CONTENTS · Lab. t Departmen of ysics Ph o oky T Institute of hnology ec T ebruary F 2001. Abstract...
Transcript of CONTENTS · Lab. t Departmen of ysics Ph o oky T Institute of hnology ec T ebruary F 2001. Abstract...
Master's Thesis
Flavor Asymmetry of the polarized
light Sea Quarks extra ted from
deep-inelasti S attering at HERMES
Fumiko Sato
Shibata Lab.
Department of Physi s
Tokyo Institute of Te hnology
February 2001
Abstra t
The spin ontent in the nu leon is an interesting subje t of nu lear physi s, and
it has been found that the quark ontribution is less than 30 % in the nu leon spin
in re ent experiments. On the other hand from the unpolarized experiment, it was
found that the unpolarized quark distributions of �u and
�
d are not equal : there is
an asymmetry of �u and
�
d.
The main subje t of this thesis is the extra tion of the polarized quark avor
distribution, espe ially the polarized avor asymmetry extra tion. The avor asym-
metry of light sea quarks is expressed as ��u��
�
d, where ��u(�
�
d) is the polarized
�u(
�
d) quark distribution. In this analysis, a new method for extra tion of avor
asymmetry of light sea quarks is proposed. It is based on the Chiral Quark Soliton
Model (CQSM) predi tion. With the new method, the single quark distributions
an be extra ted as well as the avor asymmetry.
At HERMES experiment the produ ed parti les are measured in Deep-Inelasti
S attering as well as s attered lepton. The Ring Imaging Cherenkov Counter is
operated to identify parti les sin e 1998. The study of the dete tor eÆ ien y is also
dis ussed in this thesis. The HERMES RICH an identify pions, kaons and protons.
16 types of in lusive and semi-in lusive spin asymmetry be ame available. Using
these asymmetries, the quark distributions are extra ted and their moments with
the original and the new methods are presented.
CONTENTS i
Contents
1 Introdu tion 1
2 Deep-Inelasti S attering (DIS) 3
2.1 DIS Kinemati s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Cross Se tion and Nu leon Stru ture Fun tion . . . . . . . . . . . . . 7
2.2.1 Unpolarized Cross Se tion and Stru ture Fun tion . . . . . . . 8
2.2.2 Polarized Cross Se tion and Stru ture Fun tion . . . . . . . . 8
2.3 The Quark Parton Model . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.1 The unpolarized PDF . . . . . . . . . . . . . . . . . . . . . . 10
2.3.2 The polarized PDF . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.3 Measurement of g
1
(x) . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Sum Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.1 Gottfried Sum Rule . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.2 Bjorken Sum Rule . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.3 Ellis-Ja�e Sum Rule . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 Semi-In lusive Measurement . . . . . . . . . . . . . . . . . . . . . . . 15
2.5.1 Fragmentation Fun tions . . . . . . . . . . . . . . . . . . . . . 16
2.5.2 Fragmentation Models . . . . . . . . . . . . . . . . . . . . . . 17
2.5.3 Hadron Asymmetry . . . . . . . . . . . . . . . . . . . . . . . . 19
3 HERMES Experiment 21
3.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 The HERA Storage Ring . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 The Internal Gas Target . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 The HERMES Spe trometer . . . . . . . . . . . . . . . . . . . . . . . 24
3.4.1 The Tra king System . . . . . . . . . . . . . . . . . . . . . . . 25
3.4.2 The Parti le Identi� ation System . . . . . . . . . . . . . . . . 27
4 Ring Imaging Cherenkov Counter 29
4.1 Dete tor Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.2 Study of RICH Property . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.1 Aerogel Radiator . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.2 Mirrors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.3 Photon Dete tor Plane . . . . . . . . . . . . . . . . . . . . . . 41
4.3 Monte Carlo Simulation for RICH . . . . . . . . . . . . . . . . . . . . 42
4.3.1 Development of MC for RICH . . . . . . . . . . . . . . . . . . 44
4.3.2 PID EÆ ien y after Modi� ation of MC . . . . . . . . . . . . 48
ii CONTENTS
5 Extra tion of Polarized Quark Distributions 57
5.1 The Purity Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.1.1 De�nition of Purity . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2 Evaluation of Purities . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3 Assumptions on Sea Quark Polarizations . . . . . . . . . . . . . . . . 70
5.3.1 Assumption of Polarization Symmetri Sea . . . . . . . . . . . 70
5.3.2 Assumption of Sea Quark from CQSM . . . . . . . . . . . . . 71
5.4 Interpretation of Asymmetries on the
3
He and D Targets . . . . . . . 76
5.4.1 Asymmetry on the
3
He Target . . . . . . . . . . . . . . . . . . 76
5.4.2 Asymmetry on the D Target . . . . . . . . . . . . . . . . . . . 77
5.4.3 Method to use Asymmetry on
3
He and D . . . . . . . . . . . . 77
5.5 The Set of Analyzed Asymmetries . . . . . . . . . . . . . . . . . . . . 82
5.5.1 1995 Asymmetries on the
3
He Target . . . . . . . . . . . . . . 82
5.5.2 1996 and 1997 Asymmetries on the Proton Target . . . . . . . 83
5.5.3 1998 Asymmetries on the D Target . . . . . . . . . . . . . . . 84
5.5.4 The Sets of Input Asymmetries . . . . . . . . . . . . . . . . . 85
6 Analysis and Results 87
6.1 Results for the Polarized Quark Distributions . . . . . . . . . . . . . 87
6.1.1 The Valen e De omposition . . . . . . . . . . . . . . . . . . . 90
6.1.2 The Flavor De omposition . . . . . . . . . . . . . . . . . . . . 93
6.1.3 Flavor Asymmetry of Light Sea Quarks . . . . . . . . . . . . . 96
6.1.4 Single Quark Distribution . . . . . . . . . . . . . . . . . . . . 99
6.1.5 Results for the di�erent Purities . . . . . . . . . . . . . . . . . 102
6.2 Moments of Polarized Quark Distributions . . . . . . . . . . . . . . . 104
6.2.1 The Valen e and Flavor De omposition . . . . . . . . . . . . . 105
6.2.2 Flavor Asymmetry and Single Quark Distribution . . . . . . . 106
6.2.3 Comparison of Results with Predi tions . . . . . . . . . . . . . 108
6.3 Flavor Asymmetry Extra tion with Another Method . . . . . . . . . 110
6.3.1 Another Method for Flavor Asymmetry Extra tion . . . . . . 110
6.3.2 Theoreti al Model of the Fragmentation Fun tions . . . . . . 111
6.3.3 Results and Comparison with the Purity Matrix Method . . . 114
6.4 Simple Estimation of Systemati Errors . . . . . . . . . . . . . . . . . 116
7 Con lusions and Summary 119
A Detail of MC Development 121
B Detail of Analysis 125
B.1 Covarian e Matrix V
A
. . . . . . . . . . . . . . . . . . . . . . . . . . 125
B.2 Singular Value De omposition . . . . . . . . . . . . . . . . . . . . . . 127
C Table of Results 129
C.1 De�nition of the Binning in x . . . . . . . . . . . . . . . . . . . . . . 129
C.2 Input Asymmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
C.2.1 1995 Asymmetries on the
3
He Target . . . . . . . . . . . . . . 130
C.2.2 1996 Asymmetries on the Proton Target . . . . . . . . . . . . 131
C.2.3 1997 Asymmetries on the Proton Target . . . . . . . . . . . . 132
CONTENTS iii
C.2.4 1998 Asymmetries on the D Target . . . . . . . . . . . . . . . 133
C.3 Correlation CoeÆ ients for Parti le Count Rates . . . . . . . . . . . . 135
C.4 Results of Quark Polarizations and Distributions . . . . . . . . . . . 138
C.4.1 Valen e De omposition . . . . . . . . . . . . . . . . . . . . . . 138
C.4.2 Flavor De omposition . . . . . . . . . . . . . . . . . . . . . . 140
C.4.3 Flavor Asymmetry of Light Sea Quarks . . . . . . . . . . . . . 142
C.4.4 Single Quark Distributions . . . . . . . . . . . . . . . . . . . . 144
C.5 Moments of Polarized Quark Distributions . . . . . . . . . . . . . . . 146
C.5.1 Valen e De omposition . . . . . . . . . . . . . . . . . . . . . . 146
C.5.2 Flavor De omposition . . . . . . . . . . . . . . . . . . . . . . 147
C.5.3 Flavor Asymmetry of Light Sea Quarks . . . . . . . . . . . . . 148
C.5.4 Single Quark Distributions . . . . . . . . . . . . . . . . . . . . 149
1
Chapter 1
Introdu tion
The spin stru ture of the nu leon be ame one of the fundamental questions in
physi s, when results from the EMC experiment [1℄ showed that only a fra tion
of the nu leon's spin is arried by its quarks. Before that experiment, it has been
predi ted, that spin of the nu leon, like its magneti moment, is almost ompletely
arried by the quarks. As a onsequen e of this dis overy, a more general approa h
for the spin ontents of a proton, taking quarks, gluons and orbital momenta into
a ount, has to be used:
S
z
=
1
2
=
1
2
�� +�g + L
z
; (1.1)
where �� = �u+��u+�d+�
�
d+�s+��s is the ontribution of quark spins, �g
is that of the gluon spin and L is that of the orbital angular moment.
Further experiments [2℄ on�rmed the EMC results and obtained : �u
�
=
0:8;�d
�
=
�0:4 and �s
�
=
�0:1. But the ontributions of light sea quarks (��u;�
�
d), gluons and
the orbital angular momenta are not lear yet. In most further studies to measure
the ontribution of quark spins, the ontributions of light sea quarks are assumed to
be symmetri as ��u = �
�
d = �s = ��s to ompensate for the la k of experimental
data. The ontributions of valen e quarks have been extra ted from the di�erent
experiments and their results are almost onsistent, however, the spin stru tures
of light sea quark are not made lear omparing the valen e quarks. In addition,
the asymmetry between the spin of the �u and the
�
d quarks have been not obtained
learly yet. In the unpolarized experiment at HERMES, a avor asymmetry of the
unpolarized sea quark densities (�u�
�
d 6= 0) are measured already [3℄, and this mea-
surement may be a hint that there might be an asymmetry for the polarized light
sea quark densities. This unpolarized avor asymmetry have been obtained in other
experiment [4℄ [5℄.
This polarized avor asymmetry of light sea quarks is theoreti ally predi ted by
the 1=N
-expansion of Quantum ChoromoDynami s (QCD) [6℄, where it is al ulated
in the large-N
limit by the Chiral Quark Soliton Model. This model predi ts a quite
large polarized sea quark avor asymmetry. From the theoreti al side, the estimation
of the avor asymmetry are a hieved in [7℄ [8℄. Additionally there are the extra tion
of avor asymmetry with the HERMES and SMC data in [9℄ [10℄ and it suggests the
existen e of the avor asymmetry, although it is not on�rmed be ause the statisti s
is not enough. It is very interesting and hallenging to study of the avor asymmetry
of light sea quarks, ��u��
�
d, for solving the spin stru ture of nu leon.
2 CHAPTER 1. INTRODUCTION
One of the main tasks of the HERMES Experiment at DESY is the determination
of the quark spin distributions with semi-in lusive measurements in deep-inelasti
s attering. Semi-in lusive measurement allows us to study the spin ontributions of
ea h quark in the nu leon by tagging the stru k quark with help of the identi� ation
of the leading hadron in the �nal state. The HERMES Experiment started in 1995
and has obtained many important basi results already. Sin e 1998 a Ring Imag-
ing Cherenkov Counter (RICH) is installed, and this is one of the most important
devi es to tag the stru k quark from the target. The HERMES RICH uses two
radiators, aerogel and C
4
F
10
gas. In ident photons make a dual ring image on the
photon dete tor plane, whi h allows us to dete t three types of hadrons (�, K, p) in
the momentum region of 2 � 15 GeV. This dual radiator RICH provides a urate
identi� ation of the dete ted hadrons. With this RICH data we an extra t the
ontributions of quark spins to the total spin of the nu leon.
3
Chapter 2
Deep-Inelasti S attering (DIS)
2.1 DIS Kinemati s
Deep-Inelasti S attering (DIS) is one of the most important tools to investigate
the internal stru ture of the nu leon. The DIS pro ess of the intera tion between a
nu leon N and a lepton l an be written as follows:
l +N ! l
0
+X; (2.1)
where l
0
is the s attered lepton, h is the dete ted hadron and X stands for other
undete ted parti les. The Feynman diagram of this pro ess is shown in Fig. 2.1.
The in oming lepton whi h hits the nu leon intera ts with the quarks by the
weak or ele tromagneti intera tion, by ex hanging a neutral ve tor boson Z
0
or a
virtual photon
�
. But in the energy region of HERMES the ontribution of the
Z
0
boson an be negle ted be ause the mass of the Z
0
boson is far beyond the
kinemati s of HERMES. So from here on, we onsider the ele tro-magneti pro ess
whi h the ex hange of one photon
�
only.
Figure 2.1: The diagram of DIS. The kinemati al valuables are expressed in
Tab.2.2. X expresses the produ ed parti les in the fragmentation
pro esses.
4 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
The in lusive DIS variables
k = (E;
!
k
) Four-momentum of the in oming lepton
k
0
= (E
0
;
!
k
0
) Four-momentum of the s attered lepton
s = (0;
^
!
s
)
^
!
s
=
1
m
(j
!
k
j;
^
!
k
E) ;
^
!
k
=
!
k
=j
!
k
j
Spin four-ve tor of the in ident lepton
s
0
= (0;
^
!
s
0
) Spin four-ve tor of the s attered lepton
P = (M;
!
0
) Four-momentum of the target nu leon
S = (0;
^
!
S
) ,
^
!
S
= (sin� os �; sin� sin�; os�)
Spin four-ve tor of the target nu leon
� S attering angle of the lepton in the lab system
� =
P � q
M
lab
= E � E
0
Energy transfer to the target
q = (�;
!
q
) Four-momentum transfer to the target
Q
2
lab
= 4EE
0
sin
2
�
2
Negative square of the four-momentum transfer
W
2
= (P + q)
2
lab
= M
2
+ 2M� �Q
2
Invariant of the hadron �nal state
x =
Q
2
2P � q
lab
=
Q
2
2M�
Bjorken s aling variable
y =
P � q
P � k
lab
=
Q
2
2M�
Fra tional energy transfer of the virtual photon
Table 2.1: De�nition of DIS kinemati al values. These variables of DIS are
shown in Fig.2.2 and Fig.2.3.
2.1. DIS KINEMATICS 5
Figure 2.2: De�nition of DIS kinemati variables. The angles �, �, � and �
are de�ned by the in oming lepton k, the s attered lepton k' and
the spin four-ve tor of target nu leon S.
Figure 2.3: De�nition of s attering angles in DIS. The angles whi h are
shown here are identi al with those in Fig.2.2.
6 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
We next de�ne the kinemati s of the DIS pro ess. The kinemati variables are
de�ned in Tab.2.1. In this thesis, the dis ussion are based on the laboratory (lab)
system, whi h is the spe ial ase where an in oming lepton with four-momentum
k = (E;
!
k
) is s attered by a �xed target with four-momentum P = (M; 0). M is
the rest mass of the target nu leon. All quantities are expressed in units of and �h
as � �h � 1 and skip all afterwards.
First the four-momentum transfer q between the in oming lepton and the target
nu leon is de�ned as:
q
2
= (k
0
� k)
2
lab
�
�4EE
0
2
sin
2
�
2
; (2.2)
and the negative square of the four-momentum transfer squared is de�ned:
Q
2
� �q
2
= �(k
0
� k)
2
: (2.3)
The invariant mass W of the hadron �nal state is obtained by the four-momentum
transfer of the ex hanged photon and the target four-momentum P of the initial
state:
W
2
= (P + q)
2
= M
2
+ 2Pq + q
2
= M
2
+ 2M� �Q
2
; (2.4)
where the Lorenz invariant variable � is de�ned as:
� =
P � q
M
lab
= E � E
0
: (2.5)
In the laboratory frame � gives the transfer energy from the in oming lepton to the
target by the virtual photon. Using this Q
2
the two dimensionless variables x; y are
given by:
x =
Q
2
2P � q
lab
=
Q
2
2M�
; (2.6)
y =
P � q
P � k
lab
=
�
E
: (2.7)
x is the variable of Bjorken s aling and y is the fra tional energy transfer and these
an be determined in the region of 0 to 1.
In the elasti s attering where W = M :
2M� �Q
2
= 0: (2.8)
On the other hand in the inelasti s attering where W is larger thanM the relation:
2M� �Q
2
> 0; (2.9)
is obtained. The Bjorken s aling variable x is a measure for the inelasti ity of the
s attering pro ess. In the elasti pro ess x is de�ned as:
x = 1: (2.10)
In the inelasti pro ess x takes the value from 0 to 1:
0 < x < 1: (2.11)
2.2. CROSS SECTION AND NUCLEON STRUCTURE FUNCTION 7
2.2 Cross Se tion and Nu leon Stru ture Fun -
tion
In DIS, the pro ess o urs in a large region of the square of four-momentum transfer
Q
2
and energy transfer �. The ross se tion of DIS an be written as:
d
2
�
dE
0
d
=
�
2
Mq
4
�
E
0
E
L
��
W
��
; (2.12)
where � is the �ne stru ture onstant, L
��
is the lepton tensor andW
��
is the hadron
tensor. The lepton tensor an be obtained from QED al ulation:
L
��
(k; s; k
0
; s
0
) = [�u(k
0
; s
0
)
�
u(k; s)℄[�u(k
0
; s
0
)
�
u(k; s)℄
�
; (2.13)
and we an split into symmetri (S) and anti-symmetri (A) part:
L
��
(k; s; k
0
; s
0
) = L
��
(S)
+ iL
��
(A)
; (2.14)
where
L
��
(S)
= 2
n
k
�
k
0�
+ k
0�
k
�
� g
��
(k � k
0
�m
2
)
o
; (2.15)
L
��
(A)
= 2 m �
����
s
�
(k � k
0
)
�
; (2.16)
where m is the lepton mass, s and s
0
are ovariant spin four-ve tor of the initial
and �nal lepton, respe tively. In Eq.2.15-Eq.2.16, the notations are used, whi h are
shown as:
�
0123
= �1; (2.17)
g
��
=
8
>
<
>
:
1 � = � = 0;
�1 � = � = 1; or 2 or 3;
0 � 6= �:
(2.18)
Like the lepton tensor, the hadron tensor is separated into symmetri (S) and
anti-symmetri (A) part:
W
��
(q;P; S) =W
��
(S)
(q;P ) + iW
��
(A)
(q;P; S): (2.19)
The hadron tensor re e ts the stru ture of hadrons. Although the lepton tensor
an be obtained exa tly from the al ulation, the hadron tensor annot be al ulated
be ause hadrons are not pointlike parti les. Therefore it an be parameterized by so
alled stru ture fun tions. In the following se tions, the ross se tion and stru ture
fun tion in the unpolarized and polarized ase will be dis ussed.
8 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
2.2.1 Unpolarized Cross Se tion and Stru ture Fun tion
The spin-independent part of the hadron tensor W
��
(S)
an be expressed by dimen-
sionless stru ture fun tions, F
1
(x;Q
2
) and F
2
(x;Q
2
):
W
��
(S)
(q;P ) = 2
�g
��
�
q
�
q
�
Q
2
F
1
(x;Q
2
)
!
+
P
�
+
P � q
Q
2
q
�
!
P
�
+
P � q
Q
2
q
�
!
F
2
(x;Q
2
)
P � q
: (2.20)
The stru ture fun tions F
1
(x;Q
2
) and F
2
(x;Q
2
) are Lorentz-invariant and re e t
the internal stru ture of the nu leon. With the symmetri parts of the lepton and
hadron tensors, the unpolarized di�erential ross se tion in the lab system (Eq.2.12)
is:
d
2
�
dxdQ
2
=
d�
d
!
�
1
EE
0
�
x
"
F
2
(x;Q
2
) +
2�
M
F
1
(x;Q
2
) tan
2
�
2
!#
: (2.21)
These two stru ture fun tions F
1
(x;Q
2
) and F
2
(x;Q
2
) are independent or have
very weak dependen e of the value of Q
2
for the �xed x. Therefore, F
1
(x;Q
2
) and
F
2
(x;Q
2
) an be rewritten in the limit of the large Q
2
:
F
1
(x;Q
2
) ! F
1
(x); (2.22)
F
2
(x;Q
2
) ! F
2
(x); (2.23)
where x =
Q
2
2M�
is the momentum fra tion whi h the parton arries.
The inelasti stru ture fun tions F
1
(x) and F
2
(x) depend only on the x variable,
whi h is alled the Bjorken s aling.
2.2.2 Polarized Cross Se tion and Stru ture Fun tion
In the polarized ase, the ross se tion in ludes two types stru ture fun tions. One
is the two spin-averaged stru ture fun tions F
1
(x;Q
2
) and F
2
(x;Q
2
) ( ontaining in
W
��
(S)
) like in the unpolarized ase, other is the two spin-dependent stru ture fun -
tions g
1
(x;Q
2
) and g
2
(x;Q
2
) ( ontaining in W
��
(A)
). W
��
(A)
is expressed by g
1
(x;Q
2
)
and g
2
(x;Q
2
):
W
��
(A)
(q;P; s) = �
����
�
q
�
�
"
S
�
g
1
(x;Q
2
) +
s
�
�
S � q
P � q
P
�
g
2
(x:Q
2
)
!#
: (2.24)
The spin-dependent stru ture fun tions an be obtained by measuring the dif-
feren es of the ross se tions in the opposite target spin states be ause the spin-
independent stru ture fun tions are an eled in the ross se tion di�eren es. The dif-
feren es of ross se tion has only the L
��
(A)
W
��(A)
term and are obtained by g
1
(x;Q
2
)
and g
2
(x;Q
2
) as follows:
X
s
0
"
d
2
�
ddE
0
(k; s; P;�S; k
0
; s
0
)�
d
2
�
ddE
0
(k; s; P; S; k
0
; s
0
)
#
=
�
2
2Mq
4
�
E
0
E
4L
��
(A)
W
��(A)
(2.25)
=
8��
2
Q
4
y
E
"
(E + E
0
os �)g
1
(x;Q
2
)�
Q
2
�
g
2
(x;Q
2
)
#
: (2.26)
2.3. THE QUARK PARTON MODEL 9
In the Bjorken limit, the spin-dependent stru ture fun tions g
1
(x) and g
2
(x)
whi h are dependent only on x are obtained:
g
1
(x;Q
2
)! g
1
(x); (2.27)
g
2
(x;Q
2
)! g
2
(x): (2.28)
2.3 The Quark Parton Model
The basis of the Quark Parton Model (QPM) is that the proton stru ture is ob-
tained from the longitudinal momentum of onstituents whi h are alled 'partons'.
The harged partons are the quarks and others are the gluons. We an easily treat
the deep-inelasti s attering in the Breit-Frame (Fig 2.4), where the transverse mo-
mentum and the rest mass of the partons an be negle ted.
If P is de�ned as the four-momentum ve tor of proton target and �P is the part
of the proton momentum whi h is arried by the parton the following relation is
obtained in deep-inelasti s attering by this parton and lepton:
(�P + q)
2
= m
2
q
: (2.29)
Negle ting the mass of the parton m
q
, � is given by:
� =
Q
2
2M�
�
2
1 +
q
1 +Q
2
=�
2
=
2x
1 +
q
q + 4(Mx)
2
=Q
2
: (2.30)
In ase of Q
2
>> M
2
, Eq.(2.30) simpli�es:
� �
x
1 + (Mx)
2
=Q
2
� x: (2.31)
In the Breit-Frame, � (Na htmann valuable) whi h expresses the ratio of the
longitudinal momentum of parton to the total momentum of proton is identi al
with the Bjorken x in the large-Q
2
limit. In this frame, we noti e that the photon
whi h has the four-momentum q in the lab-frame intera ts with the parton whi h
has the momentum fra tion x = Q
2
=2M�.
e+’
e+
γξ
ξ P
- P
Figure 2.4: DIS Pro ess in Breit-Frame. �P expresses the momentum whi h
is arried by the parton.
10 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
From the above onsideration, we an de�ne the stru ture of nu leon using the
parton momenta. The nu leon ontains two types of quarks, the valen e and the
sea quarks. The valen e quarks de ide the quantum number of the nu leon and the
sea quarks are pairs of quarks and anti-quarks whi h are produ ed virtually from
the gluons by the strong intera tion.
The stru ture fun tion F
2
(x) an be expressed by quark momentum distributions
in the QPM as:
F
2
(x) = x �
X
f
e
2
f
(q
v
(x) + q
s
(x))
f
; (2.32)
in whi h f denotes the quark avor, q
v
(x) is the number density of the valen e
quarks, q
s
(x) is that of the sea quarks and e
f
is the orresponding ele tri harge.
The number density of a quark with avor f whi h has the momentum fra tion x
is alled Parton Density Fun tion (PDF).
2.3.1 The unpolarized PDF
The unpolarized PDF of quark q with avor f is des ribed as follows:
q
f
(x) = (q
v
(x) + q
s
(x))
f
; (2.33)
q
s
f
(x) = �q
f
(x): (2.34)
The q(x) onsists of the density of the valen e quarks q
v
(x) and the sea quarks
q
s
(x). q
s
(x) and the anti-quark density �q(x) are identi al. The PDF re e ts the
distribution of quark momentum in the nu leon.
Assuming that the heavy quarks an be negle ted and using the isospin symmetry
between the proton and neutron, we obtain the relations:
u
p
(x) = d
n
(x) � u(x); (2.35)
d
p
(x) = u
n
(x) � d(x); (2.36)
s
p
(x) = s
n
(x) � s(x): (2.37)
The valen e quarks are de�ned as:
u
v
(x) = u(x)� �u(x); (2.38)
d
v
(x) = d(x)�
�
d(x); (2.39)
and we an reprodu e the quantum number of the proton summing up all parton
densities:
Z
1
0
[u(x)� �u(x)℄dx = 2; (2.40)
Z
1
0
[d(x)�
�
d(x)℄dx = 1; (2.41)
Z
1
0
[s(x)� �s(x)℄dx = 0: (2.42)
2.3. THE QUARK PARTON MODEL 11
2.3.2 The polarized PDF
In the polarized QPM, the quark spin distribution �q
f
(x) is de�ned as follows:
�q
f
(x) = q
+
f
(x)� q
�
f
(x); (2.43)
where q
+
f
(x) is the distribution of quarks with spins in the same dire tions as the spin
of the nu leon, q
�
f
(x) is that with the opposite dire tion. When the spin dire tion
of the virtual photon and the quarks are longitudinally parallel or anti-parallel, we
an de�ne the polarized stru ture fun tions:
g
1
(x) =
1
2
X
f
e
2
f
(q
+
f
(x)� q
�
f
(x))
=
1
2
X
f
e
2
f
�q
f
(x); (2.44)
g
2
(x) = 0: (2.45)
In the simple QPM, g
2
(x) is approximated to zero. The value of g
1
(x) is most
sensitive to the quark spin distribution. This is explained in Fig.2.5. The in oming
polarized lepton emits a polarized virtual photon. This photon an be absorbed
only by a quark with the opposite spin dire tion be ause the �nal state of a quark
must have spin 1/2. In the top pi ture of Fig.2.5 we an measure the ross se tion
for beam and target polarization anti-parallel and in the bottom pi ture we an
measure that for polarization parallel. The top pro ess is sensitive to q
+
f
(x) and the
bottom is sensitive to q
�
f
(x). Sin e g
1
(x) is the di�eren e of q
+
f
(x) and q
�
f
(x), it
gives the information of the quark spin.
Spin of photon
Spin of nucleon
Spin of photon
Spin of nucleon
e+
e+
e+’
e+’
q (x)f-
q +(x)f
Figure 2.5: Diagram polarized DIS. The top �gure shows the pro ess whi h
is sensitive to quarks polarized parallel to the nu leon's spin, and
the bottom �gure is the pro ess for quarks with the anti-parallel
spin to the nu leon.
12 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
2.3.3 Measurement of g
1
(x)
To measure g
1
(x) in the DIS experiments, we de�ne the asymmetry of the polarized
ross se tions in the ases of the dire tion of the target polarization longitudinal
(� = 0
o
; 180
o
in Fig.2.3) and transversal (� = 90
o
; 270
o
in Fig.2.3) to that of the
in oming lepton:
A
k
=
�(180
o
)� �(0
o
)
�(180
o
) + �(0
o
)
; (2.46)
A
?
=
�(270
o
)� �(90
o
)
�(270
o
) + �(90
o
)
: (2.47)
This measured ross se tion asymmetry is related to the asymmetry in absorption
of the virtual photon whi h was emitted by the in ident lepton:
A
k
= D(A
1
+ �A
2
); (2.48)
A
?
= d(A
2
� �A
1
); (2.49)
where D; �; d; � are given by:
D =
y(2� y)
y
2
+ 2(1� y)(1 +R(x;Q
2
))
;
d = D
s
2�
1 + �
;
� =
1� y
1� y +
y
2
2
;
� =
2 (1� y)
2� y
;
� = �
1 + �
2�
;
=
Q
�
=
2Mx
Q
:
R(x;Q
2
) is the ratio of longitudinal (�
T
) to transverse (�
L
) absorption ross
se tion
1
of the virtual photon:
R(x;Q
2
) =
�
L
�
T
= (1 +
2
)
F
2
(x;Q
2
)
2xF
1
(x;Q
2
)
� 1: (2.50)
In the Bjorken limit, R(x;Q
2
) be omes zero. The polarized stru ture fun tions g
1
and g
2
an be dedu ed from A
1
and A
2
by:
A
1
=
g
1
�
2
g
2
F
1
; (2.51)
A
2
=
(g
1
+ g
2
)
F
1
: (2.52)
A
2
is limited by R(x;Q
2
) be ause it ontains the interferen e term �
I
between �
T
and �
L
whi h has to obey the triangular relation �
I
�
p
�
L
�
T
[11℄, leading to:
jA
2
j �
q
R(x;Q
2
): (2.53)
1
For photons \polarization" is usually used for the! E-�eld dire tion, not for the spin dire tion.
In this sense, transverse photon polarization means longitudinal spin dire tion.
2.4. SUM RULES 13
From the results of re ent experiment [12℄, it is lear that A
2
is lose to zero. If
g
2
and A
2
an be ignored, g
1
(x) is obtained by these simple relations:
A
1
�
A
k
D
; (2.54)
g
1
� A
1
� F
1
: (2.55)
In Bjorken limit, F
1
(x) and g
1
(x) are des ribed by:
F
1
(x) �
F
2
(x)
2x
; (2.56)
g
1
(x) �
A
k
D
F
2
(x)
2x
: (2.57)
Eq.2.56 is alled the Callan-Gross relation. The re ent measurement of g
1
(x) is
shown in Fig.2.6.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
x
gp 1
HERMES ( 1997 )
SLAC E-143
Q2 = 2 GeV2
0.1 0.50.02x
HERMES ( 1997 )
SMC ( 1993+1996 )
Q2 = 10 GeV2
0.1 0.50.02
Figure 2.6: The re ent results of the spin dependent stru ture fun tion on the
proton g
p
1
(x). The solid ir le shows the �nal result for g
1
from
1997 data at HERMES, whi h are shown at Q
2
=2 GeV
2
(the
left �gure) and Q
2
=10 GeV
2
(the right �gure). The results from
SLAC-E143 (Q
2
=2 GeV
2
) and SMC (Q
2
=10 GeV
2
) are shown
for omparison.
2.4 Sum Rules
2.4.1 Gottfried Sum Rule
The Gottfried Sum Rule [13℄ is for the unpolarized stru ture fun tions of proton F
p
2
and neutron F
n
2
. From Eq.2.32, Eq.2.40 - Eq.2.42 we obtain the relation:
S
G
=
Z
1
0
1
x
(F
p
2
(x)� F
n
2
(x))dx;
=
1
3
+
2
3
Z
1
0
u
s
(x)� d
s
(x): (2.58)
14 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
If the unpolarized sea quark distributions is symmetri , S
G
=1/3 should be obtained.
The measured value by NMC is S
G
= 0.235 � 0.026 [14℄, whi h means this assump-
tion is not valid and an unpolarized sea quark avor asymmetry exists.
2.4.2 Bjorken Sum Rule
In polarized ase, the integral of g
1
(x;Q
2
) is derived from Eq.2.44:
�
1
(Q
2
) �
Z
1
0
g
1
(x;Q
2
)dx: (2.59)
With the assumption of iso-spin symmetry, the integrals of the proton and neutron
spin stru ture fun tions �
p
1
and �
n
1
are expressed as follows:
�
p
1
=
1
9
a
0
+
1
12
a
3
+
1
36
a
8
; (2.60)
�
n
1
=
1
9
a
0
�
1
12
a
3
+
1
36
a
8
; (2.61)
where a
0
, a
3
and a
8
are the SU(3) avor singlet, isotriplet and o tet isosinglet
ombinations. They are expressed by the quark distributions:
a
0
= (�u+��u) + (�d+�
�
d) + (�s+��s); (2.62)
a
3
= (�u+��u)� (�d +�
�
d); (2.63)
a
8
= (�u+��u) + (�d+�
�
d)� 2(�s+��s): (2.64)
(2.65)
Furthermore, assuming SU(3) avor symmetry, a
3
and a
8
an be obtained from the
weak oupling onstants F and D:
a
3
= F +D; (2.66)
a
8
= 3F �D: (2.67)
The values of F;D are measured in neutron and hyperon �-de ays. F and D are
obtained as follows [15℄:
F = 0:459� 0:008; (2.68)
D = 0:798� 0:008: (2.69)
From the di�eren e of the two polarized stru ture fun tions, the Bjorken sum
rule is derived as follows:
�
p
1
(Q
2
)� �
n
1
(Q
2
) =
1
6
�
�
�
�
�
�
g
A
g
V
�
�
�
�
�
� Corr (2.70)
=
1
6
� (F +D) � Corr; (2.71)
where g
A
; g
V
are the axial-ve tor and ve tor oupling onstants. Corr is the orre -
tion fa tor with higher order QCD.
The Bjorken Sum Rule [16℄ [17℄ uses the assumption of isospin invarian e only,
and it is therefore a good tool to test QCD. No existing experimental data show a
violation of this sum rule [18℄.
2.5. SEMI-INCLUSIVE MEASUREMENT 15
2.4.3 Ellis-Ja�e Sum Rule
If we assume that �s = 0 in the Bjorken Sum Rule, we an derive the moment of
g
p
1
and g
n
1
from Eq.2.60 and Eq.2.61:
�
p
1
=
1
12
�
�
�
�
�
�
g
A
g
V
�
�
�
�
�
�
�
1 +
5
3
�
3F �D
F +D
�
� Corr
=
1
18
(9F �D); (2.72)
�
n
1
=
1
12
�
�
�
�
�
�
g
A
g
V
�
�
�
�
�
�
�
�1 +
5
3
�
3F �D
F +D
�
� Corr
=
1
18
(6F � 4D): (2.73)
At Q
2
= 5GeV
2
, the following values are obtained for Ellis-Ja�e sum rule[19℄:
�
p
1
= 0:179� 0:009;
�
n
1
= �0:019� 0:009:
These al ulated values are not onsistent with the values from the experiment
[18℄. The measured values are learly smaller than the predi tion. This disagreement
shows a polarization of the strange quarks. The violation of Ellis-Ja�e Sum Rule is
the starting point of the study of the nu leon spin stru ture (the Spin Crisis). One
of the motivations for the HERMES experiment is to solve this puzzle.
2.5 Semi-In lusive Measurement
Now we return the onsideration of the DIS pro ess again. The DIS pro ess of
lepton l on nu leon N is shown in the previous se tion as follows:
l +N ! l
0
+ h+X; (2.74)
where l
0
is the s attered lepton, h is the dete ted hadron, X are other not de-
te ted parti les. In in lusive measurements, we dete t the s attered lepton only. In
semi-in lusive measurements, we dete t produ ed hadrons h in oin iden e with the
s attered lepton l
0
. In this measurement we an extra t the quark spin distributions
of ea h avor. The diagram of semi-in lusive DIS is shown in Fig.2.7. The virtual
photon emitted by the in ident lepton is absorbed by one quark in the target and
some hadrons are produ ed. We assume that the hadron with the largest momentum
is produ ed from the fragmentation of the stru k quark in the target. This hadron
is alled \leading hadron".
