meg – the experiment to search for

MEG – The Experiment to Search for μ → eγ

T. Moria∗

aInternational Center for Elementary Particle Physics, The University of Tokyo7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan

The MEG experiment, being prepared at the Paul Scherrer Institute (PSI), is soon to start its search for thelepton flavor violating decay of a muon into an electron and a gamma ray, μ → eγ, with a two orders of magnitudehigher sensitivity than the previous experiments. The supersymmetric grand unification and the seesaw model forthe neutrino masses generally predict branching ratios that are within the reach of the MEG experiment. Here,after a brief description of its concept and detectors, the status and the future plan of the MEG experiment arereported.

1. Why search for μ → eγ?

In a mysterious three generation structure, par-ticles seem to mix among different generations.Quarks mix each other according to the Cabibbo-Kobayashi-Maskawa matrix, while neutrinos os-cillate into each other at seemingly maximal lev-els. Then it appears very unnatural that onlycharged leptons are essentially forbidden to mix inthe Standard Model. A discovery of their mixingsmay therefore lead to a deeper understanding ofprinciples and symmetries behind the generationstructure which lie beyond the Standard Model.

Many processes have been experimentally stud-ied to look for the charged lepton mixings. Cur-rently the most actively studied are the leptonflavor violating (LFV) τ decays at the B facto-ries [1]. Searches for LFV muon decays, such asμ → eγ, have been continuously carried out sincethe discovery of the muon and provide the moststringent limits on many of the new physics sce-narios beyond the Standard Model.

As clarified by Masiero and Herrero in thisworkshop [2], LFV processes of charged leptonsare expected to occur through the loops of newTeV-scale particles such as those predicted by thesupersymmetric theories. Therefore the study ofthe LFV processes is of similar significance to thenew particle searches at the Large Hadron Col-

∗Supported by MEXT Grant-in-Aid for Scientific Re-

search on Priority Areas 441.

lider (LHC).On the other hand, the source of LFV in

charged leptons is thought to originate from muchhigher energy scale governed by grand unified the-ories (GUT) [3] or seesaw models [4] that pre-dict ultra-heavy right-handed neutrinos [5]. Thestudy of the LFV processes therefore could alsoprovide hints of physics at extremely high energyscales, which are not accessible even by the LHCexperiments.

The MEG experiment [6] is designed to con-siderably increase the sensitivity of the search forthe LFV decay of the muon, in order to reachbranching ratios predicted by the supersymmet-ric grand unified theories and the supersymmet-ric seesaw models. In this presentation, after abrief introduction of experimental issues concern-ing the μ → eγ search in general, an overview ofthe MEG experiment is given. The status andthe prospects of the experiment are also reportedat the end.

2. What a μ → eγ Experiment looks like

A μ → eγ event is characterized by the sim-ple 2-body final state. The decay electron andthe gamma ray are emitted in opposite directionswith the energy equal to half the muon mass, i.e.52.8 MeV. To utilize this simple kinematics in thesearch, muons are stopped in a material (a stop-ping target). Positive muons are used to avoid

Nuclear Physics B (Proc. Suppl.) 169 (2007) 166–173

0920-5632/$ – see front matter © 2007 Elsevier B.V. All rights reserved.

www.elsevierphysics.com

doi:10.1016/j.nuclphysbps.2007.03.018

Figure 1. A schematic view of the MEG experiment.

formation of muonic atoms by the target nuclei.For positive muons, so-called surface muons

can be abundantly produced by bombarding pri-mary protons into a thick production target. Thesurface muons come from the decays of pions thatstop near the surface of the production targetand have a sharp momentum spectrum around29 MeV/c. Because of this narrow momentumspread, they are very efficiently stopped by a thintarget (∼100 μm). A thin stopping target is veryimportant to achieve good resolutions in measur-ing positrons and to reduce background of anni-hilation gamma rays. Since the surface muonsare naturally 100% spin polarized, angular dis-tribution of μ → eγ can be measured once theyare discovered. The μ → eγ angular measure-ment should eventually help pin down the sourceof LFV.

