The Next Generation of IceCube Realtime Neutrino Alerts
TThe Next Generation of IceCube Realtime NeutrinoAlerts
The IceCube Collaboration ∗ http://icecube.wisc.edu/collaboration/authors/icrc19_icecubeE-mail: [email protected] In 2016, IceCube initiated a system of public real-time alerts that are typically issued within oneminute, following the detection of a neutrino candidate event that is likely to be of astrophysicalorigin. The goal of these alerts is to enable multi-messenger observations that may identify theneutrino source. Through January 31, 2019, a total of 20 public alerts have been issued, withmany of them receiving follow-up observations across multiple wavelength bands. One alert inparticular, IceCube-170922A, was found to be associated with a flaring gamma-ray blazar, TXS0506+056. This was the first >3 sigma association of a high-energy neutrino with an electro-magnetic counterpart. In 2019, the IceCube collaboration is introducing a new set of neutrinocandidate selections that expand the alert program. These new selections provide two alert chan-nels. A "Gold" channel will issue alerts for neutrino candidates at least 50% likely to be ofastrophysical origin and is expected to deliver ∼
10 alerts per year. Additionally a more frequent"Bronze" channel will provide ∼
20 alerts per year for neutrino candidates that are between 30%and 50% likely to be of astrophysical origin. We present the neutrino event selections used togenerate these alerts, the expected alert rates, and a description of the alert message.
Corresponding authors:
Erik Blaufuss , Thomas Kintscher , Lu Lu , Chun Fai Tung † University of Maryland DESY, Zeuthen Chiba University Georgia Institute of Technology36th International Cosmic Ray Conference -ICRC2019-July 24th - August 1st, 2019Madison, WI, U.S.A. ∗ For collaboration list, see PoS(ICRC2019) 1177. † Speaker. c (cid:13) Copyright owned by the author(s) under the terms of the Creative CommonsAttribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0). http://pos.sissa.it/ a r X i v : . [ a s t r o - ph . H E ] A ug ealtime Neutrino Alerts Chun Fai Tung
1. Introduction
High-energy neutrinos, being the "smoking gun" of high-energy hadronic interactions, playan important role in understanding the origin of cosmic rays. The discovery of the high-energyastrophysical neutrino flux by the IceCube Neutrino Observatory [1, 2] marked the beginning ofhigh-energy neutrino astronomy. A number of studies have been performed to locate point sourcesresponsible for the astrophysical neutrino flux [3, 4], but none has been found with 5 σ confidencelevel to date.IceCube is a cubic-kilometer neutrino detector [5] located at the South Pole. It does not ob-serve neutrinos directly, but instead it has 5160 digital optical modules (DOMs) buried in the ice ofAntarctica to capture the Cherenkov light emitted by the secondary particles produced in neutrino-nucleon interactions. The direction and energy of the neutrino can then be determined by recon-structing the secondary particles from the data collected by the DOMs. Therefore, the accuracy ofthe information of the neutrino depends greatly on the secondary particles.The majority of the neutrino signals detected by IceCube can be classified into two categories:track events and cascade events. Track events are characterized by their long paths of light whichcan span multiple kilometers. They are caused by highly penetrating muons, which are producedas secondary particles in the charged current interactions of muon-flavour neutrinos (CC ν µ ). Thelong light paths allow more accurate reconstruction of the directions of the muons, and the uncer-tainty can be below 1 degree. However, because most of the track events lie only partly inside theinstrumented volume, the energy reconstruction has large uncertainty. Cascade events are charac-terized by spheres of light. They can be caused by all kinds of neutrino interactions except CC ν µ ,when the interaction vertices are inside the instrumented volume. The reconstruction of cascadeevents is usually better in energy but worse in direction [6]. The angular uncertainties are usually10 to 15 degrees, but can also be much larger.Multi-messenger astronomy combines the information provided by photons, neutrinos, andgravitational waves together to reveal the hidden physics in the most exotic environments of the uni-verse. A potential neutrino source can be located if a counterpart in another messenger is observedin the direction from which an astrophysical neutrino candidate was detected. To achieve suchobservations, IceCube established a low-latency realtime neutrino alert system in April 2016 [7].When IceCube detects an astrophysical neutrino candidate which satisfies the selection criteria, acomputer-readable message is generated and sent out to the community automatically. During itsthree years of operation, the average alert rate was ∼ ∼ σ significance [8].IceCube also searched the archival data at TXS 0506+056’s location and found a neutrino flare inthe 2014-2015 season. This result shows that TXS0506+056 is a time-dependent neutrino sourcewith 3 . σ significance [9].The success of the follow-up observations of TXS 0506+056 demonstrated the potential ofthe multi-messenger approach to neutrino astronomy. After three years of operation, the realtime2 ealtime Neutrino Alerts Chun Fai Tung neutrino alert system has shown room for improvement, which includes: (1) provide more neu-trino candidates from a larger sample pool, (2) avoid mis-characterised events, (3) improve alertmessages’ clarity, and (4) define "signalness" for all the alerts.To address these issues, updates have been performed on most parts of the realtime alert sys-tem. These updates include an expanded and improved event selection, which is discussed inSection 2.1. In addition, the alert message format and alert streams are revamped to improve clar-ity and reduce confusion for the general astronomy community, which is discussed in Section 2.2.These updates result in a higher rate of alerts along with a higher signal purity, which is tabulatedin Section 2.3.
