V
OLUME
80, N
UMBER
11
PHYSICAL REVIEW LETTERS
16 M
ARCH
1998
Search for Inclusive
b
!
sl
1
l
2
S. Glenn,
1
S. D. Johnson,
1
Y. Kwon,
1,
* S. Roberts,
1
E. H. Thorndike,
1
C. P. Jessop,
2
K. Lingel,
2
H. Marsiske,
2
M. L. Perl,
2
V. Savinov,
2
D. Ugolini,
2
R. Wang,
2
X. Zhou,
2
T. E. Coan,
3
V. Fadeyev,
3
I. Korolkov,
3
Y. Maravin,
3
I. Narsky,
3
V. Shelkov,
3
J. Staeck,
3
R. Stroynowski,
3
I. Volobouev,
3
J. Ye,
3
M. Artuso,
4
A. Efimov,
4
M. Goldberg,
4
D. He,
4
S. Kopp,
4
G. C. Moneti,
4
R. Mountain,
4
S. Schuh,
4
T. Skwarnicki,
4
S. Stone,
4
G. Viehhauser,
4
X. Xing,
4
J. Bartelt,
5
S. E. Csorna,
5
V. Jain,
5,
†
K. W. McLean,
5
S. Marka,
5
R. Godang,
6
K. Kinoshita,
6
I. C. Lai,
6
P. Pomianowski,
6
S. Schrenk,
6
G. Bonvicini,
7
D. Cinabro,
7
R. Greene,
7
L. P. Perera,
7
G. J. Zhou,
7
B. Barish,
8
M. Chadha,
8
S. Chan,
8
G. Eigen,
8
J. S. Miller,
8
C. O’Grady,
8
M. Schmidtler,
8
J. Urheim,
8
A. J. Weinstein,
8
F. Würthwein,
8
D. W. Blis,
9
G. Masek,
9
H. P. Paar,
9
S. Prell,
9
V. Sharma,
9
D. M. Asner,
10
J. Gronberg,
10
T. S. Hill,
10
D. J. Lange,
10
S. Menary,
10
R. J. Morrison,
10
H. N. Nelson,
10
T. K. Nelson,
10
C. Qiao,
10
J. D. Richman,
10
D. Roberts,
10
A. Ryd,
10
M. S. Witherell,
10
R. Balest,
11
B. H. Behrens,
11
W. T. Ford,
11
H. Park,
11
J. Roy,
11
J. G. Smith,
11
J. P. Alexander,
12
C. Bebek,
12
B. E. Berger,
12
K. Berkelman,
12
K. Bloom,
12
D. G. Cassel,
12
H. A. Cho,
12
D. S. Crowcroft,
12
M. Dickson,
12
P. S. Drell,
12
K. M. Ecklund,
12
R. Ehrlich,
12
A. D. Foland,
12
P. Gaidarev,
12
L. Gibbons,
12
B. Gittelman,
12
S. W. Gray,
12
D. L. Hartill,
12
B. K. Heltsley,
12
P. I. Hopman,
12
J. Kandaswamy,
12
P. C. Kim,
12
D. L. Kreinick,
12
T. Lee,
12
Y. Liu,
12
N. B. Mistry,
12
C. R. Ng,
12
E. Nordberg,
12
M. Ogg,
12,
‡
J. R. Patterson,
12
D. Peterson,
12
D. Riley,
12
A. Soffer,
12
B. Valant-Spaight,
12
C. Ward,
12
M. Athanas,
13
P. Avery,
13
C. D. Jones,
13
M. Lohner,
13
C. Prescott,
13
J. Yelton,
13
J. Zheng,
13
G. Brandenburg,
14
R. A. Briere,
14
A. Ershov,
14
Y. S. Gao,
14
D. Y.-J. Kim,
14
R. Wilson,
14
H. Yamamoto,
14
T. E. Browder,
15
Y. Li,
15
J. L. Rodriguez,
15
T. Bergfeld,
16
B. I. Eisenstein,
16
J. Ernst,
16
G. E. Gladding,
16
G. D. Gollin,
16
R. M. Hans,
16
E. Johnson,
16
I. Karliner,
16
M. A. Marsh,
16
M. Palmer,
16
M. Selen,
16
J. J. Thaler,
16
K. W. Edwards,
17
A. Bellerive,
18
R. Janicek,
18
D. B. MacFarlane,
18
P. M. Patel,
18
A. J. Sadoff,
19
R. Ammar,
20
P. Baringer,
20
A. Bean,
20
D. Besson,
20
D. Coppage,
20
C. Darling,
20
R. Davis,
20
N. Hancock,
20
S. Kotov,
20
I. Kravchenko,
20
N. Kwak,
20
S. Anderson,
21
Y. Kubota,
21
S. J. Lee,
21
J. J. O’Neill,
21
S. Patton,
21
R. Poling,
21
T. Riehle,
21
A. Smith,
21
M. S. Alam,
22
S. B. Athar,
22
Z. Ling,
22
A. H. Mahmood,
22
H. Severini,
22
S. Timm,
22
