1: %%%%%%%%%%%%%% Use for CLNS preprint (hep-ex) and Paper Drafts

2: \documentclass[aps,prd,preprint,superscriptaddress,tightenlines,nofootinbib,showpacs]{revtex4}

3: 

4: %%%%%%%%%%%%%% Use for PRD submission

5: %\documentclass[aps,prd,preprint,nopreprintnumbers,nofootinbib,showpacs]{revtex4}

6: 

7: \usepackage{graphicx}

8: \usepackage{dcolumn}

9: \usepackage{bm}

10: 

11: \begin{document}

12: 

13: \preprint{\vbox{\hbox{\hfil CLNS 06/1981}\hbox{\hfil CLEO 06-21}

14: %%                        \hbox{\hfil CLEO CONF 97-08}

15: }}

16: 

17: \title{Measurement of Upper Limits for $\Upsilon\to \gamma + {\cal R}$ Decays}

18: 

19: \author{J.~L.~Rosner}

20: \affiliation{Enrico Fermi Institute, University of

21: Chicago, Chicago, Illinois 60637}

22: \author{N.~E.~Adam}

23: \author{J.~P.~Alexander}

24: \author{D.~G.~Cassel}

25: \author{J.~E.~Duboscq}

26: \author{R.~Ehrlich}

27: \author{L.~Fields}

28: \author{R.~S.~Galik}

29: \author{L.~Gibbons}

30: \author{R.~Gray}

31: \author{S.~W.~Gray}

32: \author{D.~L.~Hartill}

33: \author{B.~K.~Heltsley}

34: \author{D.~Hertz}

35: \author{C.~D.~Jones}

36: \author{J.~Kandaswamy}

37: \author{D.~L.~Kreinick}

38: \author{V.~E.~Kuznetsov}

39: \author{H.~Mahlke-Kr\"uger}

40: \author{P.~U.~E.~Onyisi}

41: \author{J.~R.~Patterson}

42: \author{D.~Peterson}

43: \author{J.~Pivarski}

44: \author{D.~Riley}

45: \author{A.~Ryd}

46: \author{A.~J.~Sadoff}

47: \author{H.~Schwarthoff}

48: \author{X.~Shi}

49: \author{S.~Stroiney}

50: \author{W.~M.~Sun}

51: \author{T.~Wilksen}

52: \author{M.~Weinberger}

53: \author{}

54: \affiliation{Cornell University, Ithaca, New York 14853}

55: \author{S.~B.~Athar}

56: \author{R.~Patel}

57: \author{V.~Potlia}

58: \author{J.~Yelton}

59: \affiliation{University of Florida, Gainesville, Florida 32611}

60: \author{P.~Rubin}

61: \affiliation{George Mason University, Fairfax, Virginia 22030}

62: \author{C.~Cawlfield}

63: \author{B.~I.~Eisenstein}

64: \author{I.~Karliner}

65: \author{D.~Kim}

66: \author{N.~Lowrey}

67: \author{P.~Naik}

68: \author{M.~Selen}

69: \author{E.~J.~White}

70: \author{J.~Wiss}

71: \affiliation{University of Illinois, Urbana-Champaign, Illinois 61801}

72: \author{R.~E.~Mitchell}

73: \author{M.~R.~Shepherd}

74: \affiliation{Indiana University, Bloomington, Indiana 47405 }

75: \author{D.~Besson}

76: \author{S.~Henderson}

77: \altaffiliation{Current address: Massachusetts Institute of Technology, 

78: Cambridge, MA 02139.}

79: \affiliation{University of Kansas, Lawrence, Kansas 66045}

80: \author{T.~K.~Pedlar}

81: \affiliation{Luther College, Decorah, Iowa 52101}

82: \author{D.~Cronin-Hennessy}

83: \author{K.~Y.~Gao}

84: \author{J.~Hietala}

85: \author{Y.~Kubota}

86: \author{T.~Klein}

87: \author{B.~W.~Lang}

88: \author{R.~Poling}

89: \author{A.~W.~Scott}

90: \author{A.~Smith}

91: \author{P.~Zweber}

92: \affiliation{University of Minnesota, Minneapolis, Minnesota 55455}

93: \author{S.~Dobbs}

94: \author{Z.~Metreveli}

95: \author{K.~K.~Seth}

96: \author{A.~Tomaradze}

97: \affiliation{Northwestern University, Evanston, Illinois 60208}

98: \author{J.~Ernst}

99: \affiliation{State University of New York at Albany, Albany, New York 12222}

100: \author{K.~M.~Ecklund}

101: \affiliation{State University of New York at Buffalo, Buffalo, New York 14260}