2. IceCube Realtime Alert System Update
The infrastructure of the realtime alert system remains largely the same as the previous system,which is described in detail in [7]. As illustrated in Figure 1, when IceCube detects an event, it
Figure 1:
Schematic overview of the realtime alert system. At South Pole, information of the events sat-isfying the selection criteria is sent North instantly through the Iridium satellite system. In the North, thesignalness of each event is assessed, and is used to decide if an alert is sent out. If the signalness is above50%, it is sent out via the Gold stream. If it is below 50% but above 30%, it is sent out via the Bronze stream.Both streams distribute the information in GCN Notice format. is first processed through the filtering system. For events that pass the filter, they are sent to theIceCube data center over the Iridum satellite. After reaching the North, the remaining selectioncriteria are applied on the event to determine if it is an astrophysical neutrino candidate. If it isselected as a candidate, it is sent as an alert through either one of the two streams, namely "Gold"or "Bronze". The choice of stream depends on the signalness of the event, which measures theprobability of the event being caused by an astrophysical neutrino. Signalness is defined as:
Signalness ( E , δ ) = N signal ( E , δ ) N signal ( E , δ ) + N background ( E , δ ) , (2.1)where N signal and N background are the number of signal events and number of background events atdeclination δ above the selection-specific energy proxy E . For example, E can be the estimatedneutrino energy. Candidates with signalness above 30% but below 50% are sent out in the Bronzestream, while candidates with signalness above 50% are sent out in the Gold stream. Alerts from3 ealtime Neutrino Alerts Chun Fai Tung both streams are distributed as Gamma-ray Coordinates Network (GCN) Notices. A more sophisti-cated reconstruction for the neutrino candidate is performed in the North to provide more accuratedirection and energy. This information is sent out in GCN circulars generally a few hours after theinitial GCN notice.
The updated event selection scheme consists of three different selections, which are namedGFU, HESE, and EHE. As a whole, they are responsible for picking out the neutrino-like trackevents most likely to be astrophysical from all of IceCube’s data. HESE and EHE selections werealready implemented in the previous realtime alert system. In the new realtime alert system, theHESE selection is updated while the EHE selection is kept unchanged. The GFU selection is a newaddition to the lineup of selection criteria, and by employing a different strategy, it is able to pickup neutrino events which are missed by the other two selections. All of the selections are based onprevious IceCube analyses and optimized for fast execution as filters in the online alert system.
GFU Track Selection
The Gamma-ray Follow Up selection is designed to select high-qualitytrack events that are similar to the ones used in IceCube’s point-source analysis [4, 10]. Boosteddecision trees (BDTs) are trained to pick out through-going track events that are caused by astro-physical neutrinos and are well reconstructed. To further improve the astrophysical purity, alerts aresent out only for the events with the highest energy. Energies of events from the Northern sky (up-going) are gauged with their reconstructed muon energies, and two thresholds are applied on themto obtain the 30% and 50% signalness. Energies of events from the Southern sky (down-going)are estimated with the total charge of the photoelectrons (PE) recorded by the detector. Since thebackground rate in this portion of the sky is strongly declination-dependent, a two-dimensionalselection criteria which depends on both the declination and charge is applied to achieve the samesignalness cut as the up-going events.
HESE Track Selection
High-Energy-Starting-Event (HESE) employs a vetoing technique to se-lect neutrino events with the interaction vertices lying inside the detector [2]. However, the majorityof starting events are cascades, which are not ideal for source-pointing due to their large angularuncertainty. Therefore, for the alert selection, additional cuts are applied to choose starting eventswith an out-going muon track. First, the likelihood of the event reconstruction must be in favorof the track-like hypothesis. Second, the measured track length must be longer than 200 meters,which also ensures the reconstruction quality. Finally, a declination-based cut on the charge ofthe recorded PE is applied to improve the astrophysical purity and to cut on the 30% and 50%signalness for the Bronze and Gold streams.
