The upgrade of the KOTO CsI calorimeter for the separation of 𝒏 and 𝜸

Nobuhiro Shimizu for the KOTO collaboration1

2

Introduction

⚫ 𝐾𝐿 → 𝜋0𝜈 ҧ𝜈⚫

⚫Detector

Signal: 2𝛾 + nothingTwo electromagnetic showers in CsI and no signal in the hermetic veto counters

KOTO experiment

3

JHEP 11 033 (2015)

𝜸

𝜸

ҧ𝜈𝜈

+NP

𝐾𝐿 𝜋0

Very rare decay:ℬSM 𝐾𝐿 → 𝜋0𝜈 ҧ𝜈 = (3.0 ± 0.3) × 10−11

CsI calorimeter of the KOTO detector

CsI crystal⚫undoped CsI (𝜆~300 nm)

#crystal = 27162240 small (25×25 mm2)

476 large (50×50 mm2)

Excellent energy resolution⚫ 𝜎𝐸/𝐸 = 0.99%⨁1.74%/ 𝐸[GeV]

PMT signals are digitized by ADC⚫14 bit 125 MHz sampling with

Gaussian filter

⚫512 ns timing window (64 samples)

⚫ Timing resolution 𝜎𝑡 ∼ 1ns

4

1m

JPS Conf. Proc. 8, 024007 (2015)

27𝑿𝟎

ADC

0 63

Neutron background 5

To achieve SM sensitivity, we need to suppress neutrons by a factor of ten

shallow distribution uniform distribution

Idea of the CsI calorimeter upgrade 6

PMT

PMT

𝜸

𝐧𝐞𝐮𝐭𝐫𝐨𝐧

𝑿𝟎 ∼ 𝟐 cm

interaction length 〜 40 cm

Attach MPPCs

S13360-6050CS (HPK)

CsI crystal

Previous

MPPCs

upstream

6×6 𝐦𝐦𝟐

PMT

Measure the depth with the time difference 𝚫𝑻 ≡ 𝑻𝑴𝑷𝑷𝑪 − 𝑻𝑷𝑴𝑻

→ Small 𝚫𝑻 implies 𝜸

7

Readout

New front-end readout

MPPC readout# of MPPCs: 4080 (>#PMT=#CsI)

Bias connection

8

To reduce # of channels..4 MPPCs are connected

“Hybrid” bias connection• adopted by MEG II upgrade• AC line: series, to read out signals• DC line: parallel, to apply bias voltagereadout

2r

r r 2r

r r

“Hybrid”

...

.

. ..

.

.

.

.

..

Crystals

5cm

5cm

Large

2.5

cm

10cm

Small

Development of front-end: amplifier 9

4 MPPCs are connected ×1/4

4 readout is summed × 1/4

→Manageable number of channels

4080 MPPCs

1020

256 channels

𝑽𝒐𝒖𝒕−

𝑽𝒐𝒖𝒕+

mixer (sum amp)+HVHybrid

H

H

H

32 hybridsare connected to one board

Segmentation 10

Read 10 cm x 10 cm region as a single channel

～typical size of EM shower

Small crystalsLarge crystals

11

Installation of MPPCs

MPPC on the CsI surface

Two problems to be addressed① Concave shape of MPPC

② Bubbles appear at low temperature

12

𝑻 ∼ 𝟓℃

Quartz

silicone

MPPC

UV-transparentGlueCsI

Quartz plate to make sure the flatness and transparency in advance

weight

Keep positive pressure inside the glue

Wait for curing strong glue

MPPC installation 13

Frames to support gluing jigs

Quartzplate

Board

MPPC instrumentation 14

Glue MPPCs on two rows in a day Start from 1st Oct. and 45 days to

finish all Installation finished as scheduled

1st Oct. 15th Nov.1st Nov.

