A search for dark matter in the center of the Earth with the IceCube neutrino detector
Jan Kunnen
Public PhD Defense
December 11th 2015
27%
Regular Matter
5%
Regular Matter 5%
Dark Energy
68%
Dark Matter 27%
Regular Matter 5%
Dark Energy
68%
The four questions I want to answer in this talk
What is Dark Matter?
How To Look for it?
Where to look for it?
How did I look for it?
� ��� = √ � � � (�) �
� ��� = √ � � � (�) �
A popular DM candidate is the WIMP
W eakly
I nteracting
M assive
P article
A popular DM candidate is the WIMP
W eakly
I nteracting
M assive
P article
The four questions I want to answer in this talk
What is Dark Matter?
How To Look for it?
Where to look for it?
How did I look for it?
The four questions I want to answer in this talk
WIMPs!?
How To Look for it?
Where to look for it?
How did I look for it?
There are three types of detection,
which are complementary
There are three types of detection,
which are complementary
There are three types of detection,
which are complementary
There are three types of detection,
which are complementary
There are three types of detection,
which are complementary
There are three types of detection,
which are complementary
No indisputable evidence for particle dark
matter has been found so far
The four questions I want to answer in this talk
What is Dark Matter?
How To Look for it?
Where to look for it?
How did I look for it?
The four questions I want to answer in this talk
What is Dark Matter?
Directly, in colliders or indirectly Where to (indirectly) look for it?
How did I look for it?
One way to search for DM is
via indirect detection
One way to search for DM is
via indirect detection
One way to search for DM is
via indirect detection
One way to search for DM is
via indirect detection
Photons easily get absorbed on their trajectory
Photons easily get absorbed on their trajectory
Charged particles don’t point back to their source
Photons easily get absorbed on their trajectory
Charged particles don’t point back to their source
Neutrinos are good messengers!
Photons easily get absorbed on their trajectory Charged particles don’t point back to their source Neutrinos are good messengers!
But difficult to detect…
Indirect detection can be done with
the IceCube Neutrino Observatory.
IceCube can determine the muon and thus the neutrino direction
Detector completion In December 2010
5160 DOMs on 86 strings Central part : DeepCore
• Deployed in deepest, clearest ice
• Lowers energy threshold to ~10 GeV
• IceCube as active veto (muon shield)
Knowing the muon direction ≈
Knowing the neutrino direction
Many WIMP searches have been done,
such as : Galactic Center
Many WIMP searches have been done,
such as : Galactic Halo
Many WIMP searches have been done,
such as : Dwarf Galaxies
Many WIMP searches have been done,
such as : Galaxy Clusters
Many WIMP searches have been done,
such as : the Sun
Many indirect WIMP searches
have been done with IceCube
No Earth WIMP search has been done with IceCube
The four questions I want to answer in this talk
What is Dark Matter?
How To Look for it?
Where to look for it?
How did I look for it?
The four questions I want to answer in this talk
What is Dark Matter?
How To Look for it?
Let’s look for a signal from the center of the Earth
How did I look for it? (Now it gets technical…)
How did I look for it?
Investigate if the Earth is an interesting source Look at the event signature in IceCube
Filter the data
Set up a statistical analysis
Look if there is a signal in the data
The last question I want
to answer in this talk
The Earth is an interesting source!
Signal neutrinos :
Coming from WIMP annihilations in the center of the Earth.
Maximum a few 10 3 events per year (more = excluded).
GeV to TeV energies.
The capture in the Earth depends
on the WIMP velocity
The existence of a Dark Disc could boost the muon flux.
— 68% contour --- 95% contour
— 68% contour
--- 95% contour
The capture in the Earth also depends
on the WIMP mass
How did I look for it?
Investigate if the Earth is an interesting source Look at the event signature in IceCube
Filter the data
Set up a statistical analysis
Look if there is a signal in the data
The last question I want
to answer in this talk
The capture in the Earth also depends
on the WIMP mass
The highest capture (and thus signal) rate is expected
for WIMPs with masses of 50 GeV
To be sensitive to a big parameter space,
2 statistically independent analyses are done.
