GERDA & the future of
76Ge-based experiments
Matteo Agostini on behalf of the GERDA Collaboration
Technische Universitat Munchen (TUM), GermanyGran Sasso Science Institute (INFN), L’Aquila, Italy
25th International Workshop on Weak Interactions and Neutrinos (WIN2015)June 8–13, 2015, MPIK Heidelberg, Germany
Double-� decays
2-neutrino double-� decay (2⌫��):
• (A,Z) ! (A,Z + 2) + 2e� + 2⌫e
• allowed in the Standard Model
• measured in several isotopes
• T
2⌫1/2 in the range 1019 � 1024 yr
Neutrinoless double-� decay (0⌫��):
• (A,Z) ! (A,Z + 2) + 2e�
• lepton number violation (�L = 2)
• ⌫ has non-null Majorana mass component
• T
0⌫1/2 limits in the range 1021 � 1026 yr (1025 yr for 76Ge)
• claim for a signal (subgroup of HdM experiment)
Matteo Agostini (TU Munich & GSSI) 1
Double-� decays
2-neutrino double-� decay (2⌫��):
• (A,Z) ! (A,Z + 2) + 2e� + 2⌫e
• allowed in the Standard Model
• measured in several isotopes
• T
2⌫1/2 in the range 1019 � 1024 yr
Neutrinoless double-� decay (0⌫��):
• (A,Z) ! (A,Z + 2) + 2e�
• lepton number violation (�L = 2)
• ⌫ has non-null Majorana mass component
• T
0⌫1/2 limits in the range 1021 � 1026 yr (1025 yr for 76Ge)
• claim for a signal (subgroup of HdM experiment)
Matteo Agostini (TU Munich & GSSI) 1
Neutrinoless double-� decay & neutrino physics
Assuming light-Majorana neutrino exchange as dominant 0⌫�� channel:
• (T 0⌫1/2)
�1 = G0⌫(Q�� ,Z)|M0⌫(A,Z)|2|m�� |2
• e↵ective Majorana mass:|m�� | ⌘
��Pi
U
2ei
m
i
�� =��c
212 c
213 m1 + s
212 c
213 m2 e
i2↵ + s
213 m3 e
i2���
10 4 0.001 0.01 0.11024
1026
1028
1030
1032
mlightest eV
T 120Ν
yrIH
NH
QD
76Ge
0vbb limits
[arXiv:1305.0056]
Matteo Agostini (TU Munich & GSSI) 2
State of the art of 0⌫�� search with
76
Ge
GERDA + HdM + IGEX
PRL 111, 122503 (2013)
GERDA Collaboration
[Phys.Lett. B586, 198 (2004)]
Klapdor-Kleingrothaus et al.
[Phys.Rev.D65,092007 (2002)]
IGEX Collaboration
[Eur. Phys. J. A 12, 147 (2001)]
HdM Collaboration
0.9 1.2 1.5 1.8 2.1 2.4 2.7 3 3.3 3.6 3.9
T1/20ν (Ge76) [1025 yr]
T1/2 > 1.9 1025 yr (90% CL)
T1/2 > 1.6 1025 yr (90% CL)
T1/2 = 1.19 -0.23+0.37 1025 yr
T1/2 > 2.1 1025 yr (90% CL)
T1/2 > 3.0 1025 yr (90% CL)
Matteo Agostini (TU Munich & GSSI) 3
GERDA collaboration
Matteo Agostini (TU Munich & GSSI) 4
Sensitivity and background goals
Phase I (Nov 2011 - May 2013):
• 15� 20 kg of target mass (87% 76Ge)
• bkg ⇠ 10�2 cts/(keV· kg· yr) at Q��
• exposure 21.6 kg·yr• sensitivity to scrutinize KK claim
Phase II (migration ongoing):
• new custom-produced BEGe detectors(additional 17 kg, 87% 76Ge)
• bkg .10�3 cts/(keV· kg· yr) at Q��
(active techniques for bkg suppression)
• exposure & 100 kg·yr• start exploring T
0⌫1/2 in the 1026 yr range
[Phys.Rev.D75, 092003 (2006)]
Matteo Agostini (TU Munich & GSSI) 5
Detectors
• HPGe detectors from materialenriched in 76Ge (⇠87%)
• detectors well established technology
• optimal spectroscopy performance:� long-term stability� �E ⇡ 0.1% at Q��
� radio purity
p+ electrode(read-out)
0 V
n+ electrode3-4 kV
p-typeGe
60-80 mm
70-1
10 m
m
65-80 mm
25-50 mm
p-typeGe
E
he
he
Coaxial-type BEGe-type
��
ee
2�⇥⇥
2�⇥⇥: 76Ge -> 76Se + 2e + 2�0�⇥⇥: 76Ge -> 76Se + 2e
ee
0�⇥⇥
Energy'(keV)'
2νββ
0νββ
Ge-76: Qββ=2039 keV
arbitrary
units
Calorimeter detectors:• source=detector• high detection e�ciency• peak at Q-value (Q��)
Matteo Agostini (TU Munich & GSSI) 6
Shielding strategy and apparatus
• bare Ge detectors in liquid Argon (LAr)• shield: high-purity LAr/H2O
• radio-pure material selection• deep underground (LNGS, 3800 m.w.e.)
