The Mu3e Experiment

Searching for the Decay µ → eee The Mu3e Experiment in Phase I

Ann-Kathrin Perrevoort

Physics Institute, Heidelberg University

VISTAS on Detector Physics December 11, 2017

The Mu3e Experiment

SINDRUM Mu3e

Bµ→eee<1.0⋅10⁻¹² at 90%CL [1988]

Single-event sensitivity of 2⋅10⁻¹⁵ in phase I^∗

Challenges

• Background-free operation

▸ Very good vertex and time resolution

▸ Excellent momentum resolution

• High muon stopping rates of 10⁸µ/s

▸ Detectors and data acquisition need to cope with high rates

∗[SES of∼10⁻¹⁶in phase II]

Signal and Background

Signal Background

Signalµ⁺ → e⁺e⁻e⁺ Accidental background Internal conversion µ⁺ → e⁺e⁻e⁺νµνe

• Common vertex

• Coincident

• ∑Ee=m_µ

• ∑p⃗e=0

• No common vertex

• Not coincident

• ∑Ee≠m_µ

• ∑p⃗e≠0

• Common vertex

• Coincident

• ∑Ee<m_µ

• ∑p⃗e≠0

Multiple Coulomb Scattering

Ω MS

θMS

B

• Decay electrons have low momentum<53 MeV/c

• Momentum resolution is dominated by multiple scattering

Scattering angle: θ_MS∝

√ x X₀

Resolution: σ_p

p ∼θ_MS Ω

→ Reduce material thickness x

→ Increase opening angle Ω

Multiple Coulomb Scattering

Ω ~ π MS

θMS

B

• Decay electrons have low momentum<53 MeV/c

• Momentum resolution is dominated by multiple scattering

Scattering angle: θ_MS∝

√ x X0

Resolution: σp

p ∼θ_MS Ω

→ Reduce material thickness x

→ Increase opening angle Ω⇒ Ω≈π

Multiple Coulomb Scattering

• Decay electrons have low momentum<53 MeV/c

• Momentum resolution is dominated by multiple scattering

Scattering angle: θ_MS∝

√ x X0

Resolution: σp

p ∼θ_MS Ω

→ Reduce material thickness x

→ Increase opening angle Ω⇒ Ω≈π

The Mu3e Detector

Target Inner pixel layers

Outer pixel layers Recurl pixel layers

Scintillator tiles μ Beam

Tracking detector:

Thin Si pixel sensors (HV-MAPS)

+ Timing detector:

Scintillating fibres and tiles B-field: 1 T

Length: 110 cm Diameter: 18 cm

Pixel Sensors: HV-MAPS

High Voltage Monolithic Active Pixel Sensors

• AMS 180 nm HV-CMOS process

• N-well in p-substrate

• Reverse bias of∼80 V

▸ Fast charge collection via drift

▸ Depletion zone of a few 10µm Thinning possible (≲50µm)

• Integrated readout electronics

• Pixel size 80×80µm² Sensor size 2×2cm²

P-substrate N-well

Particle E field

I. Peri´c, NIM A 582 (2007)

Pixel Sensors: MuPix7 Prototype

• Small-scale prototype

▸ 32×40 pixels `a 103×80µm²

▸ 2.9×3.2mm²of active area

▸ 50µm thin

• Integrated signal processing

▸ Amplification and signal shaping within the pixel

▸ Hit detection and digitisation in periphery

▸ Zero-suppressed data output:

pixel address and time stamp

▸ LVDS link at 1.25 Gbit/s

Pixel Sensors: MuPix7 Prototype

• Small-scale prototype

▸ 32×40 pixels `a 103×80µm²

▸ 2.9×3.2mm²of active area

▸ 50µm thin

• Integrated signal processing

▸ Amplification and signal shaping within the pixel

▸ Hit detection and digitisation in periphery

▸ Zero-suppressed data output:

pixel address and time stamp

▸ LVDS link at 1.25 Gbit/s

Efficiency >99% at low noise rates

Threshold [V]

