High-Voltage

(1)

High-Voltage

Monolithic Active Pixel Sensors for the

Mu3e Experiment

Niklaus Berger

Institut für Kernphysik, Johannes-Gutenberg Universität Mainz

October 2018

(2)

• The Mu3e Experiment

• High-Voltage Monolithic Active Pixel Sensors (HV-MAPS)

• The MuPix Sensors

• The Mu3e Data Acquisition

Overview

(3)

The Mu3e Experiment:

Searching for μ

⁺

→ e

⁺

e

^-

e

⁺

with a sensitivity of 10

^-16

(2∙10

^-15

in phase I)

(4)

e ⁺

e ⁺ e ^-

• μ⁺ → e⁺e^-e⁺

• Two positrons, one electron

• From same vertex

• Same time

• Sum of 4-momenta corresponds to muon at rest

• Maximum momentum: ½ m_μ = 53 MeV/c

The signal

(5)

• Combination of positrons from ordinary muon decay with electrons from:

- photon conversion, - Bhabha scattering, - Mis-reconstruction

• Need very good timing, vertex and momentum resolution

Accidental Background

e

⁺

e

⁺

e

^-

(6)

• Allowed radiative decay with internal conversion:

μ

⁺

→ e

⁺

e

^-

e

⁺

νν

• Only distinguishing feature:

Missing momentum carried by neutrinos

Internal conversion background

• Need excellent

momentum resolution

• New: NLO available from Matteo Fael and Signer et al. - now 10-20% easier

Branching Ratio

10^-12

10^-16 10^-18 10^-14

105 106 104

103 102

101

Internal conversion background

Signal

(7)

Building the

Mu3e Experiment

aiming for a branching ratio sensitivity of 10

^-16

(8)

2 Billion Muon Decays/s

50 ns, 1 Tesla field

(9)

• Apply magnetic field (e.g. 1 Tesla)

• Measure curvature of particles in field

• Limited by detector resolution and scattering in detector

Momentum measurement

(10)

• Apply magnetic field (e.g. 1 Tesla)

• Measure curvature of particles in field

• Limited by detector resolution and scattering in detector

Momentum measurement

• At ~ 30 MeV/c momentum: Scattering completely dominates

• Large pixels: 80 μm

• Very little material: 0.1% X₀ per layer

(11)

• Treat hit measurements as arbitrarily precise

• Consider scattering in each detector plane

• Two hits, two helices:

Underconstrained problem

• Minimize scattering angles

• Use multiple scattering theory to define χ²

Nucl. Instrum. Meth. A 844C, 135 (2017)

Multiple Scattering Track Fit

x y

z s

x₀

x₁

x₂

c₁ c₂

r₁ r₂

d₀₁ d₁₂

s₀₁ s₁₂

Φ₁ Φ₂ Φ_MS

ϕ₀₁

ϕ₁₂ B

x₀

x₁

x₂

z₀₁

z₁₂ ϑ₁

ϑ₂ Θ_MS B

Φ_MS

(12)

• 1 T magnetic field

• Resolution dominated by multiple scattering

• Momentum resolution to first order:

Σ

_P

/P ~ θ

^MS

/Ω

• Precision requires large lever arm (large bending angle Ω) and low multiple scattering θ_MS

Momentum measurement

Ω MS

θ

_MS

B

(13)

Precision vs. Acceptance

50 MeV/c 25 MeV/c 12 MeV/c B^→

33 cm

(14)

Precision vs. Acceptance

(15)

Precision vs. Acceptance

(16)

Precision vs. Acceptance

(17)

Precision vs. Acceptance

Ω ~ π MS

θ_MS

B

(18)

Detector Design

muon beam

target

(19)

Detector Design

muon beam

target

(20)

Detector Design

muon beam

target

inner pixel layers

(21)

Detector Design

outer pixel layers

muon beam

target

inner pixel layers

(22)

Detector Design

scintillating ﬁbres

outer pixel layers

muon beam

target

inner pixel layers

(23)

Detector Design

outer pixel layers

muon beam

target inner pixel layers recurl pixel

layers

(24)

Detector Design

outer pixel layers

muon beam

target inner pixel layers recurl pixel

layers

recurl pixel layers

Scintillating tiles

(25)

