Nuclear emulsion scanning - PowerPoint Presentation

Slide1

Nuclear emulsion scanning for double strangeness nuclei search

Junya Yoshida, Kazuma Nakazawa, Khin Than Tint, Myint Kyaw Soe, Aye Moh Moh Theint,Shinji Kinbara, Akihiro Mishina, Yoko Endo, Hiroki Ito, Hidetaka KobayashiPhysics department, Gifu University, JapanJSPS KAKENHI 23224006 and MEXT 15001001, 24105002

1

Slide2

ｎ

ｎ

ｐ

ｐ

ｎ

Ξ

-

X-ray

ｎ

ｐ

ｎ

ｐ

ｎ

ｎ

ｐ

ｎ

ｐ

ｐ

ｎ

ｎ

ｎ

ｐ

ｐ

ｎ

Λ

ｐ

π

-

ｎ

ｐ

ｐ

ｎ

Λ

Λ

PHYSICAL REVIEW C 88, 014003 (2013)

Vertex C

Vertex B

Vertex A

#1

Double

lambda

hypernucleus

p + Ξ

-

　→　

Λ + Λ +

28MeV

* Important

subject to investigate

hyperon-hyperon

interaction

.

* Tiny

characteristic shape having 3 vertexes at the rest point of

X

-

hyperon in nuclear emulsion.

2

Slide3

Nuclear emulsion plate: photographic tracking device for charged particles

after photographic development

charged particles

track

layer structure

AgBr

crystals suspended in gelatin.

nuclear emulsion plate

3

~450 micron

~450 micron

40 micron

emulsion

emulsion

base

Read-out: by optical microscope

Slide4

4Emulsion pictures

Slide5

5

Diamond target

X

-

K

+

(p>1.0GeV/c)

K

-

Double Lambda

Hypernucleus

Emulsion stack

Only 9 events are observed in 2000s in the world.

V

arious double lambda

hypernuclei

must be detected to investigate

L-L

interaction.

How we obtain more

double lambda

hypernuclear

events?

*

The experiment ten

times larger

than the previous

experiment.

J-PARC E07

->

Ekawa

-san’s talk

* PID for daughter nucleus ->

Kinbara’s

talk

Slide6

6

disassembled emulsion stack

A schematic view of

“Track following”

thin type plate,

good angular

resolution:

I

nterface between tracker and emulsion stack.

thick type plate, large angular acceptance

F

or detection of nuclear events.

PoC

. of automated tracking system is ongoing

Slide7

7

Diamond target

X

-

K

+

(p>1.0GeV/c)

K

-

Double Lambda Hypernucleus

Emulsion stack

*

Latent events

(~10 times)

may be recorded in existing emulsion.

Only 9 events are observed in 2000s in the world.

V

arious double lambda

hypernuclei

must be detected to investigate

L-L

interaction.

How we obtain more

double lambda

hypernuclear

events?

*

The experiment ten

times larger

than the previous experiment. (J-PARC E07

)

* PID for daughter nucleus ->

Kinbara’s

talk

*

New searching

method: Overall-scanning method

yield: ~30%

K

0

X

-

K

-

(K

-

, K

0

)

Slide8

8Overall scanningConcept and Key technologies;(1). 3-dimensional scanning under optical microscopic view.(2). Image

process to find multi-vertex shape in emulsion.* High speed is required* The method is also useful to find calibration sources such as, alpha decays, single lambda hypernuclei.* This method will be applied to E07 after regular analysis with hybrid method.

Slide9

9

Difference of Gaussian

Binalization

Gaussian filter

Original image

Filter for bold tracks

Thinning

Prob. Hough Trans.

Vertexing

kernel size = 17pix

x20 Objective (NA0.35)

1 pix = ~0.6 micron

threshold ~= peak*0.2

dist-edge-to-edge = 15pix

N_Tracks > 3

polygon area > 1000pix

resolution = 1degree

threshold = 16 pix / line

contour length > 5pix

area > 10pix

(smoothed – source)

if negative, brightness=0

(2).