0.7 0.71 0.72 0.73 0.74 0.75

Efficiency

0.95 0.96 0.97 0.98 0.99 1

Efficiency Noise 99 %

Preliminary

Noiserate per pixel [1/s]

−1 10 1 10 102 103 104

Pixel Sensors: MuPix8 Prototype

• First large MuPix sensor: 2×1cm²

• 128×200 pixels `a 81×80µm²

• Two different approaches for the line driver

• Analogue pulse information for time walk correction

• Different substrates:

20Ωcm and 80Ωcm

Pixel Sensors: MuPix8 Prototype

column 0 20 40 60 80 100 120

row

0 20 40 60 80 100 120 140 160 180

0 200 400 600 800 1000 1200

Preliminary hitmap of a Sr-90 source

• First large MuPix sensor: 2×1cm²

• 128×200 pixels `a 81×80µm²

• Two different approaches for the line driver

• Analogue pulse information for time walk correction

• Different substrates:

20Ωcm and 80Ωcm

Pixel Sensors: Cooling

Cooling with gaseous helium Power consumption of Si pixel sensors is 250 mW/cm²

SciFi

Pixel Layers

Scintillating Tiles

water cooled beam pipe water cooled beam pipe

Scintillating Tiles Pixel Layers

gaseous helium

Pixel Sensors: Cooling

Cooling with gaseous helium Power consumption of Si pixel sensors is 250 mW/cm²

T[°C]

60 50 40 30 20 10 0 70

400mW/cm²

Timing Detector: Scintillating Fibres

Entries 24772

Mean 0.05009 ±0.008864 Std Dev 1.394 ±0.006268 Integral 2.474e+04

/ ndf

χ2 327.1 / 54

p0 191.7 ±7.7

p1 0.09037 ±0.03579 p2 2.363 ±0.047

p3 2596 ±26.8

p4 0.05652 ±0.00492 p5 0.5724 ±0.0056

) / 2 [ns]

- tR t(L 10

− −5 0 5 10

Counts

0 500 1000 1500 2000 2500

3000 Entries 24772

Mean 0.05009 ±0.008864 Std Dev 1.394 ±0.006268 Integral 2.474e+04

/ ndf

χ2 327.1 / 54

p0 191.7 ±7.7

p1 0.09037 ±0.03579 p2 2.363 ±0.047

p3 2596 ±26.8

p4 0.05652 ±0.00492 p5 0.5724 ±0.0056

Time resolution of square fibres

• 3 layers of fibres with ∅∼250µm and length of 28 to 30 cm

• Photon detection at both ends with LHCb SiPM column array

• Readout with custom-designed MuTRiG

• Round and square fibres under investigation

• Prototype with 3 layers of square multiclad fibres:

σ_t=(572±6)ps and ϵ≳95%

Timing Detector: Scintillating Tiles

Mezzanine Board Connector 448 Channel

Module

Endring Cooling

Pipe

Scintillator Tiles

MuTRiG PCB SiPM Flexprint

Time Difference [ps]

-400 -200 0 200 400

# Entries

0 5 10 15 20 25 30

103

×

TWC No TWC ) ps

× 2 = (56 σ

) ps

× 2 = (70 σ

• 6.5×6.5×5.0mm³ tiles with individual SiPMs

• Custom-designed MuTRiG:

TDC ASIC for SiPM readout

• Prototype yields time resolution

∼70 ps and efficiencyϵ≳99.7%

Data Acquisition

2844 Pixel Sensors

up to 45 1.25 Gbit/s links

FPGA FPGA FPGA

...

86 FPGAs 1 6 Gbit/s

link each

GPU

PC GPU

PC GPU

12 PCs PC 12 10 Gbit/s

links per

8 Inputs each

3072 Fibre Readout Channels

FPGA FPGA

...

12 FPGAs

6272 Tiles

FPGA FPGA

...

