V
OLUME
80, N
UMBER
18
PHYSICAL REVIEW LETTERS
4 M
AY
1998
Observation of the Radiative Decay
D
p
1
!
D
1
g
J. Bartelt,
1
S. E. Csorna,
1
V. Jain,
1,
* K. W. McLean,
1
S. Marka,
1
R. Godang,
2
K. Kinoshita,
2
I. C. Lai,
2
P. Pomianowski,
2
S. Schrenk,
21
G. Bonvicini,
3
D. Cinabro,
3
R. Greene,
3
L. P. Perera,
3
G. J. Zhou,
3
B. Barish,
4
M. Chadha,
4
S. Chan,
4
G. Eigen,
4
J. S. Miller,
4
C. O’Grady,
4
M. Schmidtler,
4
J. Urheim,
4
A. J. Weinstein,
4
F. Würthwein,
4
D. W. Bliss,
5
G. Masek,
5
H. P. Paar,
5
S. Prell,
5
V. Sharma,
5
D. M. Asner,
6
J. Gronberg,
6
T. S. Hill,
6
D. J. Lange,
6
R. J. Morrison,
6
H. N. Nelson,
6
T. K. Nelson,
6
J. D. Richman,
6
D. Roberts,
6
A. Ryd,
6
M. S. Witherell,
6
R. Balest,
7
B. H. Behrens,
7
W. T. Ford,
7
A. Gritsan,
7
H. Park,
7
J. Roy,
7
J. G. Smith,
7
J. P. Alexander,
8
C. Bebek,
8
B. E. Berger,
8
K. Berkelman,
8
K. Bloom,
8
V. Boisvert,
8
D. G. Cassel,
8
H. A. Cho,
8
D. S. Crowcroft,
8
M. Dickson,
8
S. von Dombrowski,
8
P. S. Drell,
8
K. M. Ecklund,
8
R. Ehrlich,
8
A. D. Foland,
8
P. Gaidarev,
8
L. Gibbons,
8
B. Gittelman,
8
S. W. Gray,
8
D. L. Hartill,
8
B. K. Heltsley,
8
P. I. Hopman,
8
J. Kandaswamy,
8
P. C. Kim,
8
D. L. Kreinick,
8
T. Lee,
8
Y. Liu,
8
N. B. Mistry,
8
C. R. Ng,
8
E. Nordberg,
8
M. Ogg,
8,
†
J. R. Patterson,
8
D. Peterson,
8
D. Riley,
8
A. Soffer,
8
B. Valant-Spaight,
8
C. Ward,
8
M. Athanas,
9
P. Avery,
9
C. D. Jones,
9
M. Lohner,
9
C. Prescott,
9
J. Yelton,
9
J. Zheng,
9
G. Brandenburg,
10
R. A. Briere,
10
A. Ershov,
10
Y. S. Gao,
10
D. Y.-J. Kim,
10
R. Wilson,
10
H. Yamamoto,
10
T. E. Browder,
11
Y. Li,
11
J. L. Rodriguez,
11
T. Bergfeld,
12
B. I. Eisenstein,
12
J. Ernst,
12
G. E. Gladding,
12
G. D. Gollin,
12
R. M. Hans,
12
E. Johnson,
12
I. Karliner,
12
M. A. Marsh,
12
M. Palmer,
12
M. Selen,
12
J. J. Thaler,
12
K. W. Edwards,
13
A. Bellerive,
14
R. Janicek,
14
D. B. MacFarlane,
14
P. M. Patel,
14
A. J. Sadoff,
15
R. Ammar,
16
P. Baringer,
16
A. Bean,
16
D. Besson,
16
D. Coppage,
16
C. Darling,
16
R. Davis,
16
S. Kotov,
16
I. Kravchenko,
16
N. Kwak,
16
L. Zhou,
16
S. Anderson,
17
Y. Kubota,
17
S. J. Lee,
17
J. J. O’Neill,
17
S. Patton,
17
R. Poling,
17
T. Riehle,
17
A. Smith,
17
M. S. Alam,
18
S. B. Athar,
18
Z. Ling,
18
A. H. Mahmood,
18
H. Severini,
18
S. Timm,
18
F. Wappler,
18
A. Anastassov,
19
J. E. Duboscq,
19
D. Fujino,
19,
‡
K. K. Gan,
19
T. Hart,
19
K. Honscheid,
19
H. Kagan,
19
R. Kass,
19
J. Lee,
19
M. B. Spencer,
19
M. Sung,
19
A. Undrus,
19,
§
R. Wanke,
19
A. Wolf, M. M. Zoeller,
19
B. Nemati,
20
S. J. Richichi,
20
W. R. Ross,
20
P. Skubic,
20
M. Bishai,
21
J. Fast,
21
J. W. Hinson,
21
N. Menon,
21
D. H. Miller,
21
E. I. Shibata,
21
I. P. J. Shipsey,
21
M. Yurko,
21
S. Glenn,
22
S. D. Johnson,
22
Y. Kwon,
22,
k
S. Roberts,
22
E. H. Thorndike,
22
C. P. Jessop,
23
K. Lingel,
23
H. Marsiske,
23
M. L. Perl,
23
V. Savinov,
23
D. Ugolini,
23
R. Wang,
