z_{\rm max}=5$9 MeV range (Suliga, 2022). A 2025 review further reports that the combined Super-K analysis across all phases showed a light excess over backgrounds and a $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$0 rejection of background-only across studied models (Santos, 15 Jan 2025).

Near-term sensitivity projections are correspondingly strong. Super-Kamiokande-Gd is expected to collect dozens of DSNB events over $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$1 years, and one analysis projected a $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$2 detection within $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$3 years of operation (Gouvêa et al., 2020). JUNO can reach $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$4 for $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$5 years of data taking and better than $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$6 after $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$7 years for a reference DSNB model (Collaboration et al., 2022). Hyper-Kamiokande is expected to collect hundreds of DSNB events after a decade of running (Gouvêa et al., 2020), while DUNE will record fewer events but with unique $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$8 flavor coverage (Moller et al., 2018, Tabrizi et al., 2020). In one 20-year baseline model, Hyper-K was assigned $$H(z) = H_0\left[\Omega_m(1+z)^3+\Omega_\Lambda\right]^{1/2}.$$9 inverse-beta-decay events, DUNE $10$0 charged-current $10$1 events, and JUNO $10$2 inverse-beta-decay events, illustrating the hierarchy between statistics and flavor complementarity (Tabrizi et al., 2020).

5. Astrophysical and cosmological inference from the DSNB

Because the DSNB is sourced by all core collapses, regardless of electromagnetic visibility, it is intrinsically sensitive to the cosmic core-collapse rate and to the fraction of invisible or failed supernovae. This is the basis for using the DSNB as a census of massive-star death rather than only of optically bright explosions (Lien et al., 2010, Ando et al., 2023). Synoptic sky surveys are expected to provide direct counts of the optically visible cosmic supernova rate out to $10$3, reducing source-history uncertainty and enabling the DSNB to isolate invisible-collapse contributions (Lien et al., 2010).

The DSNB is also sensitive to cosmological expansion through the factor $10$4. A dedicated cosmology study compared $10$5CDM, the Logotropic universe, and a bulk viscous matter-dominated model, finding that $10$6CDM and the Logotropic model predict the same number of DSNB events at best fit, whereas a bulk viscous matter-dominated cosmological model predicts $10$7 times more events (Barranco et al., 2017). Within $10$8CDM, the current Super-Kamiokande limit implies $10$9, independently of $30$0 (Barranco et al., 2017). Another analysis argued that after $30$1 years at Hyper-K or Theia, $30$2 could be measured with $30$3 precision, and that DSNB uncertainties below $30$4 would permit discrimination between local and CMB-inferred values of $30$5 (Gouvêa et al., 2020, Barranco et al., 2017).

On the astrophysical side, a joint 20-year analysis of Hyper-Kamiokande, JUNO, and DUNE suggested sensitivity to the local supernova rate at the $30$6–$30$7 level, with Hyper-K driving the result (Moller et al., 2018). The same study found that a non-zero fraction of black-hole-forming supernovae would be confirmed at $30$8 C.L. if the true fraction were larger than $30$9, whereas the DSNB has extremely poor statistical sensitivity to the nuclear equation of state and mass accretion rate of progenitors forming black holes (Moller et al., 2018). This suggests that the DSNB is better suited to population-level inference than to precision discrimination among fine details of core-collapse microphysics.

6. New-physics reach and nonstandard spectral structure

The DSNB is especially powerful for physics beyond the Standard Model because its neutrinos propagate over cosmological baselines through diffuse relic backgrounds. In nonradiative decay models,

$z<1$0

the DSNB has unique sensitivity to $z<1$1 (Iváñez-Ballesteros et al., 2023). The phenomenology depends sharply on ordering and hierarchy: in normal ordering with quasi-degenerate masses, short lifetimes can enhance the $z<1$2 flux at Earth; in normal ordering with a strongly hierarchical spectrum the effect is smaller and mainly at lower energies; in inverted ordering, short lifetimes can strongly suppress the $z<1$3 flux, in some cases rendering the DSNB effectively non-detectable even at next-generation detectors (Iváñez-Ballesteros et al., 2023). Determination of mass ordering is therefore crucial to interpreting either a detection or a null result.

Several distinct mechanisms produce spectral dips or depletions rather than overall normalization shifts. Self-interacting keV sterile-neutrino dark matter can resonantly scatter DSNB neutrinos through

$z<1$4

leading to a dip at

$z<1$5

potentially in the Hyper-Kamiokande detection window (Balantekin et al., 2023). Scalar-mediated secret neutrino interactions with the cosmic neutrino background generate broad spectral depletion, and one full three-flavor calculation found projected $z<1$6 sensitivities down to $z<1$7 for $z<1$8–$z<1$9 eV, with flavor-dependent reach at JUNO, Hyper-Kamiokande with gadolinium loading, and DUNE (Bharadwaj et al., 22 Jun 2026). Unlike flavor-blind cosmological and supernova bounds, this DSNB sensitivity is flavor-discriminating (Bharadwaj et al., 22 Jun 2026).

The DSNB can also be modified by late-time neutrino mass generation, pseudo-Dirac splittings, or dark-matter-induced contamination. If neutrinos were effectively massless at $\sim 87\%$0, the $\sim 87\%$1 DSNB spectrum could be significantly different from standard expectations in normal ordering, while the $\sim 87\%$2 flux would not be significantly impacted; this makes a combined $\sim 87\%$3/$\sim 87\%$4 measurement a direct test of recent neutrino mass generation (Gouvêa et al., 2022). Pseudo-Dirac scenarios with $\sim 87\%$5 are also within DSNB reach (Gouvêa et al., 2020). Conversely, low-mass dark matter annihilation to neutrinos can pollute the Hyper-Kamiokande DSNB window, especially for $\sim 87\%$6 MeV, biasing inferred DSNB temperatures and star-formation parameters; because inverse beta decay has negligible angular sensitivity at these energies, disentangling the two contributions is difficult (Bell et al., 2022).

The broad picture is that the DSNB is simultaneously a relic supernova observable, a low-redshift cosmological probe, and a long-baseline neutrino-physics laboratory. Its first detection would constrain the cosmic core-collapse rate and the fraction of black-hole-forming collapses; its spectral measurement would test flavor conversion, decay, secret interactions, sterile sectors, and other nonstandard propagation effects. If the DSNB remains undetected as detector exposures and background rejection improve, that outcome would itself become a nontrivial statement about supernova demographics, neutrino emission, or new neutrino physics (Beacom, 2010, Iváñez-Ballesteros et al., 2023).