Image

process

J.Yoshida

et al., JPS

Conf. Proc. , (2014) 1.013070

Slide10

10

Blurred

brightness(x)

=

Blurred

-

Raw

brightness(

x,y

) = B - A

A. raw image

B. blurred

if negative, brightness = 0

Raw

x

x

brightness(x)

Slide11

11

Difference of Gaussian

Binalization

Gaussian filter

Original image

Filter for bold tracks

Thinning

Prob. Hough Trans.

Vertexing

kernel size = 17pix

x20 Objective (NA0.35)

1 pix = ~0.6 micron

threshold ~= peak*0.2

resolution = 1degree

threshold = 16 pix / line

contour length > 5pix

area > 10pix

(smoothed – source)

if negative, brightness=0

(2).

Image

process

J.Yoshida

et al., JPS

Conf. Proc. , (2014) 1.013070

dist

-edge-to-edge = 8micron

N_Tracks

> 3

polygon area >

300micron

2

tuned by

Nagara

event and some alpha decays

Slide12

12

・extracted images of vertex-like-object

Slide13

13

example :E373 mod86 pl7scanned volume: 2.2mm * 10mm * 0.9mm

detected candidates:

3496

vertex:

361

misrecognition :

3135

Beam Interaction:

329

Alpha decay

：

32

crossing‐over

of 2tracks:

<Ideas for improvement>

* Parameter tuning.

* Recognition under a higher magnificent optics after 1

st

level selection.

* And so on.

Beam Int.

Alpha decay

Eye check for extracted vertexes

SingleHyper

:

6

Slide14

14high resolution CMOS2048*358　pixel

high frame rate800fpswide F.O.V.x20 dry lens (NA0.35)1142*200　micron2 Piezoelectric drivestroke 500 micronperiod 5Hzpicture 40picts /cycleContinuously stage moving(1). A dedicated scanning system（Stage#7,8） collaborate with UNIOPT Co.

emulsion plate

objective lens

Piezoelectric drive

camera

Designed speed:

10mm*10mm*0.9mm / 2min. = ~40

hours @plate

(340*335*0.9[mm

3

])

T

o accomplish scanning of ~1000 films within a few year.

Slide15

15

Xaxis

Yaxis

Field Of View : 0.2*1.1 [mm

2

]

stage motion in

X-Y

plane (top view)

1.0 mm / sec

1.0 mm / sec

1

2

3

4

200msec /cycle

Yaxis

Zaxis

stage motion in Y-Z plane (cross sectional view)

1.0

mm / sec

40

picts

.

* Exhaustively scanning by a

computer-controlled optical

microscope.

* Continuous

stage drive

to

save damping time of vibration

.

Slide16

16

Difference of Gaussian

Binalization

Gaussian filter

Original image

Filter for bold tracks

Thinning

Prob. Hough Trans.

Vertexing

kernel size = 17pix

x20 Objective (NA0.35)

1 pix = ~0.6 micron

threshold ~= peak*0.2

dist-edge-to-edge = 15pix

N_Tracks > 3

polygon area > 1000pix

resolution = 1degree

threshold = 16 pix / line

contour length > 5pix

area > 10pix

(smoothed – source)

if negative, brightness=0

(2).

Image

process

J.Yoshida

et al., JPS

Conf. Proc. , (2014) 1.013070

* These are processed in real-time by CPU&GPU (5msec/

pict

.)

* Data

is

stored

in

HDDs

Slide17

* Stage drive, piezoelectric, camera are synchronized. Trigger timing is proper.* Missed trigger and troubles in data transfer < 3%17

Quality checking for scanned dataExhaustiveness of image taking

pict.ID

number_of_’1’_pixel

base

Tiled

binalized

images.

200micron*1140micron*3views

The same layer ID

Time or Y

Z

T

rigger timing

surface

50views ->

4

0pictures

1 10 20 30 40

trigger

trigger

trigger

Slide18

Found events; ~3000 alpha decays ~700 single hypernuclear event 7 three-vertexes candidates.A typical twin-single-lambda- hypernuclear event (Kiso)

Prototype Upgraded High-speed Since ~2011 2014 2013Stage Camera 100fps CCD 800fps CMOS Z-drive Stepping motor PiezoelectricOptics Field of View 0.11*0.13mm2 0.18*0.21mm2 0.20*1.14mm2Speed[mm3/h] 2.6 13.1 2700Site Gifu-univ. Gifu-univ. Gifu-univ. Toho-univ,Japan GNU,Korea

GNU,Korea

Scanned vol.

~5.0cc

just started

17.4cc

10

m

m

18

Operation

Slide19

summary* Overall scanningA new searching method for double lambda hypernucleus in emulsion.Concept: high speed 3D scanning & image recognition* The current statusScanning software, Image process is basically successful.

Improvement of detection efficiency and S/N ratio is being carried out.*OperationIn operation in Gifu-univ., Toho-univ., Japan, GNU, Korea. ~700 candidates of single hypernuclear ~7 3-vertexes candidates.A typical twin-single-lambda- hypernuclear event (Kiso)19

Slide20

20

Slide21

Difference of

Gaussian

Binalization

Gaussian filter

Raw

image

Filter for bold tracks

Thinning

Prob. Hough Trans.

Vertexing

kernel size = 17pix

x20 Objective (NA0.35)

1 pix = ~0.6 micron

threshold ~= peak*0.2

dist

-edge-to-edge = 15pix

N_Tracks

> 3

polygon area > 1000pix

resolution = 1degree

threshold = 16 pix / line

contour length > 5pix

area > 10pix

(smoothed – source)

if negative, brightness=0

Image process

JPS Conf. Proc. , (2014) 1.013070

Slide22

22

Piezo Ctrl.

BNC

USB

PiezoElec.

CMOS Camera

Grabber Board

Host PC

TRG

PCI & PCI-E

GPGPU

Stage Ctrl.

CameraLink Base & Full

Main Board

USB

command

Time or StageYaxis

Z

Emulsion

thickness

TRG:

start sequential shooting

optics

vibration of focal plane driven by

Piezoelectricity

Synclonization between piezoelectricity and camera

400us

　

3.5V

(LVTTL)

400us

　

5.0V

Slide23

23

Z-Axis (Optics)

Grabber Board

Host PC

PCI & PCI-E

GPGPU

Stage Ctrl.

Main Board

Serial

command

Motor Driver

(3-axes)

LED Driver

X,Y-Axis (Stage)

LED Light

Stage control

Slide24

24

CMOS Camera

Grabber Board

Host PC

PCI & PCI-E

GPGPU

(GTX680)

Stage Ctrl.

CPU

Corei7

3960X

RAM

External HDD

Chipset

X79

eSATA (100MB/s)

(1) Camera -> grabber_board -> RAM

(2) RAM -> GPGPU

(3) Difference_of_Gaussians on GPGPU

(4) GPGPU -> RAM

(5) Binalization on CPU

(6) RAM -> External HDD

(7) Image process on another computer

Data transmission

2048*358 pix

8bit-depth

2048*358 pix

1bit-depth

シリアル

Slide25

25Overall scanning

Stage size:35cm*35cm (Nikon. )45cm*45cm (Uniopt. )Field of View: 115micron*135micron(x50 objective lens)ほかにも、乾板中にランダムに存在するキャリブレーションソースアルファ崩壊、シングルハイパーも探索したい。そこで

・原子核乾板の全体積を光学顕微鏡で走査し、

　

(1)

高速顕微鏡画像取得の開発

・少なくとも

1

つの

分岐点をもつパターンを検出

　

(2)

画像解析の開発

Slide26

26Phase I :　技術開発　兼　E373乾板の再解析　先の実験（KEK E373）の乾板を用いて技術開発　E373乾板の全体積をスキャンし、　　　　従来比10

倍、70例のダブルΛハイパー核検出をねらうPhase II : E07実験解析のアシスト　キャリブレーションソース：アルファ崩壊の探索高速化Phase III : E07乾板の再解析　E07乾板の全体積をスキャンし、　　　プロポーザル比10倍、103例のダブルΛハイパー核検出をねらうOverall Scanning と解析のストラテジー