2

] [MeV/c m

rec

96 98 100 102 104 106 108 110

2

Events per 0.2 MeV/c

−3

10

−2

10

−1

10 1 10 10

2

at 10

-12

→ eee µ

at 10

-13

→ eee µ

at 10

-14

→ eee µ

at 10

-15

→ eee µ

ν ν

→ eee µ

muons/s muon stops at 10

8

10

15

Mu3e Phase I

Performance Simulations: Mass reconstruction

Work in

progress

(26)

High-Voltage

Monolithic Active Pixel Sensors

(27)

High voltage monolithic active pixel sensors - Ivan Perić

• Use a high voltage commercial process (automotive industry)

Fast and thin sensors: HV-MAPS

P-substrate

N-well E field

(28)

• Small active region, fast charge collection via drift

Fast and thin sensors: HV-MAPS

P-substrate

N-well E field

(29)

• Small active region, fast charge collection via drift

Fast and thin sensors: HV-MAPS

P-substrate N-well

Particle E field

• Implement logic directly in N-well in the pixel - smart diode array

• Can be thinned down to < 50 μm

(I.Perić, P. Fischer et al., NIM A 582 (2007) 876 )

(30)

Developed a series of HV-MAPS prototypes

• Goal: Detection and signal processing with just 50 μm silicon

• 6th chip, MuPix7, is a full system-on-a-chip

• Well characterized, working very nicely

• Next step is going big: 2 x 1 cm² MuPix8 under test

The MuPix Prototypes

(31)

MUPIX electronics (MuPix7)

(32)

MuPix7

3 m m

(33)

MuPix7

3 m m

Pixels with amplifier

40 x 32 pixels

80 x 103 μm pixel size

(34)

MuPix7

3 m m

Pixels with amplifier

40 x 32 pixels

80 x 103 μm pixel size

Comparator and digital pixel logic

(35)

• Hits are streamed out on a 1.25 Gbit/s LVDS link

• Up to 30 MHz hits

• Tested up to 2.5 MHz - no loss of efficiency beyond single pixel dead-time (~ 1 μs)

Fully digital output

Rate [1000/s]

0 500 1000 1500 2000 2500

Efficiency

0.99 0.991 0.992 0.993 0.994 0.995 0.996 0.997 0.998 0.999 1

χ

−

− χ

−

(36)

Tests done at

• CERN 250 GeV pions

• DESY 5 GeV electrons

• PSI 250 MeV pions

• Mainz 855 MeV electrons

• Thanks for all the beam time and support!

Beam tests

(37)

Introduction

Y

• X

(38)

Introduction

Y

• X

(39)

• Beam test at DESY with 4 GeV electrons

• 50 μm sensor, 90° incidence

• Using high-resolution EUDET-Telescope as reference

• All features well understood

MuPix7 Performance: Efficiency

column direction [mm]

0.5 1 1.5 2 2.5 3

row direction [mm]

0 0.5 1 1.5

2 2.5 3

0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

(40)

Digital readout: Resolution given by pixel size (plus reference telescope resolution)

MuPix7 Performance: Spatial Resolution

(41)

MuPix7 Performance: Time Resolution

Time difference [ns]

−200 −180 −160 −140 −120 −100 −80 −60 −40 −20 0

Number of entries / 2 ns

2000 4000 6000 8000 10000 12000

Sigma = 12 ns Mean = -110 ns

• Using 16 ns timestamps

• Relative to scintillator reference

• Sizeable tail: time-walk

(42)

MuPix7 Performance: Time resolution

ToT length [ns]

0 500 1000 1500 2000

• Single pixel with time-over-threshold signal (~ signal size)

• MuPix8 has signal size for all pixels and finer timestamps

• Can do time-walk correction

(43)

• Measurements of time delay

(At fixed threshold: proxy for signal size) with sub-pixel resolution

• Simulation using TCAD: All features can be reproduced

MuPix7 Simulation and Data

Column direction [µm]

0 50 100 150 200

hit - trigger time [ns]

−104

−102

−100

−98

−96

−94

Simulation Measurement

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8

charge collection time [ns]

0 50 100 150 200

112

− 111

− 110

−

−109 108

−

−107

Simulation Measurement

0.8 0.9 1 1.1 1.2 1.3 1.4

Column direction [µm]

charge collection time [ns]

hit - trigger time [ns]