[EPJ C 73 (2013) 2330]
Matteo Agostini (TU Munich & GSSI) 7
Backgrounds and mitigation techniques
Background sources:
• natural radioactivity (232Th and 238U chains):
� �-rays (e.g. 208Tl, 214Bi)
� ↵-emitting isotopes from surface contamination
(e.g. 210Po) or 222Rn in LAr
• long-lived cosmogenic Ar isotopes (39Ar,42Ar)
• cosmogenic isotopes activated in Ge (68Ge, 60Co)
Mitigation strategy:
• detector anti-coincidence
• time-coincidence (Bi-Po or 68Ge)
• pulse shape analysis
• detection of LAr-scintillation light
Matteo Agostini (TU Munich & GSSI) 8
Backgrounds and mitigation techniques
Background sources:
• natural radioactivity (232Th and 238U chains):
� �-rays (e.g. 208Tl, 214Bi)
� ↵-emitting isotopes from surface contamination
(e.g. 210Po) or 222Rn in LAr
• long-lived cosmogenic Ar isotopes (39Ar,42Ar)
• cosmogenic isotopes activated in Ge (68Ge, 60Co)
Mitigation strategy:
• detector anti-coincidence
• time-coincidence (Bi-Po or 68Ge)
• pulse shape analysis
• detection of LAr-scintillation light
Matteo Agostini (TU Munich & GSSI) 8
Phase I detector array configuration
• 3 + 1 strings• 8 enrGe coaxial detectors (2 not considered in the analysis)• 5 enrGe BEGe detectors (1 not considered in the analysis)• 1 natGe coaxial detectors
enrGe mass for physics analysis: 14.6 kg (coaxial) + 3.0 kg (BEGe)
Matteo Agostini (TU Munich & GSSI) 9
Data taking of Phase I
Matteo Agostini (TU Munich & GSSI) 10
Data taking of Phase I
Matteo Agostini (TU Munich & GSSI) 10
Background modeling
even
ts/(3
0 ke
V)
-210
-110
1
10
210
310 datamodel
��⇥2K42K40Ac228Th228
AlphasCo60HCo60inGeBi214HBi214P
GE
RD
A 1
3-06
energy (keV)2000 2500 3000 3500
data
/mod
el ra
tio
012345
68%95%99.9%
data/model
Contribution at Q�� :
• �-rays (close sources):Bi-214, Tl-208, K-42
• ↵- and �-rays (surface decays):Ra-226 daughter, Po-210, K-42
more details in [EPJ C74 (2014) 2764]
• no line expected in the blinded window
• background flat between 1930-2190 keV(excluding peaks at 2104 and 2119 keV)
• mean FWHM at Q�� (mass/exposure weighted):
coax —> 4.8±0.2 keV
BEGe —> 3.2±0.2 keV
expectedco
unts
/keV
MC simulation
Matteo Agostini (TU Munich & GSSI) 11
Background modeling
even
ts/(3
0 ke
V)
-210
-110
1
10
210
310 datamodel
��⇥2K42K40Ac228Th228
AlphasCo60HCo60inGeBi214HBi214P
GE
RD
A 1
3-06
energy (keV)2000 2500 3000 3500
data
/mod
el ra
tio
012345
68%95%99.9%
data/model
Contribution at Q�� :
• �-rays (close sources):Bi-214, Tl-208, K-42
• ↵- and �-rays (surface decays):Ra-226 daughter, Po-210, K-42