4080

#MP

PC Progress

Time lapse movie

days20 30 40100

New outer wear of the calorimeter 15

-2018 2019-

2019 run

16

Data correction in 2019 runs 17

All physics runs

Good physics runs after selection

Neutron control sample

Collected good quality data with∼ 𝟐𝟎 × 𝟏𝟎𝟏𝟖 POT

(almost same amount as 2015 data)

Neutron control sample∼ 𝟑 × 𝟏𝟎𝟏𝟖 POT

In spite of the limited time schedule, we could collect physics data with new detectors!

MPPC irradiation 18D

ark

curr

en

to

f M

PP

C (

uA

)

1st2nd3rd

1st layer

2nd layer

Increase of dark current of MPPC due to irradiation→ 20 times larger than what we had expected

Dark current after 2019 runs

Separately evaluated by another beam test at Kobe Tandem facility

0

75userbeamstart

Run81end

Run82start

Run82end

Layer of MPPCs

Date

MPPC irradiation 19D

ark

curr

en

to

f M

PP

C (

uA

)

1st2nd3rd

Increase of dark current of MPPC due to irradiation→ 20 times larger than what we had expected

Dark current after 2019 runs

Separately evaluated by another beam test at Kobe Tandem facility

0

75userbeamstart

Run81end

Run82start

Run82end

Even with >50 times the irradiation in 2019, the timing resolution will be within our specification

Layer of MPPCs

1st layer

2nd layer

Date

Analysis

20

Control sample 21

𝒎𝟑𝝅𝟎 distribution

Control of sample of 𝜸

𝐾𝐿 → 3𝜋0 decay• Clean and abundant control

sample of 𝛾

Control sample of neutron

Insert Al plate in the upstream of detector.Scattered neutrons well emulate the BG of physics data

Gamma energy (MeV)

𝚫𝑻 distribution of the control samples 22

max𝚫𝑻 ≡ 𝐦𝐚𝐱 𝚫𝑻𝟏, 𝚫𝑻𝟐

✓Use the larger 𝚫𝑻 out of two clusters (max𝚫𝑻)✓𝐾𝐿 → 3𝜋0 MC well reproduces the distribution of data

𝜸

neutron

Emulated distribution of 𝑲𝑳 → 𝝅𝟎𝝂ത𝝂 23

✓ The spectra of 𝐾𝐿 → 𝜋0𝜈 ҧ𝜈 is harder than 𝐾𝐿 → 3𝜋0 decay

✓ Weight based on 𝑤 𝐸 = 𝑃𝐷𝐹 𝜋0𝜈ഥ𝜈 (𝐸)/𝑃𝐷𝐹 3𝜋0 (𝐸)

Emulated 𝚫𝑻 distribution weighted spectrum of 𝐾𝐿 → 3𝜋0 with 𝑤 𝐸

Retaining 90% of 𝜸 from 𝑲𝑳 → 𝝅𝟎𝝂ഥ𝝂 decay, neutron contribution can be suppressed down to 1/45 !Achieve much better performance than the goal of design 1/10

𝑬𝜸 spectrum

Summary

KOTO collaboration aims to search for New Physics via very rare decay 𝐾𝐿 → 𝜋0𝜈 ҧ𝜈, ℬSM = (3.0 ± 0.3) × 10−11.

In the last autumn, we attached >4000 MPPCs on the front surface of CsI crystal to improve 𝜸/n separation power:

In 2019 run, we successfully collected physics data and confirmed⚫ irradiation of MPPCs was acceptable for future data collection⚫performance of the neutron rejection was to be 1/45

(for 90% 𝜀 of signal), which was much better than our goal of design

Detector paper is now under preparation

24

25

Thank you!