23
X. Zhou,
23
T. E. Coan,
24
V. Fadeyev,
24
I. Korolkov,
24
Y. Maravin,
24
I. Narsky,
24
V. Shelkov,
24
J. Staeck,
24
R. Stroynowski,
24
I. Volobouev,
24
J. Ye,
24
M. Artuso,
25
F. Azfar,
25
A. Efimov,
25
M. Goldberg,
25
D. He,
25
S. Kopp,
25
G. C. Moneti,
25
R. Mountain,
25
S. Schuh,
25
T. Skwarnicki,
25
S. Stone,
25
G. Viehhauser,
25
and X. Xing
25
(CLEO Collaboration)
1
Vanderbilt University, Nashville, Tennessee 37235
2
Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061
3
Wayne State University, Detroit, Michigan 48202
4
California Institute of Technology, Pasadena, California 91125
5
University of California, San Diego, La Jolla, California 92093
6
University of California, Santa Barbara, California 93106
7
University of Colorado, Boulder, Colorado 80309-0390
8
Cornell University, Ithaca, New York 14853
9
University of Florida, Gainesville, Florida 32611
10
Harvard University, Cambridge, Massachusetts 02138
11
University of Hawaii at Manoa, Honolulu, Hawaii 96822
12
University of Illinois, Urbana-Champaign, Illinois 61801
13
Carleton University, Ottawa, Ontario, Canada K1S 5B6
and the Institute of Particle Physics, Canada
14
McGill University, Montréal, Québec, Canada H3A 2T8
and the Institute of Particle Physics, Canada
15
Ithaca College, Ithaca, New York 14850
16
University of Kansas, Lawrence, Kansas 66045
17
University of Minnesota, Minneapolis, Minnesota 55455
18
State University of New York at Albany, Albany, New York 12222
19
Ohio State University, Columbus, Ohio 43210
20
University of Oklahoma, Norman, Oklahoma 73019
21
Purdue University, West Lafayette, Indiana 47907
22
University of Rochester, Rochester, New York 14627
23
Stanford Linear Accelerator Center, Stanford University, Stanford, California 94309
24
Southern Methodist University, Dallas, Texas 75275
25
Syracuse University, Syracuse, New York 13244
(
Received 19 November 1997
)
0031-9007
y
98
y
80(18)
y
3919(5)$15.00
© 1998 The American Physical Society
3919
V
OLUME
80, N
UMBER
18
PHYSICAL REVIEW LETTERS
4 M
AY
1998
We have observed a signal for the decay
D
p
1
!
D
1
g
at a significance of 4 standard deviations.
From the measured branching ratio
B
s
D
p
1
!
D
1
g
dy
B
s
D
p
1
!
D
1
p
0
d
≠
0.055
6
0.014
6
0.010
we find
B
s
D
p
1
!
D
1
g
d
≠
0.017
6
0.004
6
0.003,
where the first uncertainty is statistical and the
second is systematic. We also report the highest precision determination of the remaining
D
p
1
branching fractions.
[S0031-9007(98)05932-8]
PACS numbers: 13.20.Fc, 12.39.Fe, 13.40.Hq, 14.40.Lb
The decays of the excited charmed mesons,
D
p
1
and
D
p
0
, have been the subject of extensive theoretical [1 –
4] as well as experimental [5 – 11] investigation. The
decay of the
D
p
0
via emission of a
p
0
or a photon
has been observed and its branching ratio well measured
[12]. While the
D
p
1
hadronic decays (
D
p
1
!
D
1
p
0
and
D
p
1
!