F. Wappler,
22
A. Anastassov,
23
J. E. Duboscq,
23
D. Fujino,
23,
§
K. K. Gan,
23
T. Hart,
23
K. Honscheid,
23
H. Kagan,
23
R. Kass,
23
J. Lee,
23
M. B. Spencer,
23
M. Sung,
23
A. Undrus,
23,
k
R. Wanke,
23
A. Wolf,
23
M. M. Zoeller,
23
B. Nemati,
24
S. J. Richichi,
24
W. R. Ross,
24
P. Skubic,
24
M. Bishai,
25
J. Fast,
25
J. W. Hinson,
25
N. Menon,
25
D. H. Miller,
25
E. I. Shibata,
25
I. P. J. Shipsey,
25
and M. Yurko
25
(CLEO Collaboration)
1
University of Rochester, Rochester, New York 14627
2
Stanford Linear Accelerator Center, Stanford University, Stanford, California 94309
3
Southern Methodist University, Dallas, Texas 75275
4
Syracuse University, Syracuse, New York 13244
5
Vanderbilt University, Nashville, Tennessee 37235
6
Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061
7
Wayne State University, Detroit, Michigan 48202
8
California Institute of Technology, Pasadena, California 91125
9
University of California, San Diego, La Jolla, California 92093
10
University of California, Santa Barbara, California 93106
11
University of Colorado, Boulder, Colorado 80309-0390
12
Cornell University, Ithaca, New York 14853
13
University of Florida, Gainesville, Florida 32611
14
Harvard University, Cambridge, Massachusetts 02138
15
University of Hawaii at Manoa, Honolulu, Hawaii 96822
16
University of Illinois, Champaign-Urbana, Illinois 61801
17
Carleton University, Ottawa, Ontario, Canada K1S 5B6 and the Institute of Particle Physics, Canada
18
McGill University, Montréal, Québec, Canada H3A 2T8 and the Institute of Particle Physics, Canada
19
Ithaca College, Ithaca, New York 14850
20
University of Kansas, Lawrence, Kansas 66045
21
University of Minnesota, Minneapolis, Minnesota 55455
22
State University of New York at Albany, Albany, New York 12222
23
Ohio State University, Columbus, Ohio 43210
24
University of Oklahoma, Norman, Oklahoma 73019
0031-9007
y
98
y
80(11)
y
2289(5)$15.00
© 1998 The American Physical Society
2289
V
OLUME
80, N
UMBER
11
PHYSICAL REVIEW LETTERS
16 M
ARCH
1998
25
Purdue University, West Lafayette, Indiana 47907
(
Received 1 October 1997
)
We have searched for the effective flavor changing neutral-current decays
b
!
sl
1
l
2
using
an inclusive method. We set upper limits on the branching ratios
B
s
b
!
se
1
e
2
d
,
5.7
3
10
2
5
,
B
s
b
!
s
m
1
m
2
d
,
5.8
3
10
2
5
, and
B
s
b
!
se
6
m
7
d
,
2.2
3
10
2
5
[at 90% confidence level (C.L.)].
Combing the dielectron and dimuon decay modes we find
B
s
b
!
sl
1
l
2
d
,
4.2
3
10
2
5
(at 90% C.L.).
[S0031-9007(98)05533-1]
PACS numbers: 13.20.He, 11.30.Hv
Flavor changing neutral currents (FCNC) are forbidden
to first order in the standard model. Second order loop
diagrams, known as penguin and box diagrams, can
generate effective FCNC which lead to
b
!
s
transitions.