102: \author{H.~Severini}

103: \affiliation{University of Oklahoma, Norman, Oklahoma 73019}

104: \author{W.~Love}

105: \author{V.~Savinov}

106: \affiliation{University of Pittsburgh, Pittsburgh, Pennsylvania 15260}

107: \author{O.~Aquines}

108: \author{Z.~Li}

109: \author{A.~Lopez}

110: \author{S.~Mehrabyan}

111: \author{H.~Mendez}

112: \author{J.~Ramirez}

113: \affiliation{University of Puerto Rico, Mayaguez, Puerto Rico 00681}

114: \author{G.~S.~Huang}

115: \author{D.~H.~Miller}

116: \author{V.~Pavlunin}

117: \author{B.~Sanghi}

118: \author{I.~P.~J.~Shipsey}

119: \author{B.~Xin}

120: \affiliation{Purdue University, West Lafayette, Indiana 47907}

121: \author{G.~S.~Adams}

122: \author{M.~Anderson}

123: \author{J.~P.~Cummings}

124: \author{I.~Danko}

125: \author{D.~Hu}

126: \author{B.~Moziak}

127: \author{J.~Napolitano}

128: \affiliation{Rensselaer Polytechnic Institute, Troy, New York 12180}

129: \author{Q.~He}

130: \author{J.~Insler}

131: \author{H.~Muramatsu}

132: \author{C.~S.~Park}

133: \author{E.~H.~Thorndike}

134: \author{F.~Yang}

135: \affiliation{University of Rochester, Rochester, New York 14627}

136: \author{T.~E.~Coan}

137: \author{Y.~S.~Gao}

138: \affiliation{Southern Methodist University, Dallas, Texas 75275}

139: \author{M.~Artuso}

140: \author{S.~Blusk}

141: \author{J.~Butt}

142: \author{J.~Li}

143: \author{N.~Menaa}

144: \author{G.~C.~Moneti}

145: \author{R.~Mountain}

146: \author{S.~Nisar}

147: \author{K.~Randrianarivony}

148: \author{R.~Sia}

149: \author{T.~Skwarnicki}

150: \author{S.~Stone}

151: \author{J.~C.~Wang}

152: \author{K.~Zhang}

153: \affiliation{Syracuse University, Syracuse, New York 13244}

154: \author{G.~Bonvicini}

155: \author{D.~Cinabro}

156: \author{M.~Dubrovin}

157: \author{A.~Lincoln}

158: \affiliation{Wayne State University, Detroit, Michigan 48202}

159: \author{D.~M.~Asner}

160: \author{K.~W.~Edwards}

161: \affiliation{Carleton University, Ottawa, Ontario, Canada K1S 5B6}

162: \author{R.~A.~Briere}

163: \author{T.~Ferguson}

164: \author{G.~Tatishvili}

165: \author{H.~Vogel}

166: \author{M.~E.~Watkins}

167: \affiliation{Carnegie Mellon University, Pittsburgh, Pennsylvania 15213}

168: %\author{(CLEO Collaboration)} %FOR PRD_SPECIAL_CHANGEME

169: \collaboration{CLEO Collaboration} %FOR PRL,CLNS

170: \noaffiliation

171: 

172: \date{April 20, 2007}

173: 

174: \begin{abstract} 

175: 

176: Motivated by concerns regarding possible

177: two-body contributions to the recently-measured

178: inclusive $\Upsilon$(nS)$\to\gamma+X$ (n=1, 2, 3)

179: direct photon spectra,

180: we report on a new study of exclusive radiative decays of these narrow

181: $\Upsilon$(nS) resonances into two-body final states ${\cal R}\gamma$, with

182: ${\cal R}$ a narrow resonant hadronic state

183: decaying into four or more charged particles. 

184: Such two-body processes are not explicitly addressed in the

185: extant theoretical frameworks used to calculate the inclusive direct

186: photon spectra, and must also be explicitly inserted into Monte Carlo

187: simulations.

188: Using data collected from the CLEO~III detector at the Cornell Electron

189: Storage Ring, 

190: we present upper limits of order $10^{-4}$ for such bottomonium

191: two-body decays as a function

192: of the recoil mass $M_{\cal R}$.

193: 

194: \end{abstract}

195: 

196: \pacs{13.20.Gd,13.20.-v,13.40.Hq}

197: \maketitle

198: 

199: \section*{Introduction}

200: CLEO recently extracted $\alpha_s$ from a 

201: measurement of the direct photon

202: spectra in $\Upsilon({\rm 1S,2S,3S})\rightarrow gg\gamma$\cite{r:shawn}.  

203: That extraction was based on a comparison of the $gg\gamma$ width

204: to the dominant three-gluon width of the narrow bottomonium resonances.

205: Since the direct photon is observable above background only for

206: relatively high energies ($E_\gamma\ge E_{\mathrm{beam}}/2$), some model

207: dependence is inherent in the determination of the 

208: total $gg\gamma$ rate. Given a 

209: prescription relating the parton-level rate to 

210: $\alpha_s$, one can then use that rate to 

211: determine $\alpha_s$.

212: To extrapolate beyond the experimentally accessible direct photon energy

213: region, CLEO relies on 

214: theoretical parameterizations of the expected photon energy 

215: spectrum in the $\Upsilon$ system\cite{r:Field,r:SotoGarcia} to

216: obtain the total direct $\Upsilon\to gg\gamma$ decay width

217: relative to the dominant $\Upsilon\to ggg$ width. 

218: The theoretical calculations are generally done at the 

219: parton level, and therefore

220: avoid possible resonant

221: contributions to the photon energy spectrum due to 

222: two-body decays, e.g., $\Upsilon\to gg\gamma\to\gamma{\cal R}$, with 

223: ${\cal R}$ some resonant hadronic state. Alternatively,

224: one would like to understand hadronization in 

225: $\Upsilon\to gg\gamma\to\gamma{\cal R}$, which is measured

226: experimentally as a two-body process. 

227: For example, CLEO has recently observed signals in several

228: low-multiplicity modes\cite{r:pipi05,r:pi0pi006}, and has

229: presented preliminary

230: results consistent with the final state $\Upsilon$(1S)$\to\gamma+\eta'$, 

231: $\eta'\to\pi^+\pi^-\eta$, $\eta\to\pi^+\pi^-\pi^0$\cite{r:potlia}.

232: We point out, however, 

233: that the product of (branching ratio)$\times$(efficiency)

234: for all known exclusive

235: modes, combined with the observed level of

236: background, implies a signal yield 

237: below the statistical sensitivity required to be observed as a distinct

238: signal in the inclusive spectrum.

239: 

240: From the experimental standpoint, the

241: presence of possible digluon resonances opposite the photon leads to

242: ``bumps'' in the otherwise smooth predicted theoretical 

243: photon spectra. 

244: The inability of the current calculations to directly address

245: two-body effects, in part, restricts the applicability

246: of $\Upsilon$ decay models to the region $z_\gamma<0.92$,

247: with $z_\gamma$ defined as the scaled 

248: photon energy ($\equiv E_\gamma/E_{\mathrm{beam}}$).  $\Upsilon$

249: decay models are also not reliable above this point because one of the 

250: emitted gluons is of such low energy that 

251: a perturbative calculation cannot be trusted\cite{r:xavier}.  Given that

252: primary glueball candidates are of order 1 GeV in mass, we expect the

253: endpoint region of the photon energy spectrum ($z_\gamma>$ 0.92) to be

254: most susceptible to such contamination. 

255: Older estimates of $\alpha_s$ based on 

256: inclusive radiative photon production in $\Upsilon$ decay 

257: using the BLM\cite{r:Brod-Lep-Mack} prescription

258: have consistently

259: yielded values 

260: smaller than those obtained from different techniques\cite{r:pdg}.