D
0
p
1
) [13] have been observed and are
widely used to tag heavy quark decays, the observation
of the
D
p
1
radiative decay remained problematic. Both
D
p
mesons decay electromagnetically as the result of a
spin-flip of either the charm quark or the light quark.
In the case of the
D
p
0
, the decay amplitudes for these
two processes interfere constructively. Combined with
the phase space suppression of the hadronic decay, this
interference results in a radiative decay fraction which
competes with the hadronic decay fraction. In the case
of the
D
p
1
, the amplitudes for the two spin-flip processes
interfere destructively. Also, there is slightly more phase
space available for the hadronic decay.
These two
conditions result in a radiative decay fraction of the
D
p
1
which, in comparison to the
D
p
0
, is significantly
suppressed relative to the hadronic decay fraction.
A great deal of interest in the radiative
D
p
1
decay was
generated by an earlier Particle Data Group (PDG) av-
erage of
B
s
D
p
1
!
D
1
g
d
≠
s
18
6
4
d
%
[14]; this value
was virtually impossible to reconcile with theory without
assuming an anomalously large magnetic moment for the
charm quark [4]. Based on
780
pb
2
1
of data, a previous
CLEO II analysis [10] found an upper limit of 4.2% (90%
C.L.) for this branching fraction, a result which strongly
affected not only the
D
p
1
branching fractions but also
many
B
measurements. In addition to its importance in
measuring
B
meson decays, a precision determination of
the
D
p
1
branching fractions will provide an important test
of many quark models and other theoretical approaches to
heavy meson decays [1]. For theories built around chiral
and heavy-quark symmetry (heavy hadron chiral pertur-
bation theory) [2], this measurement will also provide a
strong constraint on the two input parameters (
g
and
b
)
allowing model-independent predictions to be made on a
wide variety of observable quantities [3].
The approach used in this analysis is to search in
the
D
M
g
;
M
s
D
1
g
d
2
M
s
D
1
d
[15] and
D
M
p
;
M
s
D
1
p
0
d
2
M
s
D
1
d
distributions for
D
p
1
events using
the decay chain
D
p
1
!
D
1
(
g
or
p
0
),
D
1
!
K
2
p
1
p
1
.
The branching ratio
R
1
g
;
B
s
D
p
1
!
D
1
g
d
B
s
D
p
1
!
D
1
p
0
d
≠
N
s
D
1
g
d
N
s
D
1
p
0
d
3
e
p
0
e
g
(1)
is then determined, where
N
s
D
1
g
dy
N
s
D
1
p
0
d
is the ratio
of the number of
D
p
1
decays observed in each mode, and
e
p
0
y
e
g
is the relative efficiency for finding the
p
0
or the
g
from the corresponding
D
p
1
decay. Assuming that the
three decay modes of the
D
p
1
add to unity and defining
R
1
p
;
B
s
D
p
1
!
D
0
p
1
dy
B
s
D
p
1
!
D
1
p
0
d
,
one finds
B
s
D
p
1
!
D
1
g
d
≠
R
1
g
ys
R
1
g
1
R
1
p
1
1
d
,
B
s
D
p
1
!
D
1
p
0
d
≠
1
ys
R
1
g
1
R
1
p
1
1
d
,
and
B
s
D
p
1
!
D
0
p
1
d
≠
R
1
p
ys
R
1
g
1
R
1
p
1
1
d
. A value for
R
1
p
can be obtained
by combining the known phase space for
D
p
1
!
D
1
p
0
and
D
p
1
!
D
0
p
1
with isospin conservation and the
expected
p
3
dependence of
p
-wave decay widths to yield
R
1
p
≠
2
μ
p
1
0
p
11
∂
3
≠
2.199
6
0.064 ,
(2)
(where
p
1
0
and
p
11
are the momenta of the
D
0
and
D
1
in the
D
p
1
rest frame, respectively). The theoretical
uncertainty in this ratio is thought to be only of the order
of 1% [4], so the error is dominated by those due to
the
M
D
p
2
M
D
mass differences [12]. This method has
the advantage of avoiding large systematic uncertainties
due to the
D
meson branching fractions and of canceling
many systematic uncertainties associated with the
D
1
reconstruction.