more details in [EPJ C74 (2014) 2764]
• no line expected in the blinded window
• background flat between 1930-2190 keV(excluding peaks at 2104 and 2119 keV)
• mean FWHM at Q�� (mass/exposure weighted):
coax —> 4.8±0.2 keV
BEGe —> 3.2±0.2 keV
expectedco
unts
/keV
MC simulation
Matteo Agostini (TU Munich & GSSI) 11
Pulse shape discrimination
Coaxial detectors:
• artificial neural network
• 0⌫�� acceptance = 90+5�9%
• background acc at Q��= ⇠45%
BEGe detectors:• A/E parameter (mono-parametric PSD)
• 0⌫�� acceptance 92±2%
• background acc at Q��20%
[Eur.Phys.J C73 (2013) 2583]
Matteo Agostini (TU Munich & GSSI) 12
Unblinding: spectrum around Q��
[PRL 111, 122503 (2013)] ⇤ w/o PSD Analysis cuts applied:
1) signals quality cuts
2) detector anti-coincidence
3) muon-vetoanti-coincidence
4) single-detectors timecoincidence (BiPo cut)
5) PSD
Survival fraction at Q�� :
1 ⇠99%2+3 ⇠60%4 ⇠100%
5 ⇠50%
w/o PSD
w/ PSD
exposure background expected cts observed ctsdata set [kg·yr] 10�2 cts/(keV· kg· yr) (Q��±5 keV) (Q��±5 keV)
golden 17.3 1.8 3.3 5BEGe 2.4 4.2 1.0 1
Matteo Agostini (TU Munich & GSSI) 13
Unblinding: spectrum around Q��
[PRL 111, 122503 (2013)] ⇤ w/o PSD⌅ w/ PSD⇤
Analysis cuts applied:
1) signals quality cuts
2) detector anti-coincidence
3) muon-vetoanti-coincidence
4) single-detectors timecoincidence (BiPo cut)
5) PSD
Survival fraction at Q�� :
1 ⇠99%2+3 ⇠60%4 ⇠100%5 ⇠50%
w/o PSD
w/ PSD
exposure background expected cts observed ctsdata set [kg·yr] 10�2 cts/(keV· kg· yr) (Q��±5 keV) (Q��±5 keV)
golden 17.3 1.8 1.1 3.3 2.0 5 2BEGe 2.4 4.2 0.5 1.0 0.1 1 0
Matteo Agostini (TU Munich & GSSI) 13
Statistical analysis
Profile likelihood analysis:
• ML fit(constant+Gauss in 1930-2190 keV range)
• multiple data sets (common T
0⌫1/2)
• T
0⌫1/2 � 0 (coverage tested)
Results (GERDA only):
• best fit for N0⌫�� = 0 signal cts
• T
0⌫1/2 > 2.1 · 1025 yr (90% C.L.)
• MC Median sensitivity (for no signal):T
0⌫1/2 > 2.4 · 1025 yr (90% C.L.)
Results (GERDA + IGEX [1] + HdM [2]):
• best fit for N0⌫�� = 0 signal cts
• T
0⌫1/2 > 3.0 · 1025 yr (90% C.L.)
PRL 111, 122503 (2013); [1] Phys.Rev. D65, 092007 (2002); [2] Eur.Phys.J. A12, 147 (2001)
Matteo Agostini (TU Munich & GSSI) 14
Statistical analysis
Profile likelihood analysis:
• ML fit(constant+Gauss in 1930-2190 keV range)
• multiple data sets (common T
0⌫1/2)
• T
0⌫1/2 � 0 (coverage tested)
Results (GERDA only):
• best fit for N0⌫�� = 0 signal cts
• T
0⌫1/2 > 2.1 · 1025 yr (90% C.L.)