The present upper limit on the branching ratioof μ → eγ is 1.2×10−11 [7]. In order to reach thebranching ratio of 10−11–10−14 predicted by thesupersymmetric theories, either GUT or seesawmodels, a very high muon rate is necessary. As-suming a detection efficiency of ε ≈ 5% and therunning time of a few years, T ≈ a few×107sec,a stopped muon rate of 1013/ε/T ≈ 107/sec is

required to achieve a single-event sensitivity of10−13.

With such a high muon rate, the major back-ground in a μ → eγ search is an accidental co-incidence of a positron, coming from the stan-dard Michel decays of muons, and a gamma ray,coming from radiative muon decays or annihi-lation of positrons. This accidental backgroundactually limits the final experimental sensitivity.On the other hand, the radiative muon decays,μ → eνν̄γ, which could mimic the μ → eγ decayswhen the neutrinos carry very small energies, canbe controlled at an order of magnitude smallerrate than the accidental background.

Since the accidental background increasesquadratically as the muon rate, a continuous DCmuon beam, that has the lowest instantaneousrate, is best suited for a μ → eγ search, ratherthan a pulsed beam.

The positron detector must have a very goodtracking capability even at a high rate of afew×107/sec. It also must be of low material,since its measurements of 52.8 MeV positrons arelimited by multiple scattering in the stopping tar-get and the detector material.

The gamma ray detector must have very good

T. Mori / Nuclear Physics B (Proc. Suppl.) 169 (2007) 166–173 167

energy resolution to strongly suppress the back-ground gamma rays whose rate drops very quicklywith the energy. In addition good coincidencetiming measurements of positrons and gammarays are needed to further reduce the accidentalbackground.

The MEG experiment, which will be describedin the following, has been carefully designed tofulfill most of these severe requirements in orderto reach a sensitivity goal of 10−13–10−14.

3. MEG Experiment

The MEG experiment [6], a μ → eγ search ex-periment, currently being prepared at the PaulScherrer Institute (PSI) in Switzerland, was pro-posed by the Japanese physicists and has sinceevolved to an international collaboration amongJapan, Italy, Switzerland, Russia, and the U.S.A.The experiment is scheduled to get ready forphysics run in 2007 and aims at a sensitivity of10−13, two orders of magnitude below the presentlimit, in a few years’ data taking.

The experimental set-up is schematicallyshown in Fig. 1. A DC surface muon beamis focused and stopped in a thin slanted CH2

target placed at the center of the experiment.Gamma rays are measured by the liquid xenonscintillation detector located just outside the su-perconducting solenoidal magnet called COBRA.Positrons are tracked by the low material driftchambers inside COBRA which provides a spe-cially graded magnetic field. Their timings aremeasured by the plastic scintillation counters inthe second turn of their trajectories.

The unprecedented sensitivity of the MEG ex-periment has been made possible by three keycomponents: (1) the highest intensity DC sur-face muon beam available at PSI; (2) a spe-cially designed COBRA positron spectrometerwith graded magnetic field; and (3) an innova-tive liquid xenon scintillation gamma ray detec-tor. These key components are described in somedetail in the following sections.

3.1. Muon Beam

The 590 MeV proton cyclotron at PSI, con-stantly operating at 50 MHz with a beam current

Figure 2. The πE5 beam line. From the bottomtoward the top: the electrostatic separator, thequadrupole magnet, the beam transport solenoid,and the COBRA magnet with the compensationcoil (a large ring).

exceeding 1.8 mA and a total beam power of morethan 1 MW, is able to produce the highest inten-sity DC surface muon beam in the world. Thisis the best and only place for a high sensitivityμ → eγ experiment.

The πE5 beam line for the MEG experimentis shown in Figure 2. An electrostatic separa-tor is placed to reject unwanted positrons in thebeam and a superconducting solenoidal magnet isused to transport and focus the beam onto a thinslanted CH2 target with a spot size of σ ≈ 10 mm.A muon stopping rate of 108/sec has been alreadydemonstrated, which is more than required forthe experiment (a few ×107/sec).

T. Mori / Nuclear Physics B (Proc. Suppl.) 169 (2007) 166–173168

Figure 3. The special graded magnetic field of COBRA has two advantages over a normal solenoid: (A)Positrons are quickly swept away from the tracking volume. (B) Positrons with the same momentumtravel with the same bending radius.