The analysis was performed using data accumulated by
the CLEO II detector [16] at the Cornell Electron Stor-
age Ring (CESR). The CLEO II detector consists of three
cylindrical drift chambers (immersed in a 1.5 T solenoidal
magnetic field) surrounded by a time-of-flight system
(TOF) and a CsI crystal electromagnetic (EM) calorime-
ter. The main drift chamber allows for charged par-
ticle identification via specific-ionization measurements
(
dE
y
dx
) in addition to providing an excellent momentum
measurement. The calorimeter is surrounded by a super-
conductor coil and an iron flux return, which is instru-
mented with muon counters.
A total of
4.7
fb
2
1
of data was collected at center-
of-mass energies on or near the
Y
s
4
S
d
resonance. The
Monte Carlo simulated events used to determine signal
shapes and detection efficiencies were produced with a
GEANT
-based full detector simulation.
Events were required to have three or more tracks and
at least 15% of the center-of-mass energy deposited in
the calorimeter. Each of the three tracks comprising a
candidate
D
1
!
K
2
p
1
p
1
decay was required to satisfy
either the
K
2
or
p
1
hypothesis at the 2.5
s
level using
dE
y
dx
alone, and then the triplet was required to satisfy
the
K
2
p
1
p
1
hypothesis, including TOF information
if available, with a
x
2
probability greater than 10%.
The three tracks were then constrained to come from a
common vertex, and the invariant mass of the triplet,
3920
V
OLUME
80, N
UMBER
18
PHYSICAL REVIEW LETTERS
4 M
AY
1998
under the
K
2
p
1
p
1
hypothesis, was required to be within
10
MeV
y
c
2
s,
1.5
s
d
of the known
D
1
mass.
Photon candidates were required to be in the best region
of the calorimeter,
j
cos
u
j
,
0.71
(where
u
is the polar
angle between the EM cluster centroid and the beam
axis), with a cluster energy of at least 30 MeV. It was
further required that no charged particle track point within
8 cm of a crystal used in the EM cluster. If the invariant
mass formed by a pair of photons was within 2.5
s
of
the
p
0
mass, taking into account the asymmetric
p
0
line
shape and the small momentum dependence of the mass
resolution, the photons were identified as being from a
p
0
. The photons were then kinematically constrained to
the
p
0
mass to improve the
p
0
momentum measurement.
Photons from
D
p
1
!
D
1
g
decays were required to
pass a lateral shower shape cut, which is 99% efficient
for isolated photons, and not to form a
p
0
when paired
with any other photon. For the momenta relevant to
D
p
1
decays at the
Y
s
4
S
d
, merging of the EM clusters from
a
p
0
decay (and the subsequent misidentification of a
radiative decay) does not occur. The decay angle
u
g
,
defined as the angle of the
g
in the
D
p
1
rest frame with
respect to the
D
p
1
’s direction in the laboratory frame,
was required to satisfy cos
u
g
.2
0.35
. This cut helps
to reduce the large combinatorial background that arises
when
D
1
mesons are combined with soft photons moving
in the opposite direction.
The combinatorial background was further reduced by
requiring
x
D
p
.
0.7,
where
x
D
p
is the fraction of the
maximum possible momentum carried by the recon-
structed
D
p
1
. This cut also removed any contribution
from
B
!
D
p
X
events. The cuts on cos
u
g
and
x
D
p
were
determined to maximize
S
2
y
B
(
S
is signal and
B
is back-
ground) by utilizing a large sample of
D
p
0
!
D
0
g
events
from the data as well as Monte Carlo simulated events.
The primary difficulty in this analysis is the small
size of the signal, due to the branching fraction, relative
to a large combinatorial background and, more impor-
tantly, relative to a background due to
D
p
1
s
radiative
decays where
D
1
s
!
K
2
K
1
p
1
. Unlike the
D
p
1
, the
D
p
1
s
almost always decays radiatively. This is a major
problem because the
M
s
D
p
1
s
d
2
M
s
D
1
s
d
mass difference
is
143.97
6
0.41
MeV [17] and the
M
s
D
p
1
d
2
M
s
D
1
d
mass difference is
140.64
6
0.09
MeV [12], so these two
processes cannot be separated in the mass difference plot
because the resolution in photon energy in the decay is
,
6
MeV.
Misidentification of
D
1
s
!
K
2
K
1
p
1
as
D
1
!
K
2
p
1
p
1
can occur because the TOF and
dE
y
dx
information used for particle identification does not
adequately separate
K
’s from
p
’s with momenta above
,
1
GeV
y
c
. When reconstructed under the
K
pp
hypoth-
esis, the two invariant mass distributions partially overlap,
and any attempt to estimate the fraction of
D
1
s
under the
D
1
peak will depend strongly on the resonant substructure
of the
D
1
s
!