EHE Track Selection
The Extreme-High-Energy (EHE) track selection is modified from theevent selection used in the analysis which led to the observation of the first PeV neutrinos [11].It targets track-like neutrino events with energy between 500 TeV and 10 PeV by requiring thepassing events’ numbers of photoelectron (NPE) to be at least 4000. A fit quality parameter ( χ ) isused to select well-reconstructed track events, and a declination-dependent cut on NPE is used toincrease the portion of astrophysical events. The NPE cut value is set to achieve 50% signalness inthe final sample, which makes EHE track events exclusively in the Gold stream.4 ealtime Neutrino Alerts Chun Fai Tung
The effective area of these selections is shown in Figure 2. It is possible that an event passesmultiple selections, and each returns a different set of values for the event. The most probableoverlap is between the GFU selection and the EHE selection, because they both target through-going track events. However, only one alert is sent out. The source of information reported followsa hierarchical order: GFU first, EHE second, and HESE last. This ordering is chosen based on thesignal purity and angular resolutions of the three selections. ( — – ) ⌫ µ E ff e c t i v e A r e a ( m ) , ] Neutrino Energy (TeV) , 30 ] , 90 ] EHE EHE + GFU Gold GFU Bronze
IceCubepreliminary
All − − − − A ll F l a vou r E ff ec ti v e A r ea ( m ) δ ∈ [ − ◦ , − ◦ ] 10 Neutrino Energy (TeV) δ ∈ [ − ◦ , 30 ◦ ] 10 IceCubepreliminary δ ∈ [30 ◦ , 90 ◦ ]HESE All HESE Gold Figure 2:
Effective areas of the new realtime neutrino alerts as a function of neutrino energy. Top: Muonneutrino effective area of selections that target through-going neutrino (EHE + GFU) in three declinationbands. The effective area of the EHE selection is also shown for reference. Bottom: Neutrino effective areaof starting track events selection (HESE) in three declination bands.
The content of the message is redesigned to improve the ease of understanding the details andsignificance of the neutrino candidate. Each GCN Notice will include the following: • Discovery time and date - 0.01 second precision expressed in UTC format. • IceCube run number and event number - unique IDs used within the IceCube Collabora-tion. 5 ealtime Neutrino Alerts
Chun Fai Tung • Right Ascension and Declination - the reconstructed direction from which the neutrinocandidate came. Values in J2000, current, and 1950 epochs are reported. • Direction Error - angular radii of the 50% and 90% containment circles. Errors are derivedfrom Monte Carlo simulations of similar events. The 50% error as a function of neutrinoenergy is shown in Figure 3. • Signalness - probability of the neutrino candidate being an astrophysical neutrino. Definitionfollows Eq 2.1. • False Alarm Rate - yearly rate of alerts that have equal or higher signalness. • Likely Neutrino Energy - the most likely energy of the neutrino deduced from the param-eters of the alert event under the astrophysical neutrino hypothesis. The spectrum of thediffuse astrophysical neutrino flux is assumed to be a power law with spectral index -2.19.All the values for the quantities listed above (except time and date and run/event number) arecalculated based on the event sample and historical observations of the selection passed by theneutrino candidate. All of the information is generated during the reconstructions performed atSouth Pole. Neutrino Energy (TeV) M e d i a n A n g u l a r R e s o l u t i o n ( d e g ) GoldBronze
IceCube preliminary
Through-going GoldThrough-going All Neutrino Energy (TeV)0.400.450.500.550.600.650.700.75 M e d i a n A ngu l a r R e s o l u ti on ( d e g ) IceCube Preliminary
HESE GoldHESE All
Figure 3:
Angular resolution of IceCube realtime neutrino alerts as a function of neutrino energy. Left:Angular resolution of through going track events (GFU and EHE selections). Right: Angular resolution ofstarting track events (HESE selection). In this figure, "all" includes all the events above the Bronze parametercut in the corresponding selection, and "Gold" includes all the events above the Gold parameter cut, so theGold events are counted in both curves. The Gold alerts have a worse angular resolution at lower energyregion. For through-going tracks, this is because the events have high declination, where the reconstruction ismore difficult . For starting tracks, this is because the events have bright cascades at the interaction vertices.Also, as a new practice introduced in this update, the angular error reported in the GCN Notice is set to beat least 0.2 degrees to account for systematic uncertainty.
Since the alerts are designed to promote follow-up observation instead of detailed analysis, thepotential systematics of the information listed above are not included in the message. For example,the reported signalness and neutrino energy vary with spectral index of the astrophysical neutrinoflux. It must also be noted that because a significant portion of a muon track can lie outside ofthe detector’s instrumented volume, the true neutrino energy can be much higher than the reportedvalue. 6 ealtime Neutrino Alerts
Chun Fai Tung
The reconstruction methods applied at South Pole are limited by the computation power, sothey employ assumptions to improve execution speed. When the IceCube data center receives thecomplete data of the events, a more sophisticated reconstruction method is applied to them to obtaina more refined direction and angular uncertainty. Once the reconstruction is finished and verified,the new information is sent out via GCN Circular. The typical time lag between the automatedGCN Notice and the GCN Circular is a few hours.