That’s all

Degradation of neutron rejection 26

Effi

cien

cy o

f n

eutr

on

s w

ith

ret

ain

ing

90

% e

ffic

ien

cy o

f𝛾

Correlation b.t.w. CSDDL vs 𝚫𝑻

high E clus

low E clusLet us separate regions into “front” and “rear” region and evaluate the reduction by applying CSDDL value>0.9.

low E clusfront

high E clusfront

high E clusrear

low E clusrear287/15304

=1.9%

599/35176=1.7%

358/39080=0.92%

85/11400=0.75%

We cannot see strong degradation of the performance of neutron rejection by CSDDL.

(Rather, E dependence can be observed.)

Q. Does the CSD see the depth of clusters?

Correlation does not look large.

27

Correlation b.t.w. PSLH vs 𝚫𝑻

Let us separate regions into “front” and “rear” region and evaluate the reduction by applying PSLH value>0.1. (pulse shape template is run74,75)

low E clus front

high E clus front high E clus rear

low E clus rear

For high energy cluster, we can observe small correlation between PSLH and 𝚫𝑻.

Nevertheless, the degradation of performance is by a factor of 1.5.

Q. Does the PSLH see the depth of clusters?

Correlation does not look serious.

6873/15304 =44.9%±0.4%

16926/35716 =48.1%±0.2%

5839/39080 =14.9%±0.2%

2683/11400=23.4%±0.4%

high E clus

low E clus

28

Effect of the irradiation

Dark current⚫ increases ×100 for

⚫Prepare irradiated sample of MPPCs

Instability of bias voltage

29

∼ 1 × 109 1MeV- 𝑛/cm2 (3-years operation)

readout

Series

.

𝑹 (cm)

Position dependence of the neutron fluence

Position dependence of different 𝐼 𝑉causes instability of bias for the series connection.

→ Solved by adopting the Hybrid connection

Beam test at RCNP-Osaka cyclotron⚫𝛾/𝑛 beam from Li target

Performance tests (𝛾/𝑛 separation) 30

Distribution𝚫𝒕 ≡ 𝑻𝑴𝑷𝑷𝑪 − 𝑻𝑷𝑴𝑻

PMT𝜸/nMPPC

p392MeV

Li target

collimatorCsI

𝜸: continuous beamup to 392 MeV

𝒏: 392 MeV

upstream downstream

Retain 90% of 𝛾 whilesuppressing 𝒏 to 34%

Performance tests (𝜎Δ𝑡) Beam test at the ELPH (Tohoku, Japan)

electron synchrotron

⚫evaluate 𝜎Δ𝑡 (as a func. of E)• Monochromatic 200, 400, 600, 800

MeV 𝑒+ beams

⚫Used setup as realistic as possible

⚫Confirmed MPPC functionality after dose→Irradiated MPPCs were used

31

◉Beam

summed100×100 𝐦𝐦𝟐

region

✓ Irradiated MPPCs worked enough✓ Readout worked well

𝝈𝜟𝒕

(ns)

𝑬𝒃𝒆𝒂𝒎

𝐾 → 𝜋𝜈 ҧ𝜈 decay 32

Suppressed by FCNC in the SM

Small QCD uncertainty⚫useful prove to the New Physics

Two compatible processes⚫𝐾+ → 𝜋+𝜈 ҧ𝜈: 𝒜 ∝ 𝑉𝑡𝑑⚫𝐾𝐿 → 𝜋0𝜈 ҧ𝜈 : 𝒜 ∝ Im𝑉𝑡𝑑

ℬ(𝐾+ → 𝜋+𝜈 ҧ𝜈) = 17.3−10.5+11.5 × 10−9 E949

ℬ(𝐾𝐿 → 𝜋0𝜈 ҧ𝜈) < 2.6 × 10−8 (90% C.L.) E391a

𝓑(𝑲+ → 𝝅+𝝂ഥ𝝂)

𝓑(𝑲

𝑳→𝝅𝟎𝝂ഥ 𝝂)

JHEP11 033 (2015).

EXP

SM prediction

ℬ(𝐾+ → 𝜋+𝜈 ҧ𝜈) = (9.11 ± 0.72) × 10−11ℬ(𝐾𝐿 → 𝜋0𝜈 ҧ𝜈) = (3.0 ± 0.3) × 10−11

JHEP11 033 (2015).