261: Recently, it has been realized that color octet contributions previously

262: ignored in the older BLM calculation result in estimates of $\alpha_s$

263: in excellent agreement with estimates made at the Z-resonance\cite{r:joan07}.

264: Nevertheless,

265: subtraction of possibly-enhanced exclusive contributions to the overall

266: rate would, in principle, result in a lowered estimate for the ratio of rates

267: $\Gamma(\Upsilon\to gg\gamma)/\Gamma(\Upsilon\to ggg)$ 

268: and a correspondingly larger

269: estimate for $\alpha_s$. 

270: To set the scale,

271: given that the typical branching fraction for $\Upsilon\to gg\gamma$ is of

272: order $10^{-2}$, a total resonant

273: enhancement at the level of $5\times 10^{-4}$ would result in

274: a $\sim$5\% reduction in the estimate of $\alpha_s$ at the Z-pole.

275: 

276: By comparison,

277: a large fraction of all $J/\psi\to gg\gamma$ decays have been

278: identified as two-body\cite{r:pdg}.

279: Two-body

280: contamination (and larger

281: relativistic corrections) also

282: makes the $J/\psi$ system

283: somewhat less reliable than the $\Upsilon$ in estimating $\alpha_s$.

284: This systematic consideration in the $gg\gamma$ analysis motivates

285: our 

286: search for radiative decays of the

287: $\Upsilon$ to resonances: 

288: $\Upsilon$(nS)$\to \gamma{\cal R}$ (n=1, 2, 3).  We concern ourselves with 

289: high multiplicity ($\ge 4$ charged tracks) final states,

290: as we employ

291: the same hadronic event selection cuts in this analysis that we did

292: in the $gg\gamma$ analysis \cite{r:shawn}.

293: We note that

294: although two-body branching fractions have been observed

295: for, e.g., $\Upsilon$(1S)$\to\gamma f_2(1270)$ at the level of

296: $10^{-4}$, the fraction of $f_2(1270)$ decays into $\ge$4 charged

297: tracks is only $\approx 3\%$\cite{r:pdg}. 

298: 

299: The analysis, in general terms, proceeds as follows. After selecting a

300: high-quality sample of $e^+e^-$ annihilations into hadrons using the

301: hadronic event selection cuts of the previous analysis \cite{r:shawn}, 

302: we construct the

303: inclusive isolated photon spectra in data taken at both

304: on-$\Upsilon$-resonance and off-$\Upsilon$-resonance energies (the latter

305: samples are used for systematic checks of the overall procedure). A two-body

306: radiative decay of the $\Upsilon$ will produce a monochromatic photon

307: in the lab frame; the energy of the radiated photon $E_\gamma$

308: is related to the

309: mass of the recoil hadron ${\cal R}$ via

310: $M_{\cal R}=2E_{\mathrm{beam}}\sqrt{1-z_\gamma}$. 

311: In the case where the intrinsic width of the recoil hadron

312: is much smaller than the experimental photon energy resolution, the

313: measured radiative photon energy should be a Gaussian centered at the

314: energy $E_\gamma$. For a 1 GeV (4.5 GeV) recoil photon, this

315: implies a recoil resonance with width 

316: typically narrower than 20 MeV (260 MeV). 

317: Not knowing \emph{a priori} the mass 

318: of the hadron

319: ${\cal R}$, we therefore perform a set of fits of the $\Upsilon$(nS) photon

320: spectrum to a Gaussian signal, centered at a series of $E_\gamma$

321: values,

322: and with resolutions corresponding to the known CLEO~III electromagnetic

323: calorimeter resolution

324: atop smooth polynomial backgrounds, over the range

325: $0.2<z_\gamma<1.0$.\footnote{It should be noted that 

326: ``bumps'' in the inclusive photon spectrum can be due not only to

327: resonant two-body decays but also to 

328: continuum threshold effects such as 

329: the crossing of the $c\overline{c}$ threshold.

330: Given our photon energy resolution,

331: all processes of the type $e^+e^-\to D^{(*)}{\overline D^{(*)}}$

332: result in photons relatively close in energy and produce an

333: apparent enhancement in the region $z_\gamma\sim0.8$.

334: In the previous analysis\cite{r:shawn}, 

335: we also identified an excess of photons in

336: data as $z_\gamma\to1$ ($\equiv E_\gamma/E_{\mathrm{beam}}$). 

337: Further examination of these events indicated that

338: they were dominated by continuum

339: production: $e^+e^-\to\gamma\pi^+\pi^-\pi^+\pi^-$, although

340: the possibility that the 4-pion state resulted from the decay

341: of an intermediate resonance ${\cal R}$ was not investigated.}

342: 

343: We construct 95\% confidence level upper limits from these fits 

344: as a function of recoil mass $M_{\cal R}$,

345: corrected for the efficiency loss due to the fiducial 

346: acceptance of the

347: detector and the event and photon-selection cuts that define our data

348: sample. In estimating this correction, we assume ${\cal R}$ has spin=0, 

349: with a  

350: corresponding $1+\cos^2\theta$ 

351: angular distribution for the recoil gamma; higher

352: spins will generally give flatter angular distributions.

353: Note that the

354: exact form of the efficiency correction due to the event and photon 

355: selection cuts varies with the decay final states considered for 

356: ${\cal R}$.  

357: In the high-momentum region typical of the particles

358: recoiling against the direct photon, our per-track charged particle

359: detection efficiency is generally $\ge$90\% over the fiducial

360: volume.

361: To be conservative, we derive our 

362: $z_\gamma$-dependent efficiency correction 

363: from the decay mode yielding the worst 

364: reconstruction efficiency.

365: 

366: This final efficiency-corrected limit is converted into an

367: $M_{\cal R}$-dependent branching ratio upper limit 

368: ${\cal B}(\gamma {\cal R})$ by dividing 

369: the resulting yield by the

370: calculated total number of resonant $\Upsilon$ events.  For the off-resonance 

371: running, the distributions are divided by the off-resonance luminosity 

372: for the sake of comparison between continuum samples.

373: An example of simulated signal

374: superimposed on background is given in 

375: Figure~\ref{fig:fakeSignal} for the 

376: hypothetical process $\Upsilon({\rm 4S})\to\gamma+{\cal R}$, 

377: ${\cal R}\to\pi^+\pi^-\pi^+\pi^-$.