ν signal μ signal
A typical signal event if m χ = 1 TeV A typical signal event
if m χ = 50 GeV
The event topology will be different for different WIMP models
μ signal
ν signal
Noise has a bigger impact in the case of low energetic events
A typical signal event if m χ = 1 TeV A typical signal event
if m χ = 50 GeV
A lot of background is constantly coming from all directions
Background :
Produced in the atmosphere by cosmic rays
Few 10 9 muons and 10 5 neutrinos per year.
GeV to PeV energies.
Look at events to optimize
the Low Energy event selection.
A more difficult to catch atmospheric muon event
This muon sneaks in and makes some SLC hits on the outer strings, that are removed by hit cleaning.
ν signal
μ signal μ atmospheric
A typical signal event
if m χ = 50 GeV
How did I look for it?
Investigate if the Earth is an interesting source Look at the event signature in IceCube
Filter the data
Set up a statistical analysis
Look if there is a signal in the data
The last question I want
to answer in this talk
Size (mm)
A m o u n t
Good stuf Bad stuf
Size (mm)
A m o u n t
Good stuf Bad stuf
Diameter of filter holes
Size (mm)
A m o u n t
Good stuf Bad stuf
Diameter of filter holes
Level 1 Level 2 Level N
Some good stuf
Lots of bad stuf
Level 1 Level 2 Level N
Some good stuf Lots of bad stuf
Still some
good stuf
Less bad stuf
Level 1 Level 2 Level N
Some good stuf Lots of bad stuf
Still some good stuf Less bad stuf
Still enough
good stuf
Almost no bad stuf
#include <filterscripts/I3VEFFilter_13.h>
#include <filterscripts/I3FilterModule.h>
I3_MODULE(I3FilterModule<I3VEFFilter_13>);
#include <dataclasses/geometry/I3Geometry.h>
#include <dataclasses/physics/I3RecoPulse.h>
#include <dataclasses/physics/I3Particle.h>
#include <dataclasses/I3Double.h>
#include <icetray/I3Units.h>
I3VEFFilter_13::I3VEFFilter_13(const I3Context& context) : I3JEBFilter(context),
linefitcut_(2.9), muonllhcut_(2.6), toplayerDOMcut_(5),
allpulseskey_("InitialPulseSeriesReco"), poleMuonLlhFit_("PoleMuonLlhFit"), poleMuonLinefit_("PoleMuonLinefit"), singleStringReq_(false)
{
AddParameter("LinefitCut",
"Remove events that have linefit zenith angle less than this",
linefitcut_);
AddParameter("MuonLlhCut",
"Remove events that have MuonLlh zenith angle less than this",
muonllhcut_);
AddParameter("ToplayerDOMcut",
"How many layers of DOM's should be used in the veto cap", toplayerDOMcut_);
AddParameter("PoleMuonLlhFit",
"The standard log likelihood linefit performed by Muon Group",
poleMuonLlhFit_);
AddParameter("PoleMuonLinefit",
"The standard muon linefit", poleMuonLinefit_);
AddParameter("RecoPulsesKey",
"Key for all the reco pulses.", allpulseskey_);
AddParameter("SingleStringRequirement",
"Reject events with pulses on multiple strings.", singleStringReq_);
}
void I3VEFFilter_13::Configure() {
GetParameter("LinefitCut",linefitcut_);
GetParameter("MuonLlhCut",muonllhcut_);
GetParameter("ToplayerDOMcut",toplayerDOMcut_);
GetParameter("RecoPulsesKey",allpulseskey_);
GetParameter("PoleMuonLlhFit",poleMuonLlhFit_);
GetParameter("PoleMuonLinefit",poleMuonLinefit_);
GetParameter("SingleStringRequirement",singleStringReq_);
nRejNOmuonllhsrt=nRejNOpulses=nRejIsDownGoingLlh=nRejNOlinefit=
nRejIsDownGoingLineFit=
nRejNOgeo=nRejTopLayers=nRejMultipleStrings=nDOMs=0;
}
bool I3VEFFilter_13::KeepEvent(I3Frame& frame) {
I3RecoPulseSeriesMapConstPtr all =
frame.Get<I3RecoPulseSeriesMapConstPtr>(allpulseskey_);
if(!all) {a
log_debug("Pulses not found. Ignoring event.");
nRejNOpulses++;
return false;
}
// MuonLLH upgoing cut
I3ParticleConstPtr muonllhsrt =
frame.Get<I3ParticleConstPtr>(poleMuonLlhFit_);
if(!muonllhsrt) {
log_debug("Could not find the MuonSRTllh reco. Ignoring event.");
nRejNOmuonllhsrt++;
return false;
}
bool IsUpGoingLlh = (muonllhsrt->GetZenith())>muonllhcut_;
if(!IsUpGoingLlh) {
log_debug("The Event's MuonSRTllh reco is downgoing. Ignoring event.");
nRejIsDownGoingLlh++;
return false;
}
// Linefit upgoing cut I3ParticleConstPtr linefit =
frame.Get<I3ParticleConstPtr>(poleMuonLinefit_);
if(!linefit) {
log_debug("Could not find the linefit. Ignoring event.");
nRejNOlinefit++;
return false;
}
AND MUCH MORE OF THIS
θ θ
On-Source Off-Source
In this analysis we cannot just define an off- source region → need to rely on simulations!