It is noted that, because of its DC beam struc-ture, the instantaneous muon rate for the MEGexperiment is about 10 times lower than the pre-vious experiment which used a pulsed beam with6− 7 % duty cycle [7]. Thus MEG is expected tosuffer much less accidental background.

3.2. Positron Spectrometer

The COBRA (COnstant Bending RAdius)positron spectrometer [8] consists of a supercon-ducting solenoidal magnet designed to form a spe-cial graded magnetic field (1.27 T at the centerand 0.49 T at both ends), in which positrons withthe same absolute momenta follow trajectorieswith a constant projected bending radius, inde-pendent of the emission angles over a wide angu-lar range of | cos θ| < 0.35 [Figure 3 (B)]. This al-lows to sharply discriminate high momentum sig-nal positrons out of a few×107 Michel positrons

emitted every second from the target. Becausethe drift chambers are placed at larger radii, onlyinteresting high momentum positrons enter thechamber volumes. The graded field also helpsto sweep away curling tracks quickly out of thetracking volume [Figure 3 (A)], thereby reducingaccidental pile-up of the Michel positrons in thedrift chambers.

High strength Al stabilized conductor is used tomake the magnet as thin as 0.197X0, so that 85%of 52.8MeV/c gamma rays traverse the magnetwithout interaction before entering the gammaray detector placed outside the magnet. A He-free, simple and easy operation of the magnet isrealized with a GM refrigerator. As the COBRAmagnet does not have a return yoke, a pair ofcompensation coils (seen as a large ring in Fig-ures 2 and 6) suppress the stray magnetic field be-low 50 Gauss in the vicinity of the gamma ray de-


tector, so that the photomultiplier tubes (PMTs)can operate.

The drift chambers are made of 12.5 μm thinfoils supported by C-shaped carbon fiber frameswhich are open for the entrance of positrons tominimize material traversed by the positrons.The foils have special “vernier” cathode pads tomeasure the hit positions along the anode wireswith an accuracy of about 500 μm. The driftchambers use a gas mixture of He:C2H6 =1:1.Outside the drift chambers, the whole interior ofthe COBRA magnet is filled with He gas to mini-mize multiple scattering. The gas pressures of thedrift chambers and inside COBRA are carefullycontrolled at a few Pa level. Altogether, the totalmaterial budget for a positron tracking is of theorder of 1.5 × 10−3 radiation lengths.

High energy positrons, after passing throughthe drift chambers, eventually enter the timingcounters in their second turns. There are two tim-ing counters at the upstream and the downstreamof the target, each of which consists of a layer ofplastic scintillator fibers and 15 plastic scintilla-tor bars of 4 × 4 × 90 cm3. The fibers measurehit positions along the COBRA axis and the barsmeasure timings of positrons with a precision ofσ = 40 psec. Under the high magnetic field up to1 Tesla, fine-mesh PMTs are used to read out sig-nals from the bars, while avalanche photo diodes(APDs) are used for the fibers. Because the coun-ters are placed at larger radii, only high energypositrons hit them, giving a total positron hit rateof a few×104/sec for each bar counter.

3.3. Gamma-Ray Detector

An innovative liquid xenon (LXe) scintillationdetector was specially devised for this experimentto make very precise measurements of energy, po-sition and timing of gamma rays [9]. The detectorholds an active LXe volume of 800 without anysegmentation. Scintillation light emitted insideLXe are viewed from all sides by 846 PMTs thatare immersed in LXe in order to maximize directlight collection.

High light yield of LXe (roughly 80 % of NaI)and its uniformity are necessary ingredients forgood energy resolution. A scintillation pulse fromxenon is very fast and has a short tail of 45 nsec,

Figure 4. Energy resolutions obtained by thelaser Compton scattering at AIST and the chargeexchange reaction at PSI. Indicated are σ of thehigher edge of the spectrum.

thereby minimizing the pile-up problem. Distri-butions of the PMT outputs enable a measure-ment of the incident position of the gamma raywith a few mm accuracy. The position of theconversion point is also estimated with an accu-racy that corresponds to a timing resolution ofσ ≈ 50 psec.

Various studies were carried out using a 100 prototype detector with 238 PMTs in order togain practical experiences in operating such a newdevice and to prove its excellent performance [9].