The expected rates of signal events passing the updated selection are calculated using the best-fit diffuse astrophysical neutrino flux, which has a spectral index of − .
19 and normalization at100 TeV of 1 . × − GeV − cm − s − sr − , as reported in [12]. The rate of background eventspassing is calculated with the simulated atmospheric contamination, which includes both neutrinosand muons. These expected values are then compared with the observed values obtained by apply-ing the same event selection to seven years of IceCube data. The expected and observed rates ofpassing each selection are tabulated in Table 1. Gold Events Bronze EventsSignal ( E − . ) 6.6 (Total) 2.8 (Total)5.1 (GFU) 2.5 (GFU)0.5 (HESE) 0.3 (HESE)2.1 (EHE)Atmospheric Backgrounds 6.1 (Total) 14.7 (Total)4.7 (GFU) 13.8 (GFU)0.4 (HESE) 0.9 (HESE)1.9 (EHE)Observed historical rate 9.9 (Total) 19.5 (Total)7.8 (GFU) 18.4 (GFU)1.1 (HESE) 0.9 (HESE)4.3 (EHE) Table 1:
Expected and observed passing rates for Gold and Bronze selections. All values shown are eventsper year. Because of the overlap between GFU and EHE, the total rate of Gold alerts is not the sum of allselections.
The majority of the alerts are expected to be through-going tracks (from GFU and EHE), witha small fraction contributed by starting tracks (from HESE). These alerts are not expected to bedistributed isotropically in declination. This is a consequence of the declination dependence ofbackground events and high-energy neutrinos’ Earth absorption effect. As shown in Figure 4, mostalerts are located within a 30-degree range centered north of the celestial equator.
3. Summary and Future Outlook
To help realize the potential of multi-messenger astronomy, IceCube has designed a new re-altime neutrino alert system. This new system has been deployed since June 17th, 2019. After7 ealtime Neutrino Alerts
Chun Fai Tung
IceCube preliminary
IceCube preliminary
All
Figure 4:
Expected yearly rate of IceCube realtime alerts distribution in declination. "Thru-going trackastro ν " includes events that passed either GFU or EHE selection; "starting track astro ν " includes eventsthat passed HESE selection. Left: Alerts passed the Gold criteria. Right: Alerts passed either the Gold orBronze criteria. its deployment, IceCube is now issuing realtime alerts at an increased rate, and these alerts havea higher astrophysical purity and more reliable information than the previous generation. Thesechanges provide more opportunities for follow-up observations to the astronomical community.As multi-messenger astronomy becomes more mature, it is reasonable to expect more frequentcoincident observations in the future. The next multi-messenger detection might help resolve themysteries of the origin of ultra-high energy cosmic rays, or it might give us another riddle aboutthe high-energy universe. References [1]
IceCube
Collaboration, M. G. Aartsen et al.,
Science (2013) 1242856.[2]
IceCube
Collaboration, M. G. Aartsen et al.,
Phys. Rev. Lett. (2014) 101101.[3]
IceCube
Collaboration, M. G. Aartsen et al.,
Eur. Phys. J.
C79 (2019) 234.[4]
IceCube
Collaboration,
PoS(ICRC2019)851 (these proceedings).[5]
IceCube
Collaboration, M. G. Aartsen et al.,
JINST (2017) P03012.[6] IceCube
Collaboration, M. G. Aartsen et al.,
JINST (2014) P03009.[7] IceCube
Collaboration, M. G. Aartsen et al.,
Astropart. Phys. (2017) 30–41.[8] IceCube, Fermi-LAT, MAGIC, AGILE, ASAS-SN, HAWC, H.E.S.S., INTEGRAL, Kanata,Kiso, Kapteyn, Liverpool Telescope, Subaru, Swift NuSTAR, VERITAS, VLA/17B-403
Collaboration, M. G. Aartsen et al.,
Science (2018) eaat1378.[9]
IceCube
Collaboration, M. G. Aartsen et al.,
Science (2018) 147–151.[10]
IceCube, MAGIC, VERITAS
Collaboration, M. G. Aartsen et al.,
JINST (2016) P11009.[11] IceCube
Collaboration, M. G. Aartsen et al.,
Phys. Rev. Lett. (2013) 021103.[12]
IceCube
Collaboration, C. Haack and C. Wiebusch,
PoS(ICRC2017)1005 (2018).(2018).