+NP

Rejection of neutron BG

Halo-neutron BGResult of 4 days run: ℬ 𝐾𝐿 → 𝜋0𝜈 ҧ𝜈 < 5.1 × 10−8 (90% C.L.)

⚫We need 3 more magnitudes of suppression two-dimensional shower envelope → 1/10 ✓done

Pulse shape likelihood → 1/10 ✓done

33

* Prog. Theor. Exp. Phys. (2017) 021C01

*

The largest contribution from BG

measure shower development (in z)

in the calorimeter →O(1/10)

Performance tests (𝛾/𝑛 separation) 34

→ suppresses halo neutron BG to 10%while retaining 90% efficiency of two 𝜸signal events!

Taking into account…

MC evaluation of performance for halo neutron events,based on the result of beam test

① Correlation of two cluster position:• the second cluster is deeper

② Other neutron cuts

𝚫𝒕𝒎𝒊𝒏

𝚫𝒕𝒎𝒂𝒙

𝒏 neutron case

𝜸 gamma case

𝚫𝒕𝒎𝒊𝒏

𝚫𝒕𝒎𝒂𝒙

The larger one → 𝚫𝒕𝒎𝒂𝒙

The smaller one → 𝚫𝒕𝒎𝒊𝒏

Quality assurance of MPPCs 35

Quartz gluingSolderingtemperature test

MPPCs

I/V inspectionLED test

Process 80 MPPCs/day

I/V curves#MPPC～500

Summed MPPCs

Individual test

Inspect all of MPPCs (#~4000) before installation → Start gluing on CsI in this summer

Development of front-end: monitor 36

0.1uF

51Ω0.1uF

510Ω

510Ω130kΩ

10kΩ

+

-

to AMP

Signal readout

to ADC

Current monitor+HV

𝑰 𝑽 =𝜶 𝑽 − 𝑽𝟎

𝟐

𝟏 − 𝜷 𝑽 − 𝑽𝟎𝟐 + 𝜸

DC dark current is continuously monitored to confirm the functionality and level of radiation damage.

Operation current increases by a factor of 100 in three snowmass year:𝑰𝒐𝒑 = 0.5𝜇A→ 50𝜇A

AC DC

Development of front-end

MPPC readout

Bias connection

37

readout

Series

.

readout

2r

r r 2r

r r

“Hybrid”

...

.

. ..

.

.

.

.

..

Cross section of CsI

1mTwo types of crystals

4 MPPCs are simultaneously read out

S13360-6050CS (HPK)

6×6 mm2 sensitive regionSi-window

☺ small time constant (~200ns) high bias voltage (220V) unstable individual operation voltage toward irradiation

large time constant (~0.5us)☺ low bias voltage (55V)☺ stable breakdown voltage

toward irradiation

“Hybrid”-connection• adopted by Meg2 upgrade• AC line: series• DC line: parallel• have both pros

readout

Parallel

adopted!

2716 crystals

Fabrication of MPPCs 38

Drop glue

1 1

Insert MPPC on jig

2

3

4

Drop glue onquartz

2

3

4 wait for cure keeping the quartz floated

5

dispense epoxy glue (araldite 2011)

6

apply weight

6

7

7

Put MPPCs into oven andwait 24 h (keeping 45 deg)

8

wait 24 h for cure

39

Summed MPPCs

PIC microcontroller

I/V inspection of MPPCs 40

16ch MUX

16ch OPAMPs (8×2)

20 cm

10

cm

+HV

to MUX

⋯

ADC

OPAMP→FET input（high impedance）Gain 100

16×3

I/V inspection front end

I/V conversion

Basic design• 16ch are chosen by MUX • DC voltage is buffered by voltage follower

after the MUX

MUX

register

4bit

Bottom view

Buffer

AMP