K
2
K
1
p
1
decay, as well as the momentum
distribution of the
D
1
s
’s. The large
D
1
s
contribution to
the lower
D
1
sideband further complicates the analysis
by preventing the use of this sideband in a subtraction
of combinatorial background. Because of its small rate
(
,
0.2%
of the signal), no correction is necessary to
address the presence of the recently observed hadronic
decay
D
p
1
s
!
D
1
s
p
0
[17] in the
D
M
p
distribution.
A means to veto
D
p
1
s
events, independent of the
decay’s resonant substructure, is to require that the
invariant mass of the three tracks reconstructed under
the
K
2
K
1
p
1
hypothesis be greater than a cut which
removes all the
D
p
1
s
events. An unwanted side effect of
vetoing
D
p
1
s
events by this method is that a cut in the
KK
p
mass distribution greatly distorts the
K
pp
mass
distribution, making the relative normalization between
the
D
1
upper sideband and the signal region uncertain.
Thus, the use of a sideband subtraction to remove the
combinatorial background from the mass difference plot is
impossible. Figure 1 shows the Monte Carlo
K
2
K
1
p
1
mass distribution found in
D
1
s
decays and that found in
D
1
decays when one of the
p
1
’s is misidentified as a
K
1
. Since there are two possible tracks to assign the
K
1
mass, both combinations are tried, and the one yielding
the smaller mass is plotted.
Figure 2(a) shows the
D
M
g
distribution for events
from the
M
s
D
1
d
signal region as well as for those from
the
M
s
D
1
d
upper sideband (a region 3 times as wide
as the signal region starting
̄
3
s
above the nominal
D
1
mass). The
D
M
g
distribution for the combinatorial
background found in the
M
s
D
1
d
sideband is quite flat
under the signal region, justifying the use of a first order
polynomial in fitting this background. No
D
p
1
s
veto has
been applied to the data in Fig. 2(a), so a fair fraction
of the events in this “signal” are
D
p
1
s
background. The
signal was fit with a modified Gaussian, the parameters
for which were obtained from a large Monte Carlo sample
of
D
p
1
!
D
1
g
events. The systematic error in the fit
parameters was estimated by studying data versus Monte
Carlo differences in the very similar decay
D
p
0
!
D
0
g
.
Figure 2(b) shows the
D
M
g
signal and sideband dis-
tributions for events satisfying the
D
p
1
s
veto require-
ment that
M
s
K
2
K
1
p
1
d
.
1.990
GeV
y
c
2
. Monte Carlo
indicates the fraction of
D
p
1
s
events passing this cut
is
0.002
1
0.003
2
0.002
, thus if the entire signal yield (
180
6
26
events) found in Fig. 2(a) were due to
D
p
1
s
decays,
0.4
1
0.6
2
0.4
events would be expected in Fig. 2(b). The fit to
FIG. 1. The
M
s
K
2
K
1
p
1
d
distributions for
D
1
s
background
(solid line) and
D
1
signal (dashed line) Monte Carlo samples.
3921
V
OLUME
80, N
UMBER
18
PHYSICAL REVIEW LETTERS
4 M
AY
1998
FIG. 2. The
D
M
g
;
M
s
D
1
g
d
2
M
s
D
1
d
distributions for (a)
data before the
D
p
1
s
veto has been applied, (b) data after
the tight
D
p
1
s
veto has been applied. The large feature on
the left of the plots is due to
D
p
1
!
D
1
p
0
, where one of
the photons from the
p
0
decay is not detected. Monte Carlo
studies indicate that this decay does not contribute to the signal
region. The dashed histograms are data taken from the upper
M
s
D
1
d
sideband.
the
D
M
g
distribution in Fig. 2(b) yields
68
6
19
events.
When these data are refit with the signal constrained to be
0.4
1
0.6
2
0.4
events, the
x
2
of the fit increases by 15.8, corre-
sponding to a significance of 4.0 standard deviations that
the signal is not due to misidentified
D
p
1
s
events. There-
fore, the peak must be due to the decay
D
p
1
!
D
1
g
.
The presence of
D
p
1
!
