IceCube Sensitivity for Low-Energy Neutrinos from Nearby Supernovae

Transcription

1 ApJ IceCube Sensitivity for Low-Energy Neutrinos from Nearby Supernovae IceCube Collaboration ABSTRACT This paper describes the response of the IceCube neutrino telescope located at the geographic South Pole to outbursts of MeV neutrinos from the core collapse of nearby massive stars. IceCube is scheduled to be completed in 2011 forming a grid of 5160 photo-multiplier tubes monitoring a volume of 1km 3 in the deep Antarctic ice for particle induced photons. Although the telescope is designed to detect neutrinos with energies ranging from to ev, large amounts of MeV neutrinos, predominantly from the ν e + p e + + n reaction, can be identified by a collective rise in the photomultiplier rate on top of a low dark noise. In case of a supernova neutrino burst near the galactic center, IceCube will be able to provide a measurement of the time profile corresponding to a 2.5 Mt proton decay and supernova search experiment. At the time of this writing, IceCube has reached 83% its final sensitivity for supernova detection and is sending potential supernova triggers to the Supernova Early Warning System. The sensitivity to neutrino properties such as the θ 13 mixing angle and the neutrino hierarchy are discussed, as well as the possibility to detect the deleptonization peak via ν e + e ν e + e scattering. The paper ends with a discussion of results obtained with IceCube s predecessor, AMANDA, after monitoring our Galaxy for more than one decade. Subject headings: neutrinos supernovae: general telescope Contents 1 INTRODUCTION 1 2 ICECUBE DETECTOR Digital Optical Module (DOM) Electronics, Data Acquisition and Artificial Deadtime

2 2 3 SUPERNOVA DETECTION METHODOLOGY Expected Signal in IceCube Data analysis method PHYSICS PERFORMANCE SIMULATIONS Expected Signal for a Supernova at the Center of Our Galaxy Expected Signficance and Galaxy Coverage Onset of Neutrino Production Sensitivity to Earth Effects DETECTOR PERFORMANCE Qualification of Modules Long Term Stability Background Significance Distribution Realtime Monitoring for Supernovae Systematics RESULTS FROM AMANDA 20 7 CONCLUSION INTRODUCTION Although optical detection of supernova explosions has a long history, nearly 99% of their energy is released by elusive neutrinos that only recently become detectable by experiments. As neutrinos may escape from an exploding supernova at much higher matter densities than photons, neutrinos will be observable several hours before their optical counterpart. The detailed observation of the onset of a supernova explosion is of much interest to astronomers. Therefore, a Supernova Early Warning System (SNEWS) has been set up to broadcast a reliable alert to the astronomy community in case a supernova has been detected by several neutrino detectors within seconds of each other (Antonioli et al. 2004). Currently, Super-Kamiokande (Fukuda et al. 2002), LVD (Aglietta et al. 2002), Borexino (Alimonti et al. 2009) and IceCube (Ahrens et al. 2004) contribute to SNEWS, with a number of other neutrino and gravitational wave detectors planning to

3 join in the near future. In spite of the huge discovery potential of Neutrino Astronomy, the only extra-terrestrial sources of neutrinos that have been detected so far are the Sun and the Supernova SN1987A. On February 23, 1987, a burst of mainly electron anti-neutrinos with energies of a few tens of MeV emitted by the Supernova SN1987A was registered simultaneously by the Baksan (Alexeyev et al. 1987), IMB (Bionta et al. 1987), and Kamiokande-II (Hirata et al. 1987, 1988) detectors, a few hours before its optical counterpart was discovered. With just 20 neutrinos collected, stringent limits on the mass of the ν e, its lifetime, its magnetic moment and the number of leptonic flavors could be derived (Kotake et al. 2006). Despite these achievements, detailed features of the gravitational collapse of a star and subtleties of neutrino properties, however, can only be studied with a much larger set of detected neutrinos. The possibility to monitor our Galaxy for supernovae with high energy neutrino telescopes was first pointed out by Halzen et al. (1996), suggesting to detect a coherent increase in the measured single photon rates arising from the Cherenkov light induced by neutrino reactions. IceCube is uniquely suited for this purpose due to its size and location at the geographic South Pole, with light sensors embedded at m depth in the highly transparent ice of the antarctic glacier. Using low radioactivity glass for the spheres housing the photomultipliers, typical dark noise rates of 540 Hz are achieved. (REMARK 1: make consistent with quoted number on subsection 5.2 and quantify the considered deadtime) The photomultipliers are sufficiently far apart so that their effective volumes for positron detection do not overlap, thus providing the equivalent of a 2.5 Mt ice detector for supernova neutrinos, once all 5160 sensors have been deployed in At present, 59 out of 86 strings have been installed for IceCube. While differences in the onset of the neutrino emission between various models are small (Kachelriess et al. 2005), the models have yet to overcome problems with the supernova explosion mechanisms. Therefore not much is known on the neutrino emission at times larger than several 100 µs after the deleptonization (Buras et al. 2003; Kitaura et al. 2006). We refer to the older Lawrence-Livermore model (Totani 1997) as it is the only one so far that extends over larger times (15 s), encompassing the complete accretion and part of the cooling phase. The newer Garching-model (Kitaura et al. 2006), which includes more detailed information on neutrino energy spectra and uses a sophisticated neutrino transport mechanism, is used as another benchmark, even though this model covers only 0.8 s following the collapse. At densities larger than 1010 gcm 3 the mean free path of neutrinos is smaller than the size of the iron core of a supernova. Already prior to the gravitational collapse of the core, neutrinos of different species are trapped in their relative neutrinospheres. The shockwave following the collapse dissociates nuclei, which suddenly increases the number of protons resulting in an increase in electron capture and the production of a burst of ν e. The timescale of this neutronization burst is of order 10 ms. The protoneutron star subsequently cools over 10 s, emitting most of its gravitational energy in neutrinos of all flavors.