378: 

379: \begin{figure}[htpb]

380: \centerline{\includegraphics[width=8cm]{3031106-019.eps}}

381: \caption{Scaled photon spectrum for

382: $\Upsilon({\rm 4S})\to \gamma{\cal R}$, ${\cal R}\to 4\pi$ simulations, for

383: various hypothetical ${\cal R}$ masses.  The lower (dashed) curve is 

384: $\Upsilon({\rm 4S})$ data, 

385: while the upper (solid) curve is $\Upsilon({\rm 4S})$ data with signal

386: Monte Carlo added on top.

387: The magnitude of 

388: ${\cal B}$($\Upsilon({\rm 4S})\to \gamma{\cal R}$, ${\cal R}\to 4\pi$) 

389: has been grossly exaggerated for the sake of presentation ($\approx 5 \times 10^{-3}$,

390: well above any observed radiative decay fraction into $\geq4$ charged tracks).

391: The six elevations correspond to masses 7.5, 6.5, 4.5, 3.5, 2.5 and,

392: for the right-most peak, 

393: the overlap of a ${\cal R}$ of mass $1.5$ GeV and a 

394: ${\cal R}$ of mass $0.5$ GeV, respectively.}

395: \label{fig:fakeSignal}

396: \end{figure}

397: 

398: \section*{Event Selection}

399: Event selection criteria in this analysis are identical to those imposed

400: in the previous analysis \cite{r:shawn}. The inclusive photon

401: spectra are therefore identical to those taken from our previous

402: analysis as well. The background shape is approximately

403: exponential in the region $0.2<z_\gamma<1.0$.

404: 

405: \section*{Fitting the Inclusive Photon Spectrum}

406: 

407: To extract the possible magnitude of a two-body

408: radiative signal, we step along the inclusive photon 

409: spectrum over the interval $0.2<z_\gamma<1.0$, fitting it to a Gaussian with width

410: equal to the detector resolution at that value of photon

411: energy, plus a background

412: parametrized by a smooth Chebyshyev polynomial.  

413: We assume that the intrinsic width of the resonance ${\cal R}$ is

414: considerably smaller than the detector resolution.

415: Our step size is

416: determined by the energy resolution of the detector $\sigma_{\rm E}$; we use steps of width 

417: $\sigma_{\rm E}/2$.

418: 

419: For photons in the central barrel

420: ($|\cos\theta_\gamma|<$0.7, with

421: $\theta_\gamma$ the polar angle of the photon momentum vector relative to the

422: beam axis) region of the CsI 

423: electromagnetic calorimeter, the energy resolution over the kinematic interval

424: relevant to this analysis is of order 2\%.

425: For two-body radiative decays from the $\Upsilon$(1S), the photon

426: energy resolution and recoil mass resolution as a function of $z_\gamma$,

427: are shown in Figure \ref{fig:xgamma_v_MR}. Curves are similar for the

428: other narrow $\Upsilon$ resonances.

429: \begin{figure}[htpb]

430: \centerline{\includegraphics[width=8cm]{3031106-020.eps}}

431: \caption{Photon energy resolution (solid)

432: and recoil mass resolution for 

433: $\Upsilon$(1S)$\to\gamma{\cal R}$ (dashed).}

434: \label{fig:xgamma_v_MR}

435: \end{figure}

436: 

437: We use a fourth-order Chebyschev polynomial to describe the background.

438: For each fit, the Gaussian fit area $A(z_\gamma)$ and fit error $\sigma_A(z_\gamma)$

439: is recorded. 

440: Note that the fits are highly correlated point-to-point, and

441: that the bin width is much finer than the detector resolution. 

442: At each step, we use a

443: $\pm$10$\sigma_{\rm E}$ fitting window; the background

444: is expected to be relatively smooth over such a limited interval.  

445: 

446: \section*{Extracting Upper Limits}

447: 

448: To convert the $A(z_\gamma)$ distribution obtained from fitting the inclusive photon spectrum

449: into a 95\% confidence-interval

450: upper limit, we add $1.645\cdot\sigma_A(z_\gamma)$ point-wise to the 

451: $A(z_\gamma)$ distribution,

452: as a function of photon energy.  In this process, since we are interested in

453: enhancements in the inclusive photon spectrum, all negative areas from the raw fits are

454: set equal to zero, and the corresponding upper limit

455: set to $1.645\cdot\sigma_A(z_\gamma)$

456: at these points.  The resulting contour for the $\Upsilon$(1S) fitting is shown in

457: Figure~\ref{fig:rawUpperLimit1S}.  The analogous contours for the other $\Upsilon$ spectra look 

458: similar.

459: 

460: \begin{figure}[htpb]

461: \centerline{\includegraphics[width=8cm]{3031106-021.eps}}

462: \caption{$A(z_\gamma)+1.645\cdot\sigma_A(z_\gamma)$ versus $z_\gamma$ for fits to the 

463: $\Upsilon$(1S) inclusive photon spectrum where negative points have been mapped to 

464: $1.645\cdot\sigma_A(z_\gamma)$.  This plot is the upper limit yield, before efficiency correction,

465: computed from statistical errors only.}

466: \label{fig:rawUpperLimit1S}

467: \end{figure}

468: 

469: We convert the limits as a function of 

470: photon energy $z_\gamma$

471: into a function of a hypothetical resonance recoil mass $M_{\cal R}$.

472: For the purposes of this conversion, the mean values

473: for each running period of $E_{\mathrm{beam}}$ 

474: are used for each data sample; we neglect the MeV-scale variation in

475: beam energies for a particular run period.  These values are given in Table~I, along with 

476: the integrated luminosities of the resonance and below-resonance $\Upsilon$ data samples used

477: in this analysis.

