Skip to main content

Full text of "Final results on the neutrino magnetic moment from the MUNU experiment"

See other formats

Final results on the neutrino magnetic moment from the MUNU 

The MUNU collaboration: Z. Daraktchieva a , C. Amsler b , M. Avenier c , C. Broggini d , J. Busto a , 
C. Cerna d , F. Juget a , D.H. Koang c , J. Lamblin c , D. Lebrun c , O. Link b , G. Puglierin d , A. Stutz c , 
A. Tadsen d , J.-L. Vuilleumier a , V. Zacek c 

a Institut de physique, A.-L. Breguet 1, CH-2000 Neuchatel, Switzerland 
b Physik-Institut, Winterthurerstr. 190, CH-8057 Zurich, Switzerland 

c Laboratoire de Physique Subatomique et de Cosmologie, IN2P3/CNRS-UJF, 53 Avenue des Martyrs, 
F-38026 Grenoble, France 

d INFN, Via Marzolo 8, 1-35131 Padova, Italy 

°Universite de Montreal, C. P. 6128, Montreal, P.Q. Canada H3C 3J7 

The MUNU detector was designed to study v e e~ elastic scattering at low energy. The central component is 
a Time Projection Chamber filled with CF4 gas, surrounded by an anti-Compton detector. The experiment was 
carried out at the Bugey (France) nuclear reactor. In this paper we present the final analysis of the data recorded 
at 3 bar and 1 bar pressure. Both the energy and the scattering angle of the recoil electron are measured. From 
the 3 bar data a new upper limit on the neutrino magnetic moment nl hort < 9 ■ 1CP 11 (ib at 90 % CL was derived. 
At 1 bar electron tracks down to 150 keV were reconstructed, demonstrating the potentiality of the experimental 
technique for future applications in low energy neutrino physics. 

1. Introduction 

The MUNU experiment was designed to study 
V e e~ elastic scattering at low energy and to probe 
the existence of a magnetic moment of the elec- 
tron antineutrino. The detector is located at 18 
m from the core of a 2.75 GWth commercial nu- 
clear reactor in Bugey (France) . The central com- 
ponent is a CF4 gas Time Projection Chamber 
(TPC). In we presented an analysis of 66.6 
days live time of reactor-on data, as well as 16.7 
days of reactor-off data, taken at a pressure of 3 
bar. In this Letter we present the final analysis, 
using the same data set, and the same event selec- 
tion, but taking better advantage of the electron 
kinematics, extending the area in which the back- 
ground is measured and achieving a more precise 
determination. Moreover we present the analysis 
of 5.3 days live time of reactor-on data taken at 

1 bar pressure. 

Technical details of the MUNU detector have 
already been presented in ref. |2!3| . Here we only 
describe the most essential features. 

2. The experiment 

The central part of the detector is a cylindri- 
cal TPC filled with CF4 gas. Measurements were 
performed at a pressure of 3 bar (11.4 kg of CF4) 
and 1 bar (3.8 kg). The gas serves as target and 
detector medium for the recoil electrons. CF4 was 
chosen because of its high density, low atomic 
number, which reduces multiple scattering, and 
its absence of free protons, which eliminates the 
background from V e p — > e + n. As shown in fig. ^ 
the gas is contained in a 1 m 3 acrylic vessel 90 
cm in diameter and 160 cm long. The drift vol- 
ume is defined on one end by a plain cathode and 


Figure 1. The MUNU detector at the Bugey re- 

Figure 2. The four kinematical cones. 

the other one by a grid made from wires. An an- 
ode plane made of 20 /im wires with a pitch of 
4.95 mm, separated by 100 /im potential wires, is 
placed behind the grid to collect and amplify the 
ionization charge. The integrated anode signal 
gives the total deposited energy. An x - y read- 
out plane is located behind the anode plane. It 
contains x strips on one side, and perpendicular y 
strips on the other one. The pitch is 3.5 mm. The 
signals induced in these strips provide the spatial 
information in the x and y directions. The third 
projection z is obtained from the time evolution 
of the signal. The drift field was selected such as 
to achieve a drift velocity of 2.15 cm/^s. 

