1
00:00:00,210 --> 00:00:02,730
Good. Okay, you can start

2
00:00:04,620 --> 00:00:07,020
Yeah, thank you. Hello everyone.

3
00:00:08,280 --> 00:00:13,290
I'm glad to give the talk about KD
performance in run to it are some

4
00:00:13,290 --> 00:00:21,060
experiments today and I would like to
introduce you quickly to the I outline of

5
00:00:21,060 --> 00:00:27,060
the magic of particle identification at
FCB. At first I will introduce the other

6
00:00:27,060 --> 00:00:32,070
subnet detector and explain the main p ID
observables. Those observables are

7
00:00:32,070 --> 00:00:38,130
computed combining information from the
CBP ID sub detectors, which are the rich

8
00:00:38,130 --> 00:00:42,330
detectors in your chambers and the
kilometer kilometer system. Then I will

9
00:00:42,330 --> 00:00:47,400
introduce the general VP ID strategy
including the key ID computing strategy.

10
00:00:47,970 --> 00:00:52,860
And after that, I will talk about key ID
calibration and calibration samples for

11
00:00:52,860 --> 00:00:57,360
charged and neutral particles. and
conclude my talk with an overview over the

12
00:00:57,360 --> 00:01:03,570
pod performance. The party identification
is a crucial ingredient of the physics

13
00:01:03,570 --> 00:01:10,110
analysis at nsep as a flavor physics
experiment, and one important field of P

14
00:01:10,110 --> 00:01:16,230
ID is the flavor tagging of neutropenia
zones. Furthermore, it allows for an

15
00:01:16,260 --> 00:01:20,070
excellent reduction of combinatorial
background and also the separation of

16
00:01:20,070 --> 00:01:26,970
different planets states with an otherwise
identical topology. Like for example, the

17
00:01:26,970 --> 00:01:35,970
zero to kk versus B zero to pi k versus
lambda B to PK or from eradicates B plus

18
00:01:35,970 --> 00:01:41,790
to K B plus two k versus B plus two k pi
pi.

19
00:01:44,280 --> 00:01:45,900
Yeah, and

20
00:01:47,430 --> 00:01:53,460
what you can see in the pop Below is a fit
to the invariant mass of the PK mu mass

21
00:01:53,460 --> 00:01:58,020
system of the latest measurement of
Neptune for universality for the ratio

22
00:01:58,020 --> 00:02:05,370
applique in blue Consider total signal
which is mainly from the signal itself and

23
00:02:05,370 --> 00:02:09,090
the other lines that are hardly visible or
the background components we are not

24
00:02:09,090 --> 00:02:15,090
interested in. And yeah, so the other
signal detector in run tool can be seen in

25
00:02:15,090 --> 00:02:19,080
the sketch on the slide where in yellow I
have marked ring imaging to rank of

26
00:02:19,080 --> 00:02:25,890
detectors. The abbreviate every variation
is rich, and they allow for p ID for K on

27
00:02:25,890 --> 00:02:31,350
pylons, protons and low momentum electrons
which carry electromagnetic charge. The

28
00:02:31,350 --> 00:02:35,880
Mune stations are colored in green and
comprised of five different immune

29
00:02:35,880 --> 00:02:41,550
stations and deliver high purity tip for
nuance. And the kilometer system is

30
00:02:41,550 --> 00:02:45,570
colored in violet and consists of
scintillating the scintillating pet

31
00:02:45,570 --> 00:02:51,150
detector, SPD, pre sharp detector PS anti
electromagnetic and hydronic for renewals,

32
00:02:51,210 --> 00:02:55,920
ecad and H color. And this carry meter
system allows for p ID for electrons,

33
00:02:55,920 --> 00:02:57,270
photons and neutral

34
00:02:58,560 --> 00:02:59,190
ions.

35
00:03:00,569 --> 00:03:05,279
To identify heavy flavor decays with large
hydronic background the rich detectors

36
00:03:05,789 --> 00:03:11,189
allowed to separate k ions pylons and
protons over a very wide momentum range.