478: 

479: \begin{table}[htpb]

480: \label{tab:beamEnergies}

481: \begin{center}

482: \begin{tabular}{|c|c|c|}\hline

483: Event Type & $E_{\mathrm{beam}}$ (GeV) & ${\cal L}$ (${\rm pb}^{-1}$) \\

484: \hline

485: Resonance $\Upsilon$(1S) & 4.73 & 1076$\pm$11 \\

486: Resonance $\Upsilon$(2S) & 5.01 & 1189$\pm$12 \\

487: Resonance $\Upsilon$(3S) & 5.18 & 1228$\pm$12 \\

488: Resonance $\Upsilon$(4S) & 5.29 & 6456$\pm$65 \\

489: Below $\Upsilon$(1S) & 4.72 & 188$\pm$2 \\

490: Below $\Upsilon$(2S) & 5.00 & 396$\pm$4 \\

491: Below $\Upsilon$(3S) & 5.16 & 158$\pm$2 \\

492: Below $\Upsilon$(4S) & 5.27 & 2278$\pm$23 \\

493: \hline

494: \end{tabular}

495: \end{center}

496: \caption{The mean values of average beam energy ($E_{\mathrm{beam}}$) and the integrated luminosity 

497: (${\cal L}$) for each data sample used in this analysis.}

498: \end{table}

499: 

500: The resulting $M_{\cal R}$-dependent contour, 

501: for the $\Upsilon$(1S), is shown in Figure~\ref{fig:rawRecoilUpperLimit1S}.

502: 

503: \begin{figure}[htpb]

504: \centerline{\includegraphics[width=8cm]{3031106-022.eps}}

505: \caption{$A(M_{\cal R})+1.645\cdot\sigma_A(M_{\cal R})$ versus $M_{\cal R}$ for fits to the 

506: $\Upsilon$(1S) inclusive photon spectrum, where negative points have been mapped to 

507: $1.645\cdot\sigma_A(M_{\cal R})$. Statistical errors included

508: only.}  

509: \label{fig:rawRecoilUpperLimit1S}

510: \end{figure}

511: 

512: \section*{Efficiency Correction}

513: 

514: We consider two efficiency corrections to the upper limit contour:

515: one due to the fiducial acceptance of the 

516: detector, and the other due to our event and shower selection cuts.

517: 

518: For photons in the barrel of the detector,

519: we assume that 

520: ${\cal R}$ is spin zero, 

521: which corresponds to a $1+\cos^2\theta$ distribution of the 

522: photons in the two-body decays we are considering; higher

523: spins will generally give flatter angular distributions \cite{r:pipi05}.  Assuming this 

524: angular distribution amounts to an $\approx0.6$

525: uniform angular acceptance efficiency correction factor

526: to our limit.  

527: 

528: In addition to this 

529: angular acceptance correction,

530: we assess an efficiency correction due to the CLEO~III detection efficiency.

531: Not knowing \emph{a priori} what the decay mode of our hypothetical 

532: resonance ${\cal R}$ will be,

533: we have

534: generated 5000-event Monte Carlo samples spanning a wide range of final

535: state multiplicities (all with $N_{\mathrm{charged}}\ge 4$)

536: and masses $M_{\cal R}$.

537: In the interests of producing a conservative upper limit, 

538: we used this study to choose the mode with the worst average efficiency.

539: In this manner, we

540: efficiency-correct, as a function of $M_{\cal R}$ mass, 

541: our upper limit 

542: as a function of $z_\gamma$ (or $E_\gamma$) 

543: before mapping the upper limit into $M_{\cal R}$.

544: A list of ${\cal R}$ decay modes considered in this analysis and their average 

545: efficiencies (averaged over the photon energy spectrum from 

546: $1.0$ GeV $\leq{E_\gamma}\leq4.5$ GeV) is given in Table~II.  

547: We find that the worst efficiency among the

548: decay modes considered

549: was obtained from ${\cal R}\to2(K^+K^-)\pi^0$

550: (Figure~\ref{fig:4Kpi0Efficiency}).  We therefore use this efficiency 

551: function to determine our upper limit contours.

552: 

553: \begin{table}[htpb]

554: \label{tab:eventeff}

555: \begin{center}

556: \begin{tabular}{|c|c|}\hline

557: Event Type & Average Efficiency ($\overline{\epsilon}$)  \\

558: \hline

559: ${\cal R}\to{K^+K^-\pi^+\pi^-}$ & 0.53 $\pm$ 0.03 \\

560: ${\cal R}\to{K^+K^-\pi^+\pi^-\pi^0}$ & 0.53 $\pm$ 0.02 \\

561: ${\cal R}\to{K^+K^-\pi^+\pi^-\pi^0\pi^0}$ & 0.54 $\pm$ 0.02 \\

562: 

563: ${\cal R}\to{K^+K^-p^+p^-}$ & 0.56 $\pm$ 0.02 \\

564: ${\cal R}\to{K^+K^-p^+p^-\pi^0}$ & 0.50 $\pm$ 0.05 \\

565: ${\cal R}\to{K^+K^-p^+p^-\pi^0\pi^0}$ & 0.57 $\pm$ 0.02 \\

566: 

567: ${\cal R}\to{p^+p^-\pi^+\pi^-}$ & 0.62 $\pm$ 0.03 \\

568: ${\cal R}\to{p^+p^-\pi^+\pi^-\pi^0}$ & 0.54 $\pm$ 0.05\\

569: ${\cal R}\to{p^+p^-\pi^+\pi^-\pi^0\pi^0}$ & 0.63 $\pm$  0.02 \\

570: 

571: ${\cal R}\to{K^+K^-K^+K^-}$ & 0.50 $\pm$ 0.02 \\

572: ${\cal R}\to{K^+K^-K^+K^-\pi^0\pi^0}$ & 0.49 $\pm$  0.02 \\

573: 

574: ${\cal R}\to{p^+p^-p^+p^-}$ & 0.67 $\pm$ 0.02 \\

575: ${\cal R}\to{p^+p^-p^+p^-\pi^0}$ & 0.65 $\pm$  0.02 \\

576: ${\cal R}\to{p^+p^-p^+p^-\pi^0\pi^0}$ & 0.63 $\pm$  0.02 \\

577: 

578: ${\cal R}\to{\pi^+\pi^-\pi^+\pi^-}$ & 0.59 $\pm$ 0.02 \\

579: ${\cal R}\to{\pi^+\pi^-\pi^+\pi^-\pi^0}$ & 0.65 $\pm$ 0.02 \\

580: ${\cal R}\to{\pi^+\pi^-\pi^+\pi^-\pi^0\pi^0}$ & 0.59 $\pm$ 0.01 \\

581: 

