Indirect search for dark matter in the
Galactic Centre with IceCube
Indirect DM Search with Neutrino Telescopes
2Dark matter halo
Milky Way surrounded by dark matter halo
→ Highest DM density expected towards the Galactic Centre
Indirect search
Look for neutrinos produced by dark matter annihilation
𝝂
𝝂
Coverage of the Galactic Centre
3Up-going events
[arXiv:astro-ph.9907432]
IceCube background:
Dominated by atmospheric muons and neutrinos For up-going events:
Atmospheric muons stopped by Earth
Galactic Centre located at dec ~−29°
→ Neutrinos coming from the GC are seen as Down-going events by IceCube
→ Consider a veto
Reduce atmospheric muons background
Event Selection
4Low energy event selection (oscNext)
§ Consider 8.03 years of IceCube data from 2012 to 2020
§ Developed for atmospheric neutrino oscillation measurements Vetoed volume
Consider only DeepCore events
𝝁
𝝁
𝝂
𝝁Theory inputs
• Dark Matter Mass: 5 GeV and 1 TeV
• Neutrino spectra from dark matter annihilation
→ 𝝌aro𝝂 spectra [1]
→ Channels: 𝑊!𝑊", 𝑏(𝑏 , 𝜈#𝜈*#, 𝜏!𝜏", 𝜇!𝜇"
Signal Expectations
5d𝜙!
𝑑𝐸! = 1 2
𝜎"𝜐 4𝜋 𝑚#$
𝑑𝑁!
𝑑𝐸! .
%
&'
𝑑Ω .
(.*.+
𝜌#$ 𝑟(𝑠, Ψ, 𝜃) 𝑑𝑠
J-factor
NFW and Burkert halo profiles
→ Computed with Clumpy [2] with parameters from [3]
Dark matter self-annihilation cross section
Neutrino flux from dark matter annihilation in the GC
Probability Density Functions
63-dimensional PDFs
→ Event topology: PID (Particle ID)
→ Opening angle to the Galactic Centre: 𝚿
→ Reconstructed energy: 𝐥𝐨𝐠𝟏𝟎(𝐄)
Binning choice
§ Binning in PID
Optimised separation of tracks and cascades
→ 3 bins with edges [0, 0.5, 0.85, 1]
§ Binning in 𝚿:
18 bins ranging from from 0° to 180°
§ Binning in 𝐥𝐨𝐠𝟏𝟎 𝐄
50 bins ranging from 0 to 3
PDF Smoothing
7Kernel Density Estimation (KDE)
§ Apply gaussian kernel implemented in Scipy [1]
§ Use KDE on 2D distributions
→ Applied on 𝚿𝐫𝐞𝐜𝐨 - 𝐥𝐨𝐠𝟏𝟎(𝐄𝐫𝐞𝐜𝒐) distributions
→ Done for each PID bin
KDE
KDE Edge Correction
8correction Need to account for eventual boundary effects
Reflect the KDE at boundary
• Evaluate function for values outside of range and reflect it
• Applied in 1 dimension for Ψ+,-.
Background PDF
9Simulations weighted with atmospheric flux
KDE
Signal PDFs
10Simulations weighted with
• Spectra: 𝜒aro𝜈 for annihilation through 𝝉!𝝉" for DM mass of 100 and 1000 GeV
• Source morphology: NFW halo model
Analysis Method
11Binned likelihood method
Likelihood formulations considered:
Poisson Likelihood
ℒ!"#$$"% 𝜉 = $
&
Poisson(n"'$& ; n"'$("( 𝑓(i; 𝜉))
where
𝑓 𝑖 ;
𝜉
=𝜉
𝑓/ 𝑖 + 1 −𝜉
𝑓01(𝑖) Interval construction methods:Frequentist construction
Sensitivities
12Comparison to previous results
13Considerable improvement with respect to:
• IC86 3y GCWIMP search [1]
• Combined ANTARES/IceCube search [2]
Enhanced sensitivities due to:
• Improved event selection
• Energy and neutrino flavour considered
[1] Eur. Phys. J. C 77 (2017) 9, 627 [2] Phys. Rev. D 102 (2020) 8, 082002
Conclusion
14§ Enhanced event selection with more years of data
→ OscNext event selection
§ Additionnal informations used in the PDFs
→ Energy and flavour information
§ Considerable improvement of the sensitivities
→ Up to ~ 2 orders of magnitude at 10 GeV
Outlooks
§ Unblinding of the analysis
15
Backup Slides
Sensitivities: Charon vs PPPC4
16OscNext Event Selection
17Compared to IC86 3y GCWIMP search sample [1]
→ Effective area ~1 order of magnitude better