37
00:03:11,609 --> 00:03:16,409
The first rich detector which one which
can be seen in the figures in the figure

38
00:03:16,409 --> 00:03:25,049
on the right hand side comes from momentum
range of two to 16 gV with CF C for F 10

39
00:03:25,049 --> 00:03:30,899
radiator and arrow drill, which was
removed in the long shutdown to have the

40
00:03:30,899 --> 00:03:34,799
Large Hadron Collider. This is why you
don't see the original in the sketch. The

41
00:03:34,799 --> 00:03:40,919
rich tool covers a high momentum rates
range of 50 to 100 gV using a CFD radiator

42
00:03:40,949 --> 00:03:47,369
the chunk of light cones are then mirrored
outside as you detect acceptance and are

43
00:03:47,369 --> 00:03:52,949
measured with a photon detector array. And
from a combination of those photon rings

44
00:03:52,949 --> 00:03:56,369
and to track momentum from detector
components to lock likelihood

45
00:03:56,369 --> 00:04:02,759
distributions for the charged mass
hypothesis. calculated. The NSE media

46
00:04:02,759 --> 00:04:06,869
ticker comprises five meal tracking
stations with headphone absorbers in

47
00:04:06,869 --> 00:04:11,939
between the first mewn station and one is
positioned before the carry meats as can

48
00:04:11,939 --> 00:04:17,159
be seen in the sketch hits in the Mune
chambers are then searched around

49
00:04:17,159 --> 00:04:22,979
extrapolated tracks for other detector
components. Yeah, and it is important the

50
00:04:22,979 --> 00:04:27,569
moon stations are very important for
particle identification of neurons and is

51
00:04:27,569 --> 00:04:33,989
they are also employed at the trigger
number. The color emitter system itself

52
00:04:33,989 --> 00:04:40,349
consists of the SPD and fnps with 6066
kilometer cells each the hydronic

53
00:04:40,379 --> 00:04:47,399
perimeter with ties of iron scintillating
material with 1400 88 kilometer cells. The

54
00:04:47,399 --> 00:04:52,529
electromagnetic kilometer with especially
Kirby's which is the current kilometer

55
00:04:53,879 --> 00:05:02,519
from Latin single letter old 6016 seconds
Carrie meter system itself identifies

56
00:05:02,549 --> 00:05:06,659
electrons, photons and neutral piles with
information from those sub detectors by

57
00:05:06,659 --> 00:05:11,879
measuring the information for neutral
objects and also triggers for electrons

58
00:05:11,879 --> 00:05:18,179
and photons. Yeah, the perimeter system is
the last building block of the particle

59
00:05:18,179 --> 00:05:22,169
identification system I want to present
today. Which brings me to the P ID

60
00:05:22,169 --> 00:05:29,159
strategy of NFC. Where for the charged
particles, the information of the signals

61
00:05:29,189 --> 00:05:36,269
detectors subsystem is combined to
powerful signal absorbers. The most

62
00:05:36,269 --> 00:05:41,249
prominent ones are the so called DLL,
which are the lock likelihood differences

63
00:05:41,279 --> 00:05:46,229
of particle hypothesis of a particle x and
pythons where the pylons act as our

64
00:05:46,229 --> 00:05:51,719
reference. The DLL observable is computed
as a product of the log likelihood of the

65
00:05:51,719 --> 00:05:55,979
similar big sub detectors detectors
mentioned before which can be seen in the

66
00:05:56,279 --> 00:06:00,419
equation on the slide. another variable
that is combining subjects textual

67
00:06:00,419 --> 00:06:04,049
information is solved with the so called
problem and observable, which is the

68
00:06:04,049 --> 00:06:08,039
output of a neural net which is trained on
simulation with input from the detector

69
00:06:08,039 --> 00:06:12,449
components and tracking information and
there are also dedicated neutral neural

70
00:06:12,449 --> 00:06:20,279
nets for neutral and as simulation is not
perfectly describing the P ID response, it

71
00:06:20,279 --> 00:06:26,849
has to be calibrated for the data room
calibration technique for For this reason,

72
00:06:27,179 --> 00:06:32,669
data driven calibration techniques with
high statistical ration data are applied.

73
00:06:32,999 --> 00:06:40,799
Yeah, if you would not use this corrective
simulation for the efficiencies you would

74
00:06:40,799 --> 00:06:46,409
like to introduce large systematic
uncertainties which should be avoided. And

75
00:06:47,249 --> 00:06:52,919
those calibration samples are selected and
on to as a part of the HA trigger which is

76
00:06:52,919 --> 00:06:57,149
a software based trigger following the
hardware trigger Zoo and this filtering

77
00:06:57,149 --> 00:07:01,829
detector data the selection of the pod
killed ration sampling is performed with a

78
00:07:01,829 --> 00:07:08,609
dedicated stream at the so called telcos
level which means total calibration. And

79
00:07:08,609 --> 00:07:13,439
this is done to save resources to to
Central computation and central control

80
00:07:13,439 --> 00:07:19,229
systematic uncertainties for Mr. samples.
Yeah, this stream delivers luscious

81
00:07:19,259 --> 00:07:23,219
statistics, calibration symbols and
symbols for high precision precision

82
00:07:23,219 --> 00:07:28,319
measurements. Yeah, to dive a bit deeper
into the calibration samples, I will focus

83
00:07:28,319 --> 00:07:30,269
now on the charge calibration samples.