582: ${\cal R}\to{\pi^+\pi^-\pi^+\pi^-4\pi^0}$ & 0.57 $\pm$ 0.02 \\

583: ${\cal R}\to{\pi^+\pi^-\pi^+\pi^-6\pi^0}$ & 0.60 $\pm$ 0.02 \\

584: ${\cal R}\to{\pi^+\pi^-\pi^+\pi^-8\pi^0}$ & 0.60 $\pm$ 0.02 \\

585: 

586: ${\cal R}\to{K^+K^-K^+K^-K^+K^-}$ & 0.68 $\pm$ 0.04 \\

587: ${\cal R}\to{p^+p^-p^+p^-p^+p^-}$ & 0.53 $\pm$ 0.04 \\

588: ${\cal R}\to{\pi^+\pi^-\pi^+\pi^-\pi^+\pi^-}$ & 0.74 $\pm$ 0.03 \\ 

589: {\boldmath ${\cal R}\to{K^+K^-K^+K^-\pi^0}$ 

590: \unboldmath} & {\bf 0.48 $\pm$  0.02} \\ \hline

591: \end{tabular}

592: \end{center}

593: \caption{Average efficiencies for the reconstruction of 

594: various decay modes 

595: that could be detected in this analysis, obtained by

596: fitting the photon energy-dependent reconstruction efficiencies to a straight line

597: in the interval $1.0$ GeV $<E_\gamma<$ 4.5 GeV (statistical

598: errors only). Quoted efficiencies correspond to 5000-event Monte Carlo

599: samples for which photons are generated with

600: a flat angular distribution within the barrel fiducial

601: volume of the calorimeter, and therefore

602: are restricted in geometry. The lowest-efficiency

603: final state ($K^+K^-K^+K^-\pi^0$ in bold above) is used for setting upper limits.}

604: \end{table}

605: 

606: \begin{figure}[htpb]

607: \centerline{\includegraphics[width=8cm]{3031106-023.eps}}

608: \caption{The efficiency for detecting an $\Upsilon\to\gamma+{\cal R}$, 

609: ${\cal R}\to2(K^+K^-)\pi^0$ event as a function of observed photon energy $E_\gamma$ as determined

610: from a 25000-event Monte Carlo sample.

611: Each point in this efficiency is obtained from a different 

612: ${\cal R}$ mass hypothesis.  This photon energy-dependent efficiency 

613: correction distribution is used to point-wise correct our upper limit, where the efficiency between

614: points in this distribution is estimated by linear interpolation.}

615: \label{fig:4Kpi0Efficiency}

616: \end{figure}

617: 

618: \section*{Results}

619: 

620: To convert the efficiency-corrected upper limit contour into an

621: upper limit on the two-body radiative branching ratio ${\cal B}(\gamma {\cal R})$,

622: we simply divide the efficiency-corrected 

623: upper limit contour by the total

624: calculated number of $\Upsilon$(nS) decays\cite{r:shawn}, as shown

625: in Table~III.  For completeness, we also include the results for the $\Upsilon$(4S),

626: for which the decay width is expected to be nearly saturated by

627: $\Upsilon$(4S)$\to B{\overline B}$\cite{r:pdg}.  The total number of $\Upsilon$(4S) 

628: events was obtained by multiplying the total luminosity of our on-resonance $\Upsilon$(4S)-running 

629: by the well-known $\Upsilon$(4S) on-resonance cross-section (which we calculate from \cite{r:4sxsect}).

630: The resulting on-resonance 

631: upper limits ${\cal B}(\gamma {\cal R})$ are shown in Figure~\ref{fig:resonanceLimit}.

632: 

633: 

634: Given the fact that we have not performed a continuum 

635: subtraction on the on-resonance inclusive photon

636: spectrum from $\Upsilon$ decays, it is interesting to compare the structure

637: observed in Figure~\ref{fig:resonanceLimit} with structure observed when we

638: apply the fitting procedure to continuum data.

639: Figures~\ref{fig:limit1sVb1s}, \ref{fig:limit2sVb2s}, 

640: \ref{fig:limit3sVb3s} and \ref{fig:limit4sVb4s}

641: show the resonances' limits of Figure~\ref{fig:resonanceLimit} 

642: separately, with the 

643: corresponding continuum distributions

644: overlaid for comparison.

645: We observe 

646: a partial

647: correlation between the continuum and resonance spectra, suggesting that

648: both spectra have large contributions from initial state radiation (ISR)

649: photons.

650: 

651: \begin{figure}[htpb]

652: \centerline{\includegraphics[width=8cm]{3031106-024.eps}}

653: \caption{The $M_{\cal R}$-dependent ${\cal B}(\gamma {\cal R})$ upper limit contours obtained for 

654: $\Upsilon\to\gamma+{\cal R}$, ${\cal R}\to\geq4$ charged tracks for the 

655: $\Upsilon$(1S), $\Upsilon$(2S), $\Upsilon$(3S)

656: and $\Upsilon$(4S). Limits are obtained by dividing upper limits on yield by

657: reconstruction efficiency and number of resonant events, and also

658: incorporating systematic uncertainties. 

659: All limits are of order ${\cal B}(\gamma {\cal R})\approx10^{-4}$.}

660: \label{fig:resonanceLimit}

661: \end{figure}

662: 

663: \begin{figure}[htpb]

664: \centerline{\includegraphics[width=8cm]{3031106-025.eps}}

665: \caption{Comparison of the 

666: $M_{\cal R}$-dependent Gaussian fit area upper limit 

667: $A(M_{\cal R})+1.645\cdot\sigma_A(M_{\cal R})$  

668: for the $\Upsilon$(1S) versus the below $\Upsilon$(1S) continuum;

669: we observe some correlation

670: between the resonance and the below-resonance structure.  Note that the normalization of the below-resonance upper limit curve is arbitrary and has been adjusted so as to allow an easy visual

671: comparison.}

672: \label{fig:limit1sVb1s}

673: \end{figure}

674: 

675: \begin{figure}[htpb]

676: \centerline{\includegraphics[width=8cm]{3031106-026.eps}}

677: \caption{Comparison of the $M_{\cal R}$-dependent Gaussian fit area 

678: upper limit $A(M_{\cal R})+1.645\cdot\sigma_A(M_{\cal R})$  

679: for the $\Upsilon$(2S) versus the below $\Upsilon$(2S) continuum;