14 FPGAs

Data Collection

Server

Mass Storage Gbit Ethernet

Switching

Board Switching

Board Switching

Board

Front-end(inside magnet)

Switching Board

Data Acquisition

2844 Pixel Sensors

up to 45 1.25 Gbit/s links

FPGA FPGA FPGA

...

86 FPGAs 1 6 Gbit/s

link each

GPU

PC GPU

PC GPU

12 PCs PC 12 10 Gbit/s

links per

8 Inputs each

Data

Collection Mass Storage Gbit Ethernet

Switching Board

Switching Board

Triggerless data acquisition Front-end board

▸ Decode and merge data of∼15(36)sensors

▸ Time-sorting Switching board

▸ Switch between front-end and filterfarm

▸ Merge data of sub-detectors GPU filterfarm

▸ Fast online track and vertex reconstruction

▸ Data rate reduction for mass storage from 60 Gbit/s to 100 MB/s

The Mu3e Simulation Framework

• Full Geant4-based detector simulation

• Detector geometry and beam

• Generators of physics processes

• Track reconstruction and vertex fit

Sensitivity Studies for µ → eee

96 98 100 102 104 106 108 1102 2 Events per 0.2 MeV/c

−4

10

−3

10

−2

10

−1

10 1 10 102

at 10-12

→ eee µ at 10-13

→ eee µ at 10-14

→ eee µ at 10-15

→ eee µ ν

ν

→ eee µ

Bhabha +Michel

muons/s muon stops at 108

1015

Mu3e Phase I

Sensitivity Studies for µ → eee

Data taking days

0 50 100 150 200 250 300

eee)→µBR(

−15

10

−14

10

−13

10

−12

10

−11

10

10-15

× 2

SES 90% C.L. 95% C.L.

Mu3e Phase I 10⁸ muon stops/s 19.7% signal efficiency

SINDRUM 1988

Summary

Mu3e Search for LFV decay µ →eee Low-material tracking detector

▸ Thin Si pixel sensors

▸ Scintillating fibres and tiles Triggerless data acquisition and online event filtering

Phase I Expected single-event sensitivity of 2⋅10⁻¹⁵ in 300 days of data taking Status Finalizing phase I detector design

Preparation for construction and commissioning

Mu3e Collaboration

University of Geneva Heidelberg University

Karlsruhe Institute of Technology JGU Mainz

Paul Scherrer Institute ETH Z¨urich

University of Z¨urich University of Bristol University of Liverpool University College London University of Oxford

The Phase II Detector

Final sensitivity of one in 10¹⁶ muon decays

Target Inner pixel layers

Outer pixel layers Recurl pixel layers

Scintillator tiles μ Beam

Thin timing detector

Increase muon stopping rate to 2⋅10⁹µ/s

Additional recurl stations increase acceptance for recurler Smaller beam profile ⇒smaller target radius

Background: µ → eeeνν

0 5 10 15 20 40 60

10^-18 10^-16 10^-14 10^-12 10^-10 10^-8

0.85 0.9 0.95 1

E//MeV

dB dE/ Kfactor

no cuts onE/ E/≤20 MeV E/≤10 MeV E/≤5 MeV K factor

NLO calculations for µ → eeeνν: Pruna, Signer, Ulrich [arXiv:1611.03617]

Background: µ → eeeνν

10⁻¹⁵ 10⁻¹⁴ 10⁻¹³ 10⁻¹² 10⁻¹¹ 10⁻¹⁰ 10⁻⁹ 10⁻⁸

0.6 0.7 0.8 0.9

101 102 103 104 105

error band×10 dB/dm123[MeV−1]

µ⁻→e⁻(e⁺e⁻)νµν¯e

BNLO/BLO

m123[MeV]

10⁻²⁰ 10⁻¹⁸ 10⁻¹⁶ 10⁻¹⁴ 10⁻¹² 10⁻¹⁰ 10⁻⁸ 10⁻⁶ 10⁻⁴

0.7 0.8 0.9

1 10 100

error band×10

B(

/ E^max

)

µ⁻→e⁻(e⁺e⁻)νµν¯e

BNLO/BLO

E/max[MeV]