4 Neutrinos streaming out of the core will meet matter densities ranging from 1010 g cm 3 to zero. They pass through a resonance layer at 103 g cm 3 associated with the atmospheric mass difference followed by a second layer at 10 gcm 3 corresponding to the solar mass difference. Although the survival probabilities depend on the details of the density profiles and generic predictions are impossible, one can anticipate the fluxes emitted in the limiting cases where sin 2 θ 13 is either large (> 10 3 ) or small (< 10 5 ). In the first case the mixing pattern further depends on whether the neutrino hierarchy is normal or inverted. In a dense environment such as a supernova core neutrino self-coupling can further modify the flavor evolution. In the case of a Fe-core collapse the presence of a high density of electrons above the core inhibits the effects of neutrino-neutrino forward scattering. The effects of neutrino refraction are therefore expected to be negligible. This may not be the case for stars of 8 10M that develop a degenerate O-Ne-Mg core; some of them produce supernovae and their occurrence may not be rare. The steeply-falling matter density above the core provides an environment where neutrino scattering modifies the flavor mixing of the deleptonization burst. Also in this case different patterns emerge for the normal and inverted hierarchies. It should be noted that additionally propagation of the flux through the Earth may further modify the observed neutrino flux We will refer in this paper to a simple generic description with a fast O(10ms) rise of the neu- 67 trino flux followed by an exponential decrease with time constant τ 3s. Neutrinos of O(10MeV) 68 energy produce positron tracks of about 0.6 E ν MeV 1 cm length via the ν e + p n + e + reaction. Considering the Eν 2 -dependence of the cross section for neutrino interactions, the total 70 light yield thus approximately scales with Eν 3. For sufficiently long light collection times, the yield 71 is approximately independent of scattering and depends linearly on absorption in the Cherenkov 72 medium The rate of galactic stellar collapses, including those obscured in the optical, is estimated by various methods to be one per years, with a best experimental upper limit of y 1 (Alexeyev et al. 1994). A compilation found in Giunti & Kim (2007) narrows this range to ( ) 10 2 y 1 by taking into account experimental and theoretical limits. The current work done on supernovae search, explained in this paper, places IceCube as a feasible Mt detector ready for the next galactic supernova explosion ICECUBE DETECTOR IceCube (Ahrens et al. 2004) is currently being installed deep in the antarctic ice sheet at the geographic South Pole. (The construction is about 69% complete.) Its final configuration will instrument a volume of about 1 km 3 of clear ice between depths of m below surface with a coarse grid of 5160 Digital Optical Modules (DOMs). Each DOM consists of a photomultiplier tube encased in a pressure-resistant glass sphere. The DOMs are installed on 86 strings of which 80 will be separated by roughly 125 m forming an hexagonal grid, with each string containing 60

5 DOMs vertically spaced by 17 m. The remaining 6 strings are separated by 60 m with the bottom 50 DOMs spaced by 7 m below an ice layer with reduced transparency at 2100 m depth. Another 10 DOMs, vertically separated by 10 m, instrument a volume above this layer. In addition, IceCube is complemented by a surface array called IceTop consisting of a pair of DOMs encased in ice tanks near the top of each string. Due to their higher noise and sensitivity to solar particles, the IceTop DOMs do not contribute to the supernova detection described in this paper. Although an overview of the main components of the detector is depicted in this section, detailed descriptions of the detector and the data acquisition can be found in Achterberg et al. (2006) and (Abbasi et al. 2009) Digital Optical Module (DOM) The IceCube Digital Optical Module (DOM) is the fundamental element in the IceCube architecture. Each DOM is housed in a 13" (33 cm) diameter, 0.5" (1.27 cm) thick borosilicate glass pressure sphere and contains a R " (25.4 cm) hemispherical photomultiplier tube manufactured by Hamamatsu as well as several electronics boards containing a processor, memory, flash file system and real time operating system that allows each DOM to operate as a complete and autonomous data acquisition system. It stores the digitized data internally and transmits the information to a surface data acquisition (DAQ) system on request. The time and amplitude information is preserved through immediate in-situ conversion to a digital format using an Analog Transient Waveform Digitizer (ATWD). The photomultiplier was chosen on the basis of low dark counts and good time and charge resolution. Its bialkali photocathode has a spectral response in the range 300 nm to 600 nm with a peak quantum efficiency of 25% ± 1% at 420 nm, well-matched to the Cherenkov signal spectrum and the optical properties of the glacial ice. Dark count rates around 540 Hz are typical for DOMs at temperatures between 35 C and 10 C in IceCube. The glass pressure sphere was selected to maximize the transmissivity in the sensitive region of the photomultiplier and to minimize the rate of background photons from intrinsic radioactivity in the glass. Optical transparency extends well into the near-uv, with 50% transmission at T 50% 340nm, but drops to a few percent at 300 nm. (see Fig. 1). The photomultiplier is mechanically attached to the glass pressure housing with a silicone elastomere (GE6156 RTV). This gel matches the refractive index of the glass to reduce optical losses at the medium interfaces. The spectral transparency of the gel extends to approximately 250 nm with T 50% 300nm. The combined response of the glass, gel, and photomultiplier is a critical input to the IceCube Monte Carlo simulation package. The response of the DOM for a head-on light source at infinity is shown in the right plot of Figure 1. (REMARK 2: Reina suggested to remove the plot because the PMT paper is almost ready, but there is no such plot in the paper...so?) Approximately 10% of incident signal photons in the spectral range from nm arriving from all directions are lost by the DOM owing to losses in the