• MC Median sensitivity (for no signal):T
0⌫1/2 > 2.4 · 1025 yr (90% C.L.)
Results (GERDA + IGEX [1] + HdM [2]):
• best fit for N0⌫�� = 0 signal cts
• T
0⌫1/2 > 3.0 · 1025 yr (90% C.L.)
PRL 111, 122503 (2013); [1] Phys.Rev. D65, 092007 (2002); [2] Eur.Phys.J. A12, 147 (2001)
Matteo Agostini (TU Munich & GSSI) 14
Double-� decay with 2⌫ or Majorons emission
• Global fit of the energy spectrum
• Most accurate measurement of:T
2⌫1/2(
76Ge) = 1.926(95)⇥ 1021 yr
(68% probability)
• Most stringent limits on exoticprocesses:
T
0⌫�1/2 > 1023 yr for n=1,3,5,7
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
250 500 750 1000 1250 1500 1750 2000
energy [keV]
arb
itra
ry u
nit
s
0νββ2νββ
n=1n=3
n=5
n=7
[Acta Phys. Polon. B 37 (2006) 1905]
data
)ββνmodel (background + 2
68% interval
backgroundββν2
(n=1) (90% C.I.)χββν0 (n=2) (90% C.I.)χββν0 (n=3) (90% C.I.)χββν0 (n=7) (90% C.I.)χββν0
energy (keV)1000 1500 2000
even
ts/(3
0 ke
V)
10
210
310
yr)⋅golden data set (17.9 kg
GER
DA
14-
12
energy (keV)1000 1500 2000
even
ts/(3
0 ke
V)
1
10
210
yr)⋅BEGe data set (2.4 kg
GER
DA
14-
12
[arxiv:1501.02345]
Matteo Agostini (TU Munich & GSSI) 15
Phase II upgrade
I Installation of additional 17 kg of BEGe detectors:
� increased array granularity (anti-coincidence cut)
� enhanced pulse shape discrimination performance
� excellent energy resolution
I PMT and fibers+SiPMto detect LAr scintillation light
I lower-mass holders
Matteo Agostini (TU Munich & GSSI) 16
Broad Energy Germanium (BEGe) detectors
0
0.5
1
1.5
2
2.5
0 500 1000 1500 2000 2500
FW
HM
[ke
V]
energy [keV]
f(x) = (a2 + b2x)1/2
a = 0.386 (8)
b = 0.0433 (2)
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.65 1.7 1.75 1.8 1.85
mass
[kg
]
FWHM at 1332 keV [keV]
enricheddepleted
natural
Matteo Agostini (TU Munich & GSSI) 17
Signal formation and development
Matteo Agostini (TU Munich & GSSI) 18
Signal formation and development
Matteo Agostini (TU Munich & GSSI) 18
Signal formation and development
Matteo Agostini (TU Munich & GSSI) 18
Signal formation and development
Matteo Agostini (TU Munich & GSSI) 18
Signal formation and development
Matteo Agostini (TU Munich & GSSI) 18
Pulse shape discrimination technique
A/E method:
E: integral of the current signal (energy)A: maximum of the current signal
0
0.2
0.4
0.6
0.8
1
0 100 200 300 400 500 600 700 800 900
curr
ent [a
.u.]
time [ns]
A
A
Multiple Interaction SiteSingle Interaction Site
A/E 0.96 0.98 1.00 1.02 1.04
no
rma
lize
d c
ou
nts
0
10
20
30
40
50
60ββν2
DEPGD32Bin Phase I
GERDA 13-06
A/E
0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
cts
(a.u
.)
0.0
0.2
0.4
0.6
0.8
1.0K n+ surface events, simulation42
Co, simulation60
Compton 1.8 - 2.2 MeV, calibration
FEP @ 1.6 MeV, calibration
GE
RD
A 1
3-0
6
[Budjas et al. JINST 4 P10007,M.A et al. JINST 6 P03005, Eur.Phys.J C73 (2013) 2583]
Matteo Agostini (TU Munich & GSSI) 19
Detection of LAr scintillation light
Design:
• low-background photo-multipliers (9top, 7 bottom)
• wave-length-shifting fibers read-out bySiPMs
• wave-length-shifting nylon mini-shroud
Matteo Agostini (TU Munich & GSSI) 20
Last commisioning results (Th-228 irradiation)
About two orders of magnitude suppression at Q�� !