These processes are of considerable interest because they
are sensitive to
V
ts
, the Cabibbo-Kobayashi-Maskawa
matrix element which will be very difficult to measure in
direct decays of the top quark. These processes are also
sensitive to non-standard-model physics [1], since charged
Higgs bosons, new gauge bosons, or supersymmetric
particles can contribute via additional loop diagrams.
The electromagnetic penguin decay
b
!
s
g
was first
observed by CLEO in the exclusive mode
B
!
K
p
g
with
B
s
B
!
K
p
g
d
≠
s
4.2
6
0.8
6
0.6
d
3
10
2
5
[2].
The inclusive rate for the decay
B
!
X
s
g
was measured
to be
B
s
b
!
s
g
d
≠
s
2.32
6
0.57
6
0.35
d
3
10
2
4
[3].
The measured inclusive
b
!
s
g
rate is consistent with
standard model calculations.
The
b
!
sl
1
l
2
decay rate is expected in the standard
model to be nearly 2 orders of magnitude lower than the
rate for
b
!
s
g
decays. Nevertheless, the
b
!
sl
1
l
2
process has received considerable attention since it offers
a deeper insight into the effective Hamiltonian describing
FCNC processes in
B
decays [4]. While
b
!
s
g
is only
sensitive to the absolute value of the
C
7
Wilson coefficient
in the effective Hamiltonian,
b
!
sl
1
l
2
is also sensitive
to the sign of
C
7
and to the
C
9
and
C
10
coefficients,
where the relative contributions vary with
l
1
l
2
mass.
These three coefficients are related to three different
processes contributing to
b
!
sl
1
l
2
: electromagnetic
and electroweak penguins, and a box diagram. Processes
beyond the standard model can alter both the magnitude
and the sign of the Wilson coefficients. The higher-order
QCD corrections for
b
!
sl
1
l
2
are smaller than for the
electromagnetic penguin and have been calculated in next-
to-leading order [5,6].
Several experiments (UA1 [7], CLEO [8], and CDF
[9]) have searched for the exclusive decays
B
!
Kl
1
l
2
and
B
!
K
p
l
1
l
2
and set upper limits at the level of
s
1
2
2
d
3
10
2
5
at 90% confidence level (C.L.). These
exclusive final states are expected to constitute about 6%
and 15% of the inclusive
X
s
l
1
l
2
rate, respectively [10].
Inclusively measured rates are more interesting because
they can be directly related to underlying quark transi-
tions without large theoretical uncertainties in formation
probabilities for specific hadronic final states. Combin-
ing electron and muon modes, the previous generation
of the CLEO experiment set an inclusive limit:
B
s
b
!
sl
1
l
2
d
,
1.2
3
10
2
3
(90% C.L.) [11]. The UA1 ex-
periment [7] searched for inclusive
b
!
s
m
1
m
2
at the
end point of the dilepton mass distribution
f
M
s
m
1
m
2
d
.
3.9
GeV
g
which comprises about a tenth of the total rate.
Extrapolating to the full phase space, UA1 claims a limit
of
,
5
3
10
2
5
(90% C.L.). However, a simulation of the
UA1 acceptance shows that UA1 overestimated their effi-
ciency by at least a factor of 3 [12].
In this Letter, we present results of the search for
inclusive
b
!
s
m
1
m
2
,
b
!
se
1
e
2
, and
b
!
se
6
m
7
.
The latter decay violates conservation of electron and
muon lepton numbers and thus can originate only from
processes beyond the standard model. The data were
obtained with the CLEO II detector at the Cornell
Electron Storage Ring. A sample with an integrated
luminosity of
3.1
fb
2
1
was collected on the
Y
s
4
S
d
resonance. This sample contains
s
3.30
6
0.06
d
3
10
6
produced
B
̄
B
pairs. For background subtraction we also
use
1.6
fb
2
1
of data collected just below the
Y
s
4
S
d
.
CLEO II is a general purpose solenoidal spectrometer
described in detail in Ref. [13].
The data selection method is very similar to the recon-
struction method presented in our previous measurement
of the
b
!
s
g
rate [3], with the
g
candidate replaced by
a lepton pair. We select events that pass general hadronic
event criteria based on charged track multiplicity, visible
energy, and location of the event vertex. The highest en-
ergy pair of oppositely charged leptons is then selected.