6 6 Fig. 1. Optical efficiencies of the three DOM components PMT, optical coupling gel, and the glass sphere (left) and overall DOM efficiency versus wavelength for head-on illumination (right). redesign pictures, black-white, bigger labels, better legend, rectangular glass, gel, and phototube. Single photoelectrons produce pulses of approximately 8 mv amplitude and 10 ns width across the load impedance. The programmable front end pulse discriminator is set to 2mV, a factor of 20 above the RMS noise level (?) (REMARK 3: check the PMT paper). A pulse crossing this voltage threshold triggers the DOM logic to start the acquisition cycle and in addition tallies the count in a series of 4-bit wide firmware scaler registers that locally buffer the rate information in ms bins (2 16 ticks of the 40 MHz DOM system clock). Previous experience from AMANDA pointed to 40 K as a major contributor to the backgroundproducing radioactivity owing to its β decay mode with a high endpoint energy of 1.4 MeV. The residual potassium content in the IceCube spheres is specified to produce less than 100 Bq in this mode. Thermal background from the photocathode is significantly reduced by the cold temperature of the ice. Three effects contribute to the prevailing rate of on average 540 Hz: a Poissonian noise contribution of 250 Hz, a contribution due to photomultiplier induced afterpulses, relevant to both neutrino signal and noise, and correlated noise from Cherenkov radiation and scintillation originating in the glass of the photomultiplier and the pressure vessel. It is suspected that residual cerium in the glass causes a series of scintillation pulses lasting for up to 3300 µs, energized by β and α decays from trace elements of the uranium and thorium series. The observed time difference between noise hits deviates from an exponential distribution as expected from a Poissonian process (see Fig. 2 left). The observed excess is due to PMT related afterpulses and correlated photon pulses originating in the glass as discussed above. Using a minimum bias data set with pulses recorded by two IceCube strings we were able to separate correlated and un-correlated noise in the observed rates. Because of the temperature gra-

7 dient in the South Polar ice (Price et al. 2002) we were able to extract the temperature dependence of correlated noise as shown in Fig. 2 right. A non-thermal empirical noise fit was applied to the data as proposed in Meyer (2009), r(t) = G A C e T Tr, (1) where A C is the cathode surface of the deployed R PMT (Hamamatsu 2007) and G and T r are constants to fit the data. The temperature dependence of the uncorrelated part of the noise rate was fitted using Richardsons law on thermal emission r(t) = A T 2 e W kt + m T + C, (2) where A and k denote Richardsons and Bolzmanns constants and W is the work function for Bialkali cathodes.(remark 4: Some questions, see Wiki) The fits match the observed temperature dependence of the noise rates fairly well. Frequency [Hz] t for String expo t [s] Noise Rate [Hz] DOM Temperature DOM Temperature vs. Correlated vs. Noise Correlated Str39 Noise & Str49Str39 & Str Temperature [ C] Fig. 2. Left: Distribution of time differences between pulses for noise (bold line) and the expectation for Poissonian process (thin line). Right: Measured average noise rates from two strings as function of ice temperature. The Poissonian contribution (dotted) is fitted to a constant and the predicted thermal noise from Richardson s law; the correlated noise contribution (solid) is fitted to the empirical model of Meyer (2009).(REMARK 5: Although one can understand the right plot, I think the text below the plot can match better what is shown. Perhaps one should include a legend on the right plot? Or put the possonian data points (with their error bars) as dotted? ) Electronics, Data Acquisition and Artificial Deadtime The rate information in ms time bins is buffered on each DOM and read out by Ice- Cube s main data acquisition system (pdaq). It transfers the data to the independently operating supernova data acquisition system (SNDAQ) which synchronizes and regroups the information in 2 ms bins before processing. Due to satellite bandwidth constraints, the data are rebinned in

8 ms intervals and then subjected to an statistical online analysis described below; the fine time information is transmitted only in case of a triggered candidate for a supernova explosion. The background noise in the optical modules has a significant correlated component as discussed above. Typical time spans between correlated pulses are O(100 µs). A field programmable gate array in the DOM can be configured to introduce an additional dead time in counting discriminated pulses to improve the signal-to-noise of the measurement in the presence of correlated noise. The significance (see section??) is proportional to the signal rate and inversely proportional to the background noise fluctuation. While roughly 18% of signal hits suffer from afterpulses, the contribution of correlated noise is about???. (REMARK 6: Complete deadtime section and its plot) Fig. 3.??? To do a proper deadtime optimization, we simulated the signal Monto Carlo according to the Lawrence-Livermore model for distances up to 75 kpc taking afterpulses into account. By increasing the applied deadtime in small steps we were able to find an optimum in the calculated significance as showed in Fig. 3 for a single string. Various schemes (paralyzed and non-paralyzed or time shifted deadtime intervals) were tested in the simulation with only slight differences observed. The optimal setting for the dead time with respect to the signal over noise ratio for supernovae has been found to be τ = 246µs (Martin 2003) close to the value chosen in IceCube (250 µs). However, this value did not take into account PMT related afterpulses that also affect particle induced pulses. Fig. 3 shows the averaged optimized deadtime setting as function of distance using the current representation of afterpulses. The optimal setting rises with distance with τ = 50 µs representing a reasonable compromise. (REMARK 7: more on artificial deadtime; perhaps move this section? ) (TG: I am not so sure if we should put this section here.) SUPERNOVA DETECTION METHODOLOGY A core collapse supernova occurs when the inactive core of a massive star grows beyond the Chandrasekhar mass limit, nuclear fusion stops and the sudden loss of radiative pressure triggers a gravitational collapse and the subsequent explosion of the star s outer layers. Less than 1% of the gravitational binding energy of a supernova is emitted as optically visible radiation and kinetic energy. The remaining 99% is released as neutrino kinetic energy of which about 1% will be electron neutrinos from the initial neutronization burst and the remainder neutrinos and antineutrinos roughly evenly distributed among all flavors from the ensuing cooling processes. Thus an estimated J (Burrows et al. 1992) is carried away by the intense neutrino burst produced