It must be noted that the detector is installed 
under the reactor core, at an angle of 45°. The 
detector axis is perpendicular to the reactor core- 
detector axis. The anode wires are parallel to the 
reactor core-detector axis, and the grid wires per- 
pendicular to it. The x strips are vertical and the 
y strips horizontal. The read-out plane is there- 
fore by construction symmetric with respect to 
four directions: reactor core to detector and op- 
posite, as well as the two orthogonal directions. 

The acrylic vessel is immersed in a steel tank 
(2 m diameter and 3.8 m long), filled with 10 
m 3 of liquid scintillator (NE235) and viewed by 

48 photomultipliers (PMT). The liquid scintilla- 
tor acts as an anti-Compton detector and as a 
veto against cosmic muons. The anti-Compton 
detector also sees the primary scintillation light 
of heavily ionizing particles such as a's, and the 
secondary light emitted during the amplification 
process around the anode wires, which provides 
a second measurement of the total deposited en- 

In addition, the detector is shielded against lo- 
cal activities by 8 cm of boron loaded polyethy- 
lene and 15 cm of Pb. The concrete and steel 
overburden of the laboratory corresponds to the 
equivalent of 20 m of water. 

Events in the TPC not in coincidence with a 
signal above 22 MeV in the scintillator within an 
80 (is time window are recorded. The selection of 
good events off-line proceeds in two steps. First 
a software filter rejects the muon related events, 
Compton electrons, discharges and uncontaincd 
events. The selection is finalized in a visual scan. 
A neutrino scattering candidate event is a con- 
tinuous single electron track fully contained in a 
42 cm fiducial radius, with no energy deposition 
above 90 keV in the anti-Compton in the pre- 
ceding 200 /is. The initial direction of the elec- 
tron track is obtained from a visual fit pQ. The 
scattering angles in the x-z and y-z projections 
are determined first, and used to calculate the 


500 1000 1500 2000 


= 200 - 




50 - 

500 1000 1500 2000 


Figure 3. 3 bar data, reactor-on, energy spec- 
tra; top: upward (dashed line), downward (solid 
line) and backward (dotted line) electrons; bot- 
tom: forward (solid line) and NB (dashed line) 

scattering angle reac with respect to the reactor- 
detector axis, which coincides quite precisely with 
the scattering angle, as well as the angle 9d e t with 
respect to the detector axis (see fig El- The angle 
ipdet between the projection of the initial track 
direction on the x-y plane and the vertical y axis 
is also determined. 

As described in more details in we apply 
the angular cut Qdet < 90° to suppress the back- 
ground from activities on the read-out plane side 
of the TPC, which was found to be noisier. This 
is presumably due to the greater complexity of 

the anode side and the larger inactive volume in 
the scintillator. 

The reactor neutrino spectrum, necessary to in- 
terpret the data, was calculated using the formal- 
ism described in |1I4 |. Above 1.5 to 2 MeV neu- 
trino spectra reconstructed from the measured (3 
spectra of the fission fragments were used. The 
uncertainty is of order 5 % or less in this en- 
ergy range, in which the neutrino spectrum was 
moreover thorougly probed in measurements of 
V e p — ■> e + n scattering at reactors. At lower en- 
ergies calculations only are available. We have 
used the calculated neutrino spectra of the fis- 
sion fragments from |5lfi| , and taken into account 
the neutron activation of 238 U, as discussed in [7], 
which is significant below 1 MeV. The uncertainty 
in the neutrino spectrum in this energy range was 
estimated to 20 %. 