Special PMTs that work at the LXe tempera-ture (-110◦C), persist under high pressures andare sensitive to the VUV scintillation light ofLXe (λ ≈ 178 nm), have been developed incollaboration with Hamamatsu Photonics. Themetal channel dynode structure makes them com-pact and robust against magnetic field up to≈ 50 Gauss. To stabilize the gain of the PMTs inthe high rate environment, while keeping the heatload inside LXe minimum, a base circuit withZener diodes was newly designed.

High power (200 W) pulse tube refrigeratorsoptimized for the LXe temperature have been de-veloped at KEK [10]. They have enabled sim-ple and carefree detector operations. The whole900 volume of LXe for the MEG experiment


Figure 5. The cryostat vessel of the LXe detectorat the manufacturer in Italy.

is safely kept in the storage tank equipped withthe pulse tube refrigerator. Many groups of re-searchers working with LXe are also starting touse these refrigerators.

Absorption of scintillation light by impuritiesinside LXe, especially water and oxygen, couldsignificantly degrade the detector performance,although there is no absorption by LXe itself. Agas-phase purification system that circulates andpurifies xenon gas was developed and an absorp-tion length longer than a few meters was achievedfor the prototype detector after a few months’operation [11]. To speed up the purification pro-cess for the final detector, a liquid-phase purifica-tion system that uses a cryogenic centrifugal fluidpump capable to flow 100 of LXe per hour wasdeveloped and successfully tested [12].

The prototype detector was tested by usinggamma rays from laser Compton scattering atNational Institute of Advanced Industrial Sci-ence and Technology (AIST) in Tsukuba, Japan.Gamma rays with the Compton edge energy of10, 20 and 40 MeV were generated via backwardscattering of laser photons by 800 MeV electronbeam in the TERAS storage ring. The gammaray beam was collimated by lead blocks into aspot size of 2 mm φ to evaluate the position res-olution.

Another test was carried out at PSI by usingthe pion charge exchange reaction, π−p → π0n,which provides two gamma rays from the π0

decay. By tagging back-to-back gamma rays,monochromatic gamma rays of 55 MeV and83 MeV are selected.

The energy resolutions obtained by these beamtests are shown in Figure 4. Similarly, the tim-ing resolution of 65 psec and the position resolu-tion of ≈ 4 mm depending on the incident posi-tion with respect to the PMT positions. Overall,these beam tests demonstrated the excellent per-formance of the detector to achieve the sensitivitygoal of the experiment.

An important issue for this large unsegmenteddetector is pile-up gamma rays. Multiple gammarays can be separated by their positions and tim-ings. To identify and separate pile-up efficiently,fast waveform digitizing is used for all the PMToutputs. A new Domino Ring Sampling (DRS)chip has been developed at PSI for this pur-pose [13]. The chip contains eight channels whichare digitized at up to 4 GHz with a resolution of12 bits. They are operated at 2 GHz for the LXedetector. The DRS chips are also used to readout all the channels of the drift chambers and thetiming counters.

Most challenging in operating this detector ishow we can maintain its high performance duringthe long period of data taking. Various calibra-tion methods have been prepared to ensure theperformance during the experiment: (1) LEDsequipped inside to make frequent PMT gain eval-uations; (2) alpha sources on thin wires [14] (aspecial feat only for liquid detectors) to evaluatequantum efficiencies of the PMTs and attenua-tion of the scintillation light; (3) laser pulses via


Figure 6. Installing the timing counter into theCOBRA magnet for the pilot run of the positronspectrometer at the end of 2006.

optical fibers for crude relative timing calibrationwith the timing counters; (4) gamma rays fromπ0 in the charge exchange reaction of π− on liq-uid hydrogen target; (5) 17.6 MeV gamma raysfrom pLi → γBe using the Cockcroft Walton ac-celerator; (6) 9 MeV gamma rays using neutronsource (Cf) on Nickel plates; and (7) radiativemuon decays, μ → eνν̄γ, where both the gammaray and the positron are detected, for absolutetiming calibration between the LXe detector andthe timing counters.

As shown in Figure 5, the LXe detector is beingconstructed at the time of writing (the beginningof 2007).

3.4. Status and Prospects

A pilot run of the positron spectrometer wascarried out at the end of 2006. The drift cham-bers and the timing counters, though not fullyequipped, were installed inside the COBRA mag-net, together with the stopping target and thegas pressure control system (Figure 6). After aquick tuning of the muon beam, one week of datataking runs with various beam conditions werecarried out.