(44)

• MuPix8, the first large sensor (2 cm x 1 cm) now available

• Currently under test

• Three sub-matrices with different signal transmission to periphery

• Results from matrix A with the Mupix7-like source follower

MuPix8

(45)

MuPix8 Architecture

(46)

MuPix8 Performance

(47)

MuPix8 Performance

(48)

MuPix8 Performance

• Charge sharing only at pixel edges

(49)

MuPix8 Performance

• Resolution given by pixel size (80 x 81 μm)

(50)

MuPix8 Performance

• 8 ns timestamps

• Some delays over the chip, large pixel- to-pixel variations: Need correction

• Further improvements possible, for matrix subset, 6 ns were obtained

(51)

MuPix8 issues

• Powering: Some voltage drop over chip, results obtained at 1.9 V or 2 V vs. 1.8 V nominal operation voltage

• Cross-talk: Long lines to the periphery have capacitive coupling

(52)

50 μm silicon is not self-supporting

• Need “no-mass” mechanics

• Also: “no-mass” connection to the outside world

See Joost’s talk

Chips are active: ~ 300 mW/cm²

• Need “no-mass” cooling

• Gaseous helium at very high flow speeds

• Prototype tests so far successful, full mock-up under construction

How to get to ~0.1 X

₀

per layer

• Note: The PANDA luminosity detector will operate MuPix in vacuum:

Cooling via diamond wafers

(53)

Data Acquisition

(54)

• 1.25 Gbit/s 8b10b encoded LVDS links

• Either three submatrices with a link each or one link multiplexing the sub-matrices

• Roughly 30 MHits/s per link maximum

• Hits are 32 bit: column, row, time, charge

• Hits are not strictly time sorted - see backup for the workings of the MuPix readout state machine

MuPix output

(55)

Phase I:

• 280 Million pixels (+ fibres and tiles)

• No trigger

• ~ 100 Gbit/s

• FPGA-based switching network

• 12 PCs with GPUs

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

4 12 Gbit/s links per

16 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

Front-end(inside magnet)

Switching Board

(56)

Front end (Altera Arria V FPGAs):

• Receive and decode data

• Correct for time-walk

• Time sorting (most resources)

• Slow control and configuration

• Send data out via 6 Gbit/s optical link

Data Acquisition

2844 Pixel Sensors

FPGA FPGA FPGA

...

86 FPGAs

GPU

PC GPU

PC

GPU 12 PCs PC

16 Inputs each

FPGA FPGA

...

12 FPGAs

6272 Tiles

FPGA FPGA

...

14 FPGAs

Data

Collection Mass Storage Gbit Ethernet

Board

Switching Board

(57)

Switching (Altera Arria 10 FPGAs):

• PCIe40 board (Marseille, LHCb and ALICE)

• Merge datastreams

• Inject pixel configuration data

• Perform monitoring tasks

Data Acquisition

2844 Pixel Sensors

FPGA FPGA FPGA

...

86 FPGAs

GPU

PC GPU

PC

GPU 12 PCs PC

16 Inputs each

FPGA FPGA

...

12 FPGAs

6272 Tiles

FPGA FPGA

...

14 FPGAs

Data Collection

Server

Mass Storage Gbit Ethernet

Board

Switching Board

(58)

PCs (Altera Arria 10 FPGAs):

• DE5a-NET board (Terasic Inc.)

• Receive data, preprocess

• DMA to GPU

• Buffering

Data Acquisition

2844 Pixel Sensors

FPGA FPGA FPGA

...

86 FPGAs

GPU

PC GPU

PC

GPU 12 PCs PC

16 Inputs each

FPGA FPGA

...

12 FPGAs

6272 Tiles

FPGA FPGA

...