3. 3-bar forward- normalized background 

The 3 bar data were taken with a TPC trig- 
ger threshold of 300 keV. Considering the entire 
event selection procedure, the live time, limited 
primarily by the data transfer rate and the anti- 
Compton, was 65 %. At 3 bar the tracks are long 
enough to be scanned with sufficient efficiency for 
electron kinetic energies T e >700 keV. As before 
we rely on kinematics to select the good candi- 
date events. For each electron track the neutrino 
energy E v is reconstructed from the scattering 
angle, taken as 9 reac , and the measured electron 
recoil energy T e . Events with E v > are de- 
clared forward events since, effectively, this crite- 
ria selects electrons with an initial track direction 
within a forward cone, the axis of which coincides 
with the reactor core-detector axis. The open- 
ing angle depends on the energy, and is larger 
than that of the kinematic cone for recoil elec- 
trons (E v > T e ). Simulations taking into account 
the angular response of the detector show that 
nearly 100 % of the recoil electrons fall in the for- 
ward category, which however also contains a con- 
tribution from the background, which is isotropic 
around the detector axis. 

To estimate the background the same proce- 
dure is applied in the three directions equiva- 


lent, considering the read-out plane, to the reac- 
tor core-detector direction, taken as reference for 
the forward events. These directions are thus also 
equivalent from the point of view of the angular 
response in (fdet, which is not completely linear, 
and of the acceptance. As depicted in figG] these 
directions define the backward cone, opposite to 
the forward cone, and the two perpendicular up- 
ward and downward cones 8 . To avoid overlap 
of the cones, which can occur for T e < 2m e c 2 , we 
require in addition that the angle tpdet is within 
less than 45° with respect to the cone axis. This 
only reduces the acceptance of the forward cone 
for recoil electrons in a negligible way. 

While the forward electrons contain recoil plus 
background events, the backward, upward and 
downward electrons are background events only. 
The energy distributions of the upward, downward 
and backward electrons (1154±34 in total) are 
presented in fig. [3J The distributions are compat- 
ible within the errors, which confirms the isotropy 
of the background inside the TPC. 

500 1000 1500 2000 


Figure 4. Energy distribution of the forward mi- 
nus NB electrons above 700 keV, 3 bar, reactor- 
on; comparison with the expected spectrum for 
weak interaction alone (dashed line) and for a 
magnetic moment of 9 ■ 10~ n /is (dotted line). 

We normalize the background to the forward 
cone by dividing by 3 the rates in the backward, 
upward and downward cones. This normalized 

background (NB) is then directly compared with 
the event rate in the forward cone. The energy 
distributions of both forward (455±21) and NB 
(384±11) electrons are shown in figGl There is 
a clear excess of forward events, from the reac- 
tor direction. The total number of events for- 
ward minus NB above 700 keV is 71±23 counts 
for 66.6 days live time reactor-on, corresponding 
to 1.07±0.34 counts per day (cpd). The forward 
minus NB spectrum representing the measured 
electron recoil spectrum is displayed in fig. 0] 

We make the same analysis with the data taken 
during the reactor-off period as a cross check (16.7 
days live time). The energy distributions of both 
forward (133±11) and NB electrons (147±7) are 
given in fig|S] The integrated forward minus NB 
rate above 700 keV is -0.8±0.8 cpd, consistent 
with zero. 

We now turn to the comparison with expecta- 
tions. Monte Carlo simulations (GEANT 3) were 
used to calculate the various acceptances of the 
event selection procedure. The detector contain- 
ment efficiency in the 42 cm fiducial radius was 
found to vary from 63 % at 700 keV, 50 % at 1 
MeV to 12 % at 2 MeV, with a typical error of 2 
%. The errors on the global acceptance, includ- 
ing track reconstruction efficiency (4 %), are of 
order of 7 %. This leads to an expected event 
rate above 700 keV of 1.02±0.10 cpd assuming 
a vanishing magnetic moment. This is in good 
agreement with the observed 1.07±0.34 cpd. 

The measured and expected recoil spectra for 
no magnetic moment are compared in fig 01 First 
we note that the large excess of events in the first 
two bins (700- 900 keV) observed in our previous 
analysis pQ has to a large extend disappeared. It 
is thus most likely due to a statistical fluctuation 
in the background, more precisely determined in 
the present analysis. And second the measured 
and expected spectra are seen to be in agreement. 