NLO calculations for µ → eeeνν: Fael, Greub arXiv:[1611.03726]

Experimental Area

Target E Infrastructure platform I Infrastructure

platform II

Access walkway

Detector control and filter farm barracks Controlled

access door

MEG II

Existing πE5 front access

Mu3e Removable access stairway

Experimental Area

Magnet and Detector Cage

Target

• Extended hollow double-cone target made of ∼80µm mylar foil

• 10 cm long with a radius of 19 mm

• High muon stopping fraction

• Vertex separation over a large surface

• Low distortion for outgoing electrons

Pixel Sensors: HV-MAPS

Pixel Periphery State Machine

readout state machine

VCO

&

PLL

8b/10b

encoder serializer LVDS ...

other pixels

sensor CSA

comparator tune

DAC

threshold baseline source

follower

test-pulse injection

readout 2^nd amplifier

integrate charge

amplification

line driver

digital output AC coupling

via CR filter per pixel threshold adjustment

Hit finding, digitisation, zero-suppression and readout on-chip Continuous and fast readout at 1.25 Gbit/s

Pixel Sensors: MuPix9 Prototype

MuPix9 submitted in August

• Small-scale prototype

• Slow control

• Serial powering

Lightweight Mechanics

• 50µm silicon sensor

• 80µm Flexible printed circuit board (FPC)

• 25µm Kapton support structure

→ ∼0.1%of radiation length

Aluminium 14 um

Aluminium 14 um Glue 5 um Glue 5 um Polyimide 25 um Polyimide 10 um

Polyimide 10 um MuPix 50 um

Kapton frame 25 um

Bus signals Power distribution Sensor signals Ground SpTAB pads Dielectric spacer layer

{

FPC

Mechanical support

Sensor

Glue 5 um Glue 5 um

MuPix Telescope

• System integration test for pixel tracker front-end

• Characterisation of prototypes

• 4 or 8 planes of MuPix7 well established

• 4 planes of MuPix8 just commissioned

Data Acquisition at the MuPix Telescope

Receiver

• Receive and de-serialize data Unpacker

• Disentangle hit and status information Hit sorter

• Merge data from all sensors to one datastream

• Sort hit data by time stamp Data transfer to PC

to PC

PCIe

MuPix MuPix MuPix MuPix

LVDS LVDS LVDS LVDS

Sorting at the Pixel Tracker Front-End

Pixel Matrix

Periphery

Readout scheme on MuPix

• Skip empty columns and rows

• Read only first hit in a column

• Hits that happened at the same time do not necessarily end up in the same readout cycle

⇒ Restore time structure at front-end using time stamp information

Sorting at the Pixel Tracker Front-End

Pixel Matrix

Periphery

read in a later cycle

Readout scheme on MuPix

• Skip empty columns and rows

• Read only first hit in a column

• Hits that happened at the same time do not necessarily end up in the same readout cycle

⇒ Restore time structure at front-end using time stamp information

Sorting at the Pixel Tracker Front-End

READING WRITING

i i-7

i-6

i-5 i-4 i-1

i-2 i-3

Sort hits by time stamp on FPGA

• Write hits in time-order to memory Address: time stamp + counter

• Dual-port RAM used as ring-buffer

• Divided into 8 blocks

either accessible for writing or reading

• Skip addresses without data

Front-End Board

• MuPix telescope:

Stratix IV development board plus custom PCBs

• First prototype of front-end board:

Stratix IV, optical transceivers, . . .

• Later prototypes will use Arria V

Reconstruction and Event Filtering

• Online reconstruction on GPU filter farm

▸ Selection cuts

▸ Fast 3D multiple scattering track fit

▸ Triplet from three consecutive hits

▸ Add fourth hit

▸ Subsequent vertex fit with three trajectories of correct charge

▸ Selectµ → eee candidates

▸ Reduction of data rate from 60 Gbit/s to 100 MB/s

• Offline Reconstruction

▸ Reconstruction of recurling tracks

▸ Refined vertex fit