680: scaling of the continuum as before.}

681: \label{fig:limit2sVb2s}

682: \end{figure}

683: 

684: \begin{figure}[htpb]

685: \centerline{\includegraphics[width=8cm]{3031106-027.eps}}

686: \caption{Comparison of the $M_{\cal R}$-dependent Gaussian fit area 

687: upper limit $A(M_{\cal R})+1.645\cdot\sigma_A(M_{\cal R})$

688: for the $\Upsilon$(3S) versus the below $\Upsilon$(3S) continuum;

689: scaling of the continuum as before.}

690: \label{fig:limit3sVb3s}

691: \end{figure}

692: 

693: \begin{figure}[htpb]

694: \centerline{\includegraphics[width=8cm]{3031106-028.eps}}

695: \caption{Comparison of the $M_{\cal R}$-dependent Gaussian fit area 

696: upper limit $A(M_{\cal R})+1.645\cdot\sigma_A(M_{\cal R})$

697: for the $\Upsilon$(4S) versus the below $\Upsilon$(4S) continuum;

698: scaling of the continuum as before.}

699: \label{fig:limit4sVb4s}

700: \end{figure}

701: 

702: \begin{table}[htpb]

703: \label{tab:eventNumber}

704: \begin{center}

705: \begin{tabular}{|c|c|}\hline

706: $\Upsilon$ Resonance & $N_{\rm total}(\Upsilon$(nS)) $(\times 10^6)$ \\

707: \hline

708: $\Upsilon$(1S) & 20.96 $\pm$ 0.06\\

709: $\Upsilon$(2S) & 8.33 $\pm$ 0.04\\

710: $\Upsilon$(3S) & 5.24 $\pm$ 0.06\\

711: $\Upsilon$(4S) & 6.8 $\pm$ 0.2 \\

712: \hline

713: \end{tabular}

714: \end{center}

715: \caption{The total number of calculated $\Upsilon$(1S), 

716: $\Upsilon$(2S), $\Upsilon$(3S) and $\Upsilon$(4S) events in our data samples\cite{r:shawn}.}  

717: \end{table}

718: 

719: Applying our fitting procedure directly to the continuum we can obtain limits on the cross-section for $e^+e^-\to\gamma+{\cal R}$, over the barrel angular acceptance region (Figure~\ref{fig:offResonanceLimit}). It is important to note here that a) the angular distribution for continuum initial state radiation processes is considerably more forward-peaked than the $1+\cos^2\theta$ distribution we have assumed for the resonance; we have therefore applied a correction based on the angular distribution appropriate to ISR, and b) the quantum numbers of particles produced in association with ISR photons are different than those produced in radiative decays of quarkonium resonances.\begin{figure}[htpb]\centerline{\includegraphics[width=8cm]{3031106-029.eps}}\caption{$M_{\cal R}$-dependent cross-section upper limit contours obtained for $e^+e^-\to\gamma+{\cal R}$, ${\cal R}\to\geq4$ charged tracks for the below $\Upsilon$(1S), $\Upsilon$(2S), $\Upsilon$(3S) and $\Upsilon$(4S) continua (nb). This plot is obtained by dividing the result of our fitting procedure on the continuum by the off-resonance luminosity. The angular correction here is based on the expected distribution appropriate for continuum initial state radiation. Systematic errors have also been incorporated into these limits.}\label{fig:offResonanceLimit}\end{figure} To set the scale of the continuum cross-section sensitivity, the raw ISR cross-section for $e^+e^-\to J/\psi+\gamma$ is  expected to be $\sim$5 pb in the 10 GeV center-of-mass regime. Taking into account the efficiency of our event selection requirements and the strong forward peaking expected for ISR processes, this corresponds to an expected observed cross-section into $\ge$4 charged tracks $\sim10^{-4}$ nb. This value is at the edge of our current statistical sensitivity. 

720: 

721: \section*{Cross-Check}

722: 

723: In order to check that we are able to identify a 

724: signal at a given sensitivity level, we embedded pure Monte Carlo signal 

725: into data, and 

726: performed our fitting procedure on the resulting 

727: distribution to ensure that we recover the correct signal magnitude in 

728: our branching ratio upper limit.  To do this, 

729: hypothetical $\Upsilon$(4S)$\to\gamma+{\cal R}$, 

730: ${\cal R}\to\pi^+\pi^-\pi^+\pi^-$ events were embedded into the 

731: $\Upsilon$(4S) inclusive

732: photon spectrum with branching ratios of the order of 

733: $10^{-5}$, $10^{-4}$, $10^{-3}$, 

734: and $10^{-2}$ under $10$ different $M_{\cal R}$ hypotheses:

735: $M_{\cal R}=0.6$ GeV, $1.5$ GeV, $2.5$ GeV, $3.5$ GeV, $4.5$ GeV, $6.5$ GeV, 

736: $7.5$ GeV, $8.5$ GeV, and $9.5$ GeV.

737: The resulting upper limit contours

738: derived from applying our procedure to these spectra are show in 

739: Figure~\ref{fig:fake_signal_upper_limit}.

740: We reconstruct all signals within our expected 

741: sensitivity (around $10^{-4}$) that

742: are within our accessible recoil mass range.

743: 

744: \begin{figure}[htpb]

745: \centerline{\includegraphics[width=8cm]{3031106-030.eps}}

746: \caption{Upper limit contours derived from applying our procedure to 

747: fabricated Monte Carlo signal spectra.  We reconstruct all input signals withing our sensitivity 

748: ($\approx10^{-4}$) that are within our accessible recoil mass range.}

749: \label{fig:fake_signal_upper_limit}

750: \end{figure}

751: 

752: 

753: \section*{Systematic errors}

754: We identify and account for systematic errors as follows:

755: 

756: \begin{enumerate}

757: \item We account for possible

758: systematics in our event and shower reconstruction efficiency by using 

759: the lowest-efficiency final state considered, and by assuming ${\cal R}$ has spin=0.  

760: The angular distributions for spin=0, 1 and 2 two-body decays have been calculated, and generally

761: yield flatter distributions for higher spins \cite{r:pipi05}.