Matteo Agostini (TU Munich & GSSI) 21
Last commisioning results (Ra-226 irradiation)
Almost two orders of magnitude suppression at Q�� !
Matteo Agostini (TU Munich & GSSI) 22
GERDA sensitivity projection for limit setting
• profile likelihood analysis
• MC-realizations of the data sets
• limit extraction performed for eachrealization
• global analyis:GERDA Phase IGERDA Phase II:(37 kg of 76Ge, 1e-3 cts/(keV· kg· yr))
• median sensitivity after 2 yr of datataking:
T
0⌫1/2& 1026 yr
|mee
| . 100meV
90
% C
.L.
limit
on
|m
ee| [
eV
]
time [yr]
4.6 < NME < 5.8
10-1
100
2012 2013 +0 +2 +4 / /
/ /9
0%
C.L
. lim
it o
n
T1
/2
[yr]
median
68% prob.
90% prob.
95% prob.
99% prob.
1025
1026
/ /
/ /
exp
osu
re [
kg •
yr]
101
102
/ /
/ /
Matteo Agostini (TU Munich & GSSI) 23
sensitivity projection for GERDA + Majorana
• profile likelihood analysis
• MC-realizations of the data sets
• limit extraction performed for eachrealization
• global analyisi of:GERDA Phase IGERDA Phase II:(37 kg of 76Ge, 1e-3 cts/(keV· kg· yr))Majorana demonstrator(30 kg of 76Ge, 8e-4 cts/(keV· kg· yr))
• median sensitivity after 4 yr of datataking:
T
0⌫1/2& 3� 4 · 1026 yr
|mee
| . 60� 80meV
90
% C
.L.
limit
on
|m
ee| [
eV
]
time [yr]
4.6 < NME < 5.8
10-1
100
2012 2013 +0 +2 +4 / /
/ /9
0%
C.L
. lim
it o
n
T1
/2
[yr]
global analysis
GERDA Phase I
GERDA Phase II
Majorana Dem
1025
1026
/ /
/ /
exp
osu
re [
kg •
yr]
101
102
/ /
/ /
Matteo Agostini (TU Munich & GSSI) 24
Conclusions
GERDA Phase I (21.6 kg·yr of exposure):• background order of magnitude lower than previous Ge experiments:
⇠0.01 cts/(keV· kg· yr) at Q�� (after PSD)
• blind analysis —> no positive 0⌫�� signal:
T 0⌫1/2 > 2.1 · 1025 yr at 90% C.L. (GERDA only)
• Long standing claim excluded at 99% C.L. (model-independent result)
• NEW: most accurate measurement of T 2⌫1/2
• NEW: stongest limits on T 0⌫�1/2 and T 2⌫
1/2 decay to excited states
GERDA Phase II:
• commissioning ongoing
• quasi background-free experiment
• start exploration of T 0⌫1/2 > 10
26yr in a ⇠2 yr of data taking
Matteo Agostini (TU Munich & GSSI) 25
Collaboration
⇠100 members, 16 institutions, 6 countries
Matteo Agostini (TU Munich & GSSI) 26
backup slides
Matteo Agostini (TU Munich & GSSI) 27
Electric field and charge collection
Contributions to theelectric field (E):
1) electrodes potentials:�p+ = 0V, �
n+ = 4 kV
2) impurity concentration:negative charges fordepleted p-type Ge
Total field (1+2):holes are pushed to thedetector central slice (2)and then collected to thep+ electrode (1)
E
E
E
Interplay between (1) and (2)results in the funnel e↵ect:
anodecathodeelectronsholesinteraction point
final part of hole tra-jectories independent ofinteraction positions
[JINST 6 (2011) P03005]
Matteo Agostini (TU Munich & GSSI) 28