For a more quantitative analysis the x 2 method 
was used as in pQ, with the same binning for 
Gaussian statistics to apply (100 keV bins from 
700 keV to 1400 keV, and then a bin from 1400 
to 2000 keV). The magnetic moment is varied 
to find the best fit. We remind that the neutri- 
nos travel over a short distance only so that the 
experiment probes the magnetic moment /j, s e hort , 


1500 2000 


500 1000 1500 2000 


Figure 5. 3 bar data, reactor-off, energy spec- 
tra; top: upward (dotted line), downward (solid 
line) and backward (dashed line) electrons; bot- 
tom: forward (solid line) and NB (dashed line) 

as described in ref. pQ, to be precise its square. 
The allowed range at the 90 % CL is (p' e h ort ) 2 = 
(-0.72 ± 1.25) • 10~ 20 fi%, with x 2 =H-5 for 7 
degrees of freedom at the central value. This re- 
sult is compatible with a vanishing magnetic mo- 
ment. To obtain a limit on n s J %ort we renormalize 
to the physical region ((nl hort ) 2 > 0) and obtain 
fif ort < 9(7) • 10" 11 [i b at 90(68) % CL. This 
constitutes a small improvement over our previ- 
ous analysis , restricted to recoil energies above 
900 keV. The improvement results from the lower 
threshold and the better background estimation. 

We can compare this with results from other 
experiments. The TEXONO collaboration is per- 
forming an experiment near the Kuo-Sheng re- 
actor in Taiwan Thanks to the use of an 
Ultra Low Background High Purity Germanium 
detector they achieve a very low threshold of 12 
keV. The reactor-on minus reactor-off rates were 
found to be identical and from that the limit 
[if ort < 1.3 • 10- 10 [i B at 90 % CL was de- 
duced. Superkamiokande |1UI5| reported the limit 
[if 1 < 1.5 • 10~ 10 [i B at 90 % CL from the study 
of the shape of the recoil spectrum of electrons in 
the scattering of solar neutrinos. Depending on 
oscillation parameters, this quantity may however 
differ from [i s e hort . 

4. 1-bar forward- normalized background 

The technical details of the 1 bar measurements 
will be presented elsewhere. Here we restrict our- 
selves to the most relevant parameters. The trig- 
ger threshold on the electron recoil was lowered 
to 100 keV. The trigger rate is 0.4 s _1 . The dead- 
time, here also mostly due to the data read-out 
and transfer time, and to the anti-Compton, is 
around 50 %, somewhat less than for the mea- 
surements at 3 bar. The energy calibration of the 
TPC is again determined with 137 Cs and 54 Mn ra- 
dioactive sources. The energy resolution is found 
to vary from 10 % (1 a) at 200 keV to 6 % at 600 
keV following an empirical E Q - 57 law. Data were 
accumulated during 5.3 days live time reactor-on. 

The selection of good events proceeds in two 
steps, as with the 3 bar data, ending with a visual 
scan. Tracks can be reconstructed with reason- 
able efficiency for energies above 150 keV. Events 
above that energy are retained. As an example a 
190 keV electron track in 1 bar of CF4 is shown 
in figEl The end of the track is easily identified 
from the increased energy deposition (blob). The 
containment efficiency of the TPC in the 42 cm 
fiducial radius for recoil electrons was calculated 
with the GEANT3 simulation code, and found to 
vary from 85 % at 200 keV, 50 % at 400 keV to 
5 % at 1 MeV. The angular resolution was de- 
termined by scanning visually simulated electron 
tracks. It varies from 15° (1 a) at 200 keV, 12° 


Figure 6. A 190 keV electron at 1 bar in the TPC; 
the xz, yz projections are displayed as well as the 
anode signal. 

at 400 keV, to 10° above 500 keV. 

The angular cut 6det <90° is also applied in 
the 1 bar analysis. The electrons are selected in 
the same four kinematical cones. Here however 
the overlap of the cones is more critical. The 
constraint in ipdet is done somewhat differently: 
downward electrons 190° to 270°, forward elec- 
trons 80° to 190°, upward electrons 0° to 80° and 
backward electrons 270°to 0°. 