9 predominantly through the thermal Helmholtz cooling reaction e e + (γ,z) ν ν. According to Thompson et al. (2003); Buras et al. (2003), the mean energy is expected to be about MeV for ν e, MeV for ν e and MeV for all other flavors (ν x ). The neutrino emission drops to zero after 15 s when neutrinos are no longer produced in the cooled down proto neutron star. More on the theory of core collapse supernova can e.g. be found in (Janka et al. 2006) and references therein. 198 Neutrinos streaming out of the core will meet matter densities ranging from kg/m 3 to 199 zero. They pass through a MSW-resonance layer at 10 7 kg/m 3 associated with m 2 23 or m2 13, depending on the neutrino mass hierarchy, followed by a second layer at 10 5 kg/m 3 associated to the quadratic neutrino mass difference m Both mix the initial fluxes of ν e, ν e and ν x depending on the survival probabilities. Although the survival probabilities depend on the details of the density profiles and generic predictions are impossible, one can anticipate the fluxes emitted in the limiting cases, where the mixing angle θ 13 is either large (sin 2 2θ 13 > 10 3 ) or small (sin 2 2θ 13 < 10 5 ). In the first case the mixing pattern further depends on whether the neutrino hierarchy is normal or inverted. Predictions are, however, complicated by the unknown density profile of the collapsing star. In addition, the forward shock or even multiple shocks and bubbles can form within the supernova. For simplicity, we will only consider a static density profile in this paper. In a dense environment such as a supernova core neutrino self-coupling can further modify the flavor evolution. To this date, these effects have not been included in our simulation and will be neglected as well in this paper. During the propagation of the flux through the Earth the observed neutrino flux is further modified (Giunti & Kim 2007; Dighe et al. 2004). Since the Earth density is constant within the mantle and the core, with sudden jump at the mantle-core transition, the treatment of neutrino oscillation is similar to the one in vacuum with effective electron and antielectron neutrino mass and a modified oscillation length L osc = 4πE ν/ m 2. Since the latter depends on the neutrino energy, the earth effect does too, leading to a deformation of the neutrino energy spectra. IceCube can neither measure individual neutrino energies nor distinguish flavor but the integral neutrino energy deposition. A combination with a second detector is necessary to quantify the influence of the oscillations within the Earth. This effect will vanish in the electron (anti electron) neutrino channel, if the spectra of ν e ( ν e ) and ν x are identical (?) Expected Signal in IceCube In the simulation of cooling the supernova is considered to be a blackbody source of neutrinos, following a modified Fermi-Dirac distribution. We adopt the following parametrization for the neutrino differential flux with time dependent luminosity L(t) at distance d for the neutrino species ν = ν e, ν e,ν x with corresponding energies E ν : dφ de ν = L(t) 4πd 2 E ν (t) f(e ν, E ν,α ν ), (3)

10 where ( ) 1 + αν (t) 1+αν(t) (E ν ) αν(t) e (1+αν(t))Eν/ Eν (t) f(e ν, E ν,α ν ) = E ν (t) Γ(1 + α ν (t)) is the normalized energy distribution depending on a shape parameter α ν (Keil et al. 2003). (4) There are still significant uncertainties in our understanding of the neutrino emission from supernova explosions. For illustrative purposes, when discussing the complete accretion and cooling phase extending to 15 s, we refer to the Lawrence-Livermore model (Totani 1997). When discussing the first hundred milliseconds of the burst we refer also to the more sophisticated Garching model (Kitaura et al. 2006). Because it assumes only half of the initial star mass (8 10 instead of 20 M ), it will produce roughly half of the signal seen by the Lawrence-Livermore model. The models describing the prompt neutronization burst appear to be robust and consistent (Kachelriess et al. 2005). Neutrinos of all species will be detected in IceCube via the interactions listed in Table 1, which includes the fractions of the total number of events without oscillations and their estimated uncertainty. The inverse beta decay is dominant and the neutrino-electron scattering gives only a small contribution which makes the detection of the deleptonization peak challenging. The production yields in the interactions with 16 O are rather uncertain due their dependence on the neutrino energy and details of the mixing. Table 1: Major reactions leading to detection of MeV neutrinos in IceCube. The fraction of the total number of events detected without (with) neutrino oscillations in stars is given in the second column. The third column gives the relative uncertainty on the cross section formulae used in the simulation. The numbers are taken from the Garching model averaged over 0.8 s. Reaction Signal Fraction Uncertainty Reference ν e + p n + e 93% (91%) <1% Strumia & Vissani (2003) ν e + e ν e + e 2% (1.7%) 1% Marciano & Parsa (2003) ν e + e ν e + e 0.8% (0.7%) 1% Marciano & Parsa (2003) ν x + e ν x + e 0.6% (0.5%) 1% Marciano & Parsa (2003) ν e + 16 O e + 16 F 1.9% (3.3%) <5% Haxton (1987) ν e + 16 O e N 1.6% (2.7%) <5% Haxton (1987) Most reactions end up with an electron or positron in the final state which radiate Cherenkov photons along their flight path x. For the inverse beta decay, the relation between the average ν e energy and the average positron energy can be approximated by E e + = (0.96±0.01) E νe (0.99± 0.46)MeV (E νe in MeV). The mean travel path for O(10MeV) electrons from ν e and positrons from ν e where simulated using GEANT-4 yielding x = (0.580 ± 0.005)cm E νe /MeV for electrons and x = (0.577 ± 0.005)cm E νe /MeV for positrons. The optical scattering and absorption in glacial ice at the South Pole has been studied extensively (Ackermann et al. 1996) by the AMANDA and IceCube collaboration by using pulsed