Electron candidates are required to have an energy depo-
sition in the calorimeter nearly equal to the measured mo-
mentum, and to have a specific ionization
s
dE
y
dx
d
in the
drift chamber consistent with that expected for an elec-
tron. Muon candidates are identified as charged tracks
with matching muon-detector hits at absorber depths of
at least three nuclear interaction lengths. In the
m
1
m
2
channel, one muon is required to penetrate at least five
interaction lengths. We then look for a combination of
hadronic particles, denoted
X
s
, with a kaon candidate and
0 – 4 pions, which together with the selected lepton pair
satisfy energy-momentum constraints for the
B
decay hy-
pothesis
B
!
X
s
l
1
l
2
. To quantify consistency with this
hypothesis, we use
x
2
B
≠
μ
M
B
2
5.279
s
M
∂
2
1
μ
E
B
2
E
beam
s
E
∂
2
,
where
M
B
≠
q
E
2
beam
2
P
2
B
,
E
B
,
P
B
are the measured
energy and momentum of the
B
candidate, and
s
M
,
s
E
are experimental errors on
M
B
and
E
B
estimated from the
2290
V
OLUME
80, N
UMBER
11
PHYSICAL REVIEW LETTERS
16 M
ARCH
1998
detector resolution and beam energy spread. The kaon
candidate is a charged track with
dE
y
dx
and time of
flight (TOF) consistent with the kaon hypothesis, or a
K
0
S
!
p
1
p
2
candidate identified by a displaced vertex
and invariant mass cut. A pion candidate is a charged
track with
dE
y
dx
and TOF consistent with the pion
hypothesis, or a
p
0
!
gg
candidate. At most one
p
0
is
allowed in the
X
s
combination. In each event, we pick the
combination that minimizes overall
x
2
, which includes
x
2
B
together with contributions from
dE
y
dx
, TOF, and
K
0
S
and
p
0
mass deviations, where relevant.
To suppress continuum background we require the
event to have
H
2
y
H
0
,
0.45
, where
H
i
are Fox-Wolfram
moments [14]. We also require
j
cos
u
tt
j
,
0.8
, where
u
tt
is the angle between the thrust axis of the candidate
B
and
the thrust axis of the rest of the event. To suppress
B
̄
B
background we require the mass of the
X
s
system to be
less than 1.8 GeV. The dominant
B
̄
B
background comes
from two semileptonic decays of
B
or
D
mesons, which
produce the lepton pair with two undetected neutrinos.
Since most signal events are expected to have zero or
one neutrino, we also require the mass of the undetected
system in the event to be less than 1.5 GeV. By excluding
the mode with
X
s
≠
K
p
1
p
2
p
0
, we reduce the expected
B
̄
B
background by an additional 21% while reducing the
signal efficiency by only 6%.
Figure 1 shows the dilepton mass
M
s
l
1
l
2
d
for the
events which pass the cuts previously described and the
B
consistency requirement
x
2
B
,
6
, in the on- and off-
resonance data samples. Unlike the
b
!
s
g
analysis,
the continuum background is small. The peaks at the
c
and
c
0
masses that are observed in the on-resonance
data are due to well known decays
B
!
X
s
c
s
0
d
,
c
s
0
d
!
l
1
l
2
involving long distance interactions in formation
of the
c
s
0
d
resonances. Using cuts on
M
s
X
s
d
to iden-
tify
K
and
K
p
, we measure the branching ratios for
B
!
K
s
p
d
c
s
0
d
and obtain results consistent with a recent
CLEO publication [15]. For further analysis, we ex-
clude events with
M
s
l
1
l
2
d
near the
c
and
c
0
masses
(
6
0.1
GeV for
m
1
m
2
,
2
0.3
,
1
0.1
GeV for
e
1
e
2
,no
cut for
e
6
m
7
), since we want to probe short distance
contributions to the production of
X
s
l
1
l
2
states. The ex-
clusion region is wider in the
e
1
e
2
channel because of
the radiative tail. After these cuts and continuum sub-
traction, we observe
10
6
5
X
s
e
1
e
2
,
12
6
6
X
s
m
1
m
2
,
and
18
6
8
X
s
e
6
m
7
events in the data, whereas from the
Monte Carlo simulation of generic
B
̄
B
events we expect
9
6
1
,
16
6
2
, and
39
6
3
(statistical errors only) back-
ground events, respectively. The generic
B
̄
B
Monte Carlo
reproduces also the number of events in the tail of the
x
B
distribution
s
6
,x
B
,
30
d
where the signal contribution
is expected to be 2.3 times smaller. Continuum-subtracted
data yield
14
6
6
,
26
6
7
, and
66
6
11
events, whereas
the Monte Carlo expectations are
24
6
2
,
29
6
2
, and
71
6
4
events, respectively. Therefore, no evidence for
signal is found in the data and we proceed to set limits on
these decay rates.