This makes the solid angle for the forward 
electrons somewhat larger. We expect from the 
Monte Carlo simulations that nearly 100 % of the 
recoil events above 200 keV will fall in it. The nor- 
malized background NB in the three other cones 
has to be scaled accordingly, with a factor which 
remains close to 3 however. The energy distri- 
butions of the upward, downward and backward 
electrons (326±18 in total) are presented in fig0 
They are seen to be compatible within the errors. 

The energy distributions of the forward 
(124±11) and NB (109±6) electrons are displayed 
in fig, [7| The difference is 15±12 events, giv- 


S 40 - 

30 - 

20 - 

10 - 

200 400 600 800 101 



g 40 - 

30 - 

20 - 

10 - 


Figure 7. 1 bar data, reactor-on, energy spec- 
tra; top: upward (dotted line), downward (solid 
line) and backward (dashed line) electrons; bot- 
tom: forward (solid line) and NB (dashed line) 

ing an indication of a signal, corresponding to 
2.89±2.39 counts per day. The energy distribu- 
tion is shown in figlBl This measured total rate 
above 200 keV agrees with the expected value 
0.49±0.12 counts per day, assuming no magnetic 
moment. The error comes mainly from the un- 
certainties in the low energy part of the neutrino 
spectrum, as described above. Due to the limited 
statistics the data do not have a competitive sen- 
sitivity to the neutrino magnetic moment. Nev- 
ertheless our experiment shows that a gas TPC 
can be used to measure the energy and direction 


a 20 - 


5 - 

-5 I 1 

200 400 600 800 1000 


Figure 8. Energy distribution of the forward mi- 
nus NB electrons above 200 keV, 1 bar, reactor- 
on; comparison with the expected spectrum for 
weak interaction alone (dashed line). 

of electrons with energies as low as 150 keV. Pro- 
vided background problems can be solved, this 
opens the way, for instance, to on-line measure- 
ments of low energy solar neutrinos from the pp 

the staff of EDF-CNPE Bugey for the hospitality. 
This work was supported by IN2P3, INFN, and 


1. The MUNU Collaboration, Z. Daraktchieva, 
ct al., Phys. Lett. B 564 (2003) 190 

2. The MUNU Collaboration, C. Amsler, et al., 
Nucl. Inst. Meth. A 396 (1997) 115. 

3. The MUNU collaboration, M. Avcnier ct al., 
Nucl. Inst, and Mcth. A482 (2002) 408 

4. G. Zacek et al., Phys. Rev. D34 (1986) 2621 

5. J. F. Beacom and P. Vogcl, Phys. Rev. Lett. 
83 (1999) 5222 

6. P. Vogel and J. Engel, Phys. Rev. D39 (1989) 

7. V.I. Kopeikin et al. Phys. of Atomic Nuclei 
Vol. 60 (1997) 172 

8. Z. Daraktchieva, Thesis, University of 
Neuchatel, 2004 

9. The TEXONO collaboration, H. B. Li, et 
al.,Phys. Rev. Lett. 90 (2003) 131802. 

10. The SupcrKamiokandc collaboration (Y. 
Fukuda et al., Phys. Rev. Lett. 82 (1999) 2430 

5. Conclusions 

The MUNU experiment studied V e e~ scatter- 
ing at low energy near a nuclear reactor, measur- 
ing both the energy and scattering angle of the 
recoil electron. Thanks to a better estimation of 
the background in a larger kinematical domain 
it was possible to reduce the statistical uncer- 
tainties, analyzing data taken at 3 bar of CF4 
during 66.6 days live time reactor-on with an en- 
ergy threshold of 700 keV. Good agreement is ob- 
served between the measured and expected recoil 
spectra assuming weak interaction alone. From 
this the limit on the neutrino magnetic moment 
Hf ort < 9 • 10" 11 lib at 90 % CL was derived. 
Moreover reactor-on data were taken at 1 bar of 
CF4 during 5.3 days live time. Electron tracks 
were reconstructed efficiently down to 150 keV 
recoil energy. This demonstrates the potentiality 
of gas TPC's for possible future applications in 
low energy neutrino physics. 

The authors want to thank the directors and 






I — [ — i 
| o i 

i t i 

i ! i 

i 1 !