11 and continuous light sources embedded with the neutrino telescope. The detectors span depths ranging from m in the ice where the scattering coefficient varies by a factor of seven and absorptivity can vary by a factor of three depending on the wavelength. The observed data are consistent with the variations in dust impurity concentration seen in ice cores obtained at other Antarctic sites to track climatological changes. The ice is considered homogeneous in the horizontal plane with a slight tilt. As the tracking of individual photons with the GEANT package is too time consuming, simulation of photon propagation within the ice relies on predetermined tables (Lundberg et al. 2007), created to track photons with in the Antarctic ice. The tables store the detection probability and the arrival time distribution for given source and detector locations as well as their orientation. It includes the source wavelength, angular and intensity information, DOM parameters such as the glass and gel transparency and the quantum efficiency of the photomultiplier tubes. It also contains information about the ice such as the depth-dependent absorption and scattering lengths. These tables were created by using the wavelength and angular distribution profiles of Cherenkov light created by 10 GeV muons, which is similar to the Cherenkov production by electrons/positrons with energy on the order of 10MeV The positrons carry approximately 90% of the ν e energy and are mostly scattered multiple times at large angles within a small volume in the ice. Therefore the angular distribution of the Cherenkov light is considered isotropic and the location is assumed to be distributed randomly over the entire detector. 271 The detection probability is characterized by the effective volume for a single photon, V γ eff, 272 which was simulated by randomly placing 10 7 photons within a sphere of radius 250 m around each 273 DOM. The probability for multiple DOMs detecting a single shower is small as is the probability 274 that more than one photon of the same shower is detected. (REMARK 8: Expand and quantify 275 (Reina?)) The single photon effective volume varies strongly with the absorption and - due to the long collection time - only little with scattering. V γ eff was calculated as function of depth in the ice (see 278 Fig. 3.1), with an average over all DOMs in one IceCube string of V γ eff = ± 0.002(stat.) ± (syst.) m3 (REMARK 9: AHA model, 80m bins? We should decide which value 280 to quote and make consistent with the plot). The statistical error is dominated by the 281 DOM sensitivities ( 10%) and the systematic error reflects the uncertainties in the modeling of 282 ice structures ( 6%) The effective volume Veff e γ for detecting an electron or positron is obtained by multiplying Veff and the average number N γ of produced Cherenkov photons. The event rate e.g. for electron neutrinos is then given by: R(t) = n t L(t) 4πd 2 E ν (t) 0 0 dσ de e (E e,e ν )V e eff (E e)f(e ν, E ν,α)de e de ν, (5) where n t is the density of targets in ice, d is the distance of the supernova and L(t) = Φ E ν its

12 V γ eff V e + eff /E e + 50 V g eff /DOM (m3 ) V e+ eff /E e + (m3 MeV -1 ) Z (m) 0 Fig. 4. The effective volume V γ eff per DOM for single photon detection is plotted as a function of depth z 1950m obtained (left axis). (REMARK 10: Update plot with only Veff gamma and consistent with the quoted average Veff cited in the text ) luminosity. Scaling the 11 events Kamiokande-II observed during the Supernova SN1987A neutrino burst to an effective volume of antarctic ice one finds a signal expectation of 80± 25 detected photons within the first 10 s per module for IceCube for a SN1987A like supernova near the galactic center at 10 kpc if one assumes an average positron energy of 14.6 MeV. Including numbers for Lawrence-Livermore and Garching model the results are consistent with earlier simulations (Feser 2004; Jacobsen 1996) that were based on the assumption of homogenous ice performed for AMANDA once correcting for the different PMT sensitive areas and glass transparencies. (REMARK 11: Discuss geant3 simulation by baton rouge as cross check? Ali Fazely s SN supplement paper?) Data analysis method The idea behind the analysis is to monitor the collective rate in all DOMs introduced by Cherenkov photons uniformly distributed in the ice that were created by charged particles produced in reactions with supernova neutrinos. The rate r i = N i / t of pulses from a given DOM i is measured by counting a number of pulses N in a given time window t. This index ranges from 1 to the total number of optical modules N DOM (3540 as of October 2009 with 59 strings deployed). With sufficiently large t s, the distributions of the r i s can be described by lognormal distributions that were approximated for simplicity by Gaussian distributions with expectation values µ i and spreads σ i. The collective rate deviation µ of all DOM noise rates r i from their individual µ i s is