There are two types of produ ed hadrons, one is the \ urrent fragment" whi h
is produ ed from the fragmentation of the stru k quark and another is the \target
fragment" whi h is produ ed from the target remnant. These two an be kinemati-
ally separated by the Feynman s aling variable x
F
. We need the urrent fragments
and therefore sele t the hadron with x
F
> 0.
The fragmentation pro esses of the stru k quark annot be al ulated exa tly,
so it is parameterized using the experimental data [20℄ [21℄, whi h are so alled frag-
mentation fun tion D
h
f
(z). These fun tions de�ne the probability that the hadron
h is with momentum fra tion z produ ed from a stru k quark with avor f .
16 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
Figure 2.7: Diagram of semi-in lusive DIS. P
h
is the momentum of a pro-
du ed hadron in the DIS pro ess. X shows other produ ed par-
ti le whi h is not identi�ed.
The semi-in lusive DIS variables
P
jj
=
!
P
h
�
!
q
j
!
q
j
longitudinal momentum of a hadron
x
F
=
P
jj
j
!
q
j
'
2P
jj
W
Feynman s aling variable
z =
P � P
h
P � q
lab
=
E
h
�
fra tional energy transfer of the virtual photon arried
by a hadron h
Table 2.2: De�nition of semi-in lusive kinemati al values.
2.5.1 Fragmentation Fun tions
In the QPM, the hadron ross se tion �
h
in semi-in lusive DIS is expressed with
fragmentation fun tions and the unpolarized quark distributions:
1
�
tot
d�
h
dz
(x;Q
2
; z) =
P
f
e
2
f
q
f
(x;Q
2
)D
h
f
(x;Q
2
; z)
P
f
e
2
f
q
f
(x;Q
2
)
; (2.75)
where �
tot
is the ross se tion of in lusive DIS, e
f
is the quark harge and q
f
are the
unpolarized PDF. The fragmentation fun tions are normalized by the multipli ity
n
h
of all produ ed hadron:
X
f
Z
1
0
D
h
f
(Q
2
; z) dz = n
h
(Q
2
); (2.76)
2.5. SEMI-INCLUSIVE MEASUREMENT 17
and energy momentum onservation leads to:
X
h
Z
1
0
zD
h
f
(Q
2
; z) dz = 1: (2.77)
The fragmentation fun tions of quarks with avor f are determined for all
hadrons. Then the n
f
� n
h
types fragmentation fun tions exist, where n
f
is the
number of the quark types and n
h
is that of the hadron types. We assume iso-spin
symmetry and harge symmetry to redu e the number of D
h
f
(x). If we use these
two assumptions, for example for pion produ tion, three types of fragmentation
fun tions are obtained as follows:
D
+
(z) � D
�
+
u
(z) = D
�
�
d
(z) = D
�
+
�
d
(z) = D
�
�
�u
(z); (2.78)
D
�
(z) � D
�
+
d
(z) = D
�
�
u
(z) = D
�
+
�u
(z) = D
�
�
�
d
(z); (2.79)
D
s
(z) � D
�
+
s
(z) = D
�
�
s
(z) = D
�
+
�s
(z) = D
�
�
�s
(z); (2.80)
where the Q
2
dependen e of the fragmentation fun tions has been omitted. Here
we onsider only three quark types (u; d; s) and � as produ ed hadron. D
+
(z) is
alled favored fragmentation fun tions and D
�
(z) is alled unfavored. In the ase
of strange quarks, we de�ne D
s
(z) separately be ause of the mass of strange quark.
In theory, some fragmentation models are established. These models are shown in
the next se tion.
2.5.2 Fragmentation Models
Fragmentation fun tions are one of the most important variables when we study
semi-in lusive DIS pro esses, but they are not al ulated in perturbative QCD. So
some phenomenologi al fragmentation models have to be used for example in Monte
Carlo programs. Three fragmentation models are des ribed below, all of whi h are
based on the QPM.
Independent Fragmentation Model
The independent fragmentation model was onsidered to des ribe the fragmentation
pro ess in 1978, whi h is one of the �rst models of fragmentation by Field and
Feynmann [22℄. This model is based on the simple assumption that every quark
fragments independently. When a quark q
0
is stru k, it ombines with other anti-
quark �q
1
and a q
0
�q
1
pair is reated whi h is the �rst meson with energy fra tion z
0
.
The energy left z
1
= (1 � z
0
) is used for another pair q
1
�q
2
. These meson reation
pro ess is iterated down to a ertain energy uto� E
min
.
One of the parameters of this model is f(1 � z)dz whi h gives the probability
that a parti ular meson is reated �rst. A further parameter is the probability
f
for the reation of a q
f
�q
f
quark pair with same avor. The isospin invarian e gives
u
+
d
+
s
= 1 and �
u
=
d
. A ording to the measured K/�-ratio the
probability of s�s reation is
s
= 0:3 .
In this model the ratio between the favored and unfavored fragmentation fun -
tions is obtained as follows:
D
�
(z)
D
+
(z)
=
(1� z)
z + (1� z)
: (2.81)
18 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
If we negle t the strange quarks ( = 0:5) this ratio is expressed as D
�
=D
+
=
(1� z)=(1 + z).
The independent fragmentation an des ribe many properties of hadron frag-
mentation, but there are some inherent problems like the violation of olor and
avor quantum number onservation. Therefore other models have been developed
and the independent fragmentation model is usually not used any longer.
LUND String Model
The LUND String model [23℄ is the most su essful model today, whi h is based on
the QCD predi tion. It is similar to the independent fragmentation model, but now
the fragmentation is not independent, but the quarks are onne ted to the olor
�eld. With growing distan e between the quarks, the olor �eld is sket hed as a
tube and is alled a string. The energy is stored in this string of olor �eld. When
the energy ex eeds the mass of a q�q-pair, a new pair is produ ed and the string is
divided in two strings. The fragmentation pro eeds until the last pair is lose to the
mass of a olorless hadron.
In this model, there are main three parameters in the fun tion whi h is alled
LUND symmetri fragmentation fun tion:
f(z) =
(1� z)
a
z
� exp
�bm
2
?
z
!
; (2.82)
where m
?
is the transverse mass squared m
2
?
= m
2
+ p
2
?
, a and b are parameters
whi h are determined from experiment.
The LUND string model is used by the JETSET [24℄ library whi h is the basis
for Monte Carlo studies for HERMES. These parameters have to be tuned for the
onditions at HERMES [25℄. The numbers are shown below (Fig.2.8 and Tab.2.3).
In the present studies, these values are used for all MC produ tions.
a b < P
2
?
>
Default Value 0.30 0.58 0.36
Fitted Value 0.82 0.24 0.34
Table 2.3: Fitted parameters of LUND model for HERMES. The parameters
of the LUND are tuned for HERMES experiment. These �tted
valued are determined to reprodu e the spe tra from all hadrons.
Cluster Model
The Cluster model des ribes the fragmentation pro esses by the QCD perturbation
theory. The highly virtual partons go into olor-singlet lusters after the intera tion
between the photon and the parton whi h is alled \pre on�nement". The lusters
de ay into hadrons from simple phase spa e arguments. The Cluster model has the
advantage that it is most theoreti al and the adjustable parameters are the QCD
s ale parameters.
2.5. SEMI-INCLUSIVE MEASUREMENT 19
0
0.5
1
1.5
2
2.5
0 0.2 0.4 0.6 0.8 1
z
dN
h/d
z
JETSET:Data (Corr.)FittedDefault
Figure 2.8: Charged hadron spe tra orre ted for dete tor in uen es. The
solid line shows the MC data with the �tted parameters in
Tab.2.3. The dashed line is the MC data with the default pa-
rameters whi h is before the tuning of MC.
This model is used in the HERWIG Monte Carlo generator [26℄. The parameters
of this model are �tted for the HERMES experiment already. However the Cluster
model fragmentation are not onsistent with the HERMES data. Sin e the LUND
model has better agreement than the Cluster model, we use the LUND model with
�tted parameters for HERMES in our study.
2.5.3 Hadron Asymmetry
In the semi-in lusive measurement, A
h
jj
is de�ned as a semi-in lusive asymmetry for
hadron h a ording to Eq.2.46:
A
h
k
=
�
h
(180
o
)� �
h
(0
o
)
�
h
(180
o
) + �
h
(0
o
)
=
N
h
"#
�N
h
""
N
h
"#
+N
h
""
; (2.83)
where N
h
"#
is the number of dete ted hadrons in the ase of the beam and target
polarizations anti-parallel and N
h
""
is for the ase of parallel polarizations.
The semi-in lusive stru ture fun tions are de�ned as:
g
h
1
(x;Q
2
; z) =
1
2
X
f
e
2
f
D
h
f
(Q
2
; z)�q
f
(x;Q
2
); (2.84)
F
h
1
(x;Q
2
; z) =
1
2
X
f
e
2
f
D
h
f
(x;Q
2
; z)q
f
(x;Q
2
); (2.85)
where the assumption that the fragmentation pro ess is spin-independent is used.
20 CHAPTER 2. DEEP-INELASTIC SCATTERING (DIS)
From Eq.2.51 and Eq.2.50, we obtain the semi-in lusive spin asymmetry A
h
1
(x;Q
2
)
from g
h
1
(x;Q
2
; z) and F
h
1
(x;Q
2
; z) under the assumption of g
2
= 0:
A
h
1
(x;Q
2
) �
R
dz g
h
1
(x;Q
2
; z)
R
dz F
h
1
(x;Q
2
; z)
=
1 +R(x;Q
2
)
1 +
2
�
P
f
e
2
f
�q
f
(x;Q
2
)
R
dz D
h
f
(Q
2
; z)
P
f
e
2
f
q
f
(x;Q
2
)
R
dz D
h
f
(Q
2
; z)
; (2.86)
where we have to apply the relation of Eq.2.50 at HERMES kinemati range. This
hadron spin asymmetry A
h
1
(x;Q
2
) provides us the possibility to extra t quark po-
larizations.
21
Chapter 3
HERMES Experiment
3.1 Experimental Setup
The HERMES experiment is arried out at DESY in Hamburg, Germany. The
dete tor is lo ated in the East Hall of the HERA lepton-proton ollider ring. The
HERMES uses only the polarized lepton beam of HERA, together with a polarized
internal gas target. The intera tion between a polarized target and a polarized beam
provides the possibility to study the quark spin. Details of the HERA lepton storage
ring in luding the lo ations of the other three HERA experiments, ZEUS, H1 and
HERA-B, are shown in Fig 3.1.
3.2 The HERA Storage Ring
The Fig.3.1 shows the HERA lepton (ele tron or positron) ring. The ring an
be operated with lepton at an energy of 27.5 GeV. The beam be omes polarized
transverse to the beam plane due to the asymmetry in the emission of syn hrotron
radiation (Sokolov-Ternov me hanism [27℄). The beam polarization P (t) is given as
fun tion of time t:
P (t) = p
max
(1� e
�t=�
); (3.1)
where � is the polarization build up time and p
max
is the equilibrium polarization.
At HERA the beam rea hes 60% polarization in about 30 minutes after the beam
inje tion.
Be ause it is ne essary to obtain the longitudinal beam polarization at HERMES,
two spin rotators are lo ated up and downstream of the HERMES Spe trometer.
The �rst rotates the lepton spin to the longitudinal dire tion the se ond one rotates
ba k to transverse. The beam polarization is measured by a transverse polarimeter
in the West Hall of HERA and the polarization is onstant in the whole storage
ring. The longitudinal polarization of the HERA lepton beam is shown in Fig. 3:2 .
22 CHAPTER 3. HERMES EXPERIMENT
Polarized HERA Beam:
� Self-polarization by Sokolov-Ternov e�ect
� Spin Rotators { Longitudinal Spin
� P
B
� 40� 60%
O
W
NSZEUS H1
HERA-B
HERMES
beam directiontransverse polarimeter
spin rotator 1
spin rotator 2
longitudinal polarimeter
(Parallel talk by M. Spengos, Section 11, Saturday 4:45 pm)
Figure 3.1: The HERA storage ring. HERMES are lo ated with the other
three experiments, ZEUS, H1 and HERA-B. Be ause the longi-
tudinal beam polarization is needed for HERMES, the two spin
rotator is lo ated up and down stream of the HERMES experi-
ment. In 1995 - 1997, the positron beam is used and the ele tron
beam is operated during 1998.
0
20
40
60
80
6 6.5 7 7.5 8Time [hours]
Pol
ariz
atio
n [%
]
Longitudinal Polarimeter
Figure 3.2: The longitudinal polarization of HERA beam. The polarization
rea hes up to 60 % in about 30 minutes.
3.3. THE INTERNAL GAS TARGET 23
3.3 The Internal Gas Target
HERMES uses a polarized internal gas target, applying the novel te hnique of an
internal storage ell. The target gas is inje ted in an open-ended aluminum tube
inside the lepton storage ring. The thi kness of this aluminum is about 100 �m. The
polarized gas oming from an atomi beam sour e (ABS), where the polarization is
a hieved by Stern-Gerla h separation, is inje ted into the storage ell via a small
feed tube. The leaking gas from both open ends is pumped away by a di�erential
pump system. Fig 3.3 shows the design of the target hamber.
We used a helium3 (
3
He) target in 1995, a proton (p) target in 1996 and 1997
and a deuteron (D) target from 1998 to 2000. In this system the target density is
1.2 � 10
15
nu leons/ m
2
for
3
He and 7.5 � 10
13
nu leons/ m
2
for p and D.
Open−Ended Storage Cell
gas injection
fixed collimator
e−beam
exit window
pumps
suppressors
wake field
vacuum chamber
thin walledbeam pipe
29 mm
9.8 mm
Figure 3.3: The internal target of HERMES.
24 CHAPTER 3. HERMES EXPERIMENT
3.4 The HERMES Spe trometer
HERMES spe trometer has a large a eptan e and is a forward angle dete tor. This
spe trometer is shown in Fig.3.4 and Fig.3.5 [28℄. With the HERMES dete tor,
s attering angles of parti les are a epted within � 170 mrad in the horizontal
dire tion and between +/(-)40 mrad and +/(-)140 mrad in the verti al dire tion.
A spe trometer magnet and several dete tors for parti le tra king allow the de-
termination of parti le momenta. The tra king system onsists of mi rostrip gas
hambers, drift hambers and three proportional hambers in the magnet.
For semi-in lusive measurements in DIS, HERMES needs a very good parti le
identi� ation (PID), not only to separate lepton and hadrons, but also to identify
the hadrons. To a hieve this high quality parti le identi� ation, the HERMES
spe trometer is equipped with four spe ial PID dete tors : an ele tro magneti lead-
glass alorimeter, a preshower hodos ope, a transition radiation dete tor (TRD) and
Cherenkov Counter. The original threshold Cherenkov has been repla ed in 1998 by
a Ring Imaging Cherenkov Counter (RICH) to allow a lean hadron identi� ation
over the momentum rage of 2 � 15 GeV.
1
0
2
-1
-2
0 1 2 3 4 5 6 7 8 m
mDRIFT CHAMBERS
LUMINOSITY
CHAMBERSDRIFT
VC 1/2
FC 1/2
VERTEX CHAMBERS
TARGETCELL
DVC
MC 1-3
HODOSCOPE H0
MONITOR
BC 1/2
BC 3/4 TRD CALORIMETER
TRIGGER HODOSCOPE H1
MAGNET
PROP.CHAMBERS
FIELD CLAMPS
PRESHOWER (H2)
CERENKOV
STEEL PLATE
Figure 3.4: The HERMES Spe trometer. Two identi al dete tors are set
upper and lower of the beam line. The Cherenkov Counter is
operated as a threshold Cherenkov Counter during 1995 - 1997
data taking. Sin e 1998, a Ring Imaging Cherenkov Counter are
repla ed to identify produ ed hadrons.
3.4. THE HERMES SPECTROMETER 25
3He Target
ChambersVertex
ChambersFront Chambers
Magnet MagnetChambers
Back Cerenkov TRD Hodoscopes
Calorimeter
10 m
Figure 3.5: A three dimensional CAD diagram of the HERMES Spe trome-
ter.
3.4.1 The Tra king System
Spe trometer Magnet
The HERMES spe trometer has a H-type magnet with a de e ting power of 1.5 Tm
and it is usually operated at 1.3 Tm for the momentum measurement. The lepton
beam lines are prote ted by a horizontal septum plate with ompensator oils whi h
is arranged in the beam plane. The angular a eptan e is 40 mrad < j�
y
j < 140
mrad verti ally, and j�
x
j < 170 mrad horizontally, where the lower verti al limit is
given by the septum plate. The total a eptan e of s attering angles is 40 mrad
< j�j < 220 mrad.
Vertex and Front Chambers
In upper region of the spe trometer magnet, the mi ro-strip gas hambers whi h is
alled vertex hambers VC1/2 and the drift vertex hambers DVC and the drift front
hambers FC1/2 are installed to re onstru t the parti le vertex and to determine the
s attering angle. The VCs onsisted of 2 modules ea h with 6 planes, these planes
are made by glass bases of 300 �m thi kness whi h are overed with 7 �m aluminum
strips. These strips are separated by 193 �m and arranged in 3 mm distan e from a
athode plane. The gap is �lled with DiMethylEther(DME)/Ne mixture (50/50%)
gas. The eÆ ien ies of vertex hambers were around 95% for 1997 running, and the
resolution of about 65 �m (�) per plane was obtained.
The FC's is a pair of drift hambers, ea h hamber has one module with 6 planes.
26 CHAPTER 3. HERMES EXPERIMENT
The size of drift ell is 7 mm (3.5 mm maximum drift length), and the used gas is
Ar/CO
2
/CF
4
(Ar:CO
2
:CF
4
90:5:5) like in all other HERMES drift hambers.
The DVC's were onstru ted and installed downstream of the VC, be ause of
problems with the VC's, The DVC a eptan e is verti ally � 35 mrad � � 270
mrad and horizontally � 200 mrad. But the DVC drift ell is smaller than the FC's.
The gas omposition of DVC is Ar/CO
2
/CF
4
, the same as for the FC's.
Magnet Chambers
The proportional wire hambers MC1/2/3 are mounted in the gap of the spe -
trometer magnet. These were originally installed to analyze multiple tra ks in high
multipli ity events. It has turned out that they are also parti ularly useful for the
analysis of parti les with low energy whi h are de e ted by the magnet so that they
do not ome in the a eptan e, pion from the �-de ays, for example. Ea h hamber
onsists of 3 modules. The distan e between anode and athode planes is 4�0.03
mm. The gas is the same as the drift hamber's but the proportions of the mixture
is di�erent (Ar:CO
2
:CF
4
65:30:5).
Ba k Chambers
In the region downstream of the spe trometer magnet two sets of the ba k drift
hambers BC1/2 and BC3/4 are mounted. The di�eren e between BC1/2 and
BC3/4 is the dete tor size. The a tive areas of these modules depend on their
z-position and the a eptan e of the spe trometer. The BC1/2 a tive area is hor-
izontally 1880 mm and verti ally 520 mm, and the BC3/4's is 2890 mm and 710
mm. Ea h set of hambers has two modules with 6 planes. The eÆ ien y of the
drift hamber depends on the drift distan e and the typi al eÆ ien y is well above
99%. The details are shown in [29℄. These dete tors are shown in Fig.3.6.
Figure 3.6: S hemati view of the HERMES BC module.
3.4. THE HERMES SPECTROMETER 27
3.4.2 The Parti le Identi� ation System
As I mentioned before, a lean parti le identi� ation is very important at HERMES.
The PID is performed in two steps : �rst a lepton-hadron separation and se ond the
identi� ation of the hadron. Espe ially the se ond point is essential for the study of
the quark spin distribution. As for the lepton-hadron separation, the hadron reje -
tion fa tor (HRF) of the system is designed to be 10
4
at least, where HRF is de�ned
as the total number of hadrons divided by the number of hadrons misidenti�ed as
leptons.
This value depends on the eÆ ien y for the lepton identi� ation. The ontamina-
tion of the s attered lepton sample by hadrons is below 1% for the whole kinemati
range. This hadron reje tion is done by the alorimeter, the preshower hodos ope
and the TRD. The se ond step, the hadron identi� ation is done by the fourth PID
dete tor, the RICH. These four dete tors are des ribed in the following.
Calorimeter
The alorimeter (Fig.3.7) provides a �rst level trigger for s attered leptons. It on-
sists of two modules whi h have 420 blo ks ea h above and below the beam line.
All 840 blo ks are made by lead-glass blo ks with size of 9 � 9 � 50 m
3
and are
arranged in a 42 � 10 array for ea h module. Ea h blo k is readout by a photomul-
tiplier tube (PMT). Both the alorimeter walls an be moved 50 m verti ally from
the beam pipe to prevent radiation damage during beam inje tion or dump. The
energy resolution of this dete tor is parameterized by the following fun tion:
�E
E
= 1:5� 0:5 +
5:1� 1:1
q
E[GeV ℄
; (3.2)
where E is energy of leptons measured by the alorimeter. Combined with the
preshower ounter, the eÆ ien y for the dete tion of lepton is about 95%. The
alorimeter is des ribed in detail in [30℄.
90 (10 cells)
Electron beam
9 x 9 x 50 cm 3
378 (42 cells) cm
cm 20
48 cm
cm90 (10 cells)
cm
Figure 1: Isometric view of the HERMES calorimeter.
The light transmittance was measured for di�erent F101 LG block: as the
working wavelengths of light exceed 500 nm, the spectral distribution well matches
the spectral response of the green extended phototube Philips XP3461 chosen.
3 Measured characteristics of the calorimeter
Measurements with 1�30 GeV/c
10)
electron beams have been performed
at CERN and DESY with a 3�3 array of counters with the addition of a preshower
detector, consisting of 1 cm of lead + 1 cm of plastic scintillator. This con�guration
provides information on the longitudinal development of the electromagnetic shower
and then allows a better rejection of pions. At DESY each lead glass block was
also calibrated and equalized to �1% in a 3 GeV/c electron beam positioned at the
module center.
3.1 Linearity of the response
The absolute energy calibration measurements showed a good linearity over the
whole measured energy range, as represented in �g. 2: the linear �t reproduces the
data better than 1%.
Figure 3.7: Isometri view of the HERMES Calorimeter.
28 CHAPTER 3. HERMES EXPERIMENT
Hodos opes
There are two hodos opes H1 and H2, whi h are plasti s intillators. H2 is behind
the 11 mm Pb radiator (two radiation lengths) sandwi hed between two 1.3 mm
stainless steel sheets and used as a preshower ounter in front of the alorimeter.
The �rst level trigger is obtained by a ombination with the alorimeter and H1,
where H2 gives a dis rimination between leptons and hadrons. Both the hodos opes
onsist of 42 verti al s intillator modules ea h in the upper and lower dete tor. The
size of the modules is 9.3 � 91 � 1 m
3
and they are read out by PMTs of 5.2 m
diameter.
Transition Radiation Dete tor
The transition radiation dete tor is used for the dis rimination between leptons and
hadrons [31℄. Only leptons emit transition radiation in the HERMES energy region.
The HERMES TRD is a multiwire proportional hamber onstru ted by 6 mod-
ules above and below the beam. Ea h module has polyethylene/polypropylene �bers
as radiators and Xe/CH
4
gas (90:10) in the proportional hamber. The �bers used
for the radiators are 17-20 �m in diameter and have a material density of 0.059
g/ m
3
. The radiators are 6.35 m thi k, and Xe/CH
4
has been hosen be ause of
its eÆ ient X-ray absorption.
The hambers are 2.54 m thi k and the wire diameter was used to be unusually
75 �m large to allow operation at high voltage while limiting the gas gain to about
10
4
. The original purpose of the TRD is to obtain the pion reje tion fa tor (PRF)
of at least 100 for 90% lepton eÆ ien y at 5 GeV and above, a ordingly the energy
averaged PRF is about 300 [31℄ for 90% lepton eÆ ien y. When we analyze the data
as the fun tion of momentum, the PRF of 130 is obtained at 5 GeV. The design
goal is therefore a omplished.
Ring Imaging Cherenkov Counter
To study the quark spin stru ture, it is ne essary that produ ed hadrons (pion, kaon,
proton) are identi�ed separately. Until 1998, threshold Cherenkov Counter was used
to identify pions whi h are produ ed in DIS s attering. This threshold Cherenkov
Counter had good performan e for pion separation from hadrons. The eÆ ien y of
this ounter is above 99%. In 1998 a Ring Imaging Cherenkov Counter (RICH) has
been installed as a substitute for the threshold Cherenkov Counter. The HERMES
RICH is designed to identify pions, kaons and protons in a momentum region of 2
� 15 GeV. This RICH has two radiators, aerogel and C
4
F
10
gas. Measuring two
ring images produ ed by Cherenkov photons on the PMT plane, we an identify and
separate ea h parti le. The RICH is des ribed in detail in the next hapter.
29
Chapter 4
Ring Imaging Cherenkov Counter
In its original version the HERMES spe trometer had a threshold Cherenkov Counter
to dete t produ ed hadrons, espe ially pions. Although this threshold ounter had
good performan e for the pion dete tion, a new Ring Imaging Cherenkov Counter
(RICH) was proposed in 1997 [32℄ to identify also other hadrons produ ed in DIS.
The RICH was installed in 1998 and is operating sin e then. This RICH is designed
to dete t three types of hadrons : pions (�
+
; �
�
), kaons (K
+
; K
�
) and (anti-)protons
(p; �p).
The HERMES RICH has a very spe ial property that it is working with two
radiator, sili a aerogel and C
4
F
10
. Thanks to re ent developments, sili a aerogel is
available with high transmission and low refra tion index. If a hadron passes the
RICH, it emits Cherenkov light at di�erent angles for sili a aerogel and C
4
F
10
Gas.
Therefore the Cherenkov light forms two rings on the photomultiplier plane. We
an identify the hadron by determining the Cherenkov angle of the two rings. The
details are dis ussed in the following.
4.1 Dete tor Design
In this se tion, the fundamental prin iple of the RICH is des ribed. If a harged par-
ti le is moving with velo ity v in an opti al medium with refra tive index n and it's
velo ity v ex eeds the velo ity of light in the medium =n, it emits Cherenkov light
with the Cherenkov radiation angle �. The Cherenkov radiation angle is determined
as:
v = �; (4.1)
os � =
1
n�
: (4.2)
From this relation, we an get the parti le velo ity v if we an measure the
Cherenkov radiation angle � and know the refra tive index n of the material. Sin e
the parti le momentum is measured by the spe trometer magnet, it is possible to
identify the parti le whi h passes the RICH if we an measure the Cherenkov angle
�. The refra tive index of the radiator material determines the Cherenkov angle
produ ed by the harged parti les. At HERMES the RICH had to be installed in
a predetermined pla e where the threshold Cherenkov Counter was mounted. The
30 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
θParticle
Cerenkov Angle
vt
ct/n
Cerenkov Photon
Figure 4.1: The diagram of Cherenkov radiation. A harged parti le with
the velo ity v whi h ex eeds the velo ity of light in the medium
an emit Cherenkov light with the Cherenkov angle �.
whole design of RICH is shown in Fig.4.2. As radiators, sili a aerogel and C
4
F
10
gas
were hosen be ause their re e tive indi es satisfy our demand.
The Cherenkov angles from aerogel and C
4
F
10
gas are shown in Fig.4.3 as a fun -
tion of parti le momentum. If the parti le momentum is above the threshold of the
Cherenkov radiation of this parti le, it emits the Cherenkov lights and makes ring
images on the photon dete tor plane whi h is then read out by arrayed photomul-
tipliers. One of the parameters whi h determines the performan e of RICH is p
max
,
whi h is the maximum separation momentum. This is the maximum momentum
for whi h the average Cherenkov angle of two parti les is separated by standard
deviations n
�
. The p
max
parameter of RICH is shown below [33℄:
p
max
=
v
u
u
t
m
2
2
�m
2
1
2k
f
n
�
; (4.3)
where m
1
; m
2
are masses of two parti les and k
f
= tan � � �
�
=
p
N is the RICH
dete tor onstant, N is the number of dete ted photons, � is the Cherenkov angle
and �
�
is the standard deviation of the Cherenkov angle distribution. The RICH at
HERMES is designed as n
�
= 4.652.
This RICH an identify parti les from two ring images made by Cherenkov pho-
tons. Besides this it an a t also as a threshold Cherenkov Counter in a parti ular
momentum region. This momentum region is indi ated in Fig.4.4. The light shaded
region in Fig.4.4 indi ates the region where the RICH a ts as a threshold Cherenkov
Counter, and the dark shaded region indi ates the region where we need to re on-
stru t the Cherenkov angle to identify parti les.
As the re onstru tion of the Cherenkov angles is most important for PID, we use
di�erent PID algorithms and ompare the re onstru ted angles. All the algorithms
need Monte Carlo simulations of RICH. Therefore a realisti MC is important. So
we have to study the properties of the RICH in detail and must in lude these into
the MC.
In the following se tion, we des ribe the properties and the MC of the RICH.
4.1. DETECTOR DESIGN 31
aluminum box
mirror array
soft steel platePMT matrix
aerogel tiles
Particle Track
Aer
ogel
Tile
Photo
n Det
ecto
r
Mirror[R=220(cm)]
6.2 (cm)
118 (cm)
86 (cm)
55 (cm)70 (cm)
Cherenkov Photon
C4F10Gas
Figure 4.2: S hemati view of Ring Imaging Cherenkov Counter at HER-
MES. Two identi al RICH is lo ated upper and lower of the beam
lime, these �gures show the upper RICH. The bottom �gure is
the side view of RICH. The aerogel tiles are assembled ba k of
the entran e window. The C
4
F
10
gas are �lled in the RICH. A
harged parti le emits Cherenkov photons by these two radia-
tors, and the photons are re e ted by the mirror to dete t the
ring images on the PMT array. The mirror are divided by eight
segments for one RICH.
32 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
0
0.05
0.1
0.15
0.2
0.25
0 2 4 6 8 10 12 14 16 18 20p (GeV)
θ (r
ad)
Aerogel (n=1.0304)
C4F10 Gas (n=1.0014)
e
e
π
π
K
K
p
p
Figure 4.3: The relation between Cherenkov angle and parti le momentum
for ele trons, pion, kaon and protons. The solid lines des ribe
the Cherenkov angle from aerogel and the dotted lines are for
that from C
4
F
10
gas. Due to using two radiators, the parti le
identi� ation in large momentum region be omes possible.
π/K aerogel
K/p aerogel
/K gasπ
K/p gas
effective threshold average angle
GeV10 155
Figure 4.4: Momentum ranges for hadron separation. Between the dashed
lines the hadrons an be separated.
4.2. STUDY OF RICH PROPERTY 33
4.2 Study of RICH Property
4.2.1 Aerogel Radiator
As mentioned before, the RICH has two radiators whi h are sili a aerogel made
from SiO
2
(Fig.4.5 and Fig.4.6) and C
4
F
10
gas. Espe ially, sili a aerogel is used as
a radiator for a RICH for the �rst time. Seeing Fig.4.2, there is a ontainer for
aerogel whi h is next to the entran e window. The entran e window is 187.7 m
wide and 46.4 m high, and the ontainer size is 200 m wide and 60 m high. As
exit window a 3.2 mm thi k UVT-lu ite is pla ed. The ontainer is made by 1mm
thi k aluminum and is �lled with dry N
2
gas. Aerogels in the shape of small tiles
are sta ked in it. The aerogel tiles are sta ked in 5 layers, 5 horizontal rows and
17 verti al olumns for the top and the bottom part of the RICH. These assemblies
of 425 aerogel tiles in total for one RICH a ts as radiator wall with a thi kness of
about 5.5 m as shown in Fig.4.9. To ut the di�used re e tion at the sides of the
aerogel tiles, bla k plasti spa ers are mounted between the tiles.
These aerogel tiles are produ ed at Matsushita Ele tri Works [34℄ [35℄. The
typi al size of aerogel tile is 11 � 11 � 1.1 m
3
, whi h is determined by the pro-
du tion pro ess. Besides, the form of one aerogel tile is not perfe t re tangular as
shown in Fig.4.8. The refra tive index of aerogel is around 1.0304, but it has some
deviations, whi h has e�e ts on the Cherenkov angle produ ed by parti les passing
through aerogel. The transmittan e of aerogel is measured already [36℄ [37℄. From
this study, the transmittan e is determined to be above 90 %. The transmittan e
T an be parameterized as a fun tion of wavelength � and aerogel thi kness t as
follow:
T = A � exp
�
�C � t
�
4
�
; (4.4)
where C and A are alled the Hunt parameters depending on the aerogel properties.
C des ribes the amount of Rayleigh s attering and A is in luded to a ount for
surfa e re e tions due to the di�eren e in the refra tive indi es between aerogel and
gas. The �tted parameters for HERMES are C � t = 0.0094 �m
4
and A = 0.964 [36℄.
The lu ite of the aerogel exit window absorbs photons whi h have wavelengths less
than 280 nm to remove Rayleigh s attered Cherenkov photons, whi h an be ome
the sour e of ba kground. As an be seen Eq.4.4, the number of Rayleigh s attered
photons in reases as the wavelength � be omes shorter. Sin e photons with short
wavelength have to be removed. This eÆ ien y is shown in Fig.4.10.
To modify the RICH Monte Carlo, we studied the details of the main aerogel
properties : the aerogel size and the refra tive indi es of the di�erent tiles, and
the des ription of the surfa e. We measured the relations between the temperature
and the index, and the dependen e of the index on the position within one aerogel
tile, be ause good knowledge on the refra tive index is most important to identify
parti les [38℄. The main results of aerogel property measurement are following.
The Distribution of the Refra tive Index
The refra tive index of aerogel is not exa tly onstant, but di�ers slightly from tile
to tile. To use aerogel as a radiator for the RICH, there are some onditions whi h
34 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
OSiSi
O
O
O
SiSi
O (SiO )2 n
~50nm1~2nm
Figure 4.5: Composition of sili a aerogel [34℄. The aerogel is made from SiO
2
hemi ally and has highly porous stru ture. The typi al parti le
size is 1 � 2 nm.
SiSi
Si
O
OO
O
OCH CHO 2 3H
(CH ) Si N Si(CH ) H
333 3
SiSi
Si
O
OO
O
OCHO 2CH3
CH3
CH3SiCH3
NH3
(SiO )2(SiO )2
Figure 4.6: Hydrophobi pro ess for sili a aerogels. Due to this new produ -
tions te hniques, lear and hydrophobi aerogels be ame avail-
able.
4.2. STUDY OF RICH PROPERTY 35
Figure 4.7: Photo of aerogel tile. Sili a aerogel produ ed with the new te h-
niques has a high transparen y and transmittan e. The typi al
size is shown in Fig.4.8.
Center
Edge
Aerogel Tile
11 (cm)
11 (cm)1 (cm)
Aerogel Tile
Center
Edge
1.5 (mm)
Figure 4.8: Typi al size of one aerogel tile. The typi al size of one aerogel
tile is 11 � 11 � 1 m. The hight of a aerogel edge is higher
than the entral part be ause of surfa e tension in the pro ess
produ ed sili a aerogels.
36 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
Aerogel Tile11cm+ 11cm+ 1cm
190cm (17 Tiles)
55cm(5 Tiles)
6.2cm(5 Tiles)
x
y
z
Figure 4.9: Aerogel wall as radiator in RICH. The assembly of aerogel tiles
as Cherenkov radiator is shown for one RICH. 850 aerogel tiles
are assembled.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 100 200 300 400 500 600 700 800λ (nm)
The
ligh
t tra
nsm
issi
on th
roug
h lu
cite
Figure 4.10: EÆ ien ies for transmittan e though lu ite. The light with the
wavelength � less than 280 nm annot be go through lu ite,
whi h is in luded in MC as a � dependen e.