Earth searches : Background estimated
by simulation Other searches :
Background estimated
by off-source data
In this analysis we cannot just define an off- source region → need to rely on simulations!
10% of 1 year of IceCube data
WimpSim : simulated signal Atmospheric muons &
atmospheric neutrinos
Blind region ν
atmosphericμ
atmosphericν
signalExp. data ν
atmosphericμ
atmosphericν
signalExp. data
Before the event selection, the data is dominated by atmospheric muons (10 10 /year).
Remember, we look for max 10 3 signal events per year
The background is removed by
making event selections based
on direction, topology, etc.
Blind region
ν
atmosphericμ
atmosphericν
signalExp. data
In this filtering process a lot of effort went into the reduction of data-MC disagreement.
This effort was essential, as the background at final level cannot be estimated by
exp. data, but needs to be calculated from the MC.
During the event selection we minimize the data-MC discrepancy
ν
atmosphericμ
atmosphericν
signalExp. data
Blind region
ν
atmosphericμ
atmosphericν
signalExp. data
In this filtering process a lot of effort went into the reduction of data-MC disagreement.
This effort was essential, as the background at final level cannot be estimated by
exp. data, but needs to be calculated from the MC.
ν
atmosphericμ
atmosphericν
signalExp. data
During the event selection we minimize
the data-MC discrepancy
Blind region
ν
atmosphericμ
atmosphericν
signalExp. data
In a next step, Machine Learning Algorithms were used to make this selection as efficient as possible, i.e.
removing as much background as possible, without removing too much signal from the sample.
After some steps of filtering, the data rate is reduced to the mHz level (10 6 /year).
ν
atmosphericμ
atmosphericν
signalExp. data
Blind region ν
μ,atmoμ
atmoν
signalExp. data ν
atmosphericμ
atmosphericν
signalExp. data
In a next step, Machine Learning Algorithms were used to make this selection as efficient as possible, i.e.
removing as much background as possible, without removing too much signal from the sample.
ν
e,atmoν
τ,atmoBackground-like Signal-like
After some steps of filtering, the data rate is reduced
to the mHz level (10 6 /year).
Blind region
ν
atmosphericμ
atmosphericν
signalExp. data
At final level, the data rate is reduced to about 10 4 /year
Remember, we look for max 10 3 signal events per year!
Blind region ν
atmosphericμ
atmosphericν
signalExp. data ν
atmosphericμ
atmosphericν
signalExp. data
The filtering process reduces the data from 10 10 /year to 10 4 /year
Remember, we look for max 10 3 signal events per year!
ν signal μ signal A typical signal event
if m χ = 1 TeV μ atmospheric μ atmospheric
A typical background event
Look at events to optimize the event selection.
(Work by Jan L. & Isabelle A.)
ν signal μ signal A typical signal event
if m χ = 1 TeV μ atmospheric μ atmospheric
A typical background event
Look at events to optimize the event selection.
(Work by Jan L. & Isabelle A.)
How did I look for it?
Investigate if the Earth is an interesting source Look at the event signature in IceCube
Filter the data
Set up a statistical analysis
Look if there is a signal in the data
The last question I want
to answer in this talk
Look which hypothesis maximizes the
likelihood (based on FC ranks)
We hope to find signal!