13 obtained by maximizing the likelihood L( µ) = N DOM i=1 1 exp ( (r i (µ i + ǫ i µ)) 2 2π σi 2σ 2 i ). (6) ǫ i denotes the module specific quantum efficiency for Cherenkov light. An analytic minimization of ln L leads to N DOM µ = σ µ 2 ǫ i (r i µ i ) σi 2, (7) with an uncertainty of σ 2 µ = i=1 ( NDOM i=1 ǫ 2 i σi 2 ) 1. (8) Assuming uncorrelated background noise and a large number of contributing DOMs, the significance ξ = µ/σ µ distribution should approximately follow a Gaussian with unit width centered at zero. Note µ has the structure of a weighted average sum: each optical module contributes with the deviation of its expected noise rate weighed by σi 2 (and ǫ i ). The likelihood that a deviation is caused by an isotropic and homogeneous illumination of the ice can be calculated from the χ 2 - probability N DOM ( ) χ 2 ri (µ i + ǫ i µ) 2 ( µ) =. (9) i=1 In order to suppress high rate deviations due to a temporary malfunction of individual detector modules, we reject events with a χ 2 -probability outside of an interval of 99.9%. To calculate a good estimator for the rate expectation value µ i and its standard deviation σ i, time intervals of 300 s before and after the investigated time interval are used to compute a sliding average, a time window of less than 5 min reduces the sensitivity of the analysis. Note that the moving average time windows start only 30s before and after the investigated bin to reduce the impact of a signal on the mean rates. At the beginning and the end of a SNDAQ-run, asymmetric moving average intervals are used. To optimize the speed of the calculations in a real time environment, an algorithm is used to reduce the n summations to 2 and from 2n to 4 in calculating the moving average and the deviation, respectively. For short time bases t the Gaussian approximation is no longer valid and Poissonian statistics needs to be applied. The collective rate deviation can then be obtained by using the equation 0 = d ln(l( µ)) d µ = N DOM i=1 { σ i n i ǫ i µ i + ǫ i µ ǫ i }, (10) which no longer can be solved analytically. The same goes for the corresponding uncertainty, which is derived by identifying a drop in ln L by 0.5. The required numerical minimization prevents an online analysis of the raw data. However, the fine time resolution of 2ms for IceCube, which

14 is recorded from 30 s before the triggering bin to 60 s after, can be used offline to determine the neutrino arrival time more precisely. With these data a more detailed analysis of a recorded supernova neutrino burst is possible. For instance, the onset of the neutrino emission t 0 can be determinined with 3.5 ms accuracy in 95% of all cases (Halzen & Raffelt 2009). This information can then be used in a triangulation of the supernova direction with other neutrino experiments. The optimal time base t depends on the expected signal shape. The generic model discussed in the introduction incorporates an instantaneous rise in neutrino luminosity to some maximum value, followed by an exponential decay as expected during proto neutron star cooling with a time constant of τ 3 s. Maximizing t µ 0 e t τ dt (11) σ µ t leads to t 1.26τ 3.8s. As the real time analysis operates on bins of 500 ms length, a time window length of 4 s has been chosen to be the best available setting for this particular model assumption. Assuming the Livermore model (Totani 1997), with a pronounced flux during the first seconds, the optimal time window is determined to be 1.6 s. To cover these model uncertainties, additional analyses with time bases of 500 ms and 10 s are run in parallel to the 4 s. The 500 ms analysis aims at short neutrino bursts (i.e. from soft gamma ray repeater sources or from supernovae collapsing into a black hole). The 10 s time base accounts for the observed time window of the detection of the neutrinos from the supervona SN1987A by Kamiokande-II (Hirata et al. 1987). By removing the cut on the χ 2, the trigger has been made sensitive to anisotropic illuminations. This gives the possibility to record hypothetical exotic particles emitting considerable amounts of light and thereby acting as a point or slowly moving source (such as ultra-heavy magnetic monopoles in some theories). The collective rate deviation µ and its uncertainty σ µ in the time bases of 4 s and 10 s are calculated using sliding windows of the finest 500 ms steps, ensuring that the signal detection efficiency is not reduced by binning effects. A supernova trigger is released whenever a ξ 6.5 fluctuation occurs, with a χ 2 confidence level of 99.9%. The soft requirement of at least 100 well behaving optical modules that need to contribute is imposed in addition PHYSICS PERFORMANCE SIMULATIONS 4.1. Expected Signal for a Supernova at the Center of Our Galaxy According to Eq. 5, a Supernova at 10 kpc distance produces a rate spectrum in the finalized IceCube configuration as seen in Fig. 5. (REMARK 12: Let s quote the total number of events over the 3 s(0.8 s) at least for the no oscillation case)with a maximal signalover-noise ratio of 55 70, the neutrino bust can clearly be detected with IceCube. Also, the (hypothetical) accretion phase lasting from 0 to 0.5 s can be separated from the following cooling