FIG. 1.
M
s
l
1
l
2
d
distributions for the on- (upper) and off-
resonance (lower) data with the
x
2
B
,
6
cut. The scaling factor
between the off- and on-resonance data is 1.9.
To avoid systematics related to absolute normalization
of the
B
̄
B
Monte Carlo, instead of counting events af-
ter the
x
2
B
,
6
cut, we loosen this cut to 30 and fit the
observed
x
2
B
distributions in the on- and off-resonance
data using a binned maximum likelihood method. We
allow for signal contribution, as well as
B
̄
B
and contin-
uum backgrounds. The relative normalization for contin-
uum background between the on- and off-resonance data
is fixed to the known ratio of integral luminosities and
cross sections. The signal is expected to peak sharply at
zero, whereas the backgrounds have flatter distributions
[as an example, we show the expected signal and
B
̄
B
background shapes for the
X
s
e
1
e
2
channel in Fig. 2(a)].
FIG. 2.
x
2
B
distributions for
x
s
e
1
e
2
data. (a) The differ-
ence between the expected distribution for the signal (solid
histogram) and
B
̄
B
background (dashed histogram). Both dis-
tributions are normalized to the same area. (b) The fit to the
on-resonance data (points with error bars). The sum of all fitted
contributions is indicated by a solid line. The fitted background
contribution (
B
̄
B
plus continuum) is indicated by a dashed line.
The estimated continuum background, indicated by a dotted
line, is simultaneously constrained to the off-resonance data (c).
2291
V
OLUME
80, N
UMBER
11
PHYSICAL REVIEW LETTERS
16 M
ARCH
1998
The shapes of all these contributions are fixed from the
Monte Carlo simulation, while the normalizations are
allowed to float. The Monte Carlo predictions for the
signal shape distribution agree well with the distribution
observed in the data for the
B
!
c
X
s
signal. We assign
a generous systematic error to the uncertainty in
x
2
B
signal
and background shapes by varying
X
s
composition in the
Monte Carlo as described below. The fitted number of
X
s
e
1
e
2
,
X
s
m
1
m
2
, and
X
s
e
6
m
7
events is
7
6
7
,
1
6
7
,
and
2
18
6
10
, respectively. As an example, the fit to the
X
s
e
1
e
2
data is displayed in Fig. 2(b) – 2(c).
To calculate the signal efficiency and to predict the
x
2
B
signal distribution we generated
b
!
sl
1
l
2
Monte Carlo
events. The parton level distributions for
b
!
se
1
e
2
and
b
!
s
m
1
m
2
are predicted from the effective Hamilton-
ian containing standard model contributions. The next-to-
leading-order calculations were used [6]. At present, the
effect of gluon bremsstrahlung on the outgoing
s
quark
is only partially included in the theoretical calculations.
After our
c
and
c
0
veto cuts, the long distance inter-
actions are expected to constructively interfere with the
short distance contributions. Estimates of these interfer-
ence effects are model dependent. The most recent calcu-
lation predicts modifications of the short distance rate by
only about 2% [16] compared to 20% predicted by some
earlier simplified models [17]. We neglect long distance
interactions in our Monte Carlo. Since no theoretical cal-
culations for the non-standard-model decay
b
!
se
6
m
7
exist, we use a phase space model for these decays. To
account for Fermi motion of the
b
quark inside the
B
me-
son we have used the spectator model by Ali
et al.
[18].
The particle content of the
X
s
system was modeled
with the conventional method quark hadronization from
JETSET [19]. For better accuracy of the simulations,
when
M
s
X
s
d
is in the
K
or
K
p
mass region, the event
is regenerated according to the theoretical predictions for
the exclusive
B
!
K
s
p
d
l
1
l
2
decays by Greub
et al.
[20].
The estimated efficiencies are 5.2%, 4.5%, and 7.3% for
e
1
e
2
,
m
1
m
2
, and
e
6
m
7
modes, respectively.