On the other hand the LXe detector awaits theconstruction of the cryostat vessels and the hon-

Figure 7. A simulated μ → eγ event.

eycomb entrance window (Figure 5), which arescheduled to be completed in April - May, 2007.After assembling the detector and liquefying andpurifying LXe, the LXe detector will be calibratedusing various calibration methods, in particular,the Cockcroft Walton accelerator in the summer,2007.

Thus the whole experiment should be readyfor commissioning and then physics runs later in2007. It is expected to take 4 × 107 sec of datataking (≈ 2 + α years) with a muon beam inten-sity of 3×107/sec to reach a 90% C.L. sensitivityof 1.2 × 10−13 with accidental background of 0.6events.

4. Conclusion

Lepton flavor violation in charged leptons isa clean and clear signal of new physics beyondthe Standard Model. It not only evidences newphysics at TeV scale, such as supersymmetryor extra dimensions, but also provides hints ofphysics at extremely high energies, such as grand


unification of forces that might have triggered in-flation of the universe, or heavy Majorana neutri-nos that might be the origin of matter.

The MEG experiment is scheduled to get readyfor physics run in 2007 and could come acrosswith some charming events such as shown in Fig-ure 7 at any time during the following years. Sostay tuned.

REFERENCES

1. T. Ohshima, “Study of LFV in Tau Decayat Belle,” these proceedings; S. Banerjee,“Searches for lepton flavor violating decaysτ± → ±γ, τ± → ±P 0 (where − = e−, μ−,and P 0 = π0, η, η′) at B-Factories: Status andCombinations,” these proceedings.

2. A. Masiero, “Probing Supersymmetric GrandUnification through Flavor Physics,” theseproceedings; M.J. Herrero, “LFV in tau andmuon decays within SUSY seesaw,” these pro-ceedings.

3. R. Barbieri and L.J. Hall, Phys. Lett. B 338(1994) 212; R. Barbieri, L.J. Hall, and A.Strumia, Nucl. Phys. B 445 (1995) 219.

4. J. Hisano and D. Nomura, Phys. Rev. D 59(1999) 116005 and references therein.

5. M. Gell-mann, P. Ramond and R. Slansky, inSupergravity, Proceedings of the Workshop,Stony Brook, New York, 1979, ed. by P. vanNieuwenhuizen and D.Freedman (North-Holland, Amsterdam); T. Yanagida, in Pro-ceedings of the Workshop on Unified The-ories and Baryon Number in the Uni-verse, Tsukuba, Japan, edited by A. Sawadaand A.Sugamoto (KEK Report No. 79-18,Tsukuba) (1979).

6. T. Mori et al., Research Proposal to PSI,R-99-5, May 1999; A. Baldini et al., Re-search Proposal to INFN, Sep. 2002. Thesedocuments and the updated informationof the MEG experiment are available athttp://meg.web.psi.ch/docs/.

7. MEGA Collaboration, M.L. Brooks et al.,Phys. Rev. Lett. 83 (1999) 1521.

8. W. Ootani et al., IEEE Trans. Applied Su-perconductivity, 14 (2005) 568-571; A. Ya-mamoto et al., Nucl. Phys. B 78 (1999) 565-

570.9. S. Mihara et al., Cryogenics, 44 (2004) 223-

228.10. T. Haruyama et al., “LN2-Free Operation of

the MEG Liquid Xenon Calorimeter by Usinga High-Power Pulse Tube Cryocooler,” KEKPreprint 64 (2005).

11. A. Baldini et al., Nucl. Instr. Meth. A 545(2005) 753-764; A. Baldini et al., IEEE Trans.on Dielectrics and Electrical Insulation, 13(2006) 547-555.

12. S. Mihara et al., Cryogenics, 46 (2006) 688-693.

13. S. Ritt, “The DRS2 Chip: A 4.5 GHz Wave-form Digitizing Chip for the MEG Experi-ment,” IEEE/NSS Conference, Rome, Italy,Oct. 16-22, 2004.

14. A. Baldini et al., Nucl. Instr. Meth. A 565(2006) 589-598.


meg – the experiment to search for

Documents