15 phase with high precision. The oscillation scenario assuming inverted neutrino mass hierarchy shows the largest signal, because the more energetic ν x -spectrum will be detected as ν e following oscillations in the progenitor star. The scenario without any oscillation is presented as a reference and leads to the weakest signal. The normal hierarchy and the case of very small mixing angle sin 2 2θ 13 < 10 5 are hard to distinguish due to very similar neutrino spectra mixings. Signal hits [20ms] no oscillation normal hierarchy inverted Hierarchy 2-5 sin < 10 2θ Time post-bounce [s] Signal hits [20ms] no oscillation normal hierarchy inverted Hierarchy 2-5 sin < 10 2θ Time post-bounce [s] Fig. 5. Expected rate distribution in 20 ms binning for the final IceCube configuration for all four oscillation scenarios assuming fluxes and energies given by the Lawrence-Livermore (left) and the Garching (right) model. The latter results in roughly half of the signal due to the smaller initial star mass. The 1σ-band of pure detector noise (hatched) was extracted from real data and is about ± 250 Hz broad Expected Signficance and Galaxy Coverage Previously shown detector response is simulated numerously and investigated every time by the analysis method introduced in Sec. 3.2, which will trigger the event. The significance of the simulated supernova signal at various distances is shown in Fig. 6. To overcome the correlation of both the 4 and 10 s binning inherent by the sliding time window approach, every group of t/ t 0 related analysis results is evaluated collectively. From each group, only the maximum significance ξ is selected. An expected signal from a Supernova within the Milky Way has to take into account the number of likely progenitor stars in the galaxy as a function of the distance from earth. This distribution is provided by Bahcall and Piran (Bahcall & Piran 1983). In case of the Magellanic Clouds a uniform star distibution along the galaxies diameter was choosen. Together, they roughly contain 5% of the Milky Way s stars. IceCube is able to detect Supernovae residing in the LMC with an average significance of 6.4 ± 1.7 for the (most sensitive) 0.5 s bin width, where the uncertainty reflects different oscillation

16 Milky Way (center) no oszillation, 0.5s no oszillation, 4s no oszillation, 10s 1000 Milky Way (center) no oszillation, 0.5s normal hierarchy, 0.5s) inverted hierarchy, 0.5s Significance y 1 false trigger rate SNEWS trigger threshold internal trigger threshold Milky Way (border) LMC SMC Significance y 1 false trigger rate SNEWS trigger threshold internal trigger threshold Milky Way (border) LMC SMC Distance [kpc] Distance [kpc] Fig. 6. Significance distribution versus distance of the Lawrence-Livermore model up to 70 kpc for all used time widths and no oscillation (left) and for three different oscillation models with fixed 0.5 s bin width (right). The Magellanic Clouds as well as the center and the border of the Milky Way are marked. The density of data points represents the star density. (REMARK 13: Check the spelling in the legend) scenarios. Supernovae in the SMC can be detected with an average significance of 3.6 ± 1.2 and will in general not trigger SNEWS as indicated in Fig. 6. The Milky Way is completely covered with minimal significances of 14 at 30 kpc distance, which is well above the limit for 1 false alert in 10 years Onset of Neutrino Production Analysis of the deleptonization peak that immediately follows the collapse is of considerable interest, since its magnitude and time profile is assumed to be almost independent of the initial star mass. The constant electron neutrino luminosity (the energy is likely not constant) may be used as a standard candle to measure the distance to the supernova. [ more! ] As the deleponization peak lasts for only 10 ms, IceCube s finest time binning of 2 ms is evaluated, as depicted in Fig. 7. The deleptonization peak is only visible via electron scattering signal and is predominantly covered by the inverse beta signal, if neutrino oscillation in the star takes place. Only for distances d 2kpc (1% of all stars within reach) the statistical error is small enough to resolve the subtle structures of the deleponization peak. Independently of the model, every oscillation scenario is showing a characteristic rate slope around the deleponization peak. Quantifying this in a series of several thousand simulations for both models and all four oscillation scenarios, it is possible to establish the inverse hierarchy and sin 2 2θ 10 3 with 3 σ (90% CL), if d 6kpc. Nevertheless, the distance has to be known from

17 Signal hits [2ms] no oscillation normal hierarchy inverted Hierarchy 2-5 sin < 10 2θ Signal hits [2ms] no oscillation normal hierarchy inverted Hierarchy 2-5 sin < 10 2θ Time post-bounce [s] Time post-bounce [s] Fig. 7. Expected average rate distribution in finest 2ms binning for all four oscillation scenarios for the Lawrence-Livermore (left) and the Garching (right) model. In the first bin the 1σ-error is shown. 403 an external measurement Sensitivity to Earth Effects The effect of neutrino oscillation becomes more be complicated, if the Earth is passed before the detector is reached. Assuming a two-component model of the density profile (Dighe et al. 2004), which separates the Earth into the mantle with a homogeneous density of 4500 kg/m 3 and the core with a homogeneous density of kg/m 3, the signal detected with IceCube varies as function of arrival direction of the neutrino as seen in Fig. 8. Signal modulation [%] normal hierarchy (accretion) inverted hierarchy (accretion) 12 sin θ 13 < 10 5 (accretion) normal hierarchy (cooling) 14 inverted hierarchy (cooling) sin θ 13 < 10 5 (cooling) Incoming angle [deg] Signal modulation [%] normal hierarchy (accretion) inverted hierarchy (accretion) sin θ 6 13 < 10 5 (accretion) normal hierarchy (cooling) inverted hierarchy (cooling) sin θ 13 < 10 5 (cooling) Incoming angle [deg] Fig. 8. Relative signal modulation versus neutrino angle for the three possible oscillation scenarios for the Lawrence-Livermore (left) and the Garching (right) model. Since the neutrino spectra change with time, a difference especially between the signal during the accretion phase and during the cooling phase is evident, adding the possibility of temporal signal analysis.