Background model – 2⌫�� half-life
energy (keV)600 800 1000 1200 1400 1600 1800
even
ts/(3
0 ke
V)
100
200
300
400
500 experimental energy spectrum model 68%
ββν 2K42 K40 Bi214
GER
DA
12-1
2
energy (keV)600 800 1000 1200 1400 1600 1800
even
ts/(3
0 ke
V)
1
10
210
energy (keV)600 800 1000 1200 1400 1600 1800
data
/mod
el ra
tio
0.51.01.52.02.5 data/model
68%95%99.9%
I Binned maximum likelihood (5 kg·yr)I Nuisance parameters:
• Active detector masses (6+1)
• Ge-76 fractions (6)
• Background contributions (3x6)
I T
2⌫1/2 common to all detectors
I After marginalizing:
T
2⌫1/2 = (1.84+0.09
�0.08 fit+0.11�0.06 syst) · 1021
[J.Phys.G 40 (2013) 035110]
publication year1990 1995 2000 2005 2010
yr)
21 (1
01/
2T
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
ITEP
-YPI
PNL-
USC
PNL-
USC
-ITEP
-YPI PN
L-U
SC-IT
EP-Y
PI
HdM
IGEX
HdM
HdM
-K
HdM
-B
this work
Barabash
NNDC
GER
DA
12-1
2
Matteo Agostini (TU Munich & GSSI) 29
Background model – 2⌫�� half-life
energy (keV)600 800 1000 1200 1400 1600 1800
even
ts/(3
0 ke
V)
100
200
300
400
500 experimental energy spectrum model 68%
ββν 2K42 K40 Bi214
GER
DA
12-1
2
energy (keV)600 800 1000 1200 1400 1600 1800
even
ts/(3
0 ke
V)
1
10
210
energy (keV)600 800 1000 1200 1400 1600 1800
data
/mod
el ra
tio
0.51.01.52.02.5 data/model
68%95%99.9%
I Binned maximum likelihood (5 kg·yr)I Nuisance parameters:
• Active detector masses (6+1)
• Ge-76 fractions (6)
• Background contributions (3x6)
I T
2⌫1/2 common to all detectors
I After marginalizing:
T
2⌫1/2 = (1.84+0.09
�0.08 fit+0.11�0.06 syst) · 1021
[J.Phys.G 40 (2013) 035110]
energy [keV]
cou
nts
/20
keV
0
2000
4000
6000
8000
10000
12000
500 1000 1500 2000
[HdM, EPJA12 (2001) 147]
Matteo Agostini (TU Munich & GSSI) 29
Background model – ↵-emitting isotopes
I fit window 3500-7500 keV
Ip-value of the fit: 0.7
I 80 bins of width 50 keV:
79% in the green band
98% in the yellow band
Matteo Agostini (TU Munich & GSSI) 30
Comparison with Phys.Lett. B586 198 (2004)
Hypothesis test: H0 (bkg only) vs H1 (T 0⌫1/2 = 1.19+0.37
�0.23 · 1025 yr + bkg)
In Q�� ± 2�E
(after PSD):
• expected 2.0±0.3 bkg cts
• expected 5.9±1.4 signal cts(assuming H1)
• observed 3 cts
energy (keV)2025 2030 2035 2040 2045 2050 2055 2060
coun
ts/k
eV
0
1
2
3 GERDA 13-07
HT = 2.1x1025 yr(90% C.L. lower limit)
1/2 1
T = 1.19x1025 yr1/2
[PRL 111, 122503 (2013)]
GERDA only:I Frequentist p-value (N0⌫�� = 0|H1) = 0.01
I Bayes factor P(H1)/P(H0)=2.4 · 10�2
GERDA + IGEX + HdM: I Bayes factor P(H1)/P(H0)=2 · 10�4
Long standing
claim strongly
disfavoured!
T
0⌫1/2 from Mod. Phys. Lett. A 21 (2006) 1547 is not considered because of inconsistencies (i.e. missing e�ciency
factors, problem in the conversion from counts to T
0⌫1/2) pointed out in Ann. Phys. 525 (2013) 269.
Matteo Agostini (TU Munich & GSSI) 31