Scenario 1 :
experimental data has higher rate than simulated background in the signal region
Some new phenomenon is going on!
Is it caused by WIMPs?
Scenario 1 :
experimental data has higher rate than simulated background in the signal region
Some new phenomenon is going on!
Is it caused by WIMPs?
Scenario 2 :
experimental data has same rate as
simulated background in the signal region
There's no muon flux caused by Earth WIMPs...
We hope to find signal!
But maybe there is no signal…
Scenario 1 :
experimental data has higher rate than simulated background in the signal region
Some new phenomenon is going on!
Is it caused by WIMPs?
Scenario 2 :
experimental data has same rate as
simulated background in the signal region
The muon flux caused
by Earth WIMPs is too small to observe.
What is the maximal allowed flux?
We hope to find signal!
If not we can always exclude models
How did I look for it?
Investigate if the Earth is an interesting source Look at the event signature in IceCube
Filter the data
Set up a statistical analysis
Look if there is a signal in the data
The last questions I want
to answer in this talk
Neutrinos from
Earth WIMPs?
The zenith angle distributions of the exp. data agree with the BG-only hypothesis
Unblinded data : IceCube collaboration approved the analysis!
The zenith angle distributions of the exp. data agree with the BG-only hypothesis
Unfortunately no evidence for a WIMP signal has been found in the full data sample...
The zenith angle distributions of the exp. data agree with the BG-only hypothesis
Unfortunately no evidence for a WIMP signal has been found in the full data...
→ Set upperlimit on μ
sThe last question I want to answer in this talk
How did I look for it?
Investigate if the Earth is an interesting source Look at the event signature in IceCube
Filter the data
Set up a statistical analysis
Look if there is a signal in the data
How did I look for it?
Investigate if the Earth is an interesting source Look at the event signature in IceCube
Filter the data
Set up a statistical analysis
Look if there is a signal in the data
Exclude Models!
The last question I want
to answer in this talk
Unfortunately no evidence for a WIMP signal has been found in the
unblinded data→ Set upperlimit on μ s and related quantities
Unfortunately no evidence for a WIMP signal has been found in the unblinded data→ Set upperlimit on μ s and related quantities
The se m
ode ls a
re r ule d o ut
Interpretation of the result in a broader perspective
No equilibrium in the Earth This search is sensitive to both WIMP capture (σ SI )
WIMP annihilation (<σ A v>)
This makes this analysis unique, being
sensitive to both these quantities
Interpretation of the result in
a broader perspective
The four questions I discussed during this talk
What is Dark Matter? WIMPS?!
How To Look for it? As many ways as possible!
Where to look for it? As many places as possible!
How did I look for it? Neutrinos from Earth WIMPs?!
How did I look for it?
Investigate if the Earth is an interesting source Look at the event signature in IceCube
Filter the data
Set up a statistical analysis
Look if there is a signal in the data
Exclude Models!
The last question I discussed during this talk
→ My work
The first search for dark matter in the center of the Earth with IceCube has been performed.
As there was no good off-source region, simulation had to be used to estimate the background.
The dataset was split in 2 statistically independent sets, each optimized independently.
No evidence for a WIMP signal has been found in the final data set, so upperlimits have been set. An improvement of a factor 10 has been found w.r.t. the AMANDA search.
This result has been interpreted in the cross-section phase space.
Under the considered assumptions, this search is complementary to the IceCube Solar WIMP search.
Summary
27%
27%
27%
Backup slides
Interpretation of the result in a broader perspective
~
There are three types of WIMP detection,
which are complementary
There are three types of WIMP detection,
which are complementary
There are three types of detection,
which are complementary
Which WIMP masses are interesting?
Look at Earth Capture Rate.
Which WIMP masses are interesting?
Look at Earth Capture Rate.
Which WIMP masses are interesting?
Look at Earth Capture Rate.
Which WIMP masses are interesting?
Look at Earth Capture Rate.
Low energy optimization
High energy optimization
Focus of
my analysis
Isabelle A. &
Jan L.
How do we split the dataset?
By cutting on the reconstructed energy.
Reconstructed energy for 50GeV WIMPs
IceCube
preliminary
IceCube