84
00:07:31,590 --> 00:07:36,720
And, as already mentioned, those samples
are from a real time selection and based

85
00:07:36,720 --> 00:07:42,510
on offline or sorry online construction
with no requirements on p ID for the probe

86
00:07:42,510 --> 00:07:47,130
techs, and they are high purity and high
statistics samples with final states as

87
00:07:47,130 --> 00:07:52,350
charged particles only. And also they are
required to be modes with low multiplicity

88
00:07:52,350 --> 00:07:58,200
and large branching ratios to have a
really good purity for the electron for

89
00:07:58,200 --> 00:08:04,140
electrons DKB plus projects K whether the
job size came to two electrons is used for

90
00:08:04,140 --> 00:08:11,700
the moon dedicated side who is employed as
can be seen as plus for the pylons the

91
00:08:11,700 --> 00:08:16,770
decay case for two pi for normal metal
tracks and East and D star to be zero

92
00:08:22,470 --> 00:08:23,430
pi pi is

93
00:08:25,350 --> 00:08:30,930
whether the zero decays to a calendar pi
on this implied and this decay is also

94
00:08:30,930 --> 00:08:36,420
used for calibrations for the calibration
for decay ones for protons the decay

95
00:08:36,420 --> 00:08:41,910
lambda 02 proton ion and lambda c two
photon k on ion for high momentum ranges

96
00:08:41,910 --> 00:08:47,790
used for the neutral PRD multivariate
classifiers combined variables which are

97
00:08:47,790 --> 00:08:53,040
describing energy deposits in the
kilometer sub detectors and allows for

98
00:08:53,040 --> 00:08:58,290
discriminating photons from head runs
electrons and high energy, neutral pylons

99
00:08:58,920 --> 00:09:05,400
and as a killer As calibration samples for
photons the case d s two prime plus d s

100
00:09:05,400 --> 00:09:12,270
two gamma d s gamma and B zero to k star
gamma used. And for neutral piles to decay

101
00:09:12,270 --> 00:09:19,350
D zero to k plus pi minus Pi Zero is used,
as can be seen in the plot here. Yeah, the

102
00:09:19,350 --> 00:09:24,000
performance of the FCB particle
identification can then be assessed by

103
00:09:24,210 --> 00:09:29,490
those pure calibration samples. And here
I'm showing the P ID efficiency over the

104
00:09:29,490 --> 00:09:35,880
track momentum range for two given cuts on
the DLL variables, I described those

105
00:09:35,880 --> 00:09:41,280
variables above as the difference of the
likelihood of the particle to select and

106
00:09:41,280 --> 00:09:47,130
the pion and only for the moon plots as an
additional requirement variable is on this

107
00:09:47,130 --> 00:09:52,320
prompt here as well. So the red empty
circles show that efficiencies for us are

108
00:09:52,320 --> 00:09:57,660
cut into solid red dots for a tighter cut.
And as a take home message it can be seen

109
00:09:57,660 --> 00:10:02,430
that those variables show an excellent
discrimination Over a very right kinematic

110
00:10:02,430 --> 00:10:07,770
range for the leptons The same can be said
for the headlines where I'm showing here

111
00:10:07,770 --> 00:10:15,810
the efficiency for K ions and protons. The
neutral performance can be measured using

112
00:10:15,810 --> 00:10:22,350
three different neural networks which are
trained with simulation to separate the

113
00:10:22,350 --> 00:10:29,610
photon signatures from our from other
species which are is not hadron for photon

114
00:10:29,610 --> 00:10:35,700
versus hadron is not electron for for
transfers electrons and is photon for

115
00:10:35,700 --> 00:10:41,100
photons versus neutral piles. And in the
plot below, you can see the signal

116
00:10:41,100 --> 00:10:46,710
distribution in blue and the background
distribution in red for the Islamic E and

117
00:10:46,710 --> 00:10:52,590
is not age training forum one based on
simulation. And from that it can be said

118
00:10:52,590 --> 00:10:57,480
that as a particle identification also
provides a good separation with neutral

119
00:10:57,480 --> 00:11:05,640
objects with dedicated towards In addition
to that, a dedicated di to run p ID,

120
00:11:07,830 --> 00:11:10,710
where a dot run consists of a proton and a
neutron

121
00:11:12,090 --> 00:11:18,840
was developed and typical use cases are
baryonic case of the kind, lambda v to

122
00:11:18,840 --> 00:11:25,290
doctrine, anti proton, or cross section
measurements of doctrine productions. And

123
00:11:25,290 --> 00:11:28,860
in the property law, I'm showing the
property and distribution, which is the

124
00:11:28,860 --> 00:11:35,490
output of a dedicated neural network to
separate the particle type. And this

125
00:11:35,490 --> 00:11:40,830
neural network gives you a value between
zero and one as the probability to be a

126
00:11:40,830 --> 00:11:46,920
doctor. It was trained with daughter runs
for London to go to an anti proton and for

127
00:11:46,920 --> 00:11:54,660
the other tracks, the decay the energetics
of decay lambda C to proton. My minutes

128
00:11:54,660 --> 00:12:00,660
left. Thank you. Yeah, in my game time,
you can see the problem I'd run

129
00:12:00,660 --> 00:12:08,340
distribution which is peaking at one as it
is expected and in contrast the

130
00:12:08,370 --> 00:12:13,230
distributions of the proper end washroom
classifier, which is then applied on other

131
00:12:13,230 --> 00:12:20,010
tracks which are proton k on pion goes
check where it goes tricks, tricks, tricks

132
00:12:20,040 --> 00:12:25,380
in which less than 70% of the hits are
from the same tricks are shown. And you

133
00:12:25,380 --> 00:12:32,910
see that for those other track types, the
peak is at zero and on an all it shows

134
00:12:33,180 --> 00:12:40,230
good separation performance for doctrines
as well. Yeah. To give a summary of my

135
00:12:40,230 --> 00:12:45,660
talk, it can be said that precise PhD
performance is crucial for LSU flavor

136
00:12:45,660 --> 00:12:52,950
physics. The particle identification is
provided by a dedicated pod system which

137
00:12:52,950 --> 00:12:58,980
is built after rich detectors, the current
meter system in New York chambers yeah and

138
00:12:58,980 --> 00:13:04,020
dedicated towards to Combine the input 
rom the signal sub detectors to powerful o

139
00:13:05,160 --> 00:13:09,150
servers also using multivariate classif
ers and machine learning for op

140
00:13:09,630 --> 00:13:16,560
imum of separation, quality are emplo
ed. And assimilation is not descr

141
00:13:16,560 --> 00:13:21,150
bing it perfectly calibration tools w
th data driven techniques are applied. 

142
00:13:21,150 --> 00:13:28,230
nd they are using high statistics pure int
gration samples with large face base cover

143
00:13:28,260 --> 00:13:34,350
ge. And probing those probing DFP efficie
cies on those samples shows that Sydney is

144
00:13:34,770 --> 00:13:39,450
showing a high pod performance for larg
 mutual articles as a key part of the MCP

145
00:13:39,450 --> 00:13:46,260
physics program. And with the increased 
CD data set on free it will be possible to

146
00:13:46,320 --> 00:13:48,060
develop and improve tools 

147
00:13:48,239 --> 00:13:49,019
or e

148
00:13:50,520 --> 00:13:56,940
en more precise p ID calibration and there
will be also have updates on the P ID d

149
00:13:56,970 --> 00:14:01,500
tectors which which will increase the
overall quality of P IDs And also the 

150
00:14:01,500 --> 00:14:08,370
omentum coverage. Yeah, there are there
are two other talks I would like to poin

151
00:14:08,370 --> 00:14:14,070
 you to which are on SCP real time recons
ruction and I'm in calibration tool by 

152
00:14:15,270 --> 00:14:19,140
he Oto. And tracking on vertices v
rtex in performance and developments 

153
00:14:19,140 --> 00:14:25,680
r on to another city by the nada, which 
re also covering parts of the productio

154
00:14:25,740 --> 00:14:30,030
 of new calibration samples and reconstru
tion of for example, el

155
00:14:30,749 --> 00:14:31,709
ctrons. Yep, thank you ve