To estimate the systematic error due to the uncertainty
in the
x
2
B
signal and background shapes, we divide
the Monte Carlo sample into low and high multiplicity
channels in the manner which produces the largest shape
variation. This shape variation changes the upper limits
by 9%, 19%, and 20% for the
e
1
e
2
,
m
1
m
2
, and
e
6
m
7
channels, respectively. Variations of the spectator model
parameters [3] result in changes of the selection efficiency
by
s
12
6
4
d
%
,
s
30
6
4
d
%
, and
s
11
6
4
d
%
, respectively.
The larger uncertainty in the
m
1
m
2
channel is the result
of the lack of muon identification for
P
m
,
1
GeV
y
c
.
Uncertainty in the modeling of the hadronization of the
X
s
system gives a contribution of 9%. Remaining systematic
error in the simulation of detector response is dominated
by charged tracking systematics and is estimated to be
14%. Adding all these sources of systematic errors in the
quadrature, we estimate the total systematic errors to be
22%, 39%, and 28%, respectively.
Using a Gaussian likelihood integrated over positive
signal values, we find upper limits using statistical errors
only.
We then loosen these limits by one unit of
systematic uncertainty.
The
final
results
are
B
s
b
!
se
1
e
2
d
,
5.7
3
10
2
5
,
B
s
b
!
s
m
1
m
2
d
,
5.8
3
10
2
5
,
and
B
s
b
!
se
6
m
7
d
,
2.2
3
10
2
5
. The results are consistent with
the standard model predictions [18],
s
0.8
6
0.2
d
3
10
2
5
,
s
0.6
6
0.1
d
3
10
2
5
, and 0, respectively. Combining the
e
1
e
2
and
m
1
m
2
results, we also set a limit on the rate
averaged over lepton flavors,
B
s
b
!
sl
1
l
2
d;f
B
s
b
!
se
1
e
2
d
1
B
s
b
!
s
m
1
m
2
dgy
2
,
4.2
3
10
2
5
(90%
C.L.).
The limit on
B
s
b
!
se
1
e
2
d
is more than an order of
magnitude more restrictive than the previous limits. The
limit on
B
s
b
!
s
m
1
m
2
d
is also significantly tighter than
the UA1 limit after correcting for the efficiency problem
(see discussion above). Furthermore, in contrast with
the UA1 measurement, the present analysis is sensitive
to a much wider range of
M
s
l
1
l
2
d
.
Therefore, our
extrapolation to the full phase space is more reliable, and
we are sensitive to a broader range of processes beyond
the standard model.
We gratefully acknowledge the effort of the CESR staff
in providing us with excellent luminosity and running
conditions. This work was supported by the National
Science Foundation, the U.S. Department of Energy, the
Heisenberg Foundation, the Alexander von Humboldt
Stiftung, the Natural Sciences and Engineering Research
Council of Canada, and the A. P. Sloan Foundation.
*Permanent address: Yonsei University, Seoul 120-749,
Korea.
†
Permanent address: Brookhaven National Laboratory,
Upton, NY 11973.
‡
Permanent address: University of Texas, Austin, TX
78712.
§
Permanent
address:
Lawrence
Livermore
National
Laboratory, Livermore, CA 94551.
k
Permanent
address: BINP, RU-630090 Novosibirsk,
Russia.
[1] See, for example, J. L. Hewett and J. D. Wells, Phys. Rev.
D
55
, 5549 (1997); G. Burdman, Phys. Rev. D
52
, 6400
(1995); N. G. Deshpande, K. Panose, and J. Trampetic
́,
Phys. Lett. B
308
, 322 (1993); W. S. Hou, R. S. Willey,
and A. Soni, Phys. Rev. Lett.
58
, 1608 (1987).
[2] CLEO Collaboration, R. Ammar
et al.,
Phys. Rev. Lett.
71
, 674 (1993); in
Proceedings of the 28th International
Conference on High Energy Physics, Warsaw, Poland,
1996
(Report No. CLEO CONF 96-05).
[3] CLEO Collaboration, M. S. Alam
et al.,
Phys. Rev. Lett.
74
, 2885 (1995).
[4] A. Ali, T. Mannel, and T. Morozumi, Phys. Lett. B
273
,
505 (1991); A. Ali, G. F. Gudice, and T. Mannel, Z. Phys.
C
67
, 417 (1995); J. L. Hewett, Phys. Rev. D
53
, 4964
(1996).
[5] M. Misiak, Nucl. Phys.