D
1
g
decays having been
established, the
D
p
1
s
veto was loosened to
1.981
GeV
y
c
2
to maximize
S
2
ys
S
1
B
d
as determined by the Monte
Carlo samples. Figure 3(a) shows the
D
M
g
distribution
for the events which passed the optimized
D
p
1
s
veto. The
fraction of
D
1
mesons passing the veto was determined
by fitting the
D
M
p
distribution before and after the veto
was applied to the data. These distributions were fit with
FIG. 3. (a)
D
M
g
distribution for data after the “optimal”
M
s
K
2
K
1
p
1
d
cut (the
D
p
1
s
veto) has been applied. (b)
D
M
g
distribution for the vetoed data. (c)
D
M
p
distribution for data
prior to the
M
s
K
2
K
1
p
1
d
cut. (d)
D
M
p
distribution for data
after the
M
s
K
2
K
1
p
1
d
cut is applied. The dashed histograms
are data taken from the upper
M
s
D
1
d
sideband.
a double Gaussian plus a background function [18] which
simulates the expected threshold behavior. Figures 3(c)
and 3(d) show the
D
M
p
distributions, along with the fits,
used to determine the
D
p
1
s
veto efficiency for
D
1
mesons.
The results of fitting the
D
M
g
distribution for events
which passed and for those which failed the
D
p
1
s
veto,
Figs. 3(a) and 3(b), respectively, were:
N
pass
g
≠
87
6
21
and
N
fail
g
≠
95
6
16
(statistical errors only). Defining
N
1
s
N
s
d
as the total number of
D
p
1
(
D
p
1
s
) in the data,
one has the following pair of equations:
s
1
2e
1
d
N
1
1
s
1
2e
s
d
N
s
≠
N
fail
g
,
e
1
N
1
1e
s
N
s
≠
N
pass
g
,
(3)
where
e
1
is the fraction of
D
1
’s which pass the veto
as determined by fitting the
D
M
p
distributions (
N
pass
p
≠
1650
6
57
and
N
total
p
≠
2265
6
66,
where the errors are
statistical only), and
e
s
≠
0.037
6
0.007
is the fraction
of
D
p
1
s
’s which escape the veto as determined by a Monte
Carlo study. Rewriting Eq. (3) in terms of the measured
quantities (
N
pass
g
,
N
fail
g
,
N
pass
p
,
N
total
p
) and
e
s
and solving
for
R
1
g
, we find
R
1
g
≠
N
pass
g
2e
s
s
N
fail
g
1
N
pass
g
d
N
pass
p
2e
s
N
total
p
3
e
p
0
e
g
≠
0.055
6
0.014
6
0.010 ,
(4)
where the ratio of efficiencies
e
p
0
y
e
g
≠
1.066
6
0.064
.
From this branching ratio we can then extract the branch-
ing fractions shown in Table I. The statistical uncertainty
is dominated by the
D
1
g
yields, and the largest systema-
tic uncertainty is due to variations in this yield, when
the mean and width of the signal shape were varied by
an amount suggested by the
D
1
0
!
D
0
g
data versus
Monte Carlo comparison. A similar comparison was used
to estimate the uncertainty introduced by the cos
u
g
cut.
Table II lists the various sources of systematic uncertainty
and gives estimates for their impact on the measurement
of
R
1
g
.
In conclusion, we have observed, with 4
s
significance,
the radiative decay of the
D
p
1
and measured
B
s
D
p
1
!
D
1
g
dy
B
s
D
p
1
!
D
1
p
0
d
≠
0.055
6
0.017
(statistical
and systematic uncertainties added in quadrature). As-
suming Eq. (2) and that the three branching fractions of
the
D
p
1
add to unity, we find the results in Table I. The
hadronic branching fractions are in good agreement with
TABLE I. The
D
p
s
2010
d
6
branching fractions determined
from the measured ratio
N
s
D
1
g
dy
N
s
D
1
p
0
d
. The first uncer-
tainty is statistical, the second is experimental systematic and
the third is that which arises from the use of Eq. (2).
Mode
CLEO
II
PDG [12]
D
1
g
s
1.68
6
0.42
6
0.29
6
0.03
d
%
s
1.1
1
2.1
2
0.7
d
%
D
1
p
0
s
30.73
6
0.13
6
0.09
6
0.61
d
%
s
30.6
6
2.5
d
%
D
0
p
1
s
67.59
6
0.29
6
0.20
6
0.61
d
%
s
68.3
6
1.4
d
%
3922
V
OLUME
80, N
UMBER
18
PHYSICAL REVIEW LETTERS
4 M
AY
1998
TABLE II. Estimates of the systematic uncertainties in the
measurement of
R
1
g
.