18 On average the signals are reduced by 4.6% for the Lawrence-Livermore and 1.5% for the Garching model during the cooling phase and about a factor of 3 smalle during accretion. However, this strongly depends on the angle of the neutrino front leading to a variation between 0 10%. It has been pointed out (?) that measuring the signal difference of two experiments, one with and one without earth effect, compensates model uncertainities and thus may help to test neutrino hierarchies DETECTOR PERFORMANCE 5.1. Qualification of Modules The stability of DOMs is crucial to the sensitivity of IceCube for the detection of supernovae. The disqualification of faulty modules (mainly modules with non-gaussian rate distributions) from consideration in the analysis is done bin-wise using automatic procedures. Only modules with stable noise rates are kept in the analysis. Additional modules are disqualified when exposing an unusual broadening factor σi 2/µ i or a high skewness in their rate distributions. Time periods where the number of qualified modules drops below a threshold of 100 are also discarded. The filter results are listed in Tab. 2, showing the excellent data quality of IceCube. Integrated over the data takiing period, only 0.3% of all DOMs are discarded. Thus, module disqualification hardly affects IceCube s significance, which scales with the square root of active DOMs. The number of booked DOMs in the IC40 configuration is 2359 (out of 2400 deployed DOMs) of which 69% are never disqualified. Nearly all (98%) disqualifications are due to just 11 DOMs, which are ignored in the analysis. Table 2: Disqualification requirements used in the online analysis were extracted from successful IC40 data taking runs covering the time from 2008/04/09 to 2009/04/19 (corresponding to an uptime of about 345 days). The table lists the percentage of filtered DOMs for all qualification parameters. Disfunctional DOMs with zero rate dominate the total filtered DOMs. Bin width Rate filtered Broadening filtered Skewness filtered Total filtered (s) (%) (%) (%) (%)

19 Long Term Stability Nearly all DOMs of IceCube show a very stable rate behavior. It is characterized by an exponential decay of the rate over long time periods and a slight annual modulation, which is represented well by r(t) = r 0 + c 1 e tk 1 + c 2 sin(tk 2 ) (12) and can be seen in Fig. 9. Decrease of the rate has a time constant of value??? and does not affect the analysis in any kind, which demands stable rate within the analysis time window of 10 min. This effect is more intense for deeper DOMs and is likely to result from of the freeze-in process causing triboluminescence in the ice itself. The average rate for a single DOM is 286 ± 20 Hz (with applied dead time), following a lognormal distribution with very high affinity to a Gaussian. [plot on this???] Rate [Hz] Annual muon amplitude [Hz] /01/01 03:53: /07/01 18:47: /12/31 09:41: DOM position Fig. 9. Left: Rate of a typical DOM as a function of time covering 556 days of lifetime. The line corresponds to a rate fit according to Eq. 12. Right: Depth dependent contribution of atmospheric muons assuming that the observed seasonal variations are due to stratospheric temperature variations modulating the muon flux The DOM rate modulation is assumed to arise from the annual change in the atmospheric muons flux. Fitting the time varying component, the depth dependent muon rate contribution can be extracted (see Fig. 9), which tracks the effect of dust layers similarly to what was seen in the determination of effective volumes. It turns out the influence of atmospheric muons, although only contributing on average about 11 Hz to the rate of individual DOMs (compared with the absolute measurement by IceCube s muon data acquisition (Tilav et. al 2009) [skip this???]), distorts the significance spectrum due to fluctuations in their rate. The most prominent effect of atmosperic muons is a broadening (σ = 1.27) and skewing of the sensitivity distribution from the expected Gaussian with σ = 1. A toy Monte Carlo, taking into account the fluctuations in the muon rates, fits the observed significance distribution well (see Fig.???). (still working on the actual, the real one... ) [ lognormal plot + plot of distributions of all rates ]

20 [ plot of significance ] 5.3. Background Significance Distribution Realtime Monitoring for Supernovae Due to the remote location of the detector in Antarctica the South Pole is out of reach for most communication satellites. Connectivity is available for only 10 h per day of which only 60% provide a high bandwidth. Therefore, a dedicated Iridium-satellite (Pratt et al. 1999) connection is used by the SNDAQ host system to transmit urgent alerts. In case of events with a signal significance above a threshold ξ, a short datagram is sent to the northern hemisphere. The receiving system parses the message and forwards information on the supernova-candidate event to the international SNEWS group. The time delay between photons hitting the optical module and the arrival of the datagram at SNEWS stands at about 6 min. A significance threshold of ξ = 6.5 satisfies the SNEWS requirement of a false background trigger of approximately once per 10 days Systematics There are three types of systematics relevant to this paper. The first type has to do with time dependent noise rate changes in the noise rate in all or a subset of DOMs that can mimic a supernova signal. This includes high voltage variations, longterm trends such PMT aging, weather effects and other external perturbances can all change the noise rate of the individual DOMs. For the supernova monitoring, longterm trends are accounted for by calculating the rate expectation values µ i and their standard deviation σ i from taking a rolling average of the noise rate for each DOM over 300 s on either side of the time of interest (see Sec. 3.2). The second type affect our understanding of the overall sensitivity of the detector. Ice properties, the wave length dependent quantum efficiency and the trigger threshold all fall under this category. The third type is due to our current knowledge of relevant cross sections, the distance to supernovae, the neutrino luminosity and energy, as well oscillation effects Detector Stability and External Disturbances The data were checked for external sources of rate changes as tracked by magnetometers, rioand photometers as well as seismometers at Pole. All have been synchronized with the rate data. Only magnetic field variations show a slight, albeit insignificant, influence on the rate deviation of