14 FPGAs

Data

Collection Mass Storage Gbit Ethernet

Board

Switching Board

(59)

• 280 Million pixels (+ fibres and tiles)

• No trigger

• ~ 1 Tbit/s

• Need to find and fit billions of tracks/s

Online reconstruction

(60)

• PCs with Graphics Processing Units (GPUs)

• Online track and event reconstruction

• 10⁹ 3D track fits/s achieved

• Data reduction by factor ~1000

• Data to tape < 100 Mbyte/s

Online filter farm

(61)

• Mu3e aims for μ → eee at the 10^-16 level

• First large scale use of HV-MAPS

• Working full prototypes MuPix7 and MuPix8

• Reconstruct 100 million tracks/s in 100 Gbit/s on ~12 GPUs

• Start data taking in 2020

• 2 billion muons/s not before 2024

Conclusion

(62)

Backup Material

(63)

• Add no material:

Cool with gaseous Helium (low scattering, high mobility)

• ~ 250 mW/cm² - total ~3 kW

• Simulations: Need ~ several m/s flow

Cooling

• Full scale heatable prototype built

• 36 cm active length

• Vibrations studied using Michelson-Interferometer

• Can keep temperature below 70°C

(64)

Cooling tests

Global helium stream

Local helium stream

(65)

Readout

(66)

Readout

(67)

Readout

(68)

Readout

(69)

Readout

(70)

Readout

(71)

Readout

(72)

Readout

(73)

Readout

(74)

Readout

(75)

Readout

(76)

Readout

(77)

Readout

(78)

Readout

(79)

Readout

(80)

Readout

(81)

Readout

(82)

Readout

(83)

Readout

(84)

Readout

(85)

Readout

(86)

Readout

(87)

Readout

(88)

Readout

(89)

Readout

(90)

Readout

(91)

Readout

(92)

Readout

(93)

Niklaus Berger – μEDM October 2018 – Slide 93

Introduction

Y

• X

State Machine

StateSync Reset

sendcounter True StateSendCounter1 StateSendCounter2

sendcounter StatePD1 True

StatePD2

StateLDCol1

StateLDCol2

StateLDPix1

PriFromDet slowdownend

True

False False

StateLDPix2 True

StateRDCol1 PriFromDet False

True

StateRDCol1

PriFromDet True maxcycles

False

StateSync Reset

sendcounter True StateSendCounter1 StateSendCounter2

sendcounter StatePD1 True

StatePD2

StateLDCol1

StateLDCol2

StateLDPix1

PriFromDet slowdownend

True

False False

StateLDPix2 True

StateRDCol1 PriFromDet False

True

StateRDCol1

PriFromDet True maxcycles

True False

False

(94)

• 90° incidence angle

• 99% efficient for less than 10 Hz noise per pixel

MuPix7 Performance: Efficiency vs. Noise

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

(95)

• MuPix8 has higher resistivity substrate:

45° signal at 90°

MuPix7 Performance: Efficiency vs. Noise

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

−1

10 1 10 102

103

104

Efficiency

0.975 0.98 0.985 0.99 0.995 1

Efficiency Noise

99 %

Preliminary

−1

1 10

High-Voltage

High-Voltage

Monolithic Active Pixel Sensors for the

Mu3e Experiment

Niklaus Berger

Institut für Kernphysik, Johannes-Gutenberg Universität Mainz

October 2018

• The Mu3e Experiment

• High-Voltage Monolithic Active Pixel Sensors (HV-MAPS)

• The MuPix Sensors

• The Mu3e Data Acquisition

Overview

The Mu3e Experiment:

Searching for μ

→ e

e

e

with a sensitivity of 10

(2∙10

in phase I)

e +

e + e -

The signal

Accidental Background

e

e

e

μ

→ e

e

e

νν

Internal conversion background

Building the

Mu3e Experiment

aiming for a branching ratio sensitivity of 10

2 Billion Muon Decays/s

Momentum measurement

Momentum measurement

Multiple Scattering Track Fit

Σ

/P ~ θ

/Ω

Momentum measurement

Ω MS

θ

B

Precision vs. Acceptance

Precision vs. Acceptance

Precision vs. Acceptance

Precision vs. Acceptance

Precision vs. Acceptance

Detector Design

muon beam

target

Detector Design

Detector Design

Detector Design

Detector Design

Detector Design

Detector Design

] [MeV/c m

96 98 100 102 104 106 108 110

Events per 0.2 MeV/c

10

10

10 1 10 10

at 10

→ eee µ

at 10

→ eee µ

at 10

→ eee µ

at 10

→ eee µ

ν ν

→ eee µ

muons/s muon stops at 10

10

Mu3e Phase I

e ⁺

e ⁺ e ^-