B393
, 23 (1993);
439
, 461(E)
(1995).
2292
V
OLUME
80, N
UMBER
11
PHYSICAL REVIEW LETTERS
16 M
ARCH
1998
[6] A. Buras and M. Münz, Phys. Rev. D
52
, 186 (1995).
[7] UA1 Collaboration, C. Albajar
et al.,
Phys. Lett. B
262
,
163 (1991).
[8] CLEO Collaboration, R. Balest
et al.,
in
Proceedings
of the 27th International Conference on High Energy
Physics, Glasgow, Scotland, 1994,
edited by P. J.
Bussey and I. G. Knowles (IOP, London, 1995) (Report
No. CLEO CONF 94 -4).
[9] CDF Collaboration, F. Abe
et al.,
Phys. Rev. Lett.
76
,
4675 (1996).
[10] We have estimated a fraction of
Kl
1
l
2
and
K
p
l
1
l
2
final states among all
X
s
l
1
l
2
decays by integrating the
1
y
G
d
G
y
dM
s
X
s
d
distribution predicted by the inclusive
theoretical model described below in the appropriate
M
s
X
s
d
ranges.
[11] CLEO Collaboration, A. Bean
et al.,
Phys. Rev. D
35
,
3533 (1987).
[12] To simulate the UA1 acceptance, we use the
PYTHIA
program [19] to generate
p
̄
p
!
b
̄
bX
at 630 GeV in the
center of mass energy. The simulation of
b
!
s
m
1
m
2
decays is described later in the text. For the kinematic
cuts used by UA1 [
P
t
s
m
d
.
3
GeV
y
c
,
P
t
s
m
1
m
2
d
.
7
GeV
y
c
, and
3.9
,
M
s
m
1
m
2
d
,
4.4
GeV
y
c
2
], we ob-
tain an efficiency of
s
0.36
6
0.03
d
%
when normalized to
b
quark production with
P
t
s
b
d
.
6
GeV
y
c
and
j
y
s
b
d
j
,
1.5
.
Further losses are expected due to inefficiencies
in the trigger and reconstruction. The overall efficiency
quoted in the UA1 publication [7] for these cuts and this
normalization is three times larger:
s
1.1
6
0.3
d
%
. A large
portion of this disagreement can be traced to the simula-
tion of
M
s
m
1
m
2
d
distribution in
B
!
X
s
m
1
m
2
decays.
When applied to
B
0
!
m
1
m
2
, our simulation of the UA1
kinematic cuts gives an acceptance consistent with the
overall efficiency quoted by UA1. Furthermore, while
our simulation of
B
!
X
s
m
1
m
2
decays predict that only
s
6.7
6
0.6
d
%
of events passing the
P
t
cuts fall into the
3.9
,
M
s
m
1
m
2
d
,
4.4
GeV
y
c
2
interval, the number ex-
tracted from Table I in the UA1 paper is 2.3 times larger:
s
15.0
6
0.3
d
%
. The
M
s
m
1
m
2
d
distribution generated by
our Monte Carlo program is in good agreement with the
distribution published by A. Ali
et al.
[18].
[13] CLEO Collaboration, Y. Kubota
et al.,
Nucl. Instrum.
Methods Phys. Res., Sect. A
320
, 66 (1992).
[14] G. Fox and S. Wolfram, Phys. Rev. Lett.
41
, 1581 (1978).
[15] CLEO Collaboration, C. S. Jessop
et al.,
Phys. Rev. Lett.
79
, 4533 (1997).
[16] M. R. Ahmady, Phys. Rev. D
53
, 2843 (1996).
[17] C. S. Lim, T. Morozumi, and A. I. Sanda, Phys. Lett. B
218
, 343 (1989); N. G. Deshpande, J. Trampetic, and K.
Panose, Phys. Rev. D
39
, 1461 (1989); P. J. O’Donnell
and H. K. K. Tung, Phys. Rev. D
43
, R2067 (1991); N.
Paver and Riazuddin, Phys. Rev. D
45
, 978 (1992).
[18] A. Ali, G. Hiller, L. T. Handoko, and T. Morozumi, Phys.
Rev. D
55
, 4105 (1997).
[19] T. Sjostrand, Comput. Phys. Commun.
82
, 74 (1994).
[20] C. Greub, A. Ioannisian, and D. Wyler, Phys. Lett. B
346
,
149 (1995).
2293