Efficiency ratio
e
p
0
y
e
g
6%
Fitting of background
9%
Fitting of signal
13%
Veto efficiency for
D
1
s
(19% on
e
s
)
1%
Veto efficiency for
D
1
(2% on
e
1
)
2%
cos
u
g
.2
0.35
5%
the current PDG averages [12], but with substantially
reduced uncertainties (which are now dominated by the
3% uncertainty in
R
1
p
). The
D
p
1
radiative branching
fraction is in good agreement with theoretical expectations
and the earlier upper limits set by CLEO II [10] and
ARGUS [11]. The uncertainty in this branching fraction
is due primarily to the large combinatorial background
under the radiative signal, so one can expect that data
taken with the new CLEO
II.5
detector, which includes
a silicon tracker, will reduce this uncertainty further in the
near future.
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, Research Corporation, the Natural Sciences and
Engineering Research Council of Canada, the A. P. Sloan
Foundation, and the Swiss National Science Foundation.
*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.
§
Permanent address: BINP, RU-630090 Novosibirsk,
Russia.
k
Permanent address: Yonsei University, Seoul 120-749,
Korea.
[1] See, for example, A. N. Kamal and Q. P. Xu, Phys. Lett.
B
284
, 421 (1992); P. J. O’Donnell and Q. P. Xu, Phys.
Lett. B
336
, 113 (1994); P. Colangelo, F. De Fazio, and
G. Nardulli, Phys. Lett. B
334
, 175 (1994); T. M. Aliev
et al.,
Phys. Rev. D
54
, 857 (1996); H. G. Dosch and
S. Narison, Phys. Lett. B
368
, 163 (1996).
[2] M. B. Wise, Phys. Rev. D
45
, R2188 (1992); G. Burdman
and J. Donoghue, Phys. Lett. B
280
, 287 (1992); H. Y.
Cheng
et al.,
Phys. Rev. D
49
, 5857 (1994);
ibid.
55
,
5851(E) (1997); R. Casalbuoni
et al.,
Phys. Rep.
281
, 145
(1997).
[3] J. Admundson
et al.,
Phys. Lett. B
296
, 415 (1992);
P. Cho and H. Georgi, Phys. Lett. B
296
, 408 (1992);
ibid.
300
, 410(E) (1993).
[4] E. Angelos and G. P. Lepage, Phys. Rev. D
45
, R3021
(1992).
[5] Mark I Collaboration, G. Goldhaber
et al.,
Phys. Lett.
69B
,
503 (1977).
[6] Mark II Collaboration, M. W. Coles
et al.,
Phys. Rev. D
26
, 2190 (1982).
[7] JADE Collaboration, W. Bartelt
et al.,
Phys. Lett.
161B
,
197 (1985).
[8] HRS Collaboration, E. H. Low
et al.,
Phys. Lett. B
183
,
232 (1987).
[9] Mark III Collaboration, J. Adler
et al.,
Phys. Lett. B
208
,
152 (1988).
[10] CLEO Collaboration, F. Butler
et al.,
Phys. Rev. Lett.
69
,
2041 (1992).
[11] ARGUS Collaboration, H. Albrecht
et al.,
Z. Phys. C
66
,
63 (1995).
[12] Particle Data Group, R. M. Barnett
et al.,
Phys. Rev. D
54
,
1 (1996).
[13] Inclusion of the charge conjugate reaction is implied
throughout this paper.
[14] Particle Data Group, J. J. Hernández
et al.,
Phys. Lett. B
239
, 1 (1990).
[15]
M(X)
refers to the invariant mass of the particles used to
form the
X
candidate.
[16] CLEO Collaboration, Y. Kubota
et al.,
Nucl. Instrum.
Methods Phys. Res., Sect. A
320
, 66 (1992).
[17] CLEO Collaboration, J. Gronberg
et al.,
Phys. Rev. Lett.
75
, 3232 (1995).
[18] The total area as well as the mean and sigma of the
primary Gaussian (62% of the signal) were allowed to
float while the ratio of the areas, means, and sigmas were
fixed to the values obtained from the Monte Carlo sample.
The background function used is
p
1
fs
D
M
p
2
M
p
0
d
1
y
2
1
p
2
s
D
M
p
2
M
p
0
d
3
y
2
g
, where
p
1
and
p
2
are parameters
determined by the fit.
3923