4.2. STUDY OF RICH PROPERTY 37
should be satis�ed. One ondition is that the average of refra tive index n
mean
is
between 1.028 and 1.032, and another is that the values of all aerogel refra tive
indi es are in the region of n
mean
� 0.002. In addition to these, there is a ondition
for the standard deviation � of refra tive indi es whi h omes from the resolution
of the Cherenkov angles. The e�e t on the resolution an be estimated as :
��
�
� 0:5%: (4.5)
From this relation, the resolution of the refra tive indi es are al ulated as :
�(n� 1)
n� 1
= 2
��
�
� 1%: (4.6)
The distribution of refra tive index has to be less than 1 %. The average of refra tive
index is about 1.03, therefore the standard deviation � of the refra tive index has
to be less than 3.0 � 10
�4
[38℄ [39℄ [40℄.
To sele t the aerogel tiles whi h an be used for the RICH, we measured the
refra tive indi es of 1332 aerogel tiles [38℄. The result of this measurement is that
the standard deviation � is 4.1 � 10
�4
, whi h didn't satisfy the above ondition.
But this problem was solved by separating the sample in two groups, be ause there
are two independent RICH dete tor in the upper and lower half of the HERMES
spe trometer. Fig.4.11 shows the distribution of the refra tive index for 1040 sele ted
aerogel tiles. The refra tive index for the two groups is shown in Fig.4.12, where
the original distribution has been divided at the entral value of refra tive index.
In this ase the standard deviations are 2.5 � 10
�4
and 2.6 � 10
�4
, whi h satisfy
the expe ted onditions. This pro edure allowed us to use the aerogel as radiator
material of the RICH.
38 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
n-1=3.02 10
RMS=4.1 10
xx
-2
-2
-
0
100
200C
ou
nts
0.028 0.03 0.032 0.034
n-1
Figure 4.11: Distribution of the refra tive index of 1040 sele ted aerogels.
The horizontal axis shows the value of n - 1, where n is the
refra tive index of aerogels. The verti al axis is for the number
of aerogel tiles.
n-1=2.99 10
RMS 2.5 10
xx
-2
-4
-
0
100
200
Cou
nts
0.028 0.03 0.032 0.034
n-1
n-1=3.06 10
RMS 2.5 10
xx
-2
-2
-
0
100
200
Cou
nts
0.028 0.03 0.032 0.034
n-1
Figure 4.12: The distribution of refra tive index n of aerogel whi h are di-
vided into two groups. The horizontal axis shows the value of n
- 1 as Fig.4.11. The left �gure is for the aerogel tiles with 1.0290
� n � 1.0302 and the right is for that with 1.0303� n � 1.0313.
4.2. STUDY OF RICH PROPERTY 39
The Geometry
The ondition on the size and the surfa e of the aerogel tiles is that the distributions
of the longitudinal and transverse sizes are within � 0.5 mm [38℄.
We measured the size of the aerogel tiles, and the distributions for all three
dimensions for the 1040 sele ted aerogel tiles are shown below in Fig.4.13.
The Des ription of Surfa e and Edge
The aerogel surfa e has to be at to obtain omplete ring images from Cherenkov
photons, whi h means that the deviation of the surfa e from at has to be less than
1.0 mm. Be ause the aerogel tile is made from sol, the edge of the tile is higher than
the enter of tile whi h was due to the surfa e tension in the produ tion pro esses.
We measured the shape of the surfa e of aerogel. We sele ted one aerogel tile to
study the surfa e. The result of this measurement is shown in Fig.4.14 together with
the �tted fun tion. From this �gure, we an see that the distan e between the edge
and the enter of the tile is about 1.5 mm. We need to tiles whi h the deviation
of the surfa e is less than 1.0 mm, although it seems that the tile is not satisfy
this ondition. But there is not so large in uen e to dete ted rings images, be ause
almost all part of aerogel surfa e is within 1.0 mm from at and only the part whi h
is distant in 5 mm from the edge has more deviation up to 1.5 mm. Beside, the
bla k plasti spa ers whi h are mounted between the tiles ut the photons whi h
pass through the edge of tile. Therefore the aerogel tiles an be satisfy our demand
and the fun tion in Fig.4.14 is used as the typi al model of the surfa e of aerogel
tile.
40 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
0
100
200
300
Cou
nts
110 112 114 116 (mm)
X dimension
Mean 114.0 mm
RMS 0.4 mm
0
100
200
300
Cou
nts
110 112 114 116 (mm)
Y dimension
Mean 114.0 mm
RMS 0.4 mm
0
100
200
Cou
nts
(mm)
Z dimension
Mean 11.3 mm
RMS 0.4 mm
8 10 12 14
Figure 4.13: Distributions of the size of the aerogel tiles. The distributions
for three dimensions are shown, the de�nition of the dimensions
are in Fig.4.9.
4.2. STUDY OF RICH PROPERTY 41
0 10 20 30 40 500
0.5
1
1.5
2
distance from the center (mm)
thic
kn
ess (
mm
)(rela
tive)
Figure 4.14: Des ription of the shape of aerogel surfa e. This is a measured
results for one tile and this �tted fun tions are used as a typi-
al shape of aerogel surfa e in MC. The horizontal axis shows
the distan e from the enter, this �gure show a half surfa e of
aerogel.
4.2.2 Mirrors
The mirror of of the both RICH dete tor halves is divided into eight segments whi h
are arrayed to make a part of spheri al surfa e. The size of one mirror segment is
252.4 m in the horizontal and 79.4 m in the verti al. It weights less than 13 kg.
The mirror re e tivity was already measured before the RICH was installed. That
study showed that the re e tivity is above 85 % for light in the 300 � 600 nm
range of wavelength (Fig.4.15). This re e tivity was measured for only one mirror
segment and we adapted this re e tivity to other segments. By this method, no
position dependen e of the mirror re e tivity is available. Besides, the mirror has
only a simple alignment be ause of the short time s ale of the RICH installation.
But the re e tivity and alignment of the mirror are important, be ause we need the
number of �red PMTs to identify parti les, and this number depends among others
on the re e tivity. Therefore we need to know the details of the mirror alignment
and have to he k the e�e ts on the PID.
I studied the details of the mirror re e tivity, in luding the e�e ts of the mirror
alignment and put the measured data into the MC. I dis uss my analysis of the
mirror re e tivity in Se .4.3.1.
4.2.3 Photon Dete tor Plane
The photomultiplier plane has a size of 120 m � 60 m and ontains an array of
PMTs, whi h are Philips XP1911 with a diameter of 18.6 mm (0.75 in h) . There
are 1934 PMTs in ea h RICH half in 73 olumns of alternately 26 and 27 PMTs
ea h. The minimum a tive photo- athode diameter is 15 mm.
This PMT has a broad quantum eÆ ien y urve that mat hes the Cherenkov
42 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 100 200 300 400 500 600 700 800λ (nm)
Mirr
or R
efle
ctiv
ity
Figure 4.15: The wavelength dependen e of the mirror re e tivity for
Cherenkov photon. This dependen e is shown as fun tions
whi h is used in MC. Before MC modi� ation, this dependen e
is in luded for the mirror property.
light spe trum of aerogel well. It is also good for Cherenkov light from the gas
be ause the eÆ ien y extends into the UV region. The PMT elementary ell of the
array is a hexagon with a PMT is at the enter. The distan e between two adja ent
ell enters is 23.3 mm (Fig. 4.16).
The PMT photo- athodes over only 38 % of the area of the fo al plane. To
in rease the overage of PMT, a light-gathering one, made of a aluminized plasti
foil funnel was inserted onto ea h photo- athode. This PMT photo- athodes with
foil funnels then over 91 % of the surfa e. These funnels minimize the dead spa e
between the PMTs. They have a high re e tivity above � = 200 nm. The photon
olle tion eÆ ien y in reased to about 80% whi h was only 40 % before inserting
the funnels. Their e�e ts on the olle tion eÆ ien y are shown in Fig.4.17.
4.3 Monte Carlo Simulation for RICH
For parti le identi� ation, we need to re onstru t the Cherenkov angles, using the
lo ation �red PMTs, the information of the parti le tra ks and the photon hit posi-
tion on PMTs. As the PMT plane does not ompletely mat h the true mirror fo al
surfa e, the photon olle tion eÆ ien y is not 100 %. We need the omparison of
Monte Carlo with experimental results to identify the parti le type with these input
parameters. So it is very important that the MC an reprodu e the experimental
data exa tly. We an do a urate parti le identi� ation if we have a reliable MC
data set. Therefore I studied the MC simulation of the RICH and in luded ertain
additional e�e ts in the MC ode.
4.3. MONTE CARLO SIMULATION FOR RICH 43
PMT
PMT
PMT15 mm23.3 mm
photo cathode
funnel
Figure 4.16: S hemati pi ture of PMTs with funnels. The minimum a tive
photo- athode diameter is 15 mm. The funnels are atta hed to
minimize the dead spa e between the PMTs. The e�e t of these
funnels are shown in Fig.4.17.
Incident angle (rad)
Eff
icie
ncy
Direct
1-bounce
2-bounce
3-bounce
Total
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Figure 4.17: The photon olle tion eÆ ien y verses angle of in iden e.
The total eÆ ien y is marked by the solid points, the other
points show the omponents of the eÆ ien y as the number of
\boun es", whi h is re e tion from the foil surfa e. The in u-
en e of the funnels is found from the di�eren e between the total
eÆ ien y and the eÆ ien y for \dire t" dete tion.
44 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
4.3.1 Development of MC for RICH
Though the MC for the RICH had been available, it was not realisti enough be ause
the des ription of the properties of aerogel and the mirror has been mu h too simple.
For example, the refra tive index and the size of all the aerogel tiles were treated as
onstant, the surfa e of the aerogel was at and the mirror alignment was assumed to
be ideal. The Cherenkov angles of photons from aerogel and gas were not onsistent
with experimental data, the Cherenkov angle from MC being smaller than that of
the experimental data. Moreover, the number of �red PMTs from MC events was
less than that of the experimental data. The Cherenkov angle and the number of
the �red PMTs are primary values for the PID. Therefore I analyzed the di�erent
e�e ts of the RICH properties on the PID and in luded these e�e ts in the MC in
order to reprodu e the experimental data.
The properties whi h a�e t the PID are the following:
1. The aerogel properties
� The realisti size (x; y; z) and the surfa e of the aerogel tile
� The a tual refra tive index of all aerogel tiles
2. The Rayleigh s attering and the boundary orre tions
� The Rayleigh s attering whi h a�e ts the number of the �red PMTs
� The lo ation of the gap between the aerogel wall and the exit window of
the aluminum ontainer
3. The position dependen e of the mirror re e tivity
� The mirror re e tivity of ea h segment whi h ontains the orre tions
due to the mirror misalignment
In the next se tion, I des ribe the orre tion of the MC and results after the
above mentioned propertied are in luded.
The Aerogel Tile Properties
In Se .4.2.1, the properties of the aerogel tiles are des ribed. As mentioned before,
425 aerogel tiles are used for one RICH, but the size of all aerogel tiles is not exa tly
the same for all the tiles. Therefore, the measured sizes in all three dimensions
(x; y; z) of 850 aerogel tiles are in luded in the MC ode. In the same way, the
values of the refra tive indi es are in luded. As the result the Cherenkov angle
be ame dependent on the refra tive index of ea h aerogel tile. Be ause the aerogel
radiator wall is assembled by a large number of aerogel tiles, there are gaps between
the tiles. These gaps are �lled with dry N
2
gas, but there are no Cherenkov photon
produ ed when parti les go through this gaps. As they ause a de rease in the
number of �red PMTs, I in luded these boundary e�e ts in the MC ode (Fig.4.18).
After the measured sizes are in luded, a more realisti value for the surfa e of the
tiles had also to be in luded. So, the fun tion from Fig.4.14 has been implemented
in the MC ode.
4.3. MONTE CARLO SIMULATION FOR RICH 45
Figure 4.18: The RICH in the MC. The aerogel wall as radiator is assembled
by 425 tiles, whi h are shown in the right �gure. The Cherenkov
photons produ ed by aerogel wall and gas are re e ted by the
mirror.
Before in luding these e�e ts, the standard deviation of the MC Cherenkov angle
from parti les whi h passed through the aerogel was smaller than that angle in the
experimental data.
Rayleigh S attering
The Cherenkov photon from a parti le whi h goes through aerogel is s attered by
Rayleigh s attering. The s attered photons are dete ted as ba kground events, so
the number of the �red PMTs be omes smaller. The number of s attered photons
is proportional to the thi kness of the radiator. The thi kness of one tile is de�ned
about 1.1 m in MC, however detailed measurement shows that the thi kness of
the main part of aerogel tile is 0.9 m, whi h is smaller than the edge thi kness,
as explained in Fig.4.8. Therefore the e�e tive distan e for Rayleigh s attering is
about 4.5 m. In the original MC version, the e�e tive distan e was 5.5 m whi h
redu ed the number of �red PMTs too mu h. Therefore I modi�ed this value of
the thi kness of the aerogel wall in the MC ode, and ompared the number of �red
PMTs with and without orre tion.
Position Dependen e of the Mirror Re e tivity
The Cherenkov photons are re e ted by the mirror of the RICH and they enter
the PMT arrays. If the mirror re e tivity is 100 %, all the produ ed Cherenkov
photons whi h hit the mirror an rea h the PMTs. The mirror is not perfe t in
reality, and the number of �red PMTs depends on the mirror re e tivity. Therefore
this re e tivity a�e ts on the performan e of the PID. As was mentioned before,
the wavelength dependen e of the mirror re e tivity is in luded in MC ode, whi h
was measured before the RICH installation, but this re e tivity is the result of
only one mirror segment and is applied to all the segments. This was obviously
46 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
not enough to reprodu e the experimental data. I studied the details of the mirror
re e tivity of all 16 segments by omparing MC data with the experimental data.
Then I in luded the position dependen e of the mirror re e tivity. Here, I de�ned
that the position dependen e of the mirror re e tivity in ludes all mirror e�e ts in
terms of the number of �red PMTs. Espe ially it should be noti ed that the position
dependen e of mirror re e tivity is not only all mirror e�e ts but also other e�e ts
from the unknown fa tors, whi h is eÆ ien y of funnels, light orre tion eÆ ien y
and PMT eÆ ien y. Beside the e�e ts among others are the mirror alignment of the
16 mirror segments, the slope of the tra k when a photon hits the mirror, and the
real mirror re e tivities. We de�ned that the eÆ ien y from all unknown fa tors are
the position dependen e of the mirror re e tivity to normalize for the experiment
data in this analysis. These ontribution of ea h e�e t should be analyzed more
a urately in the next step.
The above e�e ts on the PID, the aerogel properties and the Rayleigh s attering,
have an impa t on the Cherenkov angle of the photon whi h is produ ed by the
aerogel. On the other hand, this position dependen e of the mirror re e tivity is
relevant to the Cherenkov photons produ ed by both aerogel and C
4
F
10
gas. So both
numbers of �red PMTs have been he ked. In this analysis, the normalized number
of �red PMTs by photons from C
4
F
10
gas (N
gas
) are used sin e the Cherenkov angle
by photons from aerogel is larger than that from gas, therefore the diameter of the
Cherenkov ring by aerogel photons is larger than the size of one mirror segment.
While the diameter of the Cherenkov ring from gas are very small at the mirror,
whi h average diameter about is 5 m, hen e the number of �red PMTs by gas
photons are more suitable than that of by aerogel photons. The method of this
analysis is des ribed below [41℄.
1. divide the mirror surfa e into 5 m � 5 m bins
2. obtain the normalized number of �red PMTs by photons from C
4
F
10
gas (N
gas
)
at the point where a tra k hits the mirror in both experimental (N
gas
(EXP))
and MC (N
gas
(MC)) data
3. take the ratio of N
gas
(EXP) to N
gas
(MC) in ea h bin
4. al ulate the mean value of the ratio for ea h mirror segment
5. the mean value of the ratio is de�ned to be the re e tivity of this mirror
segment
In this analysis, I used the number of N
gas
to obtain the mirror re e tivity
be ause the Cherenkov angle of photons oming from C
4
F
10
gas is smaller than the
Cherenkov angle of photons from aerogel (see Fig.4.3). Then almost all Cherenkov
rings of photons from the gas are within one bin. I used N
gas
for ea h bin to
obtain the position dependen e of the mirror re e tivity. In addition to this, N
gas
is normalized as follows:
N
gas
=
N
pmt
gas
L
gas
=73
; (4.7)
4.3. MONTE CARLO SIMULATION FOR RICH 47
where N
pmt
gas
is the number of �red PMTs from the gas before normalization, L
gas
is
the parti le tra k length through the gas and 73 is the mean value of L
gas
whi h
is obtained from the MC data. This is ne essary be ause the number of produ ed
Cherenkov photons is proportional to the parti le tra k length through the radia-
tor. The de�nition of L
gas
and the distribution of L
gas
are shown in Appendix.A
(Fig.A.1).
There are some onditions for data sele tion. First, I sele ted �
+
and �
�
parti les
with momentum of 4.5 � 8.5 GeV/ . The resulting Cherenkov angle from the gas is
then 0.03 � 0.07 rad. In addition, I used the parti le with N
gas
larger than 2. The
reason that �
+
and �
�
are sele ted is to avoid geometri al a eptan es e�e t due
to the parti le harge. Furthermore �
+
and �
�
an be most a urately identi�ed
among all the dete ted hadrons. The sele ted momentum range, 4.5 � 8.5 GeV/ ,
allows to obtain the purest �
+
and �
�
samples (see Fig.4.3). The photons from these
parti les are well distributed in a large region of the mirror. The sele tion uts of
the Cherenkov angle from the gas and N
gas
are made to assure that the dete ted
photons are really generated by a pion.
To obtain the mirror re e tivity in ea h bin, I needed the ratio between the
MC data, whi h have no photon absorption (all photon whi h hit the mirror are
re e ted, as the mirror re e tivity is 100 %) ex ept for the wavelength dependen e
of the re e tivity (Fig.4.15) and the experimental data. These ratios are all obtained
within the 5 � 5 m bins of the mirror. I determined the mean values of the ratios
in ea h mirror segment, whi h is the re e tivity of that segment in luding other
eÆ ien ies from unknown fa tors. Using this method, the results with the ratio of
N
gas
(EXP) to N
gas
(MC) at the mirror surfa e are shown in Appendix.A (Fig.A.2).
From these ratios, the mean value of the ratios for ea h mirror segment are evaluated,
as shown in Appendix.A (Fig.A.3).
I in luded these value for ea h mirror segment as the position dependen e of the
re e tivity into the MC ode to normalize the experimental data. The e�e ts of this
on the PID are shown in the next se tion (Se . 4.3.2).
48 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
4.3.2 PID EÆ ien y after Modi� ation of MC
In this se tion, the eÆ ien y for the PID and the omparison between this eÆ ien y
before and after in luding the RICH properties, whi h I dis ussed above. The \eÆ-
ien y" whi h is used in this se tion is de�ned below.
At �rst, the omparison of the Cherenkov angle between experimental data and
MC is shown. The aerogel tile properties whi h are in luded in MC a�e t the
Cherenkov angle. And the omparison of the number of �red PMTs, whi h is a�e ted
by the Rayleigh s attering and the Mirror re e tivities is dis ussed.
PID algorithm
The dete ted ring on the PMTs are not a omplete ir le be ause of the RICH
geometry. Therefore the re onstru ted parti le tra ks are needed to identify the
parti le type. At HERMES, a method whi h is alled \Inverse Ray Tra ing" (IRT)
is used for the re onstru tion of the Cherenkov angle from the tra k parameters and
the position of the PMTs. A detailed dis ussion of IRT is beyond the subje t of this
thesis. It is found in [42℄ [43℄ [44℄.
The MC data is used for the estimation of the eÆ ien y of the IRT method. To
do this estimation, it is ne essary to ompare the Cherenkov angle and the number
of �red PMTs between MC and experimental data. I'll show this omparison in the
next se tion, and it be omes obvious that in luding the RICH properties is ne essary
to reprodu e the experimental data.
Cherenkov Angle Resolution
The aerogel properties des ribed in Se .4.3.1 have been in luded step by step in
several MC produ tion to he k their e�e t on the Cherenkov angle. The MC
produ tions whi h I made are listed below.
1. Simplest simulation
This produ tion is the base for the others. In this MC, the aerogel radiator is
onsidered as one wall, whi h is not divided into tiles.
2. Rayleigh s attering
The Rayleigh s attering e�e t is added to the above.
3. Aerogel tile e�e ts
The aerogel tiles are in luded as 425 separate tiles, ea h di�erent size and
refra tive index. The surfa e of the aerogel tiles is also simulated.
4. Boundary orre tion of photon refra tion
The e�e t of the gap between aerogel tiles and the exit window is onsidered.
5. Mirror re e tivity
The position dependen e of the mirror re e tivity whi h ontain all mirror
alignments e�e ts is in luded.
The Cherenkov angles from aerogel and gas are shown Tab.4.1 for all above
MC produ tions and the experimental data. In Tab.4.1, �
�
aero
and �
�
gas
is the
4.3. MONTE CARLO SIMULATION FOR RICH 49
type �
aero
� �
�
aero
[mrad℄ �
gas
� �
�
gas
[mrad℄
EXP 244 � 9.3 49.0 � 9.2
MC 1 235 � 6.3 54.0 � 7.6
MC 2 233 � 6.0 54.2 � 7.5
MC 3 234 � 6.5 54.0 � 7.6
MC 4 241 � 7.1 54.2 � 7.7
MC 5 241 � 7.1 54.0 � 7.5
Table 4.1: Comparison of the Cherenkov angle for ea h MC produ tion and
the experimental data. The deviation are found between the type
1 and type 3 of MC data.
standard deviation of the Cherenkov angle of photons from aerogel (gas). Be ause
the properties of the RICH are mainly dominated by the aerogel, the Cherenkov
angle from the gas was not a�e ted by these MC modi� ations. The Cherenkov
angle from aerogel be ame lose to the experimental value after in luding all e�e ts.
Fig.4.19 shows the omparison between the experimental data and the MC data.
The dark shaded histogram are for the MC data, and the light shaded histogram is
the experimental data. The MC data of the top �gure of Fig.4.19 is the produ tion
of type 1, the se ond from the top shows type 2, the third shows type 3, the forth
shows type 4 and the bottom shows type 5. We an see that the e�e t from the
boundary orre tion (type 4) has large in uen e to Cherenkov angles. The MC
produ tion an reprodu e the experimental data for the Cherenkov angle of the
photon from aerogel and gas rather well though some dis repan y remains. We an
see better agreement of the Cherenkov angle in the re ent results, whi h is shown
in Appendix.A (Fig.A.4).
50 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
0
100
200
300
400
500
600
0.2 0.225 0.25 0.275 0.3
θaero (rad)
coun
ts
0
100
200
300
400
500
0.02 0.04 0.06 0.08
θgas (rad)
coun
ts
0
100
200
300
400
500
600
0.2 0.225 0.25 0.275 0.3
θaero (rad)
coun
ts
0
100
200
300
400
500
0.02 0.04 0.06 0.08
θgas (rad)
coun
ts
0
100
200
300
400
500
600
0.2 0.225 0.25 0.275 0.3
θaero (rad)
coun
ts
0
100
200
300
400
500
0.02 0.04 0.06 0.08
θgas (rad)
coun
ts
0
100
200
300
400
500
600
0.2 0.225 0.25 0.275 0.3
θaero (rad)
coun
ts
0
100
200
300
400
500
0.02 0.04 0.06 0.08
θgas (rad)
coun
ts
0
100
200
300
400
500
600
0.2 0.225 0.25 0.275 0.3
θaero (rad)
coun
ts
0
100
200
300
400
500
0.02 0.04 0.06 0.08
θgas (rad)
coun
ts
Figure 4.19: Comparison of the Cherenkov angles between the experimental
data (light shaded histograms) and the MC data (dark shaded
histograms). The left �gure shows the Cherenkov angle from
aerogel and the right �gure shows that from the gas.
4.3. MONTE CARLO SIMULATION FOR RICH 51
Number of Fired PMTs
As same as the Cherenkov angle, the number of �red PMTs is an essential value for
the PID. I he ked the number of �red PMTs by photons from aerogel (N
pmt
aero
) and
from gas (N
pmt
gas
) with MC and experimental data.
The used MC data are the same as in the previous se tion. The ontribution of
ea h e�e t to the number of �red PMTs is shown in Tab.4.2.
type N
pmt
aero
N
pmt
gas
EXP 11.3 11.4
MC 1 11.2 15.6
MC 2 7.0 13.2
MC 3 13.8 15.1
MC 4 16.0 15.2
MC 5 11.2 13.1
Table 4.2: Comparison of the number of �red PMTs for ea h MC produ tion
and experimental data.
Comparing the experimental data and the MC data of type 5, the MC data after
in luding all e�e ts are lose to the experimental data, espe ially the e�e t of the
mirror re e tivity (see the di�eren e between type 4 and type 5) has largest in uen e
on the number of PMTs. Fig.4.20 shows the omparison of the experimental data
and the MC data for the number of PMTs. Again, the dark shaded histograms are
the MC data. The order of �gures are same as Fig.4.19.
52 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
0
100
200
300
400
500
600
0 5 10 15 20
Naero
coun
ts
0
100
200
300
400
500
600
700
800
0 5 10 15 20
Ngas
coun
ts
0
100
200
300
400
500
600
0 5 10 15 20
Naero
coun
ts
0
100
200
300
400
500
600
700
800
0 5 10 15 20
Ngas
coun
ts
0
100
200
300
400
500
600
0 5 10 15 20
Naero
coun
ts
0
100
200
300
400
500
600
700
800
0 5 10 15 20
Ngas
coun
ts
0
100
200
300
400
500
600
0 5 10 15 20
Naero
coun
ts
0
100
200
300
400
500
600
700
800
0 5 10 15 20
Ngas
coun
ts
0
100
200
300
400
500
600
0 5 10 15 20
Naero
coun
ts
0
100
200
300
400
500
600
700
800
0 5 10 15 20
Ngas
coun
ts
Figure 4.20: Comparison the number of �red PMTs between the experi-
mental data (light shaded histograms) and the MC data (dark
shaded histograms). The left �gure shows N
pmt
aero
and the right
�gure shows N
pmt
gas
4.3. MONTE CARLO SIMULATION FOR RICH 53
The mirror re e tivity a�e ts not only N
pmt
aero
but also N
pmt
gas
. In this respe t the
mirror re e tivity is di�erent from other e�e ts. I studied the ontribution of the
mirror re e tivity to N
pmt
aero
and N
pmt
gas
in more detail. I de�ne the \EÆ ien y", as the
ratio between the number of �red PMTs in the experiment and in the MC. Therefore,
the eÆ ien y of aerogel is de�ned as N
pmt
aero
(EXP)/N
pmt
aero
(MC) and the eÆ ien y of
gas is de�ned as N
pmt
gas
(EXP)/N
pmt
gas
(MC). These eÆ ien ies must be almost 1 if the
MC reprodu es the experimental data a urately. This value is one riterion for
the quality of the MC. Additionally, I he ked the �
x
and �
y
dependen e of the
eÆ ien ies, where �
x
and �
y
are the slopes of the tra ks at the target.
Fig.4.21 shows the eÆ ien y versus the slopes for aerogel and gas before the
MC modi� ation, where the simple MC produ tion is used for the omposition with
experimental data. In Fig.4.21, the regions for �
x
and �
y
are -0.14 < �
x
< 0.14 and
0.04 < �
y
< 0.16. It an be seen from Fig.4.21 that the ratio of the number of
�red PMTs is far from 1, about 0.4 � 0.7, and there are large u tuations in the
eÆ ien ies depending on �
x
and �
y
. This shows that some e�e ts whi h are on both
N
pmt
aero
and N
pmt
gas
are missing in MC. Therefore I ompared the experimental data and
the MC produ tion of type 5. Fig.4.22 shows this results.
Fig.4.22 shows that the mirror re e tivity has a strong e�e t to the number of
�red PMTs. The u tuations observed in the eÆ ien ies before the MC modi� ation
be ame small, and the both eÆ ien ies in reased by the MC modi� ation although
their value is only 0.7 � 0.9. Even after in luding the mirror re e tivity, there are
some di�eren e between N
pmt
aero(gas)
(EXP) and N
pmt
aero(gas)
(MC).
-0.15-0.1-0.05 0 0.05 0.1 0.15
0.040.06
0.080.1
0.120.14
0.160
0.2
0.4
0.6
0.8
1
θx (rad)
θy (rad)
Eff
icie
ncy
(ae
rog
el)
-0.15-0.1-0.05 0 0.05 0.1 0.15
0.040.06
0.080.1
0.120.14
0.160
0.10.20.30.40.50.60.70.8
θx (rad)
θy (rad)
Eff
icie
ncy
(g
as)
Figure 4.21: The �
x
and �
y
dependen e of the eÆ ien y for aerogel and gas
before MC modi� ation. The left �gure shows eÆ ien ies for
aerogel and the right �gure shows that for gas.
54 CHAPTER 4. RING IMAGING CHERENKOV COUNTER
-0.15-0.1-0.05 0 0.05 0.1 0.15
0.040.06
0.080.1
0.120.14
0.160
0.10.20.30.40.50.60.70.80.9
θx (rad)
θy (rad)
Eff
icie
ncy
(ae
rog
el)
-0.15-0.1-0.05 0 0.05 0.1 0.15
0.040.06
0.080.1
0.120.14
0.160
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
θx (rad)
θy (rad)
Eff
icie
ncy
(g
as)
Figure 4.22: The �
x
and �
y
dependen e of the eÆ ien y for aerogel and gas
after MC modi� ation. The left �gure shows eÆ ien ies for
aerogel and the right �gure shows that for gas.
I onsidered that the ause of this is the limited a eptan e of the PM tubes.
I tried to use the Cherenkov photons from the gas to study the mirror re e tivity,
but it is impossible to get the real number of dete ted photons from N
pmt
gas
be ause
the Cherenkov ring made by the gas photons is very small (the average diameter is
about 5 m at the mirror). As it is possible that some Cherenkov gas photons hit
the same PMT, the number of �red PMTs is not dire tly proportional to the mirror
re e tivities (see Fig.4.23).
To prove this spe ulation, I studied the relation between the number of �red
PMTs and the mirror re e tivities with MC. I produ ed the MC data hanging
the mirror re e tivities of all mirror segments from 0 to 1, and made plots of the
mean value of the number of �red PMTs versus the mirror re e tivity. In these
produ tions, I used �
+
and �
�
parti les. The results of this study are shown below
in Fig.4.24. Fig.4.24 shows that N
pmt
aero
is proportional to the mirror re e tivity, but
N
pmt
gas
isn't. From these results, 2 or more photons have to hit the same PMT if the
number of photons from gas is more than 12.
From the above detailed study on the number of �red PMTs, it is ne essary that
this non-linearity has to be onsidered to obtain the a urate mirror re e tivities.
However, the a urate number of dete ted photons, whi h is not the number of �red
PMTs, annot be obtain from experimental data. Therefore this non-linearity has
to be onsidered in future improvements of the MC. In spite of the last mentioned
restri tion, the MC have be ame lose to the experimental one by in luding the
RICH properties. This is bene� ial for the analysis of the RICH data, espe ially for
the extra tion of the spin asymmetries whi h is dis ussed in the following hapter.
4.3. MONTE CARLO SIMULATION FOR RICH 55
Figure 4.23: The Cherenkov rings of aerogel and gas on the PMT plane.
The smaller ring is made by photon from gas. Some Cherenkov
photons from gas an hit the same PMT. Therefore the mirror
re e tivity is not proportional to the number of �red PMTs.
0
1
2
3
4
5
6
7
8
9
10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1.94+8.07*R
(R=Reflectivity)
Mirror Reflectivity
Nae
ro
0
2
4
6
8
10
12
14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2.86+13.5*R
7.18+6.41*R
(R=Reflectivity)
Mirror Reflectivity
Nga
s
Figure 4.24: The mirror re e tivity versus the number of �red PMTs by pho-
ton from aerogel and gas with MC. The left �gure shows that
the re e tivity vs mean value of N
pmt
aero
and the right �gure shows
that the re e tivity vs mean value of N
pmt
gas
.
57
Chapter 5
Extra tion of Polarized Quark
Distributions
This hapter des ribes the formalism for the extra tion of polarized quark distribu-
tions, espe ially the avor asymmetry of the light sea quark : ��u � �
�
d. In this
analysis, spin asymmetries whi h are obtained from experiment and the parameter-
ization of unpolarized quark distributions (Parton Distribution Fun tion) whi h are
used for the purity generation are needed. All available in lusive and semi-in lusive
asymmetries measured by HERMES are used in this study and various types of
quark distributions are extra ted.
5.1 The Purity Method
5.1.1 De�nition of Purity
For the extra tion of quark polarizations, we need semi-in lusive spin asymmetries
A
h
1
(x) obtained from the experiment. In the limit of largeQ
2
, A
h
1
(x) an be expressed
from Eq.2.86 as follows:
A
h
1
(x;Q
2
) =
1 +R(x;Q
2
)
1 +
2
�
P
f
e
2
f
�q
f
(x;Q
2
)
R
dzD
h
f
(Q
2
; z)
P
f
e
2
f
q
f
(x;Q
2
)
R
dzD
h
f
(Q
2
; z)
; (5.1)
where the spin asymmetry is related to the polarized and the unpolarized PDF.
As several in lusive and semi-in lusive spin asymmetries are now available from
HERMES, we an extra t polarized quark distributions with help of the unpolarized
PDF and fragmentation fun tions.
Spin asymmetries from experiment are obtained for ea h x
i
bin and are on-
sidered to be independent of Q
2
. F
1
and g
1
are dependent on Q
2
. However, for
the measured spin asymmetries whi h is approximate to the ratio of g
1
and F
1
, no
signi� ant Q
2
dependen e has been observed [45℄ [46℄. Therefore the relation whi h
is used in pra ti al analysis reads:
A
h
1
(x
i
) = C
i
�
X
f
P
h
f
(x
i
)
�q
f
(x
i
)
q
f
(x
i
)
; (5.2)
58 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
where C
i
is de�ned as C
i
= (1+R(x
i
))=(1+
2
i
) and the kinemati variablesR(x
i
; Q
2
)
and
i
denote the mean value in the x
i
bin. P
h
f
(x
i
) is de�ned as:
P
h
f
(x
i
) =
e
2
f
q
f
(x
i
)
~
D
h
f
(x
i
; z)
P
f
e
2
f
q
f
(x
i
)
~
D
h
f
(x
i
; z)
; (5.3)
where
~
D
h
f
(x
i
; z) is the e�e tive fragmentation fun tion whi h in ludes the dete tor
a eptan e. They are the probability that one hadron h with energy fra tion z is
dete ted when the virtual photon is s attered by a quark with avor f . The values
of z
i
are determined for ea h x
i
bin. Therefore the purities depend on a single
parameter x
i
in this de�nition. This is alled the \quark avor purity" introdu ed
in [47℄. Introdu ing this purities P
h
f
(x
i
), we an express the spin asymmetry A
h
1
(x
i
)
as a linear ombination of quark polarizations. They represent the probabilities that
a hadron h is produ ed from a quark with avor f . A ording to the de�nition of
purity, the relation holds for every x
i
bin:
X
f
P
h
f
(x
i
) = 1: (5.4)
A tually, we use the equation Eq.5.2 to extra t the quark polarizations. If we
measure spin asymmetries for several types of hadrons, the quark polarizations an
be al ulated by solving the matrix equation:
A = C
i
� P Q; (5.5)
where A is the ve tor of measured spin asymmetries, Q is the ve tor of ea h quark
polarization and P is the purity matrix whi h ontains the purities of ea h quark
avor as matrix elements. They are de�ned as:
A =
0
B
�
A
h
1
1
(x
i
)
.
.
.
A
h
m
1
(x
i
)
1
C
A
; Q =
0
B
�
�q
f
1
=q
f
1
(x
i
)
.
.
.
�q
f
n
=q
f
n
(x
i
)
1
C
A
; (5.6)
P =
0
B
B
�
P
h
1
f
1
(x
i
) : : : P
h
1
f
m
(x
i
)
.
.
.
.
.
.
.
.
.
P
h
m
f
1
(x
i
) : : : P
h
m
f
n
(x
i
)
1
C
C
A
: (5.7)
Here f
i
is the quark avor : f
i
= fu; �u; d;
�
d; s; �sg , and h
i
is the hadron type:
h
n
= f�
+
; �
�
K
+
; K
�
; p; �p:::g. In present measurements of HERMES pions, Kaon
and protons an be dete ted, and we onsider these three types of hadrons in the
following. In the ideal ase, Eq.5.5 an be solved when the number of hadron types
(the number of asymmetries) m is equal to the number n of avor types. But, as we
will see later, we have to use more asymmetries than avor types be ause there are
orrelations between some asymmetries. These orrelations mean that the number
of independent asymmetries are smaller than the number of available asymmetries
m. In this analysis all available asymmetries for the extra tion of quark distributions
are used. A detailed dis ussion of the asymmetries will follow in Se .5.5.
5.1. THE PURITY METHOD 59
The formalism of quark polarization extra tion with purities an be extended to
the in lusive measurement, where we dete t the s attered lepton only. In this ase
the purity for the s attered lepton is des ribed as:
P
e
f
(x
i
) =
e
2
f
q
f
(x
i
)
P
f
e
2
f
q
f
(x
i
)
; (5.8)
be ause the fragmentation fun tion of the in lusive ele tron D
e
f
(x
i
) is equal to 1 in
every x
i
bin :
R
dzD
e
f
(x
i
; z) = 1. Therefore we an in lude the in lusive asymmetry
in this analysis.
To extra t the quark distributions we have to solve above matrix equation Eq.5.5.
With a ve tor of measured asymmetries A, �
2
an be de�ned as follows:
�
2
= (A� C
i
� P Q)
T
V
�1
A
(A� C
i
� P Q): (5.9)
�
2
is minimized if Q be omes a solution of the matrix equation Eq.5.5. In Eq.5.9,
V
A
is a ovarian e matrix whi h is introdu ed to onsider the orrelations between
the measured asymmetries in A, and is des ribed in detail in Appendix B.1 . V
A
an be expressed by orrelation oeÆ ients between asymmetries as follows:
(V
A
)
ij
= �(A
h
i
1
; A
h
j
1
) ÆA
h
i
1
ÆA
h
j
1
; (5.10)
where �(A
h
i
1
; A
h
j
1
) is a orrelation oeÆ ient between asymmetry of hadron h
i
(A
h
i
1
)
and hadron h
j
(A
h
j
1
). ÆA
h
i
1
and ÆA
h
j
1
denote the un ertainties of the asymmetries of
types i and j.
Under some assumptions, �(A
h
i
1
; A
h
j
1
) an be repla ed by the respe tive orrela-
tion oeÆ ient between the number of hadrons, whi h is des ribed in Appendix B.1.
A ording to this relation, the ovarian e matrix reads:
(V
A
)
ij
= �(N
h
i
; N
h
j
) ÆA
h
i
1
ÆA
h
j
1
; (5.11)
�(N
h
i
; N
h
j
) =
hn
h
i
n
h
j
i
q
h(n
h
i
)
2
ih(n
h
j
)
2
i
; (5.12)
where hn
h
i
i is an averaged parti le multipli ity for hadron h
i
, whi h is the number
of hadrons h
i
in one DIS event.
The statisti al errors on the extra ted quark distributions Q is expressed by
ovarian e matrix of Q. This is shown below:
V
Q
= C
�1
i
(P
T
V
�1
A
P)
�1
: (5.13)
With this ovarian e matrix, the statisti al errors on quark distributions are
found to be :
Æ
�q
f
i
q
f
i
!
=
q
(V
Q
)
ii
: (5.14)
The quark distributions are extra ted with this method whi h I have mentioned
above. To do this analysis, the Singular Value De omposition (SVD) method [48℄
60 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
is used, whi h is the method for solving most linear least-squares problems and
obtaining singular values. In this analysis the number of asymmetries is not stable
be ause I tried to use several sets of input asymmetries for extra tion of quark
distributions. In the ase where the number of input asymmetries is larger than
that of the extra ted quark distributions, SVD an solve the matrix equation Eq.5.5.
Therefore this is a powerful tool for solving this matrix. The detail of the SVD
method is des ribed in Appendix.B.2.
An a urate determination of the systemati errors has been omitted in this
study, but it will be mentioned only summarily later in Se .6.4.
5.2 Evaluation of Purities
To solve the matrix equation Eq.5.5, I need measured spin asymmetries and purity
matrix. In this analysis, asymmetries whi h have seen obtained by other studies
[49℄ [50℄ are used as input, and the details of all used asymmetries are des ribed in
Se .5.5. The purities onsist of unpolarized quark distributions and fragmentation
fun tions whi h in lude of the a eptan e the HERMES dete tors. They are pro-
du ed by MC simulation of the unpolarized DIS pro ess, whi h is made by three
steps. The �rst two steps generate the a tual DIS events and the unpolarized PDF
q(x). The fragmentation fun tions D
h
q
(x; z) whi h are independent of the a eptan e
of HERMES are ne essary as input for MC. The third step gives e�e ts from the
dete tor a eptan e. In the following, the detail of ea h step is des ribed.
In a �rst step, the PEPSI [51℄ generator, whi h is based on LEPTO [52℄, produ ed
DIS events with unpolarized PDF, based on the CTEQ4LQ [53℄ and the GRV95 [54℄
parameterizations. In the main analysis, CTEQ4LQ parameterizations are used.
The data whi h have been produ ed with GRV95 are used to estimate the systemati
errors from the purities. LEPTO sele ts the DIS kinemati s plane a ording to the
lepton-nu leon s attering ross se tion and de ides whi h quark is s attered by using
the unpolarized PDF.
In the se ond step, the JETSET generator [24℄, whi h is part of LEPTO [52℄,
pro esses hadronization and gives the �nal state of hadrons in DIS. JETSET ontains
di�erent kind of hadronization programs whi h are based on the Independent model
(Se .2.5.2), the LUND String model (Se .2.5.2) and the Cluster model (Se .2.5.2)
and there are default model parameters for ea h model. As I mentioned before in
Se .2.5.2, the LUND String model is the best to simulate hadronization nowadays,
so it has been used to produ e the �nal state hadrons in MC and the parameters for
LUND model have been tuned for HERMES (Fig.2.8 and Tab.2.3) [25℄ to reprodu e
the experimental hadron spe tra. The produ tion with other hadronization models
an be used to estimate systemati errors from fragmentation fun tions, but this
hasn't been done in this study.
In the last step, the simulation of e�e ts from the HERMES dete tors is per-
formed. A detailed model of the HERMES dete tors is available in the dete tor
simulation program, whi h is named HMC [55℄ and is based on GEANT [56℄. I
modi�ed some parts of this HMC for the RICH as mentioned in the previous se -
tions, but to simulate the response of all dete tors is time onsuming, so it takes
long to produ e DIS events. Therefore I used a simple box a eptan e model for
5.2. EVALUATION OF PURITIES 61
simulation of the dete tors, whi h uses a momentum look-up table. It in ludes the
momentum ut-o�s, an generate a urate geometry of parti le tra ks behind the
magnet, and the same uts of event sele tion whi h is used in analysis of experimen-
tal data are applied.
From this MC data, the purities are obtained from the equation below:
P
h
f
(x
i
) =
N
h
f
(x
i
)
P
f
N
h
f
(x
i
)
; (5.15)
where N
h
f
(x) is the number of hadron h whi h is produ ed when the quark in the
nu leon with avor f is s attered. The z dependen e of N
h
f
(x) is ignored by sum-
ming up yields for this variable. This relation in ludes the fragmentation fun tions
whi h ontain also the dete tor a eptan e of HERMES. This e�e tive fragmenta-
tion fun tions are:
~
D
h
f
(x
i
; z) =
N
h
f
(x
i
; z)
re
P
h
N
h
f
(x
i
; z)
gen
; (5.16)
where
P
h
N
h
f
(x; z)
gen
is the number of events whi h a quark with avor f is generated
and N
h
f
(x; z)
re
is the number of re onstru ted events with quark avor f .
A ording to this relation, I generated purities P
h
f
for a proton and a neutron
target, where the type of hadrons h is h = fh
+
; h
�
; �
+
; �
�
K
+
; K
�
g and the type
of quark avors f = fu; d; �u;
�
d; s; �sg. The hadron types of h
+
and h
�
in lude all
produ ed hadrons as f�
+
; �
�
; K
+
; K
�
; p; �p; ::::g. Similarly the in lusive purities P
e
f
are generated. The reason that I need purities for two di�erent targets proton and
neutron, is to use all available asymmetries on proton,
3
He and deuterium target.
The details of this topi are des ribed in Se .5.4. As I have mentioned before,
the unpolarized parameterization whi h is used for purity generation is CTEQ4LQ.
Fig.5.1-Fig.5.7 (Fig.5.8-Fig.5.14) show the purities for proton (neutron) target gen-
erated with CTEQ4LQ, ompared to the purities generated with GRV95.
These �gures show the probability that one hadron h is produ ed tagging a
quark with avor f in ea h x bin. For all hadrons, the high probabilities of hadron
produ tion from the valen e quarks (u; d) are shown, in parti ular a high probability
for u quarks. Beside we an see whi h hadron is sensitive to the sea quarks : �u;
�
d; s; �s.
For the ase of the proton target, �
+
is sensitive to
�
d be ause �
+
onsists of u and
�
d (Fig.5.4). Similarly, �
�
is sensitive to �u (Fig.5.5). Fig.5.6 and Fig.5.7 show that
kaons are important to study of strange quarks. K
+
, whi h ontains u and �s quarks,
has a high probability that they are produ ed from �s quarks (Fig.5.6). Espe ially,
K
�
is so sensitive to �u and s quarks than other hadrons (Fig.5.7) be ause it is made
of two types sea quarks �u and s.
For the ase of the neutron target, the purities have a high value for d quarks
than that of the proton target. Fig.5.11 - Fig.5.14 show a high probability of
�
d
quarks. As the proton target, we an see that K
�
is so sensitive to s quarks from
Fig.5.14.
Comparing the purities generated with CTEQ4LQ and GRV, there are almost
no di�eren e for the u; d; �u and
�
d quarks in all the �gures. On the other hand, the
signi� ant deviations are shown for the purities of s and �s quarks. These deviations
ome from the di�eren e of s and �s quark distributions between CTEQ4LQ and
62 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
GRV. The CTEQ4LQ parameterization is a set with a low starting value of Q
2
and
it has more information of the sea quarks than GRV. The deviations of the purities
of sea quarks show these informations. The extra tion of sea quark distributions is
one of the main purposes, hen e the purities generated with CTEQ4LQ are used in
this study.
10-3
10-2
10-1
1
GRVPu
e+ CTEQ Pde+ Ps
e+
10-2
10-1
10-3
10-2
10-1
1
Pu–e+
10-2
10-1
Pd–e+
10-2
10-1 x
Ps–e+
Figure 5.1: In lusive (e
+
) purities for proton target
5.2. EVALUATION OF PURITIES 63
10-3
10-2
10-1
1
GRVPu
h+ CTEQ Pdh+ Ps
h+
10-2
10-1
10-3
10-2
10-1
1
Pu–h+
10-2
10-1
Pd–h+
10-2
10-1 x
Ps–h+
Figure 5.2: h
+
purities for proton target
10-3
10-2
10-1
1
GRVPu
h- CTEQ Pdh- Ps
h-
10-2
10-1
10-3
10-2
10-1
1
Pu–h-
10-2
10-1
Pd–h-
10-2
10-1 x
Ps–h-
Figure 5.3: h
�
purities for proton target
64 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
10-3
10-2
10-1
1
GRVPu
π+ CTEQ Pdπ+ Ps
π+
10-2
10-1
10-3
10-2
10-1
1
Pu–π+
10-2
10-1
Pd–π+
10-2
10-1 x
Ps–π+
Figure 5.4: �
+
purities for proton target
10-3
10-2
10-1
1
GRVPu
π- CTEQ Pdπ- Ps
π-
10-2
10-1
10-3
10-2
10-1
1
Pu–π-
10-2
10-1
Pd–π-
10-2
10-1 x
Ps–π-
Figure 5.5: �
�
purities for proton target
5.2. EVALUATION OF PURITIES 65
10-3
10-2
10-1
1
GRVPu
K+ CTEQ PdK+ Ps
K+
10-2
10-1
10-3
10-2
10-1
1
Pu–K+
10-2
10-1
Pd–K+
10-2
10-1 x
Ps–K+
Figure 5.6: K
+
purities for proton target
10-3
10-2
10-1
1
GRVPu
K- CTEQ PdK- Ps
K-
10-2
10-1
10-3
10-2
10-1
1
Pu–K-
10-2
10-1
Pd–K-
10-2
10-1 x
Ps–K-
Figure 5.7: K
�
purities for proton target
66 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
10-3
10-2
10-1
1
GRVPu
e+ CTEQ Pde+ Ps
e+
10-2
10-1
10-3
10-2
10-1
1
Pu–e+
10-2
10-1
Pd–e+
10-2
10-1 x
Ps–e+
Figure 5.8: In lusive (e
+
) purities for neutron target
5.2. EVALUATION OF PURITIES 67
10-3
10-2
10-1
1
GRVPu
h+ CTEQ Pdh+ Ps
h+
10-2
10-1
10-3
10-2
10-1
1
Pu–h+
10-2
10-1
Pd–h+
10-2
10-1 x
Ps–h+
Figure 5.9: h
+
purities for neutron target
10-3
10-2
10-1
1
GRVPu
h- CTEQ Pdh- Ps
h-
10-2
10-1
10-3
10-2
10-1
1
Pu–h-
10-2
10-1
Pd–h-
10-2
10-1 x
Ps–h-
Figure 5.10: h
�
purities for neutron target
68 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
10-3
10-2
10-1
1
GRVPu
π+ CTEQ Pdπ+ Ps
π+
10-2
10-1
10-3
10-2
10-1
1
Pu–π+
10-2
10-1
Pd–π+
10-2
10-1 x
Ps–π+
Figure 5.11: �
+
purities for neutron target
10-3
10-2
10-1
1
GRVPu
π- CTEQ Pdπ- Ps
π-
10-2
10-1
10-3
10-2
10-1
1
Pu–π-
10-2
10-1
Pd–π-
10-2
10-1 x
Ps–π-
Figure 5.12: �
�
purities for neutron target
5.2. EVALUATION OF PURITIES 69
10-3
10-2
10-1
1
GRVPu
K+ CTEQ PdK+ Ps
K+
10-2
10-1
10-3
10-2
10-1
1
Pu–K+
10-2
10-1
Pd–K+
10-2
10-1 x
Ps–K+
Figure 5.13: K
+
purities for neutron target
10-3
10-2
10-1
1
GRVPu
K- CTEQ PdK- Ps
K-
10-2
10-1
10-3
10-2
10-1
1
Pu–K-
10-2
10-1
Pd–K-
10-2
10-1 x
Ps–K-
Figure 5.14: K
�
purities for neutron target
70 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
5.3 Assumptions on Sea Quark Polarizations
As I mentioned before, I need at least six asymmetries whi h must be independent of
ea h other in order to extra t quark distributions of all avors : f = fu; d; �u;
�
d; s; �sg.
At HERMES the asymmetries from 1995 � 1998 data taking are available, but
the number of independent asymmetries is not suÆ ient to extra t all the types of
quark polarizations a urately. Therefore two kinds of assumptions on the sea quark
polarizations or distributions are used to de rease the number of extra ted quark
types and avoid this problem. The �rst type assumes to be the same polarizations
for all sea quark avors, whi h has already been used in other thesis [57℄ [49℄ [58℄
[59℄. The se ond type, whi h is introdu ed �rst time in this analysis, is based on
the Chiral Quark Soliton Model [6℄. The details are shown in Se 5.3.2.
In the following, I will des ribe these two kinds of assumptions and types of
quark polarizations or distributions whi h are extra ted in ea h ase.
5.3.1 Assumption of Polarization Symmetri Sea
At �rst, the term of \quark polarization" and \quark distribution" have to be de�ned
learly. The quark polarization is de�ned that the ratio between the unpolarized
quark distributions and the polarized quark distributions as �q
f
=q
f
(x). The quark
distribution expresses the x distribution of the polarized quarks as �q
f
(x).
The �rst assumption is that all sea quark polarizations are the same as follows:
�q
s
(x)
q
s
(x)
=
��u(x)
�u(x)
=
�
�
d(x)
�
d(x)
=
�s(x)
s(x)
=
��s(x)
�s(x)
: (5.17)
With this assumption, we an extra t two types of quark polarizations. One is the
extra tion of the valen e quark polarizations, named \valen e de omposition" and
another is the extra tion of ea h quark avor polarizations, named \ avor de om-
position". They are shown next.
Valen e De omposition
The spin asymmetry of hadron h an be de omposed into ontributions from valen e
quarks with Eq.5.17:
A
h
1
(x) = P
h
1
(x) �
�u
v
(x)
u
v
(x)
+ P
h
2
(x) �
�d
v
(x)
d
v
(x)
+ P
h
3
(x) �
�q
s
(x)
q
s
(x)
; (5.18)
where the oeÆ ient C
i
= (1+R)=(1+
2
i
) whi h is multiplied to the purity matrix
(Eq.5.5) is omitted for brevity. The purities in Eq.5.18 are de�ned as:
P
h
1
(x) = P
h
u
(x) �
u(x)� �u(x)
u(x)
; (5.19)
P
h
2
(x) = P
h
d
(x) �
d(x)�
�
d(x)
d(x)
; (5.20)
P
h
3
(x) = P
h
u
(x) �
�u(x)
u(x)
+ P
h
�u
(x) + P
h
d
(x) �
�
d(x)
d(x)
+ P
h
�
d
(x)
+ P
h
s
(x) + P
h
�s
(x): (5.21)
5.3. ASSUMPTIONS ON SEA QUARK POLARIZATIONS 71
where P
h
q
(x) (q = u; �u; d;
�
d; s; �s) are the original purities de�ned in Eq.5.3.
The extra ted quark polarization ve tor Q is as follows:
Q =
�u
v
(x)
u
v
(x)
;
�d
v
(x)
d
v
(x)
;
�q
s
(x)
q
s
(x)
!
: (5.22)
Therefore with these de�nition of purities of Eq.5.19-Eq.5.21, I obtained the
valen e quark polarizations : �u
v
=u
v
and �d
v
=d
v
.
Flavor De omposition
Similarly, the spin asymmetry an be de omposed into ontributions from di�erent
quark avors with Eq.5.17:
A
h
1
(x) = P
h
1
(x) �
�u(x) + ��u(x)
u(x) + �u(x)
+ P
h
2
(x) �
�d(x) + �
�
d(x)
d(x) +
�
d(x)
(5.23)
+ P
h
3
(x) �
�q
s
(x)
q
s
(x)
; (5.24)
where the purities are de�ned as:
P
h
1
(x) = P
h
u
(x) �
u(x) + �u(x)
u(x)
; (5.25)
P
h
2
(x) = P
h
d
(x) �
d(x) +
�
d(x)
d(x)
; (5.26)
P
h
3
(x) = �P
h
u
(x) �
�u(x)
u(x)
+ P
h
�u
(x)� P
h
d
(x) �
�
d(x)
d(x)
+ P
h
�
d
(x)
+ P
h
s
(x) + P
h
�s
(x): (5.27)
In this ase, the extra ted quark polarization ve tor Q is as follows:
Q =
�u(x) + ��u(x)
u(x) + �u(x)
;
�d(x) + �
�
d(x)
d(x) +
�
d(x)
;
��q(x)
q(x)
!
: (5.28)
With these relations, the ontributions of the quark avors u and d to the spin of
the nu leon an be extra ted.
Together with the assumption on the sea quark polarization of Eq.5.17, three
independent asymmetries are needed at least.
5.3.2 Assumption of Sea Quark from CQSM
Another new assumption is about the sea quark distributions, whi h is based on
the Chiral Quark Soliton Model (CQSM). A simple relation between the polarized
anti-quark distributions is:
��u(x) + �
�
d(x)� 2��s(x) =
3F �D
F +D
h
��u(x)��
�
d(x)
i
; (5.29)
where F and D are the weak oupling onstants in neutron and hyperon �-de ays
(Eq.2.69), and whi h is based on CQSM in the large-N
limit (N
is the number of
olors) [6℄.
72 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
One of the main purposes of this study is the extra tion of avor the asymmetry
of the light sea quark : ��u��
�
d. Therefore it is not desirable to use the assumptions
in Eq.5.17 to extra t ��u � �
�
d. I introdu ed the �rst part of Eq.5.30 as a new
assumption to substitute Eq.5.17.
So the new assumptions are as follows:
��s(x) =
1
2
�
��u(x) + �
�
d(x)�
3F �D
F +D
h
��u(x)��
�
d(x)
i
�
;
�s(x)
s(x)
=
��s(x)
�s(x)
: (5.30)
With these new assumptions, I an extra t quark distributions dire tly, espe ially
the avor asymmetry of the light sea quarks.
Flavor Asymmetry of Light Sea Quarks
With Eq.5.30, the spin asymmetry an be expressed as follows:
A
h
1
(x) = P
h
1
(x) [�u(x) + ��u(x)℄ + P
h
2
(x)
h
�d(x) + �
�
d(x)
i
+P
h
3
(x)
h
��u(x) + �
�
d(x)
i
+ P
h
4
(x)
h
��u(x)��
�
d(x)
i
;
where the purities are de�ned as:
P
h
1
(x) = X
h
u
(x); (5.31)
P
h
2
(x) = X
h
d
(x); (5.32)
P
h
3
(x) = X
h
s
(x) +
3
2
X
h
0
(x); (5.33)
P
h
4
(x) = X
h
3
(x)�
3F �D
F +D
�
X
h
s
(x) +
1
2
X
h
0
(x)�X
h
8
(x)
�
; (5.34)
(5.35)
where oeÆ ients of X
h
(x) are shown below:
X
h
u
(x) =
P
h
u
(x)
u(x)
; (5.36)
X
h
d
(x) =
P
h
d
(x)
d(x)
; (5.37)
X
h
0
(x) =
1
3
(
�
P
h
u
(x)
u(x)
+
P
h
�u
(x)
�u(x)
�
P
h
d
(x)
d(x)
+
P
h
�
d
(x)
�
d(x)
�
P
h
s
(x)
s(x)
+
P
h
�s
(x)
�s(x)
)
;(5.38)
X
h
3
(x) =
1
2
(
�
P
h
u
(x)
u(x)
+
P
h
�u
(x)
�u(x)
+
P
h
d
(x)
d(x)
�
P
h
�
d
(x)
�
d(x)
)
; (5.39)
X
h
8
(x) =
1
6
�
�
P
h
u
(x)
u(x)
+
P
h
�u
(x)
�u(x)
�
P
h
d
(x)
d(x)
+
P
h
�
d
(x)
�
d(x)
+ 2
P
h
s
(x)
s(x)
�
P
h
�s
(x)
�s(x)
!
�
: (5.40)
The ratio of F and D are obtained from [60℄ : F=D = 0:492� 0:083. The P
h
(x) in
the de�nitions of X
h
(x) above are the purities that are de�ned originally in Eq.5.3.
5.3. ASSUMPTIONS ON SEA QUARK POLARIZATIONS 73
In this analysis, the elements of the purity matrix P is a ombination of the original
purities (Eq.5.3) and the unpolarized PDF.
Here the extra ted quark ve tor Q is as follows:
Q =
�
�u(x) + ��u(x); �d(x) + �
�
d(x); ��u(x) + �
�
d(x); ��u(x)��
�
d(x)
�
:(5.41)
One of the hara teristi of this analysis with the new assumptions is that the
quark distributions an be extra ted dire tly, not only the polarization whi h is
the ratio between the polarized and the unpolarized distributions. Additionally,
using the results of the avor asymmetry extra tion Eq.5.41, it is possible to extra t
almost all avor quark distributions : �u;�d;��u;�
�
d;�s = ��s . This analysis is
des ribed in the next.
It has to be mentioned that due to the ombination with the unpolarized PDF,
whi h are Q
2
dependent, also the matrix P and the quark ve tor Q be ome Q
2
dependent in this avor asymmetry extra tion. As the measured asymmetries are
onsidered to be in the �rst order Q
2
independent, the PDF in the ombination of
the original purities (Eq.5.36 - Eq.5.40) at a �xed Q
2
of Q
2
=2.5 GeV
2
has been
used to get the Q ve tor also at this �xed Q
2
. The Q
2
dependen e of the purities
whi h are used to extra t the avor asymmetry (Eq.5.31 - Eq.5.34) are shown in
Fig.5.15 and Fig.5.16. The purities are al ulated at �xed value Q
2
of Q
2
=1.2, 2.5
and 10 GeV
2
. The solid ir le shows the purities extra ted using the mean value
of Q
2
in ea h x bin. At small x, there are no Q
2
dependen e for all ombination
purities P
h
1
� P
h
4
. The small deviation an be observed in the high x bins, although
these deviations don't have in uen e on the extra ted quark distributions in this
HERMES pre ision.
Single Quark Distribution
With results of the above analysis, as has been suggested, I an obtain the quark
distributions of almost all avors, where the distribution of �s(x) and ��s(x) are
de�ned to be the same.
First, I extra t �u(x);��u(x);�d(x);�
�
d(x) from the results of Eq.5.41. With
��u+�
�
d and ��u��
�
d of Eq.5.41 and the �rst relation of Eq.5.30 the strange quark
distributions an be extra ted : ��s(x) = �s(x). With this method, the quark
distributions of ea h other avor an be obtained.
74 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1Pe+
P1 ∆(u+u–)
Q2 dep
Q2=1.2Q2=2.5Q2=10
P2 ∆(d+d–) P3 ∆(u
–+d
–) P4 ∆(u
–-d
–)
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1Ph
+
10-2
10-1
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1Ph
-
x10
-210
-1
x10
-210
-1
x10
-210
-1
x
Figure 5.15: The Q
2
dependen e of the purities for avor asymmetry extra -
tion for the proton target. These purities are ombination of the
original purities (Eq.5.3) and unpolarized PDF, whi h is used
CTEQ4LQ.
5.3. ASSUMPTIONS ON SEA QUARK POLARIZATIONS 75
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1Pe+
P1 ∆(u+u–)
Q2 dep
Q2=1.2Q2=2.5Q2=10
P2 ∆(d+d–) P3 ∆(u
–+d
–) P4 ∆(u
–-d
–)
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1Ph
+
10-2
10-1
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1Ph
-
x10
-210
-1
x10
-210
-1
x10
-210
-1
x
Figure 5.16: The Q
2
dependen e of the purities for avor asymmetry extra -
tion for the neutron target. These purities are ombination of
the original purities (Eq.5.3) and unpolarized PDF, whi h is
used CTEQ4LQ.
76 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
5.4 Interpretation of Asymmetries on the
3
He and
D Targets
At HERMES, proton (p), helium3 (
3
He) and deuteron (D) have been used as targets.
The quark distributions are de�ned on the proton and on the neutron are related
by isospin symmetry : u
p
= d
n
and �u
p
=
�
d
n
. As a polarized neutron target is not
available,
3
He and D targets are used as neutron targets. To analyze the asymmetries
on the
3
He and the D targets, we have to translate these experimental asymmetries
into the asymmetries on the neutron target. In this study, two methods are used to
translate the measured asymmetries on the
3
He and the D target, whi h are based
on the same relation and are identi al essentially. The number of input asymmetries
whi h an be used in ea h method is di�erent. The details are des ribed in the
following.
5.4.1 Asymmetry on the
3
He Target
The
3
He onsists of two protons and one neutron, If the protons have anti-parallel
spins so that the total proton spin is zero (S-state), whi h is the most probable
on�guration, the spin of
3
He is entirely arried by the neutron. But the e�e tive
spin of the neutron in
3
He seen in the measurement is diluted by the two protons
be ause of the angular momentum in the
3
He. So for a ompletely polarized
3
He,
the e�e tive polarizations of the protons and the neutron have to be onsidered.
The e�e tive polarization of the proton is p
p
3
He
= - 0.028 and that of the neutron is
p
n
3
He
= 0.86 [61℄.
With the value of these e�e tive polarizations, the measured asymmetry on the
3
He target are expressed by the asymmetry on the proton and the neutron in every
x bin as:
A
h
1;
3
He
= f
h
p;
3
He
p
p
3
He
A
h
1;p
+ f
h
n;
3
He
p
n
3
He
A
h
1;n
; (5.42)
where f
h
p;
3
He
and f
h
n;
3
He
are the dilution fa tors whi h give the probability of s attering
on one of the two protons or the neutron. From this de�nition, f
h
p;
3
He
+ f
h
n;
3
He
= 1 is
obvious. These fa tors an be expressed as follows:
f
h
p;
3
He
=
2�
h
p
�
h
3
He
= 2
n
h
p
n
h
3
He
�
�
p
�
3
He
= 2
n
h
p
n
h
3
He
�
F
p
2
2F
D
2
+ F
p
2
; (5.43)
where �
h
p
and �
h
3
He
are the integrated unpolarized ross se tions of the hadron h on
the proton target and on the
3
He target, respe tively. �
p
and �
3
He
are the in lusive
ross se tions on ea h target. n
h
p
and n
h
3
He
are the multipli ities of hadron h on ea h
target. In Eq.5.43, these relations are used:
�
h
3
He
= 2�
h
p
+ �
h
n
; (5.44)
F
n
2
= 2F
D
2
� F
p
2
; (5.45)
5.4. INTERPRETATION OF ASYMMETRIES ON THE
3
HE AND D
TARGETS 77
where �
h
n
is the integrated unpolarized ross se tion of the hadron h on the neutron
target.
With the relation of Eq.5.42 and the oeÆ ients from Eq.5.43, we an trans-
late the asymmetry on the
3
He target to a ombination of that on the proton and
the neutron. In the followings, I will des ribe the pra ti al way of using the
3
He
asymmetries as input asymmetries with Eq.5.42.
5.4.2 Asymmetry on the D Target
In ase of the D target, we an onsider the translation of the asymmetry on the D
target similarly to the ase of the
3
He target. When the deuteron is in the S-state, the
spins of the proton and the neutron are aligned parallel with no angular momentum.
But the proton and the neutron polarization are diluted by the existen e of D-state
in whi h one of the nu leons is anti-aligned. This e�e tive polarizations of the proton
and the neutron are p
p
D
= p
n
D
= 0.925
1
[62℄ [63℄. This e�e t has to be onsidered in
the treatment of the asymmetry on the D target. Here f
h
p;D
+ f
h
n;D
= 1 is also valid.
In this ase, the asymmetry is expressed by the asymmetry on the proton and the
neutron as:
A
h
1;D
= f
h
p;D
p
p
D
A
h
1;p
+ f
h
n;D
p
n
D
A
h
1;n
; (5.46)
and the dilution fa tor f
h
p;D
an be de�ned as:
f
h
p;D
=
�
h
p
�
h
D
=
n
h
p
n
h
D
�
�
p
�
D
=
n
h
p
n
h
D
�
F
p
2
2F
D
2
; (5.47)
where �
h
D
is the integrated unpolarized ross se tion of the hadron h on the D target
and �
D
is the in lusive ross se tion. n
h
D
is the multipli ity of hadron h on the D
target. Similarly, Eq.5.45 and this relation are used:
�
h
D
= �
h
p
+ �
h
n
; (5.48)
whi h is the substitute for Eq.5.44.
These relations are used in the next se tion.
5.4.3 Method to use Asymmetry on
3
He and D
In the previous se tions, the expressions of the asymmetry on the
3
He and the D
target are shown. Here I des ribe how to apply these relations to extra t the quark
distributions.
When I use two types of asymmetries on the proton and the neutron target as
input of Eq.5.5, the matrix equation is expanded as follows:
�
A
p
A
n
�
= C
i
�
�
P
p
P
n
�
�Q; (5.49)
1
They are almost equal the D-State probability, !
D
=0.058 � 0.01.
78 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
whereA
p
andA
n
are asymmetries on the proton and the neutron target respe tively,
and P
p
and P
n
are purity matri es on the respe tive target. Sin e the asymmetries
on a free neutron target are not available experimentally, as mentioned before, asym-
metries on the
3
He and the D target should be used, whi h are substitutes for that
on the neutron with the relations of Eq.5.42 and Eq.5.46. Be ause the MC gen-
erator on the
3
He and the D target annot reprodu e the experimental data, the
asymmetries have to be al ulated for that on the neutron.
There are two methods to apply the above relations to Eq.5.49. The one is
the method to use the mixing matrix whi h is shown in the following [57℄ [49℄ [58℄
[59℄. Another is the new method, whi h is introdu ed in this study, making new
de�nitions of the purity for the
3
He and the D target. Both methods are based
on the same relation, however, the kind of asymmetries whi h an be used in ea h
method is di�erent. With the mixing matrix method, the in lusive, h
+
and h
�
asymmetries on all target (p,
3
He and D) an be used. On the other hand, with
the new de�nition of purity, �
+
, �
�
, K
+
and K
�
asymmetries on the D target
also an be in luded to extra t quark distributions. Therefore, we an use up to
13 types of asymmetries with the new de�nition of purity. Although there is a
ondition to use the new de�nition of purity, whi h is des ribed later. To study
of quark distributions, in parti ular sea quarks, the pion and kaon asymmetries
are onsidered to be important, be ause they ontain sea quarks. Espe ially K
�
onsists of two sea quarks, �u and s, so it is seems to be sensitive to sea quark
distributions. Therefore, we sele t the method with the new de�nition of purity to
al ulate moment �nally, but both results are shown as follows. The details of both
methods are des ribed in the following se tions.
The Mixing Matrix
From Eq.5.42 and Eq.5.46, the asymmetries on the neutron target are obtained from
the measured asymmetries as follows:
� the ase of the
3
He target
A
h
1;n
=
1
f
h
n;
3
He
p
n
3
He
�
�
A
h
1;
3
He
� f
h
p;
3
He
p
p
3
He
A
h
1;p
�
=
1
p
n
3
He
(1� f
h
p;
3
He
)
�
�
A
h
1;
3
He
� f
h
p;
3
He
p
p
3
He
A
h
1;p
�
: (5.50)
� the ase of the D target
A
h
1;n
=
1
f
h
n;D
p
n
D
�
�
A
h
1;D
� f
h
p;D
p
p
D
A
h
1;p
�
=
1
p
n
D
(1� f
h
p;D
)
�
�
A
h
1;D
� f
h
p;D
p
p
D
A
h
1;p
�
: (5.51)
If I use Eq.5.50 and Eq.5.51 in the matrix equation of Eq.5.49, the following
expansion is a hieved:
0
B
�
A
p
A
3
He
A
D
1
C
A
= C
i
� N �
0
B
�
P
p
P
n
P
n
1
C
A
�Q; (5.52)
5.4. INTERPRETATION OF ASYMMETRIES ON THE
3
HE AND D
TARGETS 79
where A
3
He
and A
D
are the asymmetries on the
3
He and the D target, and N is
alled \the mixing matrix" whi h is introdu ed to al ulate A
3
He
and A
D
to A
n
with Eq.5.50 and 5.51.
The elements of the mixing matrix N are shown below:
N =
0
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
�
1
.
.
.
1 0
a
h
1
3
He
b
h
1
3
He
.
.
.
.
.
.
a
h
m
3
He
b
h
m
3
He
a
h
1
D
b
h
1
D
.
.
.
0
.
.
.
a
h
m
D
b
h
m
D
1
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
A
(5.53)
where
a
h
i
3
He
� f
h
i
p;
3
He
p
p
3
He
; b
h
i
3
He
� (1� f
h
i
p;
3
He
) p
n
3
He
; (5.54)
a
h
i
D
� f
h
i
p;D
p
p
D
; b
h
i
D
� (1� f
h
i
p;D
) p
n
D
; (5.55)
where h
i
is the hadron type of asymmetries. With the mixing matrixN , asymmetries
on the
3
He and the D target an be in luded in this analysis, although there is a
ondition to use this method. This method is allowed only if the orresponding
asymmetries on the proton target exists for all used asymmetries on the helium3
and deuteron target. Now the available data on the proton target are the in lusive
and the harged hadron(h
+
, h
�
) asymmetries, and there is no data for identi�ed
pions, kaons and protons, sin e the RICH had not yet been installed during the data
taking time in the proton target. However, sin e the �, K and p asymmetries are
onsidered to be sensitive to the sea quark polarization, the in lusion of deuteron
data, where �, K and p are identi�ed with RICH, are quite important. Be ause
these pion and kaon asymmetries annot be used in this mixing matrix method, I
onsidered another method to use these additional asymmetries. This new method
is des ribed in the next part.
The New De�nition of Purity for the
3
He and the D Target
Here a new de�nition of the purities on the
3
He and the D target is introdu ed to use
the purities for the proton and the neutron target. As the mixing matrix method,
the Eq.5.42 and Eq.5.46 are based on this new method and expressed as follows:
� the ase of the
3
He target
A
h
1;
3
He
= f
h
p;
3
He
p
p
3
He
A
h
1;p
+ f
h
n;
3
He
p
n
3
He
A
h
1;n
= a
h
3
He
A
h
1;p
+ b
h
3
He
A
h
1;n
= a
h
3
He
X
f
P
h
f;p
�q
f
q
f
+ b
h
3
He
X
f
P
h
f;n
�q
f
q
f
=
X
f
�
a
h
3
He
� P
h
f;p
+ b
h
3
He
� P
h
f;n
�
�q
f
q
f
80 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
=
X
f
P
h
f;
3
He
�q
f
q
f
; (5.56)
where P
h
f;p
, P
h
f;n
and P
h
f;
3
He
are purities of parti le h for the proton, the neutron
and the helium3 target respe tively. To obtain Eq.5.56, the purities for the
helium3 target are de�ned as:
P
h
f;
3
He
�
�
a
h
3
He
� P
h
f;p
+ b
h
3
He
� P
h
f;n
�
: (5.57)
� the ase of the D target is similarly to the ase of the
3
He target
A
h
1;D
= f
h
p;D
p
p
D
A
h
1;p
+ f
h
n;D
p
n
D
A
h
1;n
= a
h
D
A
h
1;p
+ b
h
D
A
h
1;n
= a
h
D
X
f
P
h
f;p
�q
f
q
f
+ b
h
D
X
f
P
h
f;n
�q
f
q
f
=
X
f
�
a
h
D
� P
h
f;p
+ b
h
D
� P
h
f;n
�
�q
f
q
f
=
X
f
P
h
f;D
�q
f
q
f
; (5.58)
where P
h
f;D
are the purities for the deuteron target, and they are de�ned as:
P
h
f;D
�
�
a
h
D
� P
h
f;p
+ b
h
D
� P
h
f;n
�
: (5.59)
With the Eq.5.56 and Eq.5.58, all available asymmetries in luding pion and
kaon asymmetries an be used. We need to know the multipli ities of parti le h
on all targets to estimate the new rede�ned purities on the
3
He and the D target.
However, sin e there are no pion and kaon data for the proton target, we assume on
the multipli ities on the proton as follows:
n
h
p
n
h
3
He
=
n
h
p
n
h
D
= 1: (5.60)
Using Eq.5.60, the oeÆ ients are hanged like below with Eq.5.43 and Eq.5.47:
a
h
3
He
=
2F
p
2
2F
D
2
+ F
p
2
� p
p
3
He
; (5.61)
b
h
3
He
=
1�
2F
p
2
2F
D
2
+ F
p
2
!
� p
n
3
He
; (5.62)
a
h
D
=
F
p
2
2F
D
2
� p
p
D
; (5.63)
b
h
D
=
1�
F
p
2
2F
D
2
!
� p
n
D
: (5.64)
These above oeÆ ients don't depend on the experimental variables, but they
depend on the parameterizations of the stru ture fun tions : F
p
2
and F
d
2
. There is
5.4. INTERPRETATION OF ASYMMETRIES ON THE
3
HE AND D
TARGETS 81
no fundamental di�eren e from the original method with the mixing matrix, but
this method has more exibility. In this study, the stru ture fun tions are obtained
from [64℄.
Fig.5.17 shows the multipli ity ratios for the proton and the deuteron targets.
The ratios are lose the one, whi h justi�es the assumptions in Eq.5.60. As a
omparison of the result with this method and that with the mixing matrix method
shows, the deviations of the ratios from one an be negle ted. In addition, with the
preliminary data of the pion and kaon multipli ities on the proton target, whi h are
obtained from the unpolarized data [65℄, the results with the assumption Eq.5.60
are ompared with that using real multipli ities. The deviation between two results
are not observed and they are almost identi al in the present statisti s.
With the assumption Eq.5.60, the new de�nition of the purity for the
3
He and
the D target is available for all hadron asymmetries. Now the asymmetries of pion
and kaon on the deuteron target whi h have been identi�ed by the RICH an be
used to extra t quark polarizations. The results of the analysis performed with
all available asymmetries on all the targets for both assumptions on the sea quark
distributions (Eq.5.17 and Eq.5.30) are shown in Se .6.1.
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
nph+ / nD
h+
Rat
io o
f h
+ Mul
tiplic
ity
10-2
10-1
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
nph- / nD
h-
x
Rat
io o
f h
- Mu
ltip
licit
y
Figure 5.17: The ratio of the multipli ities on the proton and the deuteron
target. The top �gure shows the ratio of h
+
multipli ities and
the bottom that of h
�
. The data on the proton target are from
1997 [66℄ and on the D target from 1998 [67℄.
82 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
5.5 The Set of Analyzed Asymmetries
In this se tion, all asymmetries measured by HERMES for di�erent parti les and
di�erent targets whi h have been used as an input in this analysis are presented.
5.5.1 1995 Asymmetries on the
3
He Target
In 1995, only a threshold Cherenkov Counter with a limited momentum range for the
hadron identi� ation was operated. Therefore, in lusive, h
+
and h
�
asymmetries
on the
3
He target are available. They are shown in Fig.5.18. At HERMES, the
binning in x onsists of nine bins and the de�nition of the binning in x is shown in
Appendix.C.1. Here, the values obtained from [49℄ have been used, whi h are given
in Tab.C.2.
10-2
10-1
-0.3
-0.2
-0.1
0
0.1
0.2
0.3A1
A1e+ (3He 95) A1
h+ (3He 95)
10-2
10-1
x
A1h- (3He 95)
Figure 5.18: The in lusive and semi in lusive asymmetries on the
3
He target
from 1995. The harged hadron asymmetries are shown as the
semi in lusive data. For ea h point the error bars represent the
statisti al un ertainty. The systemati un ertainty is not shown
in this �gure.
In 1995, the positron beam are used at HERA. The harged hadron asymmetries
on the
3
He target are measured for the �rst time. The in lusive asymmetry in
Fig.5.18 shows slightly negative, and the hadron asymmetries are almost zero. The
S-state, the two proton spin is zero, is dominated the wave fun tion of the
3
He
nu leus, therefore the measured asymmetries are almost aused by the neutron.
5.5. THE SET OF ANALYZED ASYMMETRIES 83
5.5.2 1996 and 1997 Asymmetries on the Proton Target
In 1996 and 1997, the threshold Cherenkov Counter was also used, So only in lusive,
h
+
and h
�
asymmetries are available on the proton target. Fig.5.5.2 and Fig.5.5.2
show the 1996 and 1997 asymmetries. These asymmetries have been also obtained
from [49℄. The numeri al values of these asymmetries are shown in Tab.C.3 and
Tab.C.4. Both asymmetries are on the proton target, although, they are treated
separately be ause their systemati un ertainty are di�erent.
10-2
10-1
-0.2
0
0.2
0.4
0.6
0.8
1A1
A1e+ (p 96) A1
h+ (p 96)
10-2
10-1
x
A1h- (p 96) 10
-210
-1
-0.2
0
0.2
0.4
0.6
0.8
1A1
A1e+ (p 97) A1
h+ (p 97)
10-2
10-1
x
A1h- (p 97)
Figure 5.19: The in lusive and semi in lusive asymmetries on the proton
target from 1996 and 1997. The harged hadron asymmetries
are shown as the semi in lusive data. For ea h point the er-
ror bars represent the statisti al un ertainty. The systemati
un ertainty is not shown in this �gure.
The in lusive asymmetries on the proton target show positive value and in rease
as x in rease. In the se ond bin from the high x, the h
+
asymmetry in 1996 shows
a strange behavior, however, it is not physi al phenomenon. This behavior doesn't
be shown in 1997, the statisti in 1997 is more than that in 1996. The same is true
for h
�
asymmetries in high x bins. The negative hadron asymmetries show smaller
positive value than that of orresponding positive hadron asymmetries. It gives a
hint that u quarks have positive and d quarks have negative polarizations.
84 CHAPTER 5. EXTRACTION OF POLARIZED QUARK DISTRIBUTIONS
5.5.3 1998 Asymmetries on the D Target
In 1998, the RICH was started to dete t the s attered hadrons, So in lusive, h
+
,
h
�
, �
+
, �
�
, K
+
and K
�
asymmetries are available on the D target. Fig.5.5.3 shows
these asymmetries. These asymmetries have been extra ted in [50℄. The numeri al
values of these asymmetries are shown in Tab.C.5 and Tab.C.6.
10-2
10-1
-0.2
0
0.2
0.4
0.6
0.8
1A1
A1e- (D 98) A1
h+ (D 98)
10-2
10-1
x
A1h- (D 98)
-0.2
0
0.2
0.4
0.6
0.8
1A1
A1π+ (D 98) A1
K+ (D 98)
10-2
10-1
-0.2
0
0.2
0.4
0.6
0.8
1A1
π- (D 98)
10-2
10-1
x
A1K- (D 98)
Figure 5.20: The in lusive and semi in lusive asymmetries on the D target
from 1998. The harged hadron, pion and kaon asymmetries
are shown as the semi in lusive data. For ea h point the error
bars represent the statisti al un ertainty. The systemati un er-
tainty is not shown in this �gure. The harged kaon asymmetries
are measured �rst time at HERMES.
The in lusive asymmetry on the D target shows positive and slightly in reasing
as x in rease. The positive value of the in lusive asymmetry is smaller than that on
the proton target. As the data in 1996, the positive hadron asymmetry shows a small
value in the se ond high x bin, although, it is aused by the statisti s. The positive
hadron, pion and kaon asymmetries show a slightly positive, and they with negative
harge are near to zero. The pion asymmetries are similar to that of hadrons, it
shows that the ontribution of pions to hadrons are most dominate. The proton and
anti-proton asymmetries on the D target an be measured with RICH, and they
are already obtained. Though they are not used to extra t the quark distributions,
sin e the knowledge of the fragmentation fun tion of proton and anti-proton are
poor and the protons whi h are identi�ed with RICH have a ontamination by the
proton de ayed from the lambda parti le. Additionally the proton and anti-proton
is not so large statisti s, hen e we don't use the proton and anti-proton asymmetries
in this study.
5.5. THE SET OF ANALYZED ASYMMETRIES 85
5.5.4 The Sets of Input Asymmetries
The following shows the set of asymmetries whi h has been used for ea h method
on the proton, the
3
He and the D target. In 1995-1997, HERA has been operated
with positron and in 1998 with ele tron.
The Method with the Mixing Matrix
In the mixing matrix method, the asymmetries of parti le h on the proton target
is ne essary for that on the
3
He and the D target, respe tively. As I mentioned
before, be ause only the in lusive, h
+
and h
�
asymmetries on the proton target are
available, I an use only the asymmetries of these parti le for all targets. With the
expanded matrix of Eq.5.52, the sets of input asymmetries are shown below:
A
p
=
0
B
B
B
B
B
B
B
B
B
�
A
e
+
1;p
(96)
A
h
+
1;p
(96)
A
h
�
1;p
(96)
A
e
+
1;p
(97)
A
h
+
1;p
(97)
A
h
�
1;p
(97)
1
C
C
C
C
C
C
C
C
C
A
; A
3
He
=
0
B
�
A
e
+
1;
3
He
(95)
A
h
+
1;
3
He
(95)
A
h
�
1;
3
He
(95)
1
C
A
; A
D
=
0
B
�
A
e
�
1;D
(98)
A
h
+
1;D
(98)
A
h
�
1;D
(98)
1
C
A
; (5.65)
where A
j
1;i
(k) is the asymmetry A
1
of parti le j on the target i of the year k. The
asymmetries of the year 1996 and 1997 are treated separately be ause the systemati
errors oming from the target polarization are di�erent.
The Method with the New De�nition of Purity for the
3
He and D Target
With the new de�nition of the purities for the
3
He and the D targets, I an use more
asymmetries as input be ause it is not ne essary to have the asymmetries on the
proton target of all the parti le types. This gives the following set of asymmetries:
A
p
=
0
B
B
B
B
B
B
B
B
B
�
A
e
+
1;p
(96)
A
h
+
1;p
(96)
A
h
�
1;p
(96)
A
e
+
1;p
(97)
A
h
+
1;p
(97)
A
h
�
1;p
(97)
1
C
C
C
C
C
C
C
C
C
A
; A
3
He
=
0
B
�
A
e
+
1;
3
He
(95)
A
h
+
1;
3
He
(95)
A
h
�
1;
3
He
(95)
1
C
A
; A
D
=
0
B
B
B
B
B
B
B
�
A
e
�
1;D
(98)
A
�
+
1;D
(98)
A
�
�
1;D
(98)
A
K
+
1;D
(98)
A
K
�
1;D
(98)
1
C
C
C
C
C
C
C
A
: (5.66)
The sets of asymmetries on the proton and the
3
He targets are the same as the
previous ones, but additionally I an use the asymmetries of �
+
, �
�
, K
+
and K
�
on
the D target. Here I didn't use the h
+
and h
�
asymmetries of the D target be ause
there are large orrelations between the asymmetries of the hadrons and that of
pions and kaons.
87
Chapter 6
Analysis and Results
6.1 Results for the Polarized Quark Distributions
In this se tion, the results for quark polarizations �q=q and distributions �q are
shown whi h have been obtained with the two methods des ribed in Se .5.4.3. There
are four types of extra ted quark polarizations and distributions (see Se .5.3.1 and
Se .5.3.1).
Before we extra t quark polarization from experimental data, we he ked the
extra tion method with MC data. This means that the quark polarizations extra ted
with this method should reprodu e the polarized PDF whi h is put into MC as input
data. First, the input data su h as in lusive and harged hadron asymmetries is
produ ed by using polarized PDF with MC. We use GRSV parameterization [68℄
as polarized PDF. Next, the purity matrix is produ ed by using CTEQ4LQ [53℄
or GRV [54℄ as unpolarized PDF. The quark polarization is determined from the
simulated asymmetries and purities with this method. If the method is orre t, the
obtained results from MC data should reprodu e the GRSV polarized PDF whi h
is used as input of MC.
Additionally, the omparison between the mixing matrix method and the new
de�nition of purity are shown, where in lusive and harged hadrons asymmetries on
all targets are used as input (see Eq.5.65) for both method. Comparing the both
results, the in uen e from the assumption of Eq.5.60 an be he ked, and we an see
that the assumption on the ratio of the multipli ities (Eq.5.60) does not have large
in uen e on the extra ted quark distributions. The results of this study is shown in
the following.
Here Fig.6.1 shows the results of the avor de omposition from MC. There are
two types of quark polarizations, whi h are the result with the mixing matrix and
that with the new de�nition of purity. In Fig.6.1, the solid line shows the GRSV
parameterization [68℄ whi h is used for produ tion of the MC data. We an see that
the extra ted quark distributions in Fig.6.1 reprodu e the input GRSV parameter-
izations for the both results. From the omparison between the results with the
mixing matrix method and the new de�nition of purity, it be omes lear that the
deviations between these two methods are very small. Therefore the result with the
assumption on the ratio of the multipli ities is onsistent with that without this
assumption, and almost an re e t the realisti quark distributions.
The �gure of the extra ted avor asymmetry from the MC is shown in Fig.6.2.
88 CHAPTER 6. ANALYSIS AND RESULTS
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
(∆u
+∆u–
)/(u
+u–)
mixing matrixnew definition of purity
GRSV
-1
-0.8-0.6
-0.4
-0.2
00.2
0.4
0.6
0.81
(∆d
+∆d–
)/(d
+d–)
GRSV
10-2
10-1
-1-0.8
-0.6
-0.4
-0.20
0.2
0.4
0.60.8
1
x
∆qs/
qs
GRSV
Figure 6.1: The avor de omposition from MC. The two results with the
mixing matrix method (Se .5.4.3) and the new de�nition of pu-
rity (Se .5.4.3) are shown for omparison. q
s
is de�ned as sea
quarks : q
s
= f�u;
�
d; s; �sg.
From this results, it is shown that the reprodu tion of the input GRSV PDF is
a hieved for the quark distribution extra tion. For both the avor de omposition
and the avor asymmetry extra tion, there is no large di�eren e between the results
with the mixing matrix and the new de�nition of purity.
From Fig.6.1 and Fig.6.2, the methods whi h are des ribed in Se .5.3.1 and
Se .5.3.2 are reliable to obtain the quark distributions and polarizations. In addition,
it is lear that the systemati un ertainties whi h ome from the analysis method is
small. In the next se tion, the extra ted quark distributions from the experimental
results are shown.
6.1. RESULTS FOR THE POLARIZED QUARK DISTRIBUTIONS 89
-1-0.5
00.5
11.5
22.5
33.5
∆u+∆
u–
mixing matrixnew definition of purity
GRSV
-3-2.5
-2-1.5
-1-0.5
00.5
1
∆d+∆
d–
GRSV
-3
-2
-1
0
1
2
3
∆u–+∆
d–
GRSV
10-2
10-1
-3
-2
-1
0
1
2
3
x
∆u–-∆
d–
GRSV
Figure 6.2: The avor asymmetry extra tion of light sea quarks from MC.
The two results with the mixing matrix method (Se .5.4.3) and
the new de�nition of purity (Se .5.4.3) are shown for omparison.
90 CHAPTER 6. ANALYSIS AND RESULTS
6.1.1 The Valen e De omposition
Fig.6.3 and Fig.6.4 are the results of the valen e de omposition whi h are dis ussed
in Se .5.3.1. Fig.6.3 is obtained by solving the matrix equation Eq.5.5. Fig.6.4
shows the values times x and unpolarized PDF. The purities whi h were produ ed
from CTEQ4LQ PDF are used as input for Eq.5.5, like for all the following results.
The numeri al values of these results are listed in Tab.C.14 and Tab.C.16.
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
∆uv/
uv
mixing matrixnew definition of purity
-1
-0.8-0.6
-0.4
-0.2
00.2
0.4
0.6
0.81
∆dv/
dv
10-2
10-1
-1-0.8
-0.6
-0.4
-0.20
0.2
0.4
0.60.8
1
x
∆qs/
qs
Figure 6.3: The extra ted quark polarizations in valen e de omposition from
experimental data. The solid ir les show the results with the
mixing matrix using in lusive and harged hadron asymmetries.
The open squares is that with the new de�nition of purities us-
ing in lusive, harged hadron, harged pion and harged kaon
asymmetries. In this de omposition, assumption of polarization
symmetri sea is used.
6.1. RESULTS FOR THE POLARIZED QUARK DISTRIBUTIONS 91
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6x
∆uv
mixing matrixnew definition of purity
-0.2
-0.1
0
0.1
0.2
x ∆d
v
10-2
10-1
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
x
x ∆q
s
Figure 6.4: The quark distributions obtained from the valen e de omposition
result at Q
2
= 2.5 GeV
2
. The solid lines express the positivity
limits whi h are unpolarized PDF, where CTEQ4LQ parameter-
izations are shown. The two results with the mixing matrix and
the new de�nition of purity are obtained.
The solid ir le shows the result with the mixing matrix to treat asymmetries
on the
3
He and D target (Se .5.4.3) , and the open square shows the result with the
new de�nition of purity (Se .5.4.3). In Fig.6.4, the solid urves show the positivity
limit j�q=qj � C
�1
i
obtained from Eq.5.2, the quark polarizations don't ex eed this
limit. q
s
expresses the sea quarks whi h in lude f�u;
�
d; s; �sg.
Fig.6.3 shows that the polarization of u
v
quarks is positive over the full x region,
and there is a slight in rease with in reasing x. The polarization of d
v
quarks is
92 CHAPTER 6. ANALYSIS AND RESULTS
negative in the small x region, but it is not obvious in the highest x bins. In ase
of the sea quarks q
s
, the polarization is at and onsistent with zero over the whole
in x range. Be ause of the de reasing ontribution of the sea quarks in the high x
range, our data are not sensitive in this region and therefore the highest two x bins
have been omitted.
Comparing the two methods of treating the asymmetries on the
3
He and the
D target (solid ir le and open square), there are only small di�eren es, whi h are
in the same order of magnitude as the di�eren e seen in the MC simulation. It it
no large e�e t on the valen e quarks to use the asymmetries of pions and kaons
instead of general hadron asymmetries. For the polarization of the valen e quarks,
the extra ted values are stable and not sensitive to the extra tion method.
6.1. RESULTS FOR THE POLARIZED QUARK DISTRIBUTIONS 93
6.1.2 The Flavor De omposition
The results for the avor de omposition are presented in Fig.6.5 and Fig.6.6, where
the purities obtained from CTEQ4LQ [53℄ have been used. The numeri al values of
these results are listed in Tab.C.18 and Tab.C.20. We an obtain the ontributions
of ea h quark avor in this de omposition espe ially that of u(= u+�u) and d(= d+
�
d)
quarks. As same as the valen e de omposition results, they show a omparison of
the two extra tion methods.
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
(∆u
+∆u–
)/(u
+u–)
mixing matrixnew definition of purity
-1
-0.8-0.6
-0.4
-0.2
00.2
0.4
0.6
0.81
(∆d
+∆d–
)/(d
+d–)
10-2
10-1
-1-0.8
-0.6
-0.4
-0.20
0.2
0.4
0.60.8
1
x
∆qs/
qs
Figure 6.5: The extra ted quark polarizations in avor de omposition from
experimental data. The solid ir les show the results with the
mixing matrix using in lusive and harged hadron asymmetries.
The open squares is that with the new de�nition of purities us-
ing in lusive, harged hadron, harged pion and harged kaon
asymmetries. In this de omposition, assumption of polarization
symmetri sea is used.
94 CHAPTER 6. ANALYSIS AND RESULTS
Looking at the results of u quark in Fig.6.5, the positive value of the u quark
polarization is shown and there is almost no di�eren e between the two methods.
The statisti errors are also small for both results, therefore it is lear that the
polarization of the u quark is positive and in reasing as x in reases.
For the d quark polarization, the two results are almost the same and there is
not obvious di�eren e. They show negative values in almost all x bins, but it is not
lear for the highest bin of x. From Fig.6.6, the d quark distribution almost ex eeds
the positivity limit in the highest bin of x.
In this ase, q
s
are de�ned as q
s
= �u +
�
d + s + �s. About the sea quarks of
q
s
, it shows zero or a slightly negative value in the small x region. The valen e
quark polarization an be extra ted without depending the extra tion method and
the type of the used asymmetries. I also al ulated the moments of �u + ��u and
�d+�
�
d. They are shown in Se .6.2.
In the valen e and the avor de omposition, the ontributions of the valen e
quarks to the nu lear spin an be extra ted separately, but that of the sea quarks
annot be obtained learly. Hen e I analyzed the sea quark distributions with the
new assumptions of Eq.5.30. The results are shown in the following se tions.
6.1. RESULTS FOR THE POLARIZED QUARK DISTRIBUTIONS 95
-0.2
0
0.2
0.4
0.6
x (∆
u+∆
u–)
mixing matrixnew definition of purity
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
x (∆
d+∆
d–)
10-2
10-1
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
x
x ∆q
s
Figure 6.6: The quark distributions obtained from the avor de omposition
result at Q
2
= 2.5 GeV
2
. The solid lines express the positivity
limits whi h are unpolarized PDF, where CTEQ4LQ parameter-
izations are shown. The two results with the mixing matrix and
the new de�nition of purity are obtained.
96 CHAPTER 6. ANALYSIS AND RESULTS
6.1.3 Flavor Asymmetry of Light Sea Quarks
Here I show the results for the avor asymmetry extra tion of light sea quarks, whi h
are the main subje t of this study based on the new extra tion method assuming
CQSM, we obtained the polarized sea quark distribution in luding ��u � �
�
d as
shown in Fig.6.7 and Fig.6.8. It should be noti ed that the results with the new
de�nition of purity in lude pion and kaon asymmetries measured with RICH of the
�rst time at HERMES, although that of the mixing matrix only use the in lusive
and harged hadron asymmetries. These results are obtained by solving Eq.5.5,
although in this ase the purities are di�erent between the previous analysis of the
valen e and the avor de omposition. The purities whi h are used here in lude the
unpolarized PDF from CTEQ4LQ (see Eq.5.31-Eq.5.40), therefore it is possible to
extra t the polarized quark distributions dire tly, not the quark polarizations. As I
mentioned earlier, the quark distributions are extra ted at Q
2
=2.5 GeV
2
.
When we look at the distributions of the u and d quarks, �u+��u and �d+�
�
d,
of Fig.6.7, they are onsistent in two extra tion methods and there is almost no
di�eren e between the solid ir le and the open square points. The distribution of
�u + ��u is positive values in all measured region, and whi h de reases with the x
in reasing. While the result of the �d + �
�
d is negative, whi h is opposite to the
ase of u quark, and it in reases with the x in reasing. From Fig.6.8, all valen e
and sea quark distributions, multiplied x, lie within the positivity limit.
However, there are some di�eren es in the distributions of the sea quarks in
Fig.6.7 between the two extra tion methods (see the s ale of Fig.6.7 arefully, whi h
is large omparing the �gures for the valen e quarks). For the distribution of the
��u + �
�
d, there might be a deviation between the two methods in small x region.
The result with the mixing matrix, whi h is shown with the solid ir le, tells us that
this distribution is slightly negative in the middle x region but it in reases to zero
in the low x bin. The result with the new de�nition of purities whi h is indi ated
with the open square. However, it is almost negative in the middle x and does not
in rease in lowest x bin.
The avor asymmetry of light sea quark ��u � �
�
d an be extra ted dire tly in
this analysis. The distribution of ��u ��
�
d has a slight positive value. Comparing
the solid ir le and the open square, the small di�eren e is observed. In the lowest x
bin, the both result shows almost zero for the avor asymmetry, although using the
new de�nition of purity, the avor asymmetry in reases sightly in the lowest x bin.
Using the pion and kaon asymmetries whi h ontain additional information about
the sea quarks seems to have the biggest in uen e in the low x region as expe ted.
From the result with the new de�nition of purity (the open square), we an see that
��u � �
�
d is almost zero, but slightly positive. However, statisti pre ision is not
enough yet to really �nd a signi� ant di�eren e between the two approa hes.
For both the distribution of ��u+�
�
d and ��u��
�
d, it might be possible that the
asymmetries of pion and kaon ause these di�eren e of the sea quark distribution
between the two extra tion methods in the low x region. However, the more number
of asymmetries and high statisti s are ne essary to obtain a urate sea distributions,
in parti ular for pions and kaons.
6.1. RESULTS FOR THE POLARIZED QUARK DISTRIBUTIONS 97
-1-0.5
00.5
11.5
22.5
33.5
∆u+∆
u–mixing matrixnew definition of purity
-3-2.5
-2-1.5
-1-0.5
00.5
1
∆d+∆
d–
-3
-2
-1
0
1
2
3
∆u–+∆
d–
10-2
10-1
-3
-2
-1
0
1
2
3
x
∆u–-∆
d–
Figure 6.7: The extra ted quark distributions in luding avor asymmetry
of light sea quarks with the new assumptions based on CQSM
(Se .5.3.2) from experimental data, whi h is obtained at Q
2
=
2.5 GeV
2
. The solid ir les show the results with the mixing ma-
trix using in lusive and harged hadron asymmetries. The open
squares is that with the new de�nition of purities using in lusive,
harged hadron, harged pion and harged kaon asymmetries.
98 CHAPTER 6. ANALYSIS AND RESULTS
-0.2
0
0.2
0.4
0.6x
(∆u
+∆u–
)mixing matrixnew definition of purity
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
x (∆
d+∆
d–)
-0.4
-0.2
0
0.2
0.4
x (∆
u–+∆
d–)
10-2
10-1
-0.4
-0.2
0
0.2
0.4
x
x (∆
u–-∆
d–)
Figure 6.8: The quark distributions times x from the avor asymmetry ex-
tra tion, whi h is obtained at Q
2
= 2.5 GeV
2
. The solid lines
express the positivity limits whi h are unpolarized PDF, where
CTEQ4LQ parameterizations are shown. The two results with
the mixing matrix and the new de�nition of purity are obtained.
6.1. RESULTS FOR THE POLARIZED QUARK DISTRIBUTIONS 99
6.1.4 Single Quark Distribution
From the results of avor asymmetry extra tion, whi h is shown in the previous
se tion (Se .6.1.3), the quark distribution for ea h single quark avor an be ex-
tra ted (see Se .5.3.2). The results are shown in Fig.6.9. For the valen e quarks,
�u and �d, there is almost no di�eren e between the two extra tion methods. The
moments of the both results are onsistent, whi h is shown in Se .6.2.
It makes lear that the absolute value of �u is positive in the measured x region
and that of �d is negative. in the same x region. From Fig.6.10, the d quark distri-
bution times x ex eeds the positivity limit in the se ond highest x bin. Therefore
we an say that the se ond point of �d in Fig.6.9 is a little o� the statisti al u -
tuation. The statisti al deviation of the data in this x bin an already be seen in
the asymmetries on all targets. The systemati un ertainty needs to be evaluated
in the next step. Even with this un ertainty, the distribution of u and d quarks an
be extra ted stable.
The distribution of ��u is almost zero over the measured x region. The om-
parison of the two methods shows that there is almost no di�eren e in any x bin.
The distributions of �
�
d and ��s seem to be a little di�erent for the two extra tion
methods, espe ially there might be di�eren e in low x region, whi h shows that the
di�eren e in ��u ��
�
d and ��u + �
�
d between the methods is totally driven by the
di�eren e in �
�
d. It is also lear that the ontribution of the
�
d quark dominates the
avor asymmetry, ��u � �
�
d, sin e ��u itself is almost zero in all the measured x
region.
There is another point to mention on erning the ��s quark distribution. In this
study, it is assumed that the distributions of �s and ��s are equal. Hen e the result
for the ��s distribution also shows that of �s quarks. From the results with the new
de�nition of purity (the open square), we an see that it is slightly negative in all
x bins (negle ting the three bins of high x). The moment of �s quark distribution
shows this e�e t learly, whi h is shown in Se .6.2. Fig.6.10 shows that no data
points, within their errors, lie outside the positivity limit.
There are some previous determinations of the distribution of ��s [69℄ before our
analysis. Some results show slight negative values of ��s distribution and moments.
In the present study, it also shows the slight negative value of the x dependen e and
moments whi h are des ribed in Se .6.2. With the two methods, the mixing matrix
method and the new de�nition of purity, I obtained the result onsistent with the
previous studies. The omparison between this results and the theoreti al predi tion
are shown in the following se tion (Se .6.2.3).
100 CHAPTER 6. ANALYSIS AND RESULTS
-1-0.5
00.5
11.5
22.5
33.5
4∆u mixing matrixnew definition of purity
-3-2.5
-2-1.5
-1-0.5
00.5
1∆d
-2-1.5
-1-0.5
00.5
11.5
2∆u–
-2-1.5
-1-0.5
00.5
11.5
2∆d–
10-2
10-1
-2-1.5
-1-0.5
00.5
11.5
2
x
∆s–
Figure 6.9: The extra ted single distributions from the results of the a-
vor asymmetry, whi h is obtained at Q
2
= 2.5 GeV
2
. The solid
ir les show the results with the mixing matrix using in lusive
and harged hadron asymmetries. The open squares is that with
the new de�nition of purities using in lusive, harged hadron,
harged pion and harged kaon asymmetries.
6.1. RESULTS FOR THE POLARIZED QUARK DISTRIBUTIONS 101
-0.1
00.1
0.2
0.30.4
0.5
0.6
x ∆u mixing matrix
new definition of purity
-0.2
-0.1
0
0.1
0.2
x ∆d
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
x ∆u
–
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
x ∆d
–
10-2
10-1
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
x
x ∆s
–
Figure 6.10: The single quark distributions times x, whi h is obtained at Q
2
= 2.5 GeV
2
. The solid lines express the positivity limits whi h
are unpolarized PDF, where CTEQ4LQ parameterizations are
shown. The two results with the mixing matrix and the new
de�nition of purities are obtained.
102 CHAPTER 6. ANALYSIS AND RESULTS
6.1.5 Results for the di�erent Purities
In the analysis des ribed above, I used the purities whi h are produ ed from CTEQ4LQ
[53℄ unpolarized PDF. However, it is important to he k the stability of the result
for extra ted quark polarization. Therefore, I tested with the value of the purities
whi h is produ ed by several di�erent unpolarized PDF, su h as CTEQ4LQ and
GRV. Then quark polarization is determined by using two purities based on CTEQ
and GRV. Fig.6.11 shows the result of the avor de omposition with the mixing
matrix by the CTEQ4LQ and the GRV purities. There is almost no deviation
whi h omes from the di�eren e of PDF be ause the used purities by CTEQ4LQ
and GRV are onsistent, espe ially for the valen e quarks. This omparison tells us
that the produ ed purities don't give an signi� ant ontribution to the systemati
un ertainty. This is also lear for Fig.6.12, whi h shows that the avor asymmetry
of light sea quarks does not depend on the purities used.
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
(∆u
+∆u–
)/(u
+u–)
mixing matrix (CTEQ)mixing matrix (GRV)
-1
-0.8-0.6
-0.4
-0.2
00.2
0.4
0.6
0.81
(∆d
+∆d–
)/(d
+d–)
10-2
10-1
-1-0.8
-0.6
-0.4
-0.20
0.2
0.4
0.60.8
1
x
∆qs/
qs
Figure 6.11: Comparison of the avor de omposition using purities from
CTEQ4LQ and GRV. The solid ir les are results using pu-
rities produ ed from CTEQ4LQ. The open ir les are that from
GRV.
6.1. RESULTS FOR THE POLARIZED QUARK DISTRIBUTIONS 103
-1-0.5
00.5
11.5
22.5
33.5
∆u+∆
u–
mixing matrix (CTEQ)mixing matrix (GRV)
-3-2.5
-2-1.5
-1-0.5
00.5
1
∆d+∆
d–
-3
-2
-1
0
1
2
3
∆u–+∆
d–
10-2
10-1
-3
-2
-1
0
1
2
3
x
∆u–-∆
d–
Figure 6.12: Comparison of the avor asymmetry extra tion using purities
from CTEQ4LQ and GRV. The solid ir les are results using
purities produ ed from CTEQ4LQ. The open ir les are that
from GRV.
104 CHAPTER 6. ANALYSIS AND RESULTS
6.2 Moments of Polarized Quark Distributions
In the previous se tions, the results of the extra tion of the quark polarizations and
distributions are shown for ea h extra tion type. To determine the moments of the
extra ted quark distributions, they are �tted with parameterized fun tions. The
used fun tional forms are as follows:
�q(x) = N
q
x
�
q
� q(x); (6.1)
�q
s
(x) = N
q
s
x
�
q
s
(1� x)
�
q
s
� q
s
(x); (6.2)
where q is the valen e quark : u; u
v
; d; d
v
, and q
s
is the sea quark : �u;
�
d; s; �s. These
forms are for the GRSV parameterizations [68℄. N
q
, �
q
and �
q
are �t parameters.
The CTEQ4 PDF at Q
2
0
= 2:5 GeV
2
, whi h is the Q
2
mean value at HERMES, are
used for the unpolarized PDF in Eq.6.1 and Eq.6.2.
The n-th moment is de�ned as:
�
(n)
q
f
(Q
2
0
) �
Z
1
0
x
n�1
�q
f
(x;Q
2
0
)dx: (6.3)
Eq.6.3 is integral over the full x range, 0 < x < 1, though the experimental data are
available only in a limited x region whi h is 0.023 < x < 0.6 at HERMES.
The pra ti al al ulation of the moments is done as follows. There is a little
di�eren e between the ase of the valen e and avor de omposition (�rst ase) and
that of avor asymmetry and ea h quark distribution extra tion (se ond ase), be-
ause in the �rst ase, the results are obtained as the quark polarizations �q
f
=q
f
but in the se ond ase, the results are the quark distributions �q
f
.
The �rst moment in the measured x region is obtained as:
�q
f
(Q
2
0
) =
X
i
�q
f
q
f
�
�
�
�
�
i
Z
x
i+1
x
i
q
f
(x;Q
2
0
)dx
!
; (6.4)
where, in the �rst ase, (�q
f
=q
f
)j
i
is the results from the experiment, and in the
se ond ase, it is the value of the results divided by the unpolarized CTEQ4LQ PDF
[53℄. (�q
f
=q
f
)j
i
is onstant within ea h bin (x
i
; x
i+1
) and q
f
(x) is obtained by the
CTEQ4LQ parameterization. Outside the measured region, the �tted parameters
of N
q
, �
q
and �
q
are used to determine the moments. In the high x region of 0.6
< x < 1, extrapolations of the �tted fun tions (Eq.6.1 or Eq.6.2) are required, and
the moments are al ulated as this integrals. In the low x region of 0 < x < 0.023,
there is no lear predi tion to determine polarized quark distributions. Therefore
we applied a simple Regge parameterization [70℄ [71℄ of �q
f
(x) / x
��
with �=0
�tted to the data for x < 0.075. The moments in the low x region are al ulated as
the integral of extrapolations with this Regge parameterization, whi h is onstant
in this study.
To obtain the moments for the sea quarks, there is a point that we should noti e.
As I have mentioned before, almost all sea quark polarizations and distributions
ex eed the positivity limit in the high x region, hen e I negle ted three points in the
high x region to al ulate the moment. Therefore the method of determination of the
moment in high x region, whi h is des ribed above, is only applied to the moment
al ulation for valen e quarks. In this study, the results with the new de�nition of
6.2. MOMENTS OF POLARIZED QUARK DISTRIBUTIONS 105
purities are used to al ulate the moments for extra ted quarks be ause these results
in lude informations from pion and kaon asymmetries.
The parameterized quark distributions and moments are shown in the next se -
tions for the results with the new de�nition of purity, and the omparison with the
theoreti al predi tion will be dis ussed. Only the �tted fun tions by Eq.6.1 are used
to obtain the moments in the high x region, 0.6 < x. Therefore the best �t for
valen e quarks by Eq.6.1 are shown in the following.
6.2.1 The Valen e and Flavor De omposition
The �t results for valen e quarks extra ted from the valen e and avor de omposi-
tion are shown in Fig.6.13, whi h are the results with the new de�nition of purity.
�u + ��u and �d + �
�
d extra ted from the avor de omposition in lude the sea
quarks, although they are �tted by Eq.6.1 be ause u and d quarks are dominant.
For the results with the mixing matrix, the moments an be seen in Appendix.C.5.
And the numeri al values of the �rst moment, extra ted from the results of Fig.6.13
by using Eq.6.4, are shown in Tab.6.1 and Tab.6.2, respe tively. The �tted pa-
rameters, N
q
; �
q
and �
q
whi h are used to al ulate the moments are shown in
Appendix.C.5 (Tab.C.29 and Tab.C.33).
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
x ∆u
v
Best fit
GRSV / (1+R)
10-2
10-1
-0.2
-0.1
0
0.1
0.2
x
x ∆d
v
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
x (∆
u+∆
u–)
Best fit
GRSV / (1+R)
10-2
10-1
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
x
x (∆
d+∆
d–)
Figure 6.13: The polarized valen e quark distributions whi h are �tted by the
fun tion of Eq.6.1. The left �gures are shown the valen e quarks
extra ted by the valen e de omposition. The right �gures are
extra ted by the avor de ompositions. The thi k lines are the
result of a best �t and the thin lines are obtained from GRSV
parameterization orre ted (1 +R).
In Fig.6.13, the thi k line shows the result of a �t with the fun tions of Eq.6.1 and
the thin line represents the GRSV PDF [68℄ whi h is orre ted by the term (1+R) for
the omparison with this results. The �t of the extra ted u quark distributions, x�u
and x(�u+��u) show good agreement with the orre ted GRSV PDF, espe ially the
�u+��u is onsistent in the full x region. For the d quark distributions, x(�d+�
�
d),
whi h is extra ted by the avor de omposition, is onsistent with GRSV, but x�d
v
106 CHAPTER 6. ANALYSIS AND RESULTS
does not agree in the ase of the valen e de omposition. They are still ompatible
with GRSV within the statisti errors.
measured region low-x high-x total integral
�u
v
0.56 � 0.09 0.04 0.02 0.62 � 0.09
�d
v
-0.14 � 0.15 -0.02 0.00 -0.17 � 0.15
�q
s
-0.03 � 0.05 -0.00 0.00 -0.04 � 0.05
Table 6.1: The �rst moment for the valen e de omposition
measured region low-x high-x total integral
�u+��u 0.52 � 0.04 0.04 0.02 0.57 � 0.04
�d+�
�
d -0.21 � 0.07 -0.03 0.00 -0.24 � 0.07
�q
s
-0.03 � 0.05 -0.00 0.00 -0.04 � 0.05
Table 6.2: The �rst moment for the avor de omposition
The moments of the valen e and the avor de omposition results are al ulated
for three x regions, one is the measured x region 0.023 < x < 0.6, others are the
low x and high x region. The moments for ea h region are given by the method
whi h is explained before. The total moment in the region of 0 < x < 1, are
obtained summing up the moments of the three regions. These total moments an
be ompared with the theoreti al predi tions and the result by SMC [72℄. This is
dis ussed in Se .6.2.3.
6.2.2 Flavor Asymmetry and Single Quark Distribution
Here, the �t results for valen e quarks from the avor asymmetry extra tion and the
single quark distribution are shown in Fig.6.14, whi h have been a hieved with the
new de�nition of the purity. For the results with the mixing matrix, the numeri al
values of the �rst moment and the �tted parameters are shown Appendix.C.5. The
�tted parameters for this results also are shown in Appendix.C.5 (Tab.C.37 and
Tab.C.41).
The de�nitions of the urves in Fig.6.14 are the same as for the �gure for the
previous valen e and the avor de omposition. In this ase, the �ts to the u quark
distributions �u and �u + ��u show onsisten y with the GRSV PDF in the full
x region. �d and �d + �
�
d are in agreement with GRSV PDF within the statisti
errors, similarly to the �t for the valen e and avor de omposition.
The al ulated moments for the avor asymmetry and all quark distributions
are shown for three x regions in Tab.6.3 and Tab.6.4. The total moments of �u +
��u and �d + �
�
d show a good onsisten y with the moments extra ted with the
avor de omposition in Tab.6.2. These moments are ompared with the theoreti al
predi tions in the following se tion.
6.2. MOMENTS OF POLARIZED QUARK DISTRIBUTIONS 107
-0.1
0
0.1
0.2
0.3
0.4
0.5
x (∆
u+∆
u–)
Best fit
GRSV / (1+R)
10-2
10-1
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
x
x (∆
d+∆
d–)
-0.1
0
0.1
0.2
0.3
0.4
0.5
x ∆u Best fit
GRSV / (1+R)
10-2
10-1
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
xx
∆d
Figure 6.14: The polarized valen e quark distributions whi h are �tted by
the fun tion of Eq.6.1. The left �gures show the valen e quarks
extra ted by the avor asymmetry extra tion. The right �gures
are results for the single quark distribution. The thi k lines
are the result of a best �t and the thin lines are obtained from
GRSV parameterization orre ted by (1 +R).
measured region low-x high-x total integral
�u+��u 0.53 � 0.05 0.04 0.02 0.59 � 0.05
�d+�
�
d -0.19 � 0.08 -0.03 0.00 -0.22 � 0.08
��u+�
�
d -0.09 � 0.08 -0.01 0.00 -0.10 � 0.08
��u��
�
d 0.09 � 0.09 0.02 0.00 0.11 � 0.09
Table 6.3: The �rst moment for avor asymmetry extra tion
measured region low-x high-x total integral
�u 0.54 � 0.12 0.04 0.02 0.60 � 0.12
�d -0.18 � 0.13 -0.03 0.00 -0.21 � 0.13
��u 0.00 � 0.06 0.00 0.00 0.00 � 0.06
�
�
d -0.09 � 0.06 -0.02 0.00 -0.10 � 0.06
��s -0.06 � 0.04 -0.01 0.00 -0.07 � 0.04
Table 6.4: The �rst moment for single quark distribution
108 CHAPTER 6. ANALYSIS AND RESULTS
6.2.3 Comparison of Results with Predi tions
The obtained moments of quark distributions an be ompared to the theoreti al
predi tions or other experimental results. Here the moment of this study whi h has
been extra ted with the new de�nition of purity is used.
There is one experimental measurement of polarized quark distributions and mo-
ments by SMC [72℄. From this experiment, the results of the valen e de omposition
are available, with the assumption of SU(3) symmetri sea : ��q(x) = ��u(x) =
�
�
d(x) = �s(x) = ��s(x) . It is possible to ompare our results and the SMC results
whi h are extrapolated to Q
2
= 2.5 GeV
2
and integrated over the HERMES x range,
0.023 < x < 0.6. This omparison is shown in Tab.6.5, where the systemati errors
are omitted. Only the statisti al errors are shown. For this study in Tab.6.5, �u
v
and �d
v
are obtained by the results of the valen e de omposition in Tab.6.1, in
addition to ��u and �
�
d are the results from Tab.6.4. There is good agreement with
the SMC results, espe ially for �u
v
. There are some deviations in the sea quark
distribution, ��u and �
�
d, be ause the assumptions on the sea quarks whi h are used
in ea h analysis are signi� antly di�erent.
In the next, the omparison with the theoreti al predi tions is dis ussed. There
are two predi tions whi h an be ompared with our results, whi h are a hieved
by the latti e QCD and the SU(3) analysis within errors. Tab.6.6 shows these
omparisons. Here the values of the latti e QCD are al ulated for Q
2
=5 GeV
2
[73℄.
The present results show a lower value than the predi tion by the latti e QCD. The
al ulation of latti e QCD is performed in the quen hed approximation.
This study SMC
�u
v
0.56 � 0.09 0.59 � 0.08
�d
v
-0.14 � 0.15 -0.33 � 0.11
��u 0.00 � 0.06 0.02 � 0.03
�
�
d -0.09 � 0.06 0.02 � 0.03
Table 6.5: Comparison of the �rst moment between this study and the results
by SMC. These value are for the HERMES measured region, 0.023
< x < 0.6, and obtained at Q
2
= 2.5 GeV
2
This study Latti e QCD
�u
v
0.62 � 0.09 0.84 � 0.05
�d
v
-0.17 � 0.15 -0.25 � 0.02
Table 6.6: Comparison of the �rst moment between this study and the pre-
di tion by the latti e QCD. The theoreti al value by the latti e
QCD is obtained at Q
2
= 5 GeV
2
.
6.2. MOMENTS OF POLARIZED QUARK DISTRIBUTIONS 109
This study SU(3)
1. �u+��u 0.57 � 0.04 0.66 � 0.03
�d+�
�
d -0.24 � 0.07 -0.35 � 0.03
2. �u+��u 0.59 � 0.05 0.66 � 0.03
�d+�
�
d -0.22 � 0.08 -0.35 � 0.03
3. �u+��u 0.60 � 0.13 0.66 � 0.03
�d+�
�
d -0.31 � 0.14 -0.35 � 0.03
�s+��s -0.14 � 0.06 -0.08 � 0.02
Table 6.7: Comparison of the �rst moment between this study and the pre-
di tion by the SU(3) analysis. These theoreti al values are ob-
tained at Q
2
= 2.5 GeV
2
.
For the moments of the avor de omposition results, the predi tion of the SU(3)
analysis an be ompared. In Tab.6.7, the extra tion method is shown in the �rst
olumn. The moments in row No.1 are obtained with the avor de omposition
results (Tab.6.2), and in row No.2 the results from the avor asymmetry extra tion
are shown (Tab.6.3). In row No.3 the moments al ulated from the extra ted results
of the single quark distribution (Tab.6.4) are shown. For row No.3, the moment of
the strange quarks is listed, whi h is obtained from the ��s moment in Tab.6.4.
In the extra tion of single quark distributions, the assumption is used that the
distributions of �s and ��s are equal, hen e �s + ��s = 2 ��s, whi h is used to
al ulate the value in Tab.6.7.
For row No.1 of the avor de omposition results, both moments are onsistent
with the predi tion of SU(3) analysis within errors. The same is true for the moments
extra ted by the avor asymmetry extra tion method. Additionally, the results
obtained by single quark distribution in row No.3 are in good agreement with the
SU(3) predi tions. Espe ially the valen e quarks show the best agreement, and
the moment of the strange quarks shows a signi� antly negative value and this is
onsistent with the value of the predi tion. From these, the sum of the quark spin
ontributions �� = �u +��u +�d +�
�
d + �s +��s = 0.23 � 0.11 in the ase of
No.3.
110 CHAPTER 6. ANALYSIS AND RESULTS
6.3 Flavor Asymmetry Extra tion with Another
Method
In the previous se tion, the avor asymmetry of light sea quarks with the purity
matrix method is extra ted. This avor asymmetry was predi ted by the CQSM.
Besides it was extra ted by another method from the asymmetries of HERMES [9℄
[10℄. In the other method, we don't need to use assumptions on the polarized quark
distributions, and use the relations whi h is based on QPM only (Eq.2.78-Eq.2.80
and Eq.2.86). We estimated the avor asymmetry of light sea quarks with this
method using the asymmetries of HERMES, and ompared it with the results ob-
tained in the previous se tion. Additionally, the dependen e of the avor asymmetry
on the fragmentation fun tions is studied. Only the parameterized fragmentation
fun tions de�ned as the favored one and the unfavored one are available. Therefore
the fragmentation fun tions of strange quarks D
s
are onsidered to be the same as
the unfavored ones, but I estimated the ontribution of D
s
on the avor asymmetry
with the theoreti al model of the fragmentation fun tions in [74℄.
6.3.1 Another Method for Flavor Asymmetry Extra tion
A ording to [9℄ [10℄, the asymmetry of hadron h on the proton target is written by:
A
h
1;p
(x) =
1 +R(x)
1 +
2
�
P
f
e
2
f
n
�q
f
(x)D
h
f
+�q
�
f
(x)D
h
�
f
o
P
f
e
2
f
n
q
f
(x)D
h
f
+ q
�
f
(x)D
h
�
f
o
; (6.5)
where f shows the quark avor : f = fu; d; sg. In Eq.6.5, the avor of quarks and
anti-quarks is shown separately. D
h
f
(x) is the integrated fragmentation fun tion in
the measured region, whi h is D
h
f
=
R
1
z
min
dzD
h
f
(z).
The fragmentation fun tions are lassi�ed in the following:
(D
�
)
+
� D
�
+
u
= D
�
+
�
d
= D
�
�
d
= D
�
�
�u
; (6.6)
(D
�
)
�
� D
�
+
d
= D
�
+
�u
= D
�
�
u
= D
�
�
�
d
= D
�
+
s
= D
�
+
�s
= D
�
�
s
= D
�
�
�s
; (6.7)
(D
K
)
+
� D
K
+
u
= D
K
+
�s
= D
K
�
�u
= D
K
�
s
; (6.8)
(D
K
)
�
� D
K
+
d
= D
K
+
s
= D
K
+
�u
= D
K
+
�
d
= D
K
�
u
= D
K
�
d
= D
K
�
�
d
= D
K
�
�s
; (6.9)
(D
p
)
+
� D
p
u
= D
p
d
= D
�p
�u
= D
�p
�
d
; (6.10)
(D
p
)
�
� D
p
s
= D
p
�u
= D
p
�
d
= D
p
�s
= D
�p
u
= D
�p
d
= D
�p
s
= D
�p
�s
; (6.11)
where (D
h
)
+
and (D
h
)
�
are the favored and the unfavored fragmentation fun tions
of hadron h, respe tively. Returning to Eq.6.5, this relations an be rewritten as:
X
f
e
2
f
n
�q
f
(x)D
h
f
+�q
�
f
(x)D
h
�
f
o
=
1 +
2
1 +R(x)
� A
h
1;p
(x)
2
4
X
f
e
2
f
n
q
f
(x)D
h
f
+ q
�
f
(x)D
h
�
f
o
3
5
� �N
h
p
(x); (6.12)
6.3. FLAVOR ASYMMETRY EXTRACTION WITH ANOTHER METHOD 111
where �N
h
p
(x) refers to the spin-dependent produ tion pro esses of harged hadrons
with the proton target. Similarly, we an get �N
h
n
(x) for the neutron target. Form-
ing a ombination of �N
h
p
(x) and �N
h
n
(x), we obtain the following relation:
��u(x)��
�
d(x) =
�N
h
+
p
(x)��N
h
+
n
(x) + �N
h
�
p
(x)��N
h
�
n
(x)
2I
1
�
�N
h
+
p
(x)��N
h
+
n
(x)��N
h
�
p
(x) + �N
h
�
n
(x)
2I
2
;(6.13)
where I
1
and I
2
are de�ned as the ombination of the fragmentation fun tions:
I
1
= 3(D
�
)
+
+ 4(D
K
)
+
+ 3(D
p
)
+
+ 3(D
�
)
�
+ 2(D
K
)
�
+ 3(D
p
)
�
; (6.14)
I
2
= 5(D
�
)
+
+ 4(D
K
)
+
+ 3(D
p
)
+
� 5(D
�
)
�
� 4(D
K
)
�
� 3(D
p
)
�
: (6.15)
From Eq.6.13, if we have N
h
p
(x) and N
h
n
(x) whi h an be obtained by experi-
mental asymmetries on the respe tive target, the avor asymmetry of light sea the
quarks an be extra ted without the help of MC and the additional assumptions.
Experimentally, I have to use the asymmetry on the D target to obtain �N
h
n
, so
it is ne essary to hange Eq.6.12 for the pro ess on the D target. In ase of the D
target, Eq.6.12 reads as follows:
X
f
e
2
f
n
�q
f
(x)D
h
f
+�q
�
f
(x)D
h
�
f
o
=
1 +
2
(1 + R(x))(1� 1:5 !
D
)
� A
h
1;D
(x)
2
4
X
f
e
2
f
n
q
f
(x)D
h
f
+ q
�
f
(x)D
h
�
f
o
3
5
� �N
h
D
(x); (6.16)
where !
D
is the D-state probability, !
D
= 0.058 [62℄ [63℄. Using Eq.6.16, Eq.6.13 is
rewritten as:
��u(x)��
�
d(x) =
2 �N
h
+
p
(x)��N
h
+
D
(x) + 2 �N
h
�
p
(x)��N
h
�
D
(x)
2I
1
�
2 �N
h
+
p
(x)��N
h
+
D
(x)� 2 �N
h
�
p
(x) + �N
h
�
D
(x)
2I
2
:(6.17)
With the relation Eq.6.17 and the asymmetries of hadrons at HERMES on the pro-
ton and the D target, I an extra t ��u��
�
d dire tly. The detail of the fragmentation
fun tions whi h are used in this analysis is des ribed in the following se tion.
In this method, the Q
2
dependen e of the avor asymmetry have been negle ted
be ause no signi� ant Q
2
dependen e has been observed in the measured spin asym-
metries. Therefore the unpolarized PDF at the �xed Q
2
=4 GeV
2
is used, whi h is
the same value as the previous analysis in [9℄ [10℄.
6.3.2 Theoreti al Model of the Fragmentation Fun tions
For the extra tion of the avor asymmetry in [9℄ [10℄, the fragmentation fun tions
parameterized from the EMC experiment [21℄ are used. These are the favored and
112 CHAPTER 6. ANALYSIS AND RESULTS
the unfavored fragmentation fun tions of pion, kaon and proton : (D
�
)
+
; (D
�
)
�
;
(D
K
)
+
; (D
K
)
�
; (D
p
)
+
; (D
p
)
�
. The fragmentation fun tions of strange quarks are
onsidered to be the same as the unfavored fragmentation fun tions. They are
not obtained from experiment yet be ause the poor statisti s of data. But the
ontributions of strange quark to the fragmentation fun tions should be treated
separately from the unfavored fragmentation fun tions, like in Eq.2.80. Therefore
the spe tator model for the strange quark fragmentation fun tion is used whi h will
be des ribed in the next se tion. In spite of using the strange quark fragmentation
fun tion, Eq.6.14 and Eq.6.15 are not hanged.
The Fragmentation Fun tions in the Spe tator Model
Be ause the fragmentation fun tions of strange quarks are not obtained from the
experiment, I used that from the spe tator model des ribed in [74℄. In this paper,
the model parameters of pion fragmentation are obtained by �tting the pion data of
EMC [21℄. This way, the fragmentation fun tions of the strange quarks are obtained
whi h are di�erent from the unfavored fragmentation fun tions. In this model,
the fragmentation fun tions of pion and kaon are estimated. These fragmentation
fun tions are shown in Fig.6.3.2.
With these parameterized fragmentation fun tions, I studied the avor asym-
metry of light sea quarks from Eq.6.17. To study the ontribution of the strange
quarks to the avor asymmetry, I used several sets of fragmentation fun tions as
input. These sets are shown below:
1. The EMC parameterizations
(D
�
)
+
; (D
�
)
�
; (D
K
)
+
; (D
K
)
�
; (D
p
)
+
; (D
p
)
�
: [ EMC parameterization ℄
2. The Spe tator model parameterizations
(D
p
)
+
; (D
p
)
�
: [ EMC parameterization ℄
(D
�
)
+
; (D
�
)
�
; (D
�
)
�
s
; (D
K
)
+
; (D
K
)
�
; (D
K
)
+
s
; (D
K
)
�
s
: [ Spe tator model ℄
3. The ontribution of the strange quarks
(D
�
)
+
; (D
�
)
�
; (D
K
)
+
; (D
K
)
�
; (D
p
)
+
; (D
p
)
�
: [ EMC parameterization ℄
(D
�
)
�
s
; (D
K
)
+
s
; (D
K
)
�
s
: [ Spe tator model ℄
4. The pion ontributions
(D
K
)
+
; (D
K
)
�
; (D
p
)
+
; (D
p
)
�
: [ EMC parameterization ℄
(D
�
)
+
; (D
�
)
�
: [ Spe tator model ℄
5. The kaon ontributions
(D
�
)
+
; (D
�
)
�
; (D
p
)
+
; (D
p
)
�
: [ EMC parameterization ℄
(D
K
)
+
; (D
K
)
�
: [ Spe tator model ℄
where (D
�
)
s
and (D
K
)
s
are the pion and kaon fragmentation fun tions of the strange
quarks respe tively. Comparing the set 1 and the set 2, we an estimate the frag-
mentation model dependen e of the avor asymmetry, and omparing the set 1 and
the set 3, the dependen e on the fragmentation fun tions of the strange quarks is
obtained. Additionally, omparing the set 1 and set 4(5), we obtain the ontribu-
tions of the pion(kaon) fragmentation fun tions. This analysis is useful to estimate
the systemati error from the fragmentation model.
6.3. FLAVOR ASYMMETRY EXTRACTION WITH ANOTHER METHOD 113
Using the asymmetries on the proton ('97) and the deuteron ('98) and the above
sets of the fragmentation fun tions, the avor asymmetry with Eq.6.17 is studied.
The result are shown in the following and omparison with that from the purity
method with the mixing matrix. It should be noted that the geometri al a eptan e
of HERMES dete tor, whi h ould be z dependent, is not taken into a ount in the
present analysis.
Figure 6.15: Fragmentation fun tions in the Spe tator model. The pion and
kaon fragmentation fun tions of strange quarks are extra ted by
�tting model parameters to EMC data.
114 CHAPTER 6. ANALYSIS AND RESULTS
6.3.3 Results and Comparison with the Purity Matrix Method
When I extra ted the avor asymmetry by solving Eq.6.17, I applied 5 sets of frag-
mentation fun tions, whi h have been already des ribed. The extra ted ��u � �
�
d
distributions are shown in Fig.6.16.
-4
-2
0
2
4
∆u–-∆
d–
EMC parametrizations
-4
-2
0
2
4
∆u–-∆
d–
Spectator model parametrizations
10-2
10-1
-4
-2
0
2
4
x
∆u–-∆
d–
contribution of the strange quarks
-4
-2
0
2
4
∆u–-∆
d–
pion contributions
10-2
10-1
-4
-2
0
2
4
x
∆u–-∆
d–
kaon contributions
Figure 6.16: Flavor asymmetry of light sea quarks extra ted from Eq.6.17.
These �gures are obtained by di�erent sets of fragmentation
fun tions.
In this analysis, only the asymmetries of the two years are used as input. There-
fore the statisti errors be ame larger than the results of the purity methods. From
Fig.6.16, in spite of using di�eren e fragmentation fun tions, these �gures don't
show any di�eren e, but they are very onsistent. From these omparisons, it an
be on luded that the fragmentation fun tions don't have large in uen e on the
avor asymmetry of light sea quarks. It also shows that the systemati errors whi h
ome from the un ertainty on the fragmentation fun tions is not dominant in the
total systemati errors.
In Fig.6.17, the omparison of the result of the purity method with the mixing
matrix based on the in lusive and hadron asymmetries and this dire t extra tion
is shown. From Fig.6.17, the un ertainty from the MC and from the assumptions
on the sea quarks is estimated in the purity method be ause the dire t extra tion
method does not depend on the MC or any assumptions, only on fragmentation
fun tions.
Fig.6.17 indi ates that the avor asymmetry extra ted by the dire t extra tion
shows good agreement with the result of the purity method. For the dire t extra tion
6.3. FLAVOR ASYMMETRY EXTRACTION WITH ANOTHER METHOD 115
10-2
10-1
-3
-2
-1
0
1
2
3
x
∆u–-∆
d–
purity method (mixing matrix)direct extraction (EMC)
10-2
10-1
-0.4
-0.2
0
0.2
0.4
x
x (∆
u–-∆
d–)
purity method (mixing matrix)direct extraction (EMC)
Figure 6.17: Comparison between the dire t extra tion method and the pu-
rity matrix method. The result of the dire t extra tion method
with the EMC parameterization are shown. The upper �gure
is shown as avor asymmetry of light sea quarks extra ted two
di�erent method. The lower �gure is the results times x.
results, the statisti errors are larger than the purity method. Be ause the dire t
extra tion is less statisti s than the purity matrix method. The bottom �gure of
Fig.6.17 shows x times the avor asymmetry. In two bins of the high x, they ex eed
the positivity limit for the result of the dire t extra tion method. Hen e the negle t
of the sea quark distribution in high x bins an be justi�ed by the other analysis.
Comparing the result of the purity method and that of the dire t extra tion
method, the existen e of a avor asymmetry of light sea quarks ��u��
�
d is on�rmed.
It is indi ated that it is positive and there are no large ontributions to the systemati
un ertainties by MC, the fragmentation fun tions and the assumptions on the sea
quark distributions.
116 CHAPTER 6. ANALYSIS AND RESULTS
6.4 Simple Estimation of Systemati Errors
In this study, the a urate extra tion of the systemati un ertainty has been skipped,
although the simple estimations are studied. In this se tion, these estimations are
summarized.
The systemati errors of the extra ted quark polarizations and distributions an
be separated into two parts, one omes from the un ertainty of the experiment and
the other omes from that of the analysis method. The experimental un ertainty
arises by di�erent auses as follows:
� Beam polarization
� Target polarization
� Yield u tuations
� Cross se tion ratio R
� Spin stru ture fun tion g
2
� Smearing orre tions
� Radiative orre tions
� Error on the measured asymmetries
The analysis method un ertainty are given as follows:
� Unpolarized PDF
� Method of using asymmetries on the
3
He and the D target
� Fragmentation Fun tions
In this study, we only dis ussed the analysis method un ertainty. For the un er-
tainty from the unpolarized PDF, we an estimate by omparing the purities gen-
erated with CTEQ4LQ and GRV. Additionally, the results for the di�erent purities
are shown in Fig.6.11 and Fig.6.12. From these omparisons, the un ertainty from
PDF are not observed signi� antly. Some deviations an be seen in the omparison
of the purities. However, the extra ted quark distributions are almost identi al.
The un ertainty from the interpretation of asymmetries on the
3
He and the D
target is already dis ussed. There is almost no in uen e on the systemati un er-
tainty from the di�erent method, the mixing matrix method and the new de�nition
of purity. We an see it omparing the both results with MC data in Fig.6.1 and
Fig.6.2. The deviation annot be observed learly.
In Se .6.3, the un ertainty from the fragmentation models are shown. The re-
sults with the di�erent fragmentation fun tions are almost the same. We an see
that the di�eren e of fragmentation fun tions doesn't have large ontribution to the
systemati un ertainty. This is shown in Fig.6.16. From these observations, it an
be on luded that the systemati un ertainty from the analysis method are small.
The main ontribution to the systemati un ertainty omes from the experimen-
tal un ertainty. Therefore the range of the systemati errors are shown from the
previous analysis [69℄ for referen e. In [69℄, the systemati errors are obtained as:
� for valen e de omposition
0:026 <
�u
v
u
v
< 0:071; (6.18)
0:098 <
�d
v
d
v
< 0:437: (6.19)
6.4. SIMPLE ESTIMATION OF SYSTEMATIC ERRORS 117
� for avor de omposition
0:007 <
�u+��u
u+ �u
< 0:039; (6.20)
0:016 <
�d+�
�
d
d+
�
d
< 0:129: (6.21)
For the present analysis, the systemati un ertainty an be expe ted to be not
so di�erent from these values.
119
Chapter 7
Con lusions and Summary
The main subje t of this thesis is the extra tion of the avor asymmetry of light sea
quarks, whi h is the asymmetry between the polarized �u and
�
d quark distribution :
��u��
�
d. Another subje t is to extra t of the sea quark distribution.
The purity matrix method has been used to extra t the quark distributions,
whi h needs several spin asymmetries from the experiment as input. The quark
distributions have been studied by this method, espe ially for the valen e quarks.
To study the sea quark distributions, more input asymmetries are ne essary. Fortu-
nately at HERMES, there are several in lusive asymmetry, h
+
and h
�
asymmetries
on the proton and helium3 target available. Sin e 1998, a Ring Imaging Cherenkov
Counter repla ed the threshold Cherenkov Counter and more asymmetries be ame
available. In the present study, these pion and kaon asymmetries on the deuteron
target in 1998 have been used.
In this work, the eÆ ien ies of the HERMES RICH have been studied �rst.
This RICH has two Cherenkov radiators, sili a aerogel and C
4
F
10
gas. I estimated
the e�e ts oming from the radiators and dete tor geometries. The results for the
eÆ ien ies have been in luded in the Monte Carlo simulation, whi h is essential
to study the experimental data a urately. It be ame possible to reprodu e the
experimental data by the updated MC mu h better than before. To a hieve a more
a urate analysis of the experimental data, it is ne essary to understand the RICH
performan e in detail. This update is an important ontribution to the extra tion
of the pion and kaon asymmetries in 1998.
Se ondly, using the asymmetries available at HERMES, I extra ted the quark
distributions by the purity matrix method. To use this method, I introdu ed new
assumptions to study the sea quark distributions, whi h are based on the Chiral
Quark Soliton Model. Additionally, I proposed a new method to treat the asym-
metries on the helium3 and the deuteron target. Therefore it be ame possible to
in lude the pion and kaon asymmetries in this analysis. Finally, the valen e and
the sea quark distributions, the avor asymmetry of the light sea quarks and their
moments were obtained. It is on�rmed that �u and �u
v
are positive and �d
shows negative values. For the avor asymmetry, ��u��
�
d is slightly positive and
its moment has also a slight positive value. ��u is almost zero, and both �
�
d and
��s(= �s) show slight negative values. The moments of this results are almost on-
sistent with the theoreti al predi tion and the other experimental results. However,
more statisti s is needed to really larify the sea quark distributions. The present
120 CHAPTER 7. CONCLUSIONS AND SUMMARY
method to extra t the avor asymmetry is a good tool for the study of the nu leon
spin ontent and gave a hint to several quark distributions. Both results from the
mixing matrix method and from the new de�nition of purities are onsistent. The
extra ted results don't depend on the extra tion methods. It has been on�rmed
that the avor asymmetry is not sensitive to the un ertainties whi h ome from
the parameterizations of the unpolarized quark distributions and the fragmentation
fun tions. The main on lusions of this thesis are summarized below:
1. An Analysis method to extra t avor asymmetry of light sea quarks ��u��
�
d
is proposed and examined. The single quark distributions �u;�d; ��u;�
�
d;
�s = ��s are extra ted at the same time.
2. Flavor asymmetry of light sea quarks is extra ted, whi h shows almost zero or
slightly positive. Their �rst moments are also slight positive values.
3. Single quark distributions are obtained. Espe ially the x dependen e of ��u;�
�
d;
and �s = ��s are extra ted. ��u is almost zero while �
�
d and �s = ��s and
their moments are slightly negative.
4. It is found that data of the pion and kaon asymmetries are important to study
of the polarized sea quark distribution.
5. There are small systemati un ertainties whi h ome from the unpolarized
PDF, the extra tion methods and the fragmentation fun tions.
As an outlook to future the following omments an be made. Be ause the data
has been taken with RICH sin e 1998, asymmetries with more statisti s will be ome
available in the near future, espe ially for the pion and kaon asymmetries. These
pion and kaon asymmetries are sensitive to the sea quarks. Therefore it is expe ted
that an a urate study of the sea quark distributions be omes possible. Also the
avor asymmetry of light sea quarks will be determined more pre isely. If enough
di�erent types of asymmetries are obtained, the single quark distributions an be
extra ted without any assumptions on the quark distributions. These studies based
on the experimental data olle ted with the RICH will give more a urate results
to the polarized quark distributions and have a large ontribution to solve the spin
problem of the nu leon.
121
Appendix A
Detail of MC Development
The details of MC development is des ribed in following. To normalize the exper-
imental data, the photon olle tion eÆ ien ies are in luded as the position depen-
den e of the mirror re e tivity. The analysis method to obtain the mirror re e tivity
for ea h segment is des ribed in Se .4.3.1. The L
gas
is used for normalization of the
number of �red PMTs. The de�nition and distribution of L
gas
are shown in Fig.A.1.
0
200
400
600
800
1000
30 40 50 60 70 80 90 100
Figure A.1: De�nition and Distribution of L
gas
for MC data. The left �gure
shows the de�nition of L
gas
, whi h is determined as the distan e
between the position that the parti le tra k goes into gas and
the hit point of the tra k on the mirror. The right �gure is the
distribution of L
gas
. The mean value of L
gas
is 73, whi h is used
for the normalization of the parti le tra k length.
122 APPENDIX A. DETAIL OF MC DEVELOPMENT
In Fig.A.2 the ratio between N
gas
(EXP) and N
gas
(MC) is shown, whi h is used to
al ulate the mirror re e tivity for ea h segment as the photon olle tion eÆ ien y.
20
30
40
50
60
70
80
-100 -80 -60 -40 -20 0 20 40 60 80 1000
0.2
0.4
0.6
0.8
1
1.2
x
y
x
y
x
y
-80
-70
-60
-50
-40
-30
-20
-100 -80 -60 -40 -20 0 20 40 60 80 10000.10.20.30.40.50.60.70.80.91
x
y
x
y
x
y
Figure A.2: The ratio between N
gas
(EXP) and N
gas
(MC). The parti les are
�
+
and �
�
with momentum of 4.5 < p < 8.5 (GeV/ ). This ratio
gives the informations of eÆ ien ies from the unknown fa tors,
for example, the position dependen e of the mirror re e tivity,
the mirror alignment, the slope of the in ident photon, the e�e t
from funnels, the PMT eÆ ien ies.
123
The mean value of the ratio whi h is shown in Fig.A.2 for ea h mirror segment
al ulated as Fig.A.3. These values are obtained in order to reprodu e the exper-
imental data. They an be onsidered as the photon olle tion eÆ ien ies. These
values are in luded in MC as the mirror re e tivities for ea h segment. The mirror
re e tivities and alignments, the funnels and other unknown fa tors an ontribute
to these values.
20
30
40
50
60
70
80
90
-100 -75 -50 -25 0 25 50 75 100
0.42± 0.06
0.78± 0.03
0.78± 0.03
0.66± 0.04
0.57± 0.05
0.67± 0.04
0.65± 0.04
0.58± 0.04
-90
-80
-70
-60
-50
-40
-30
-20
-100 -75 -50 -25 0 25 50 75 100
0.59± 0.04
0.76± 0.03
0.60± 0.04
0.55± 0.05
0.51± 0.05
0.63± 0.04
0.71± 0.04
0.45± 0.06
Figure A.3: The position dependen e of upper and lower mirror re e tivity.
These values in lude not only the mirror re e tivities but also
all other eÆ ien ies from unknown fa tors.
124 APPENDIX A. DETAIL OF MC DEVELOPMENT
The omparison between the experimental data and MC data after modi� ation
is shown in Se .4.3.2. We an see that the new MC gives a good reprodu tion om-
paring before modi� ation. The latest omparison of the Cherenkov angle between
the experimental data and the MC give a better result as shown in Fig.A.4.
0
0.01
0.02
0.03
0.04
0.05
0.2 0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.28 0.29 0.3Θ [rad ]
0
0.01
0.02
0.03
0.04
0.05
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1Θ [rad ]
Figure A.4: Re ent Comparison of the Cherenkov angles between the experi-
mental data (points) and the MC data (histogram). They shows
the normalized distributions of re onstru ted angles for single,
low ba kground ele trons (p > 5). The left �gure shows a om-
parison of the Cherenkov angle from aerogel, and the right �gure
shows that from gas.
Fig.A.4 shows a good agreement in the entral part of the distribution. The tails
of the �gure for aerogel angles are due to several e�e ts from the ba kground. The
auses of ba kgrounds are dis ussed in [44℄ and are in luding in MC.
125
Appendix B
Detail of Analysis
B.1 Covarian e Matrix V
A
As the number of dete ted parti les is Poisson distributed, the varian e of the in-
lusive leptons �
2
N
e
+
is al ulated as:
�
2
N
e
+
= hN
2
e
+
i � hN
e
+
i
2
= N
2
e
+
; (B.1)
where N
e
+
is the number of in lusive leptons. For the semi-in lusive measurement,
the varian e of the semi-in lusive hadrons �
2
N
h
is obtained as:
�
2
N
h
= hN
2
h
i � hN
h
i
2
= N
2
e
+
h(n
h
i
)
2
i = N
h
h(n
h
i
)
2
i
hn
h
i
i
; (B.2)
where n
h
i
is the number of hadrons h per event and N
h
is the number of dete ted
hadrons.
Using the varian e, the statisti al ovarian es Cov(N
h
1
; N
h
2
) for the parti le
ount rates are al ulated as:
Cov(N
h
1
; N
h
2
) � (V
N
)
12
= hN
h
1
N
h
2
i � hN
h
1
ihN
h
2
i (B.3)
= N
e
+
hn
h
1
i
n
h
2
i
i; (B.4)
Cov(N
e
+
; N
h
) � (V
N
)
eh
= N
+
e
n
h
; (B.5)
where n
e
+
i
=1 is used. The orrelation oeÆ ients of the parti le ount rate are :
Cor(N
h
1
; N
h
2
) � �(N
h
1
; N
h
2
) =
(V
N
)
12
�
N
h
1
�
N
h
2
(B.6)
=
hn
h
1
i
n
h
2
i
i
q
h(n
h
1
i
)
2
(n
h
2
i
)
2
i
; (B.7)
Cor(N
e
+
; N
h
) � �(N
e
+
; N
h
) =
n
h
i
q
h(n
h
i
)
2
i
: (B.8)
The ovarian es (V
A
)
ij
of the asymmetries A
h
i
1
and A
h
j
1
are obtained from the
ovarian es (V
N
)
ij
of the parti le ount rates as follows:
126 APPENDIX B. DETAIL OF ANALYSIS
(V
A
)
ij
=
X
�;�
�A
h
i
1
�N
h
�
�A
h
j
1
�N
h
�
(V
N
)
��
=
�A
h
i
1
�N
h
i
+
�A
h
j
1
�N
h
j
+
(V
N
)
ij
+
+
�A
h
i
1
�N
h
i
�
�A
h
j
1
�N
h
j
�
(V
N
)
ij
�
'
1
2
2
4
�A
h
i
1
�N
h
i
+
�A
h
j
1
�N
h
j
+
+
�A
h
i
1
�N
h
i
�
�A
h
j
1
�N
h
j
�
3
5
(V
N
)
ij
'
�A
h
i
1
�N
h
i
�
�A
h
j
1
�N
h
j
�
(V
N
)
ij
; (B.9)
where Greek indi es run over both spin states. In Eq.B.9, two fa ts are used, one is
that the asymmetry for one parti le type only depends on the ount rates for this
parti le in both spin states whi h are denoted as N
h
i
+
and N
h
i
�
, another is that the
two spin states are statisti ally un orrelated su h as (V
N
)
i
+
i
�
. Furthermore two
approximations are used whi h are N
h
i
�
�
1
2
N
h
i
and
�
�
�
�
�A
h
i
1
�N
h
i
+
�
�
�
�
�
�
�
�
�
�A
h
i
1
�N
h
i
�
�
�
�
�
. With these
assumptions, the errors on the asymmetries �
A
h
i
1
be omes:
�
A
h
i
1
2
=
�A
h
i
1
�N
h
i
�
!
2
�
N
h
i
2
; (B.10)
where �
N
h
i
is the varian e of the number of dete ted hadrons N
h
i
.
From Eq.B.10, the ovarian es of the asymmetries (V
A
)
ij
is written as:
(V
A
)
ij
'
�
A
h
i
1
�
A
h
j
1
�
N
h
i
�
N
h
j
(V
N
)
ij
: (B.11)
From the de�nition of orrelations of the di�erent types of asymmetries,
�(A
h
i
1
; A
h
j
1
) � (V
A
)
ij
=(�
A
h
i
1
�
A
h
j
1
) and Eq.B.11, the relation that the orrelations of
asymmetries and parti le ount rates are equal:
�(A
h
i
1
; A
h
j
1
) ' �(N
h
i
; N
h
j
); (B.12)
has been dedu ed.
Therefore in the pra ti al analysis, the orrelations of parti le ount rates are
substitutes for that of the asymmetries. The orrelations of parti le ount rates
whi h are used in this study are shown in the next Appendix.C.3, where the ount
rates are given for in lusive and semi-in lusive DIS events.
B.2. SINGULAR VALUE DECOMPOSITION 127
B.2 Singular Value De omposition
\Singular Value De omposition"(SVD) is the method for solving the matrix equa-
tion (Eq.5.5). This method an obtain the solution whi h minimizes �
2
in Eq.5.9.
However the Cholesky de omposition is ne essary to apply the SVD method. The
Cholesky de omposition is expressed as:
V
�1
A
= L
T
L; (B.13)
where L is a triangular matrix. This de omposition an be applied be ause the
ovarian e matrix V
A
is symmetri and positive. With L, A
0
and P
0
de�ned from
A and P as follows:
A
0
= LA;
P
0
= LP: (B.14)
With the relations Eq.B.14, �
2
is expressed as:
�
2
= (A
0
� C
i
� P
0
Q)
T
(A
0
� C
i
� P
0
Q): (B.15)
The SVD method is applied to Eq.B.15. From a mathemati al theorem, any
m�n matrix P an be written as a produ t of an m�n olumn-orthogonal matrix
U , an m � n diagonal matrix W whi h has positive or zero elements. Therefore P
is expressed with a n� n orthogonal matrix Z as:
P
ij
=
n
X
k=1
U
ik
W
k
Z
jk
; (B.16)
where the diagonal of W is denoted by the ve tor W . From Eq.B.16, the solution
an be obtained as:
Q = Z � [diag(1=w
i
)℄ � (U
T
�A); (B.17)
whi h minimizes �
2
. This method an be applied to the ase in whi h there are
more equations than unknowns without modi� ation.
The the ovarian es between the extra ted quark distributions and the asymme-
tries are obtained as:
V
Q
= C
�1
i
� (P
�1
V
A
(P
T
)
�1
)
= C
�1
i
� (P
T
V
�1
A
P)
�1
(B.18)
= C
�1
i
� (P
0�1
(P
0�1
)
T
);
whi h is obtained by Eq.B.13 and the following relation:
V
Mx
=M V
x
M
T
; (B.19)
where V
Mx
(V
x
) is a ovarian e matrix of Mx(x). And from Eq.B.17,
(V
Q
)
ij
=
n
X
k=1
Z
0
ik
Z
0
jk
W
0
k
: (B.20)
129
Appendix C
Table of Results
C.1 De�nition of the Binning in x
Bin number Range in x hxi
1 0.023 � 0.040 0.033
2 0.040 � 0.055 0.047
3 0.055 � 0.075 0.065
4 0.075 � 0.100 0.087
5 0.100 � 0.140 0.119
6 0.140 � 0.200 0.168
7 0.200 � 0.300 0.244
8 0.300 � 0.400 0.342
9 0.400 � 0.600 0.464
Table C.1: x binning de�nition. The �rst olumn shows the number of ea h
bin, their limits are de�ned in the se ond olumn. The mean
value hxi of x for the in lusive asymmetry on the D target is
shown in the third olumn.
130 APPENDIX C. TABLE OF RESULTS
C.2 Input Asymmetries
C.2.1 1995 Asymmetries on the
3
He Target
Bin x Q
2
A
e
+
1
� stat. � sys.
1 0.033 1:220 -0.036 � 0.013 � 0.004
2 0.048 1:470 -0.009 � 0.014 � 0.004
3 0.065 1:730 -0.028 � 0.015 � 0.004
4 0.087 2:000 -0.025 � 0.018 � 0.004
5 0.118 2:310 -0.034 � 0.019 � 0.004
6 0.166 2:660 -0.038 � 0.023 � 0.005
7 0.239 3:070 -0.006 � 0.030 � 0.008
8 0.338 3:790 0.078 � 0.051 � 0.012
9 0.450 5:250 -0.024 � 0.074 � 0.019
Bin x Q
2
A
h
+
1
� stat. � sys.
1 0.033 1:210 -0.051 � 0.032 � 0.006
2 0.048 1:450 -0.011 � 0.033 � 0.006
3 0.065 1:750 -0.030 � 0.034 � 0.006
4 0.087 2:140 -0.035 � 0.039 � 0.006
5 0.118 2:710 -0.024 � 0.041 � 0.007
6 0.165 3:680 0.006 � 0.049 � 0.008
7 0.238 5:180 -0.155 � 0.067 � 0.009
8 0.337 7:210 -0.092 � 0.138 � 0.013
9 0.447 9:810 -0.110 � 0.289 � 0.020
Bin x Q
2
A
h
�
1
� stat. � sys.
1 0.033 1:210 -0.079 � 0.037 � 0.006
2 0.048 1:450 0.021 � 0.039 � 0.006
3 0.065 1:750 -0.021 � 0.042 � 0.006
4 0.087 2:140 0.027 � 0.049 � 0.006
5 0.118 2:710 0.009 � 0.053 � 0.006
6 0.165 3:680 -0.022 � 0.066 � 0.007
7 0.238 5:180 0.064 � 0.092 � 0.009
8 0.337 7:210 -0.229 � 0.199 � 0.012
9 0.447 9:810 0.308 � 0.648 � 0.019
Table C.2: The 1995 asymmetries on the
3
He target. These data are from
[49℄.
C.2. INPUT ASYMMETRIES 131
C.2.2 1996 Asymmetries on the Proton Target
Bin x Q
2
A
e
+
1
� stat. � sys.
1 0.033 1:210 0.086 � 0.012 � 0.006
2 0.047 1:470 0.097 � 0.015 � 0.008
3 0.065 1:720 0.121 � 0.015 � 0.011
4 0.087 2:000 0.169 � 0.017 � 0.013
5 0.119 2:310 0.214 � 0.019 � 0.018
6 0.168 2:660 0.234 � 0.022 � 0.025
7 0.244 3:080 0.326 � 0.028 � 0.032
8 0.342 3:770 0.375 � 0.048 � 0.039
9 0.465 5:250 0.574 � 0.068 � 0.050
Bin x Q
2
A
h
+
1
� stat. � sys.
1 0.033 1:200 0.083 � 0.031 � 0.008
2 0.047 1:450 0.100 � 0.032 � 0.011
3 0.065 1:750 0.149 � 0.034 � 0.013
4 0.087 2:140 0.172 � 0.036 � 0.016
5 0.118 2:720 0.251 � 0.038 � 0.020
6 0.166 3:680 0.227 � 0.047 � 0.027
7 0.239 5:180 0.480 � 0.064 � 0.032
8 0.338 7:260 0.234 � 0.130 � 0.041
9 0.449 9:810 0.891 � 0.261 � 0.052
Bin x Q
2
A
h
�
1
� stat. � sys.
1 0.033 1:210 0.090 � 0.039 � 0.004
2 0.047 1:460 0.106 � 0.042 � 0.004
3 0.065 1:750 0.064 � 0.043 � 0.006
4 0.087 2:140 0.165 � 0.049 � 0.008
5 0.118 2:700 0.256 � 0.052 � 0.011
6 0.166 3:670 0.284 � 0.068 � 0.017
7 0.239 5:160 0.276 � 0.093 � 0.024
8 0.338 7:230 0.438 � 0.197 � 0.034
9 0.449 9:750 0.108 � 0.413 � 0.050
Table C.3: The 1996 asymmetries on the proton target. These data are from
[49℄.
132 APPENDIX C. TABLE OF RESULTS
C.2.3 1997 Asymmetries on the Proton Target
Bin x Q
2
A
e
+
1
� stat. � sys.
1 0.033 1:210 0.075 � 0.008 � 0.005
2 0.047 1:470 0.107 � 0.009 � 0.007
3 0.065 1:720 0.116 � 0.010 � 0.009
4 0.087 1:990 0.162 � 0.011 � 0.012
5 0.119 2:300 0.191 � 0.011 � 0.015
6 0.168 2:660 0.252 � 0.014 � 0.021
7 0.245 3:060 0.325 � 0.018 � 0.027
8 0.342 3:740 0.486 � 0.030 � 0.034
9 0.465 5:160 0.640 � 0.042 � 0.050
Bin x Q
2
A
h
+
1
� stat. � sys.
1 0.033 1:210 0.089 � 0.018 � 0.006
2 0.047 1:460 0.116 � 0.020 � 0.009
3 0.065 1:750 0.133 � 0.020 � 0.011
4 0.087 2:140 0.185 � 0.022 � 0.014
5 0.118 2:700 0.250 � 0.024 � 0.017
6 0.165 3:670 0.258 � 0.029 � 0.024
7 0.238 5:160 0.400 � 0.040 � 0.028
8 0.339 7:230 0.490 � 0.081 � 0.035
9 0.447 9:750 0.556 � 0.164 � 0.047
Bin x Q
2
A
h
�
1
� stat. � sys.
1 0.033 1:210 0.034 � 0.024 � 0.003
2 0.047 1:460 0.090 � 0.025 � 0.004
3 0.065 1:750 0.067 � 0.026 � 0.005
4 0.087 2:140 0.021 � 0.030 � 0.007
5 0.118 2:700 0.155 � 0.033 � 0.010
6 0.165 3:670 0.166 � 0.041 � 0.015
7 0.238 5:160 0.197 � 0.059 � 0.021
8 0.339 7:230 0.656 � 0.122 � 0.030
9 0.447 9:750 0.120 � 0.249 � 0.045
Table C.4: The 1997 asymmetries on the proton target. These data are from
[49℄.
C.2. INPUT ASYMMETRIES 133
C.2.4 1998 Asymmetries on the D Target
Bin x Q
2
A
e
+
1
� stat. � sys.
1 0.033 1:216 0.010 � 0.010 � 0.001
2 0.047 1:474 0.014 � 0.011 � 0.002
3 0.065 1:727 0.022 � 0.012 � 0.004
4 0.087 2:002 0.037 � 0.014 � 0.007
5 0.119 2:319 0.084 � 0.015 � 0.012
6 0.168 2:682 0.136 � 0.018 � 0.021
7 0.244 3:085 0.186 � 0.023 � 0.037
8 0.342 3:770 0.309 � 0.038 � 0.062
9 0.464 5:215 0.269 � 0.052 � 0.092
Bin x Q
2
A
h
+
1
� stat. � sys.
1 0.033 1:209 -0.009 � 0.026 � 0.001
2 0.048 1:459 0.053 � 0.027 � 0.005
3 0.065 1:772 0.011 � 0.028 � 0.003
4 0.087 2:169 0.036 � 0.032 � 0.006
5 0.118 2:744 0.124 � 0.035 � 0.013
6 0.165 3:728 0.091 � 0.042 � 0.013
7 0.238 5:216 0.269 � 0.058 � 0.028
8 0.337 7:298 0.041 � 0.120 � 0.022
9 0.449 9:974 0.562 � 0.237 � 0.063
Bin x Q
2
A
h
�
1
� stat. � sys.
1 0.033 1:207 0.008 � 0.030 � 0.001
2 0.047 1:467 -0.008 � 0.032 � 0.002
3 0.064 1:779 -0.046 � 0.034 � 0.005
4 0.087 2:186 -0.042 � 0.040 � 0.006
5 0.118 2:787 -0.007 � 0.043 � 0.007
6 0.165 3:742 -0.011 � 0.054 � 0.009
7 0.238 5:226 0.129 � 0.076 � 0.018
8 0.337 7:241 0.029 � 0.160 � 0.021
9 0.449 10:002 0.302 � 0.314 � 0.044
Table C.5: The 1998 asymmetries on the D target. These data are from [50℄.
134 APPENDIX C. TABLE OF RESULTS
Bin x Q
2
A
�
+
1
� stat. � sys.
1 0.033 1:209 -0.024 � 0.030 � 0.002
2 0.047 1:460 0.048 � 0.032 � 0.005
3 0.065 1:775 0.005 � 0.033 � 0.003
4 0.087 2:176 -0.016 � 0.038 � 0.005
5 0.118 2:753 0.156 � 0.041 � 0.016
6 0.165 3:746 0.086 � 0.050 � 0.012
7 0.238 5:215 0.206 � 0.070 � 0.024
8 0.337 7:301 0.032 � 0.145 � 0.021
9 0.449 10:062 0.535 � 0.291 � 0.060
Bin x Q
2
A
�
�
1
� stat. � sys.
1 0.033 1:206 0.020 � 0.032 � 0.002
2 0.047 1:466 -0.001 � 0.035 � 0.002
3 0.064 1:774 -0.038 � 0.037 � 0.005
4 0.087 2:173 -0.045 � 0.043 � 0.006
5 0.118 2:773 -0.009 � 0.047 � 0.007
6 0.165 3:712 -0.016 � 0.059 � 0.010
7 0.237 5:189 0.093 � 0.082 � 0.016
8 0.337 7:210 -0.027 � 0.172 � 0.021
9 0.451 10:032 0.095 � 0.337 � 0.037
Bin x Q
2
A
K
+
1
� stat. � sys.
1 0.034 1:209 0.050 � 0.071 � 0.005
2 0.048 1:457 0.113 � 0.072 � 0.007
3 0.065 1:760 0.066 � 0.075 � 0.006
4 0.087 2:185 0.179 � 0.085 � 0.015
5 0.118 2:856 0.012 � 0.088 � 0.006
6 0.165 3:904 0.111 � 0.105 � 0.014
7 0.239 5:455 0.390 � 0.146 � 0.034
8 0.336 7:480 0.029 � 0.305 � 0.024
9 0.450 9:704 0.530 � 0.617 � 0.070
Bin x Q
2
A
K
�
1
� stat. � sys.
1 0.034 1:212 -0.173 � 0.102 � 0.011
2 0.047 1:463 -0.012 � 0.104 � 0.002
3 0.064 1:776 -0.103 � 0.110 � 0.008
4 0.086 2:242 -0.056 � 0.126 � 0.008
5 0.117 2:881 -0.025 � 0.137 � 0.006
6 0.165 3:942 -0.065 � 0.174 � 0.008
7 0.237 5:450 0.183 � 0.254 � 0.019
8 0.341 7:590 -0.081 � 0.555 � 0.033
9 0.432 9:461 2.757 � 0.960 � 0.217
Table C.6: The 1998 asymmetries on the D target. These data are from [50℄.
C.3. CORRELATION COEFFICIENTS FOR PARTICLE COUNT RATES 135
C.3 Correlation CoeÆ ients for Parti le Count
Rates
Bin �(N
e
+
; N
h
+
) �(N
e
+
; N
h
�
) �(N
h
+
; N
h
�
)
1 0.446 0.395 0.128
2 0.492 0.417 0.137
3 0.507 0.411 0.130
4 0.497 0.386 0.118
5 0.451 0.336 0.105
6 0.364 0.260 0.096
7 0.260 0.177 0.083
8 0.183 0.120 0.067
9 0.127 0.081 0.054
Table C.7: The orrelation oeÆ ients of ount rates for the
3
He target in
1995. These data are from [49℄.
Bin �(N
e
+
; N
h
+
) �(N
e
+
; N
h
�
) �(N
h
+
; N
h
�
)
1 0.452 0.394 0.130
2 0.500 0.414 0.140
3 0.517 0.406 0.134
4 0.509 0.379 0.120
5 0.464 0.328 0.107
6 0.375 0.253 0.098
7 0.267 0.171 0.083
8 0.188 0.115 0.067
9 0.130 0.076 0.051
Table C.8: The orrelation oeÆ ients of ount rates for the proton target in
1997. These data are from [49℄.
136 APPENDIX C. TABLE OF RESULTS
Bin �(N
e
+
; N
h
+
) �(N
e
+
; N
h
�
) �(N
e
+
; N
�
+
)
1 0.374 0.325 0.324
2 0.402 0.344 0.345
3 0.401 0.337 0.343
4 0.385 0.313 0.326
5 0.353 0.281 0.298
6 0.292 0.226 0.244
7 0.215 0.166 0.179
8 0.155 0.119 0.129
9 0.106 0.081 0.085
Bin �(N
e
+
; N
�
�
) �(N
e
+
; N
K
+
) �(N
e
+
; N
K
�
)
1 0.299 0.140 0.097
2 0.316 0.155 0.107
3 0.310 0.157 0.106
4 0.290 0.149 0.097
5 0.261 0.135 0.085
6 0.210 0.113 0.067
7 0.155 0.082 0.047
8 0.110 0.060 0.033
9 0.076 0.043 0.024
Table C.9: The orrelation oeÆ ients of ount rates of in lusive events for
the D target in 1998. These data are from [67℄.
Bin �(N
h
+
; N
h
�
) �(N
h
+
; N
�
+
) �(N
h
+
; N
�
�
) �(N
h
+
; N
K
+
) �(N
h
+
; N
K
�
)
1 0.129 0.868 0.116 0.140 0.097
2 0.129 0.861 0.113 0.155 0.107
3 0.127 0.858 0.111 0.157 0.106
4 0.115 0.852 0.101 0.149 0.097
5 0.102 0.848 0.088 0.135 0.085
6 0.088 0.839 0.078 0.113 0.067
7 0.070 0.837 0.062 0.082 0.047
8 0.051 0.832 0.042 0.060 0.033
9 0.050 0.814 0.039 0.043 0.024
Table C.10: The orrelation oeÆ ients of ount rates of semi-in lusive h
+
for the D target in 1998. These data are from [67℄.
C.3. CORRELATION COEFFICIENTS FOR PARTICLE COUNT RATES 137
Bin �(N
h
�
; N
�
+
) �(N
h
�
; N
�
�
) �(N
h
�
; N
K
+
) �(N
h
�
; N
K
�
)
1 0.115 0.924 0.048 0.297
2 0.112 0.919 0.053 0.313
3 0.111 0.922 0.051 0.313
4 0.102 0.926 0.045 0.309
5 0.090 0.927 0.042 0.304
6 0.077 0.928 0.037 0.296
7 0.062 0.931 0.029 0.287
8 0.048 0.928 0.024 0.279
9 0.041 0.936 0.017 0.282
Table C.11: The orrelation oeÆ ients of ount rates for semi-in lusive h
�
for the D target in 1998. These data are from [67℄.
Bin �(N
�
+
; N
�
�
) �(N
�
+
; N
K
+
) �(N
�
+
; N
K
�
)
1 0.116 0.016 0.020
2 0.111 0.016 0.023
3 0.110 0.017 0.024
4 0.101 0.016 0.023
5 0.088 0.017 0.019
6 0.076 0.018 0.016
7 0.061 0.012 0.013
8 0.045 0.011 0.013
9 0.040 0.016 0.015
Bin �(N
�
�
; N
K
+
) �(N
�
�
; N
K
�
) �(N
K
+
; N
K
�
)
1 0.027 0.010 0.063
2 0.030 0.011 0.067
3 0.028 0.008 0.072
4 0.025 0.010 0.059
5 0.023 0.011 0.057
6 0.023 0.010 0.043
7 0.016 0.009 0.044
8 0.009 0.000 0.051
9 0.009 0.000 0.030
Table C.12: The orrelation oeÆ ients of ount rates for semi-in lusive �
and K for the D target in 1998. These data are from [67℄.
138 APPENDIX C. TABLE OF RESULTS
C.4 Results of Quark Polarizations and Distribu-
tions
C.4.1 Valen e De omposition
Results with the Mixing Matrix
Bin
�u
v
u
v
�d
v
d
v
�q
s
q
s
1 0.184 � 0.104 -0.437 � 0.220 0.029 � 0.082
2 0.201 � 0.075 -0.341 � 0.177 0.024 � 0.081
3 0.292 � 0.060 -0.158 � 0.157 -0.098 � 0.087
4 0.376 � 0.056 -0.106 � 0.166 -0.208 � 0.114
5 0.309 � 0.050 -0.119 � 0.174 -0.116 � 0.153
6 0.345 � 0.048 0.037 � 0.206 -0.311 � 0.259
7 0.367 � 0.038 -0.211 � 0.209 -0.107 � 0.444
8 0.472 � 0.050 0.289 � 0.369 |
9 0.694 � 0.066 -0.996 � 0.665 |
Table C.13: Results of valen e de omposition with the mixing matrix. Only
statisti errors are shown.
Bin
�
�u
v
u
v
;
�d
v
d
v
� �
�u
v
u
v
;
�q
s
q
s
� �
�d
v
d
v
;
�q
s
q
s
�
�
2
=ndf
1 0.863 -0.981 -0.935 0.701
2 0.797 -0.966 -0.911 0.675
3 0.748 -0.952 -0.896 0.500
4 0.745 -0.948 -0.899 1.525
5 0.743 -0.941 -0.906 1.564
6 0.723 -0.926 -0.910 1.583
7 0.510 -0.839 -0.859 1.805
8 0.370 | | 2.421
9 0.353 | | 0.748
Table C.14: Correlations among Results of the valen e de omposition with
the mixing matrix. The last olumn shows the minimum value
of �
2
in Eq.5.9.
C.4. RESULTS OF QUARK POLARIZATIONS AND DISTRIBUTIONS 139
Results with the New De�nition of Purity
Bin
�u
v
u
v
�d
v
d
v
�q
s
q
s
1 0.230 � 0.111 -0.332 � 0.233 -0.010 � 0.088
2 0.193 � 0.078 -0.358 � 0.182 0.033 � 0.085
3 0.304 � 0.060 -0.133 � 0.156 -0.114 � 0.088
4 0.371 � 0.055 -0.134 � 0.159 -0.193 � 0.110
5 0.318 � 0.048 -0.070 � 0.160 -0.153 � 0.143
6 0.344 � 0.045 0.029 � 0.184 -0.301 � 0.233
7 0.373 � 0.037 -0.175 � 0.196 -0.194 � 0.422
8 0.470 � 0.049 0.168 � 0.347 |
9 0.630 � 0.061 -1.517 � 0.579 |
Table C.15: Results of valen e de omposition with the new de�nition of pu-
rity. Only statisti errors are shown.
Bin
�
�u
v
u
v
;
�d
v
d
v
� �
�u
v
u
v
;
�q
s
q
s
� �
�d
v
d
v
;
�q
s
q
s
�
�
2
=ndf
1 0.878 -0.983 -0.943 1.058
2 0.810 -0.969 -0.916 0.656
3 0.749 -0.953 -0.895 0.374
4 0.729 -0.946 -0.890 1.686
5 0.708 -0.935 -0.889 1.651
6 0.673 -0.914 -0.890 1.221
7 0.482 -0.832 -0.847 1.666
8 0.342 | | 1.943
9 0.267 | | 1.275
Table C.16: Correlations among Results of the valen e de omposition with
the new de�nition of purity. The last olumn shows the mini-
mum value of �
2
in Eq.5.9.
140 APPENDIX C. TABLE OF RESULTS
C.4.2 Flavor De omposition
Results with the Mixing Matrix
Bin
�u+��u
u+�u
�d+�
�
d
d+
�
d
�q
s
q
s
1 0.098 � 0.009 -0.114 � 0.025 0.029 � 0.082
2 0.118 � 0.010 -0.109 � 0.028 0.024 � 0.081
3 0.140 � 0.011 -0.123 � 0.030 -0.098 � 0.087
4 0.195 � 0.012 -0.160 � 0.036 -0.208 � 0.114
5 0.212 � 0.013 -0.118 � 0.041 -0.116 � 0.153
6 0.252 � 0.016 -0.100 � 0.053 -0.311 � 0.259
7 0.334 � 0.019 -0.180 � 0.075 -0.107 � 0.444
8 0.416 � 0.033 -0.080 � 0.168 |
9 0.661 � 0.047 -1.173 � 0.341 |
Table C.17: Results of avor de omposition with the mixing matrix. Only
statisti errors are shown.
Bin
�
�u+��u
u+�u
;
�d+�
�
d
d+
�
d
� �
�u+��u
u+�u
;
�q
s
q
s
� �
�d+�
�
d
d+
�
d
;
�q
s
q
s
�
�
2
=ndf
1 -0.778 0.027 -0.238 0.701
2 -0.745 -0.037 -0.254 0.675
3 -0.724 -0.067 -0.267 0.500
4 -0.703 -0.107 -0.286 1.525
5 -0.686 -0.132 -0.303 1.564
6 -0.715 -0.100 -0.234 1.583
7 -0.741 0.094 0.111 1.805
8 -0.636 | | 2.421
9 -0.546 | | 0.748
Table C.18: Correlations among Results of the avor de omposition with the
mixing matrix. The last olumn shows the minimum value of
�
2
in Eq.5.9.
C.4. RESULTS OF QUARK POLARIZATIONS AND DISTRIBUTIONS 141
Results with the New De�nition of Purity
Bin
�u+��u
u+�u
�d+�
�
d
d+
�
d
�q
s
q
s
1 0.097 � 0.009 -0.109 � 0.025 -0.010 � 0.088
2 0.118 � 0.010 -0.109 � 0.028 0.033 � 0.085
3 0.141 � 0.011 -0.122 � 0.030 -0.114 � 0.088
4 0.196 � 0.012 -0.165 � 0.035 -0.193 � 0.110
5 0.211 � 0.013 -0.108 � 0.040 -0.153 � 0.143
6 0.252 � 0.016 -0.101 � 0.052 -0.301 � 0.233
7 0.333 � 0.019 -0.181 � 0.074 -0.194 � 0.422
8 0.424 � 0.032 -0.102 � 0.163 |
9 0.662 � 0.045 -0.866 � 0.318 |
Table C.19: Results of avor de omposition with the new de�nition of purity.
Only statisti errors are shown.
Bin
�
�u+��u
u+�u
;
�d+�
�
d
d+
�
d
� �
�u+��u
u+�u
;
�q
s
q
s
� �
�d+�
�
d
d+
�
d
;
�q
s
q
s
�
�
2
=ndf
1 -0.775 0.034 -0.262 1.058
2 -0.740 -0.050 -0.254 0.656
3 -0.722 -0.093 -0.243 0.374
4 -0.707 -0.136 -0.241 1.686
5 -0.696 -0.170 -0.230 1.651
6 -0.722 -0.131 -0.160 1.221
7 -0.734 0.067 0.147 1.666
8 -0.630 | | 1.943
9 -0.566 | | 1.275
Table C.20: Correlations among Results of the avor de omposition with the
new de�nition of purity. The last olumn shows the minimum
value of �
2
in Eq.5.9.
142 APPENDIX C. TABLE OF RESULTS
C.4.3 Flavor Asymmetry of Light Sea Quarks
Results with the Mixing Matrix
Bin �u+��u �d+�
�
d ��u+�
�
d ��u��
�
d
1 1.982 � 0.327 -2.102 � 0.476 0.149 � 1.264 0.245 � 1.409
2 1.941 � 0.235 -1.337 � 0.369 -0.565 � 0.756 1.307 � 0.909
3 1.675 � 0.185 -1.068 � 0.292 -0.748 � 0.538 0.582 � 0.659
4 1.677 � 0.158 -1.025 � 0.255 -0.553 � 0.455 -0.005 � 0.539
5 1.503 � 0.124 -0.493 � 0.204 -0.557 � 0.349 0.680 � 0.401
6 1.289 � 0.096 -0.256 � 0.160 -0.605 � 0.270 0.706 � 0.335
7 0.931 � 0.073 -0.196 � 0.124 0.125 � 0.149 -0.377 � 0.242
8 0.957 � 0.085 -0.343 � 0.141 | |
9 0.497 � 0.055 -0.138 � 0.105 | |
Table C.21: Results of avor asymmetry extra tion with the mixing matrix.
Only statisti errors are shown.
Bin
�
�u+��u;�d+�
�
d
� �
�u+��u;��u+�
�
d
� �
�u+��u;��u��
�
d
�
1 -0.219 -0.693 0.820
2 -0.284 -0.657 0.757
3 -0.296 -0.659 0.733
4 -0.260 -0.681 0.733
5 -0.256 -0.670 0.711
6 -0.434 -0.589 0.652
7 -0.791 -0.376 0.619
8 -0.842 | |
9 -0.815 | |
Bin
�
�d +�
�
d;��u+�
�
d
� �
�d+�
�
d;��u��
�
d
� �
��u+�
�
d;��u��
�
d
�
�
2
=ndf
1 -0.367 0.156 -0.714 0.786
2 -0.355 0.143 -0.709 0.493
3 -0.353 0.151 -0.743 0.472
4 -0.376 0.176 -0.764 1.685
5 -0.396 0.195 -0.767 1.446
6 -0.272 0.090 -0.763 1.287
7 0.230 -0.389 -0.690 1.723
8 | | | 1.282
9 | | | 0.391
Table C.22: Correlations among Results of the avor asymmetry extra tion
with the mixing matrix. The last olumn shows the minimum
value of �
2
in Eq.5.9.
C.4. RESULTS OF QUARK POLARIZATIONS AND DISTRIBUTIONS 143
Results with the New De�nition of Purity
Bin �u+��u �d+�
�
d ��u+�
�
d ��u��
�
d
1 2.178 � 0.349 -1.897 � 0.484 -0.979 � 1.405 1.202 � 1.520
2 1.911 � 0.237 -1.360 � 0.366 -0.435 � 0.760 1.173 � 0.907
3 1.655 � 0.181 -1.088 � 0.286 -0.676 � 0.496 0.451 � 0.618
4 1.604 � 0.148 -1.096 � 0.245 -0.257 � 0.381 -0.342 � 0.475
5 1.487 � 0.114 -0.495 � 0.193 -0.504 � 0.272 0.635 � 0.340
6 1.234 � 0.089 -0.300 � 0.154 -0.350 � 0.203 0.412 � 0.278
7 0.929 � 0.072 -0.189 � 0.123 0.091 � 0.133 -0.357 � 0.230
8 0.914 � 0.081 -0.273 � 0.132 | |
9 0.465 � 0.052 -0.045 � 0.091 | |
Table C.23: Results of avor asymmetry extra tion with the new de�nition
of purity. Only statisti errors are shown.
Bin
�
�u+��u;�d+�
�
d
� �
�u+��u;��u+�
�
d
� �
�u+��u;��u��
�
d
�
1 -0.137 -0.736 0.843
2 -0.291 -0.664 0.760
3 -0.369 -0.636 0.713
4 -0.401 -0.623 0.686
5 -0.444 -0.593 0.644
6 -0.578 -0.493 0.576
7 -0.795 -0.359 0.616
8 -0.827 | |
9 -0.793 | |
Bin
�
�d +�
�
d;��u+�
�
d
� �
�d+�
�
d;��u��
�
d
� �
��u+�
�
d;��u��
�
d
�
�
2
=ndf
1 -0.405 0.211 -0.760 1.101
2 -0.340 0.129 -0.708 0.546
3 -0.294 0.081 -0.704 0.384
4 -0.276 0.052 -0.685 1.851
5 -0.255 0.024 -0.660 1.514
6 -0.158 -0.055 -0.635 1.190
7 0.251 -0.414 -0.649 1.595
8 | | | 1.397
9 | | | 0.520
Table C.24: Correlations among Results of the avor asymmetry extra tion
with the new de�nition of purity. The last olumn shows the
minimum value of �
2
in Eq.5.9.
144 APPENDIX C. TABLE OF RESULTS
C.4.4 Single Quark Distributions
Results with the Mixing Matrix
Bin �u �d
1 1.785 � 1.002 -2.299 � 1.060
2 1.570 � 0.636 -1.709 � 0.697
3 1.758 � 0.464 -0.986 � 0.516
4 1.956 � 0.387 -0.746 � 0.435
5 1.442 � 0.293 -0.555 � 0.335
6 1.239 � 0.236 -0.306 � 0.268
7 1.057 � 0.160 -0.070 � 0.189
8 0.639 � 0.184 -0.661 � 0.215
9 0.623 � 0.123 -0.012 � 0.151
Bin ��u �
�
d ��s
1 0.197 � 0.947 -0.048 � 0.947 0.035 � 0.671
2 0.371 � 0.591 -0.936 � 0.591 -0.491 � 0.405
3 -0.083 � 0.425 -0.665 � 0.425 -0.467 � 0.289
4 -0.279 � 0.353 -0.274 � 0.353 -0.276 � 0.243
5 0.061 � 0.266 -0.618 � 0.266 -0.387 � 0.186
6 0.051 � 0.215 -0.655 � 0.215 -0.415 � 0.145
7 -0.126 � 0.142 0.251 � 0.142 0.123 � 0.084
8 | | |
9 | | |
Table C.25: Results of single quark distribution with the mixing matrix.
Only statisti errors are shown.
C.4. RESULTS OF QUARK POLARIZATIONS AND DISTRIBUTIONS 145
Results with the New De�nition of Purity
Bin �u �d
1 2.067 � 1.092 -2.008 � 1.142
2 1.541 � 0.637 -1.729 � 0.696
3 1.768 � 0.435 -0.975 � 0.489
4 1.903 � 0.339 -0.797 � 0.391
5 1.421 � 0.246 -0.561 � 0.291
6 1.203 � 0.194 -0.331 � 0.231
7 1.062 � 0.151 -0.056 � 0.181
8 0.668 � 0.164 -0.519 � 0.195
9 0.620 � 0.103 0.109 � 0.127
Bin ��u �
�
d ��s
1 0.112 � 1.035 -1.091 � 1.035 -0.681 � 0.743
2 0.369 � 0.592 -0.804 � 0.592 -0.405 � 0.406
3 -0.112 � 0.396 -0.564 � 0.396 -0.410 � 0.267
4 -0.300 � 0.305 0.042 � 0.305 -0.074 � 0.205
5 0.066 � 0.218 -0.570 � 0.218 -0.353 � 0.147
6 0.031 � 0.172 -0.381 � 0.172 -0.241 � 0.111
7 -0.133 � 0.133 0.224 � 0.133 0.103 � 0.076
8 | | |
9 | | |
Table C.26: Results of single quark distribution with the new de�nition of
purity. Only statisti errors are shown.
146 APPENDIX C. TABLE OF RESULTS
C.5 Moments of Polarized Quark Distributions
C.5.1 Valen e De omposition
Results with the Mixing Matrix
N
q
�
�q
�
2
/ndf
�u
v
0.77 � 0.10 0.42 � 0.08 1.53/9
�d
v
-0.01 � 0.01 -1.24 � 0.28 0.57/9
Table C.27: Fit parameters of the valen e quarks in the valen e de omposi-
tion with the mixing matrix.
measured region low-x high-x total integral
�u
v
0.56 � 0.09 0.04 0.02 0.62 � 0.09
�d
v
-0.14 � 0.16 -0.03 0.00 -0.16 � 0.16
�q
s
-0.03 � 0.05 -0.00 0.00 -0.03 � 0.05
Table C.28: First moment of polarized quark distributions in the valen e
de omposition with the mixing matrix.
Results with the New De�nition of Purity
N
q
�
�q
�
2
/ndf
�u
v
0.71 � 0.09 0.37 � 0.08 1.08/9
�d
v
-0.01 � 0.01 -1.01 � 0.28 1.17/9
Table C.29: Fit parameters of the valen e quarks in the valen e de omposi-
tion with the new de�nition of purity.
measured region low-x high-x total integral
�u
v
0.56 � 0.09 0.04 0.02 0.62 � 0.09
�d
v
-0.14 � 0.15 -0.02 0.00 -0.17 � 0.15
�q
s
-0.03 � 0.05 -0.00 0.00 -0.04 � 0.05
Table C.30: First moment of polarized quark distributions in the valen e
de omposition with the new de�nition of purity.
C.5. MOMENTS OF POLARIZED QUARK DISTRIBUTIONS 147
C.5.2 Flavor De omposition
Results with the Mixing Matrix
N
q
�
�q
�
2
/ndf
�u+��u 0.90 � 0.06 0.67 � 0.03 1.89/9
�d+�
�
d -0.21 � 0.07 0.19 � 0.12 1.53/9
Table C.31: Fit parameters of the valen e quarks in the avor de omposition
with the mixing matrix.
measured region low-x high-x total integral
�u+��u 0.52 � 0.04 0.04 0.02 0.57 � 0.04
�d+�
�
d -0.22 � 0.08 -0.03 0.00 -0.26 � 0.08
�q
s
-0.03 � 0.05 -0.00 0.00 -0.03 � 0.05
Table C.32: First moment of polarized quark distributions in the avor de-
omposition with the mixing matrix.
Results with the New De�nition of Purity
N
q
�
�q
�
2
/ndf
�u+��u 0.92 � 0.06 0.68 � 0.03 1.94/9
�d+�
�
d -0.20 � 0.07 0.19 � 0.12 1.05/9
Table C.33: Fit parameters of the valen e quarks in the avor de omposition
with the new de�nition of purity.
measured region low-x high-x total integral
�u+��u 0.52 � 0.04 0.04 0.02 0.57 � 0.04
�d+�
�
d -0.21 � 0.07 -0.03 0.00 -0.24 � 0.07
�q
s
-0.03 � 0.05 -0.00 0.00 -0.04 � 0.05
Table C.34: First moment of polarized quark distributions in the avor de-
omposition with the new de�nition of purity.
148 APPENDIX C. TABLE OF RESULTS
C.5.3 Flavor Asymmetry of Light Sea Quarks
Results with the Mixing Matrix
N
q
�
�q
�
2
/ndf
�u+��u 0.89 � 0.08 0.63 � 0.05 1.31/9
�d+�
�
d -0.17 � 0.06 0.12 � 0.14 0.81/9
Table C.35: Fit parameters of the valen e quarks in the avor asymmetry
extra tion with the mixing matrix.
measured region low-x high-x total integral
�u+��u 0.54 � 0.05 0.04 0.02 0.60 � 0.05
�d+�
�
d -0.21 � 0.09 -0.03 0.00 -0.25 � 0.09
��u+�
�
d -0.09 � 0.09 -0.01 0.00 -0.11 � 0.09
��u��
�
d 0.10 � 0.10 0.02 0.00 0.12 � 0.10
Table C.36: First moment of polarized quark distributions in the avor asym-
metry extra tion with the mixing matrix.
Results with the New De�nition of Purity
N
q
�
�q
�
2
/ndf
�u+��u 0.83 � 0.08 0.61 � 0.05 1.03/9
�d+�
�
d -0.18 � 0.07 0.16 � 0.14 0.52/9
Table C.37: Fit parameters of the valen e quarks in the avor asymmetry
extra tion with the new de�nition of purity.
measured region low-x high-x total integral
�u+��u 0.53 � 0.05 0.04 0.02 0.59 � 0.05
�d+�
�
d -0.19 � 0.08 -0.03 0.00 -0.22 � 0.08
��u+�
�
d -0.09 � 0.08 -0.01 0.00 -0.10 � 0.08
��u��
�
d 0.09 � 0.09 0.02 0.00 0.11 � 0.09
Table C.38: First moment of polarized quark distributions in the avor asym-
metry extra tion with the new de�nition of purity.
C.5. MOMENTS OF POLARIZED QUARK DISTRIBUTIONS 149
C.5.4 Single Quark Distributions
Results with the Mixing Matrix
N
q
�
�q
�
2
/ndf
�u 0.81 � 0.18 0.53 � 0.12 0.56/9
�d -0.13 � 0.07 -0.09 � 0.21 1.11/9
Table C.39: Fit parameters of the valen e quarks in the single quark distri-
bution with the mixing matrix.
measured region low-x high-x total integral
�u 0.54 � 0.13 0.04 0.02 0.59 � 0.13
�d -0.22 � 0.14 -0.03 0.00 -0.25 � 0.14
��u 0.01 � 0.07 0.00 0.00 0.01 � 0.07
�
�
d -0.10 � 0.07 -0.02 0.00 -0.11 � 0.07
��s -0.06 � 0.05 -0.01 0.00 -0.07 � 0.05
Table C.40: First moment of polarized quark distributions in the single quark
distribution with the mixing matrix.
Results with the New De�nition of Purity
N
q
�
�q
�
2
/ndf
�u 0.83 � 0.17 0.54 � 0.11 0.68/9
�d -0.12 � 0.07 -0.12 � 0.21 0.98/9
Table C.41: Fit parameters of the valen e quarks in the single quark distri-
bution with the new de�nition of purity.
measured region low-x high-x total integral
�u 0.54 � 0.12 0.04 0.02 0.60 � 0.12
�d -0.18 � 0.13 -0.03 0.00 -0.21 � 0.13
��u 0.00 � 0.06 0.00 0.00 0.00 � 0.06
�
�
d -0.09 � 0.06 -0.02 0.00 -0.10 � 0.06
��s -0.06 � 0.04 -0.01 0.00 -0.07 � 0.04
Table C.42: First moment of polarized quark distributions in the single quark
distribution with the new de�nition of purity.
LIST OF FIGURES 151
List of Figures
2.1 The diagram of DIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 De�nition of DIS kinemati variables . . . . . . . . . . . . . . . . . . 5
2.3 De�nition of s attering angles in DIS . . . . . . . . . . . . . . . . . . 5
2.4 DIS Pro ess in Breit-Frame . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Diagram of polarized DIS . . . . . . . . . . . . . . . . . . . . . . . . 11
2.6 The re ent results of g
1
(x) on the proton . . . . . . . . . . . . . . . . 13
2.7 Diagram of semi-in lusive DIS . . . . . . . . . . . . . . . . . . . . . . 16
2.8 Charged hadron spe tra orre ted for dete tor in uen es . . . . . . . 19
3.1 The HERA storage ring . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 The longitudinal polarization of HERA beam . . . . . . . . . . . . . 22
3.3 The internal target of HERMES . . . . . . . . . . . . . . . . . . . . . 23
3.4 The HERMES Spe trometer . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 A three dimensional CAD diagram of the HERMES Spe trometer . . 25
3.6 S hemati view of the HERMES BC module . . . . . . . . . . . . . . 26
3.7 Isometri view of the HERMES Calorimeter . . . . . . . . . . . . . . 27
4.1 The diagram of Cherenkov radiation . . . . . . . . . . . . . . . . . . 30
4.2 S hemati view of Ring Imaging Cherenkov Counter at HERMES . . 31
4.3 The relation between Cherenkov angle and parti le momentum . . . . 32
4.4 Momentum ranges for hadron separation . . . . . . . . . . . . . . . . 32
4.5 Composition of sili a aerogel . . . . . . . . . . . . . . . . . . . . . . . 34
4.6 Composition of sili a aerogel . . . . . . . . . . . . . . . . . . . . . . . 34
4.7 Photo of aerogel tile . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.8 Typi al size of one aerogel tile . . . . . . . . . . . . . . . . . . . . . . 35
4.9 Aerogel wall as radiator in RICH . . . . . . . . . . . . . . . . . . . . 36
4.10 EÆ ien ies for transmittan e though lu ite . . . . . . . . . . . . . . . 36
4.11 Distribution of the refra tive index of 1040 sele ted aerogels . . . . . 38
4.12 The distribution of refra tive index n of aerogel in two groups . . . . 38
4.13 Distributions of the size of the aerogel tiles . . . . . . . . . . . . . . . 40
4.14 Des ription of the shape of aerogel surfa e . . . . . . . . . . . . . . . 41
4.15 The wavelength dependen e of the mirror re e tivity . . . . . . . . . 42
4.16 S hemati pi ture of PMTs with funnels . . . . . . . . . . . . . . . . 43
4.17 The photon olle tion eÆ ien y verses angle of in iden e . . . . . . . 43
4.18 The RICH in the MC . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.19 Comparison of the Cherenkov angles between the experimental and
the MC data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
152 LIST OF FIGURES
4.20 Comparison of the number of �red PMTs between the experimental
data and the MC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.21 The �
x
and �
y
dependen e of the eÆ ien y before MC modi� ation . 53
4.22 The �
x
and �
y
dependen e of the eÆ ien y after MC modi� ation . . 54
4.23 The Cherenkov rings of aerogel and gas on the PMT plane . . . . . . 55
4.24 The mirror re e tivity versus the number of �red PMTs with MC . . 55
5.1 In lusive (e
+
) purities for proton target . . . . . . . . . . . . . . . . . 62
5.2 h
+
purities for proton target . . . . . . . . . . . . . . . . . . . . . . . 63
5.3 h
�
purities for proton target . . . . . . . . . . . . . . . . . . . . . . . 63
5.4 �
+
purities for proton target . . . . . . . . . . . . . . . . . . . . . . . 64
5.5 �
�
purities for proton target . . . . . . . . . . . . . . . . . . . . . . . 64
5.6 K
+
purities for proton target . . . . . . . . . . . . . . . . . . . . . . 65
5.7 K
�
purities for proton target . . . . . . . . . . . . . . . . . . . . . . 65
5.8 In lusive (e
+
) purities for neutron target . . . . . . . . . . . . . . . . 66
5.9 h
+
purities for neutron target . . . . . . . . . . . . . . . . . . . . . . 67
5.10 h
�
purities for neutron target . . . . . . . . . . . . . . . . . . . . . . 67
5.11 �
+
purities for neutron target . . . . . . . . . . . . . . . . . . . . . . 68
5.12 �
�
purities for neutron target . . . . . . . . . . . . . . . . . . . . . . 68
5.13 K
+
purities for neutron target . . . . . . . . . . . . . . . . . . . . . . 69
5.14 K
�
purities for neutron target . . . . . . . . . . . . . . . . . . . . . . 69
5.15 The Q
2
dependen e of the purities for the proton target . . . . . . . . 74
5.16 The Q
2
dependen e of the purities for the neutron target . . . . . . . 75
5.17 The ratio of the multipli ities on the proton and the deuteron target . 81
5.18 The in lusive and semi in lusive asymmetries on the
3
He target . . . 82
5.19 The in lusive and semi in lusive asymmetries on the proton target . . 83
5.20 The in lusive and semi in lusive asymmetries on the D target . . . . 84
6.1 The avor de omposition from MC . . . . . . . . . . . . . . . . . . . 88
6.2 The avor asymmetry extra tion from MC . . . . . . . . . . . . . . . 89
6.3 The extra ted quark polarizations in valen e de omposition from ex-
perimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.4 The quark distributions obtained from the valen e de omposition result 91
6.5 The extra ted quark polarizations in avor de omposition from ex-
perimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.6 The quark distributions obtained from the avor de omposition result 95
6.7 The extra ted quark distributions in luding avor asymmetry of light
sea quarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.8 The quark distributions times x from the avor asymmetry extra tion. 98
6.9 The extra ted single quark distributions. . . . . . . . . . . . . . . . . 100
6.10 The single quark distributions times x from the avor asymmetry
extra tion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.11 Comparison of the avor de omposition using purities from CTEQ4LQ
and GRV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.12 Comparison of the avor asymmetry extra tion using purities from
CTEQ4LQ and GRV . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.13 The �tted polarized valen e quark distributions . . . . . . . . . . . . 105
LIST OF FIGURES 153
6.14 The �tted polarized valen e quark distributions . . . . . . . . . . . . 107
6.15 Fragmentation fun tions in the Spe tator model . . . . . . . . . . . . 113
6.16 Flavor asymmetry of light sea quarks extra ted from Eq.6.17 . . . . . 114
6.17 Comparison between the dire t extra tion method and the purity
matrix method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
A.1 De�nition and Distribution of L
gas
for MC data . . . . . . . . . . . . 121
A.2 The ratio between N
gas
(EXP) and N
gas
(MC) . . . . . . . . . . . . . . 122
A.3 The position dependen e of mirror re e tivity . . . . . . . . . . . . . 123
A.4 Re ent Comparison of the Cherenkov angles between the experimen-
tal and the MC data . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
LIST OF TABLES 155
List of Tables
2.1 De�nition of DIS kinemati al values . . . . . . . . . . . . . . . . . . . 4
2.2 De�nition of semi-in lusive kinemati al values . . . . . . . . . . . . . 16
2.3 Fitted parameters of LUND model for HERMES . . . . . . . . . . . . 18
4.1 Comparison of the Cherenkov angle for ea h MC produ tion and the
experimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2 Comparison of the number of �red PMTs for ea h MC produ tion
and experimental data . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.1 The �rst moment for the valen e de omposition . . . . . . . . . . . . 106
6.2 The �rst moment for the avor de omposition . . . . . . . . . . . . . 106
6.3 The �rst moment for avor asymmetry extra tion . . . . . . . . . . . 107
6.4 The �rst moment for single quark distribution . . . . . . . . . . . . . 107
6.5 Comparison of the �rst moment between this study and the results
by SMC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.6 Comparison of the �rst moment between this study and the predi tion
by the latti e QCD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.7 Comparison of the �rst moment between this study and the predi tion
by the SU(3) analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
C.1 x binning de�nition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
C.2 The 1995 asymmetries on the
3
He target . . . . . . . . . . . . . . . . 130
C.3 The 1996 asymmetries on the proton target . . . . . . . . . . . . . . 131
C.4 The 1997 asymmetries on the proton target . . . . . . . . . . . . . . 132
C.5 The 1998 asymmetries on the D target . . . . . . . . . . . . . . . . . 133
C.6 The 1998 asymmetries on the D target . . . . . . . . . . . . . . . . . 134
C.7 The orrelation oeÆ ients of ount rates for the
3
He target in 1995 . 135
C.8 The orrelation oeÆ ients of ount rates for the proton target in 1997135
C.9 The orrelation oeÆ ients of ount rates of in lusive events for the
D target in 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
C.10 The orrelation oeÆ ients of ount rates of semi-in lusive h
+
for the
D target in 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
C.11 The orrelation oeÆ ients of ount rates for semi-in lusive h
�
for
the D target in 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
C.12 The orrelation oeÆ ients of ount rates for semi-in lusive � and K
for the D target in 1998 . . . . . . . . . . . . . . . . . . . . . . . . . 137
C.13 Results of valen e de omposition with the mixing matrix . . . . . . . 138
156 LIST OF TABLES
C.14 Correlations among Results of the valen e de omposition with the
mixing matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
C.15 Results of valen e de omposition with the new de�nition of purity . . 139
C.16 Correlations among Results of the valen e de omposition with the
new de�nition of purity . . . . . . . . . . . . . . . . . . . . . . . . . . 139
C.17 Results of avor de omposition with the mixing matrix . . . . . . . . 140
C.18 Correlations among Results of the avor de omposition with the mix-
ing matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
C.19 Results of avor de omposition with the new de�nition of purity . . . 141
C.20 Correlations among Results of the avor de omposition with the new
de�nition of purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
C.21 Results of avor asymmetry extra tion with the mixing matrix . . . . 142
C.22 Correlations among Results of the avor asymmetry extra tion with
the mixing matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
C.23 Results of avor asymmetry extra tion with the new de�nition of purity143
C.24 Correlations among Results of the avor asymmetry extra tion with
the new de�nition of purity . . . . . . . . . . . . . . . . . . . . . . . 143
C.25 Results of single quark distribution with the mixing matrix . . . . . . 144
C.26 Results of single quark distribution with the new de�nition of purity . 145
C.27 Fit parameters of the valen e quarks in the valen e de omposition
with the mixing matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 146
C.28 First moment of polarized quark distributions in the valen e de om-
position with the mixing matrix . . . . . . . . . . . . . . . . . . . . . 146
C.29 Fit parameters of the valen e quarks in the valen e de omposition
with the new de�nition of purity . . . . . . . . . . . . . . . . . . . . . 146
C.30 First moment of polarized quark distributions in the valen e de om-
position with the new de�nition of purity . . . . . . . . . . . . . . . . 146
C.31 Fit parameters of the valen e quarks in the avor de omposition with
the mixing matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
C.32 First moment of polarized quark distributions in the avor de ompo-
sition with the mixing matrix . . . . . . . . . . . . . . . . . . . . . . 147
C.33 Fit parameters of the valen e quarks in the avor de omposition with
the new de�nition of purity . . . . . . . . . . . . . . . . . . . . . . . 147
C.34 First moment of polarized quark distributions in the avor de ompo-
sition with the new de�nition of purity . . . . . . . . . . . . . . . . . 147
C.35 Fit parameters of the valen e quarks in the avor asymmetry extra -
tion with the mixing matrix . . . . . . . . . . . . . . . . . . . . . . . 148
C.36 First moment of polarized quark distributions in the avor asymmetry
extra tion with the mixing matrix . . . . . . . . . . . . . . . . . . . . 148
C.37 Fit parameters of the valen e quarks in the avor asymmetry extra -
tion with the new de�nition of purity . . . . . . . . . . . . . . . . . . 148
C.38 First moment of polarized quark distributions in the avor asymmetry
extra tion with the new de�nition of purity . . . . . . . . . . . . . . . 148
C.39 Fit parameters of the valen e quarks in the single quark distribution
with the mixing matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 149
C.40 First moment of polarized quark distributions in the single quark
distribution with the mixing matrix . . . . . . . . . . . . . . . . . . . 149
LIST OF TABLES 157
C.41 Fit parameters of the valen e quarks in the single quark distribution
with the new de�nition of purity . . . . . . . . . . . . . . . . . . . . . 149
C.42 First moment of polarized quark distributions in the single quark
distribution with the new de�nition of purity . . . . . . . . . . . . . . 149
BIBLIOGRAPHY 159
Bibliography
[1℄ J. Ashman et al., An investigation of the spin stru ture of the proton in deep
inelasti s attering of polarized muons on polarized protons, Nu l. Phys. B328
(1989) 1.
[2℄ D. Adams et al., Spin stru ture of the proton from polarized in lusive deep-
inelasti muon proton s attering, Phys. Rev. D56 (1997) 5330{5358.
[3℄ K. A kersta� et al., The avor asymmetry of the light quark sea from semi-
in lusive deep inelasti s attering, Phys. Rev. Lett. 81 (1998) 5519{5523.
[4℄ J. C. Peng et al., anti-d/anti-u asymmetry and the origin of the nu leon sea,
Phys. Rev. D58 (1998) 092004.
[5℄ E. A. Hawker et al., Measurement of the light antiquark avor asymmetry in
the nu leon sea, Phys. Rev. Lett. 80 (1998) 3715{3718.
[6℄ B. Dressler, K. Goeke, M. V. Polyakov, and C. Weiss, Flavor asymmetry of
polarized antiquark distributions and semi-in lusive DIS, Eur. Phys. J. C14
(2000) 147.
[7℄ M. Wakamatsu, Light avor sea-quark asymmetry for the spin-dependent dis-
tribution fun tions of the nu leon, OU-HEP-364 (2000).
[8℄ M. Wakamatsu and T. Watabe, Do we expe t light avor sea-quark asymmetry
also for the spin-dependent distribution fun tions of the nu leon?, Phys. Rev.
D62 (2000) 017506.
[9℄ T. Morii and T. Yamanishi, Polarized light sea quark avor asymmetry in po-
larized semiin lusive pro esses, Phys. Rev. D61 (2000) 057501.
[10℄ T. Morii and T. Yamanishi, Polarized light avor sea-quark asymmetry in po-
larized semi-in lusive pro esses, KOBE-FHD-99-01A (1999).
[11℄ D. Adams et al., Spin asymmetry in muon proton deep inelasti s attering on
a transversely-polarized target, Phys. Lett. B336 (1994) 125{130.
[12℄ K. Abe et al., Measurement of the proton and deuteron spin stru ture fun tion
g2 and asymmetry A2, Phys. Rev. Lett. 76 (1996) 587{591.
[13℄ Kurt Gottfried, Sum rule for high-energy ele tron - proton s attering, Phys.
Rev. Lett. 18 (1967) 1174.
160 BIBLIOGRAPHY
[14℄ M. Arneodo et al., A Reevaluation of the Gottfried sum, Phys. Rev. D50 (1994)
1{3.
[15℄ C. Caso and A. Gurtu, Note on the Mass of the W Boson: in Review of Parti le
Physi s (RPP 1998), Eur. Phys. J. C3 (1998) 223{224.
[16℄ J. D. Bjorken, Inelasti s attering of polarized leptons from polarized nu leons,
Phys. Rev. D1 (1970) 1376{1379.
[17℄ J. D. Bjorken, Appli ations of the hiral u(6) x (6) algebra of urrent densities,
Phys. Rev. 148 (1966) 1467{1478.
[18℄ B. Adeva et al., Spin asymmetries A(1) and stru ture fun tions g1 of the proton
and the deuteron from polarized high energy muon s attering, Phys. Rev. D58
(1998) 112001.
[19℄ John Ellis and Robert Ja�e, A sum rule for deep inelasti ele troprodu tion
from polarized protons, Phys. Rev. D9 (1974) 1444.
[20℄ J. J. Aubert et al., A determination of the fragmentation fun tions of u quarks
into harged pions, Phys. Lett. B160 (1985) 417.
[21℄ M. Arneodo et al., Measurements of the u valen e quark distribution fun tion
in the proton and u quark fragmentation fun tions, Nu l. Phys. B321 (1989)
541.
[22℄ R. D. Field and R. P. Feynman, A parametrization of the properties of quark
jets, Nu l. Phys. B136 (1978) 1.
[23℄ B. Andersson, G. Gustafson, G. Ingelman, and T. Sjostrand, Parton fragmen-
tation and string dynami s, Phys. Rept. 97 (1983) 31.
[24℄ Torbjorn Sjostrand, High-energy physi s event generation with PYTHIA 5.7
and JETSET 7.4, Comput. Phys. Commun. 82 (1994) 74{90.
[25℄ Philipp Geiger, Measurement of Fragmentation Fun tions at HERMES, phD
Thesis.
[26℄ G. Mar hesini et al., HERWIG: A Monte Carlo event generator for simulating
hadron emission rea tions with interfering gluons. Version 5.1 - April 1991,
Comput. Phys. Commun. 67 (1992) 465{508.
[27℄ A. A. Sokolov and I. M. Ternov, On Polarization and spin e�e ts in the theory
of syn hrotron radiation, Phys. Dokl. 8 (1964) 1203{1205.
[28℄ K. A kersta� et al., HERMES spe trometer, Nu l. Instrum. Meth. A417 (1998)
230.
[29℄ S. Bernreuther et al., The HERMES ba k drift hambers, Nu l. Instrum. Meth.
A416 (1998) 45{58.
[30℄ H. Avakian et al., Performan e of the ele tromagneti alorimeter of the HER-
MES experiment, Nu l. Instrum. Meth. A417 (1998) 69.
BIBLIOGRAPHY 161
[31℄ F. M. Menden R. Kaiser and M. C. Vetterli, Probability Analysis of the TRD
Response and the Contribution of Flus Ratios to HERMES Parti le, HERMES
Internal 97-036 (1997).
[32℄ E. Cisbani et al., Proposal for a Dual Radiator RICH for HERMES, HERMES
Internal 97-003 (1997).
[33℄ T. Ypsilantis and J. Seguinot, Theory of ring imaging Cerenkov ounters, Nu l.
Instrum. Meth. A343 (1994) 30{51.
[34℄ H. Yokogawa and M. Yokoyama, Hydrophobi sili a aerogels, Journal of Non-
Crystalline Solids 186 (1995) 23.
[35℄ I. Ada hi et al., Study of a threshold Cerenkov ounter based on sili a aerogels
with low refra tive indi es, Nu l. Instrum. Meth. A355 (1995) 390{398.
[36℄ E. As henauer et al., Opti al hara terization of n = 1.03 sili a aerogel used as
radiator in the RICH of HERMES, Nu l. Instrum. Meth. A440 (2000) 338.
[37℄ R. De Leo et al., Ele troni dete tion of fo used Cherenkov rings from aerogel,
Nu l. Instrum. Meth. A401 (1997) 187.
[38℄ Y. Sakemi K. Suetsugu F. Sato T.-A. Shibata J. Kanesaka, L.f. Zhang,Measure-
ments of the Properties of Sili a Aerogel Tiles for HERMES RICH, HERMES
Internal 98-061 (1998).
[39℄ P.W. Carter, The HERMES Experiment: I. Analyzing Powers in Pion Ele tro-
produ tion II. The Aerogel Radiator of the HERMES RICH, HERMES Internal.
[40℄ L.F. Zhang, Ring Imaging Cerenkov Counter with Aerogel Radiator for HER-
MES, HERMES Internal 99-064 (1999).
[41℄ F. Sato, Study of the RICH Mirror RE e tivity - Non-linearlity between the
re e tivity and the PMT hit numbers -, HERMES Internal 00-011 (2000).
[42℄ K. Negodaeva, The PID EÆ ien y of the HERMES RICH dete tor for the IRT
Method, HERMES Internal 00-028 (2000).
[43℄ R. Kaiser, Likelihood Analysis for the RICH IRT Method, HERMES Internal
00-026 (2000).
[44℄ N. Akopov et al., The HERMES Dual-Radiator Ring Imaging Cherenkov De-
te tor, Nu l. Instrum. Meth. in press.
[45℄ D. Adams et al., The spin-dependent stru ture fun tion g1(x) of the deuteron
from polarized deep-inelasti muon s attering, Phys. Lett. B396 (1997) 338{
348.
[46℄ B. Adeva et al., The spin-dependent stru ture fun tion g1(x) of the proton from
polarized deep-inelasti muon s attering, Phys. Lett. B412 (1997) 414{424.
[47℄ J. M. Ni zyporuk and E. E. W. Bruins, Quark avor tagging in polarized
hadroni pro esses, Phys. Rev. D58 (1998) 091501.
162 BIBLIOGRAPHY
[48℄ W. H. Press et al, Numeri al Re ipes in C: The Art of S ienti� Computing,
Se ond Edition, Cambridge University Press.
[49℄ M. Ruh, Measurement of the Polarised Quark Distributions of the Nu leon at
HERMES, phD Thesis.
[50℄ S. Yoneyama, Spin Asymmetry of Kaon Produ tions in polarized deep-Inelasti
S attering on Deuterium Target by HERMES, phD Thesis.
[51℄ L. Mankiewi z, A. S hafer, and M. Veltri, PEPSI: A Monte Carlo generator
for polarized leptoprodu tion, Comput. Phys. Commun. 71 (1992) 305{318.
[52℄ G. Ingelman, A. Edin, and J. Rathsman, LEPTO 6.5 - A Monte Carlo Gen-
erator for Deep Inelasti Lepton-Nu leon S attering, Comput. Phys. Commun.
101 (1997) 108{134.
[53℄ H. L. Lai et al., Improved parton distributions from global analysis of re ent deep
inelasti s attering and in lusive jet data, Phys. Rev. D55 (1997) 1280{1296.
[54℄ M. Glu k, E. Reya, and A. Vogt, Dynami al parton distributions of the proton
and small x physi s, Z. Phys. C67 (1995) 433{448.
[55℄ HERMES Collaboration, The HERMES Monte Carlo, http://www-
hermes.desy.de/hm /.
[56℄ CERN-CN Appli ation Software Group, GEANT Dete tor Des ription and
Simulation Tool, http://www-hermes.desy.de/hm / (1993).
[57℄ M. A. Funk, A measurement of the polarised parton densities of the nu leon in
deep-inelasti s attering at HERMES, DESY-THESIS-1998-017.
[58℄ M. Be kmann, Extra tion of Polarised Quark Distributions of the Nu leon from
Deep Inelasti S attering at the HERMES Experiment, HERMES Internal 00-
020 (2000).
[59℄ H. Kobayashi, Study of Spin Stru ture of the Proton with Hadron Coin iden e
Measurement of Polarized Deep Inelasti S attering, phD Thesis.
[60℄ B. Ehrnsperger and A. S hafer, Reanalysis of hyperon beta de ay data on F/D,
Phys. Lett. B348 (1995) 619{622.
[61℄ C. Cio� degli Atti, S. S opetta, E. Pa e, and G. Salme, Nu lear e�e ts in deep
inelasti s attering of polarized ele trons o� polarized He-3 and the neutron spin
stru ture fun tions, Phys. Rev. C48 (1993) 968{972.
[62℄ B. Desplanques, Deuteron d state probability and energy dependent N N inter-
a tions, Phys. Lett. B203 (1988) 200.
[63℄ K. Abe et al.,Measurements of the proton and deuteron spin stru ture fun tions
g1 and g2, Phys. Rev. D58 (1998) 112003.
[64℄ P. Amaudruz et al., The ratio F2(n) / F2(p) in deep inelasti muon s attering,
Nu l. Phys. B371 (1992) 3{31.
BIBLIOGRAPHY 163
[65℄ T. Lindemann, Private Communi ation (2001).
[66℄ M. Be kmann, Private Communi ation (2001).
[67℄ S. Yoneyama, Private Communi ation (2001).
[68℄ M. Glu k, E. Reya, M. Stratmann, and W. Vogelsang, Next-to-leading order
radiative parton model analysis of polarized deep inelasti lepton - nu leon s at-
tering, Phys. Rev. D53 (1996) 4775{4786.
[69℄ K. A kersta� et al., Flavor de omposition of the polarized quark distributions in
the nu leon from in lusive and semi-in lusive deep- inelasti s attering, Phys.
Lett. B464 (1999) 123{134.
[70℄ R. L. Heimann, Spin dependent high frequen y inelasti ele tron s attering and
heli ity ip ouplings, Nu l. Phys. B64 (1973) 429{463.
[71℄ John Ellis and Marek Karliner, An analysis of the angular momentum of the
proton, Phys. Lett. B213 (1988) 73.
[72℄ B. Adeva et al., Polarised quark distributions in the nu leon from semi- in lusive
spin asymmetries, Phys. Lett. B420 (1998) 180{190.
[73℄ M. Go keler et al., A latti e determination of the se ond moment of the polarised
valen e quark distribution, Phys. Lett. B414 (1997) 340{346.
[74℄ H. Kitagawa and Y. Sakemi, Pion and Kaon Fragmentation Fun tions in the
Spe tator Model, Prog. Theor. Phys. in press.
165
A knowledgements
During my study at the Tokyo Institute of Te hnology, I was helped by a lot of
people. I'd like to express my thanks to all people who supported me.
First of all, I would like to thank Prof. Toshi-Aki Shibata who gave me a han e
to study the spin physi s at HERMES. I am indebted to Dr. Yasuhiro Sakemi for
his valuable advi e and suggestions during in my entire analysis. I wish to express
my thanks to Dr. Stefan Bernreuther for reading the entire text in its original form
and giving helpful omments. My thanks are also due to the members of Tokyo
group with whom I worked. Espe ially, Dr. Hideyuki Kobayashi gave me many
advi e, whi h was helpful for me. Thanks are due to Mr. Shigenobu Yoneyama who
provide me with his asymmetries data on the D target in 1998. I'd like to thank Mr.
Kentaro Suetsugu who analyzed the RICH performan e. His analysis and advi e
have been helpful for my study. I am indebted to all members in my laboratory.
Additionally, I thank the members of the HERMES Collaboration. Parti ularly,
the members of the RICH group and the �q analysis group gave me valuable sug-
gestions and riti isms. Thanks to all of them.