762: \item We have assessed fitting systematic uncertainties by varying the

763: recoil mass bin width (from 20\% to 50\% of the resolution $\sigma$) and the

764: order of the background polynomial used to parametrize the background

765: (from second-order to fifth-order). Observing no statistically significant

766: variation between these extremes, we have used as defaults

767: $\sigma$=5 bins and a fourth-order background, based on the goodness

768: of fit of the pull distributions to a unit Gaussian on the continuum.

769: \item For continuum

770: measurements, we assess a uniform 1\% degradation of the limit 

771: due to the luminosity uncertainty

772: as calculated in the previous analysis \cite{r:shawn}.

773: \item For on-resonance measurements,

774: we degrade the limit uniformly by the uncertainty in the calculated number of total resonant events given in Table~III.  

775: \end{enumerate}

776: 

777: \subsection*{Summary}

778: As shown in Figure~\ref{fig:resonanceLimit}, our sensitivity

779: is of order $10^{-4}$ across the mass range 

780: corresponding to 0.2$<z_\gamma<$1.0,

781: well above the tabulated branching ratios for any known

782: $\Upsilon\to\gamma$+X, X$\to{h^+h^-h^+h^-}$+neutrals

783: process.

784: We measure upper limits of:

785: \begin{center}

786: ${\cal B}(\Upsilon$(1S)$\to\gamma+{\cal R}, {\cal R}\to\geq4$ charged tracks$) < 1.26\times10^{-3}$, \\

787: ${\cal B}(\Upsilon$(2S)$\to\gamma+{\cal R}, {\cal R}\to\geq4$ charged tracks$) < 9.16\times10^{-4}$, \\

788: ${\cal B}(\Upsilon$(3S)$\to\gamma+{\cal R}, {\cal R}\to\geq4$ charged tracks$) < 9.69\times10^{-4}$ \\

789: \end{center}

790: for all kinematically allowed 

791: masses $M_{\cal R}$, under the assumption that ${\cal R}$ is a pseudoscalar.

792: Constraining $1.5$ GeV $<M_{\cal R}<5.0$ GeV we set a more stringent limit of:

793: \begin{center}

794: ${\cal B}(\Upsilon$(1S)$\to\gamma+{\cal R}, {\cal R}\to\geq4$ charged tracks$) < 1.78\times10^{-4}$, \\

795: ${\cal B}(\Upsilon$(2S)$\to\gamma+{\cal R}, {\cal R}\to\geq4$ charged tracks$) < 1.95\times10^{-4}$, \\ 

796: ${\cal B}(\Upsilon$(3S)$\to\gamma+{\cal R}, {\cal R}\to\geq4$ charged tracks$) < 2.20\times10^{-4}$.

797: \end{center}

798: Additionally, we report these upper limits as a function of the

799: mass recoiling against the photon, as shown

800: in Figure~\ref{fig:resonanceLimit}.  

801: 

802: We limit the branching ratio 

803: for two-body radiative decays to narrow resonances

804: ($<$20 MeV in width)

805: to be $\le 10^{-4}$.

806: We conclude that

807: distortion of the inclusive photon spectrum in our 

808: previous extraction of $\alpha_s$ due to the possible

809: contribution of such events 

810: is negligible.

811: The possibility of resonances with widths greater than our experimental

812: resolution has yet to be completely addressed.

813: Further work on exclusive multiparticle final states (e.g., $\gamma 2\pi^+ 2\pi^-$,

814: $\gamma 2K^+ 2K^-$, $\gamma K^0K^0$ and $\gamma K^0 K^\pm \pi^\mp$) would help elucidate

815: the nature of such radiative decays.

816: 

817: \section{Acknowledgments}

818: We gratefully acknowledge the effort of the CESR staff

819: in providing us with excellent luminosity and running conditions.

820: D.~Cronin-Hennessy and A.~Ryd thank the A.P.~Sloan Foundation.

821: This work was supported by the National Science Foundation,

822: the U.S. Department of Energy, and

823: the Natural Sciences and Engineering Research Council of Canada.

824: 

825: 

826: \begin{thebibliography}{99}

827: 

828: \bibitem{r:shawn}D. Besson {\it et al.} (CLEO Collaboration), 

829: [arXiv:hep-ex/0512061],

830: Phys. Rev. {\bf D74}, 012003 (2006).

831: \bibitem{r:Field} R.D. Field, Phys. Lett. {\bf B133}, 248 (1983).

832: \bibitem{r:SotoGarcia}X. Garcia and J. Soto, 

833: [arXiv:hep-ph/0511167],

834: Phys. Rev. Lett. {\bf 96}, 111801 (1996).

835: \bibitem{r:pipi05}S. B. Athar {\it et al.} (CLEO Collaboration), 

836: [arXiv:hep-ex/0510015],

837: Phys. Rev. {\bf D73}, 032001 (2006).

838: \bibitem{r:pi0pi006}D. Besson {\it et al.} (CLEO Collaboration), 

839: [arXiv:hep-ex/0512003], Phys. Rev. {\bf D75}, 072001 (2007).

840: \bibitem{r:potlia} B. Athar {\it et al.} (CLEO Collaboration), 

841: submitted to Phys. Rev. D.

842: \bibitem{r:pdg} W.-M. Yao {\it et al.}, Journal of Physics G {\bf 33}, 1 

843: (2006).

844: \bibitem{r:xavier}X. Garcia i Tormo, Ph. D. Thesis, University of Barcelona, 2007 (unpublished).

845: \bibitem{r:Brod-Lep-Mack} S. J. Brodsky, G. P. Lepage and P. B. Mackenzie,

846:                           Phys. Rev. {\bf D28}, 228 (1983).

847: \bibitem{r:joan07}N. Brambilla, X. Garcia i Tormo, J. Soto, A. Vario, 

848: [arXiv:hep-ph/0702079].

849: \bibitem{r:4sxsect}S. B. Athar {\it et al.} (CLEO Collaboration), 

850: [arXiv:hep-ex/0202033], Phys. Rev. {\bf D66}, 052003 (2002).

851: 

852: \end{thebibliography}

853: 

854: \end{document}

855: 

856: 

857: 

858: 

859: 

860: 

861: