J. BELZtex2html_wrap_inline389[(a)], R.D. COUSINStex2html_wrap_inline391, M.V. DIWANtex2html_wrap_inline393[(b)], M. ECKHAUSEtex2html_wrap_inline395, K.M. ECKLUNDtex2html_wrap_inline393[(c)], A.D. HANCOCKtex2html_wrap_inline395, V.L. HIGHLANDtex2html_wrap_inline389[(d)], C. HOFFtex2html_wrap_inline395, G.W. HOFFMANNtex2html_wrap_inline405, G.M. IRWINtex2html_wrap_inline393, J.R. KANEtex2html_wrap_inline395, S.H. KETTELLtex2html_wrap_inline389[(b)], J.R. KLEINtex2html_wrap_inline413[(e)], Y. KUANGtex2html_wrap_inline395, K. LANGtex2html_wrap_inline405, R. MARTINtex2html_wrap_inline395, M. MAYtex2html_wrap_inline421, J. McDONOUGHtex2html_wrap_inline405[(e)], W.R. MOLZONtex2html_wrap_inline425, P.J. RILEYtex2html_wrap_inline405, J.L. RITCHIEtex2html_wrap_inline405, A.J. SCHWARTZtex2html_wrap_inline413, A. TRANDAFIRtex2html_wrap_inline389, B. WAREtex2html_wrap_inline405[(f)], R.E. WELSHtex2html_wrap_inline395, S.N. WHITEtex2html_wrap_inline421, M.T. WITKOWSKItex2html_wrap_inline395[(g)], S.G. WOJCICKItex2html_wrap_inline393, and S. WORMtex2html_wrap_inline405[(h)]
(BNL E888 Collaboration)

(1) Brookhaven National Laboratory, Upton, NY 11973
(2) University of California, Irvine, California 92717
(3) University of California, Los Angeles, California 90024
(4) Princeton University, Princeton, NJ 08544
(5) Stanford University, Stanford, California 94309
(6) Temple University, Philadelphia, Pennsylvania 19122
(7) University of Texas at Austin, Austin, Texas 78712
(8) College of William and Mary, Williamsburg, Virginia 23187

We have searched for a neutral H dibaryon decaying via tex2html_wrap_inline449 and tex2html_wrap_inline451. Our search has yielded two candidate events from which we set an upper limit on the H production cross section. Normalizing to the inclusive tex2html_wrap_inline455 production cross section, we find tex2html_wrap_inline457 at 90% C.L., for an H of mass tex2html_wrap_inline4612.15 GeV/tex2html_wrap_inline463.

The theory of quantum chromodynamics imposes no specific limitation on the number of quarks composing hadrons other than that they form color singlet states. Although only qqq and tex2html_wrap_inline467 states have been observed, other combinations can form color singlets. Jaffe[1] has proposed that a six-quark state uuddss may have sufficient color-magnetic binding to be stable against strong decay. Such a state, which Jaffe named H, would decay weakly, and the resultant long lifetime would allow the possibility of observing such particles in neutral beams. Theoretical estimates[2] of tex2html_wrap_inline471 have varied widely, ranging from a deeply bound state with tex2html_wrap_inline471<2.10 GeV/tex2html_wrap_inline463 to a slightly unbound state with tex2html_wrap_inline471 near the tex2html_wrap_inline481 threshold, 2.23 GeV/tex2html_wrap_inline463. In this mass range the H would decay almost exclusively to tex2html_wrap_inline487, tex2html_wrap_inline489, and tex2html_wrap_inline491[3]. Several previous experiments have searched for H's but with no compelling success[4]. The search described here is sensitive to H's having mass and lifetime in a previously unexplored range.

We have searched for tex2html_wrap_inline449 and tex2html_wrap_inline499 decays by looking in a neutral beam for tex2html_wrap_inline501 decays in which the tex2html_wrap_inline455 momentum vector does not point back to the production target. The experiment, E888, was performed in the B5 beamline of the Alternating Gradient Synchotron (AGS) of Brookhaven National Laboratory. A second phase of the experiment searched for long-lived H's by using a diffractive dissociation technique[5]. The detector used for the decay search (Fig. 1) was essentially that used for the E791 rare kaon decay experiment and has been described in detail elsewhere[6].

Figure 1: The E888 detector and beamline.

In brief, a neutral beam was produced using the 24 GeV/c proton beam from the AGS incident on a 1.4 interaction length Cu target. The targeting angle was 48 mr. After passing through a series of collimators and two successive sweeping magnets, the neutral beam entered a 10 m long vacuum decay tank within which candidate tex2html_wrap_inline455's decayed. Downstream of the tank was a two arm spectrometer consisting of two magnets with approximately equal and opposite tex2html_wrap_inline511 impulses and 5 drift chamber (DC) stations located before, after, and in between the magnets. Downstream of the spectrometer on each side of the beam were a pair of trigger scintillator hodoscopes (TSCs), a threshold Cherenkov counter (CER), a lead-glass array (PbG), 0.91 m of iron to filter out hadrons, a muon-detecting hodoscope (MHO), and a muon rangefinder (MRG) consisting of marble and aluminum slabs interspersed with streamer tubes. For the first half of the run the Cherenkov counters were filled with a He-N mixture (n = 1.000114) to identify electrons; for the second half the left-side counter was filled with freon (n = 1.0011) to identify protons from tex2html_wrap_inline501 (due to lack of light). Only the left counter was used for this purpose as the soft pion from tex2html_wrap_inline501 decay is accepted only when on the right; when it is on the left, the first magnet bends it back across the beamline and it is not reconstructed. The lead-glass array (PbG) consisted of two layers: a layer of front blocks 3.3 radiation lengths (r.l.) deep and a layer of back blocks 10.5 r.l.\ deep. The PbG was used to identify electrons by comparing the total energy deposited (tex2html_wrap_inline521) with the track's momentum. A minimum bias trigger was defined as a coincidence between all 4 TSC counters and signals from the 3 most upstream DC stations. A Level 1 trigger (L1) was formed by putting minimum bias triggers in coincidence with veto signals from the Cherenkov counters and muon hodoscope. All events passing L1 were passed to a Level 3 software trigger which used hit information from the first 3 DC stations to calculate an approximate two-body mass. Events with tex2html_wrap_inline523<1.131 GeV/tex2html_wrap_inline463 were written to tape.

Offline, all events containing two opposite-sign tracks forming a loose vertex were kinematically fit[6] and subjected to the following cuts: there could be at most one extra track-associated hit or one missing hit in the ten DC planes which measure the x (bending) view of each track; the tex2html_wrap_inline531's per degree of freedom resulting from the track and vertex fits had to be of good quality; the tex2html_wrap_inline455 vertex had to be within the decay tank and downstream of the fringe field of the last sweeper magnet; both tracks had to be accepted by CER, PbG, MHO, and MRG detectors and have p>1 GeV/c; neither track could intersect significant material such as the flange of the vacuum window; to reject background from tex2html_wrap_inline541, tex2html_wrap_inline543 had to be tex2html_wrap_inline545; and to reject background from tex2html_wrap_inline547 resulting from secondary interactions, tex2html_wrap_inline549 had to be >4 times the mass resolution of tex2html_wrap_inline553 decays (1.55 MeV/tex2html_wrap_inline463).

Events passing these cuts were subjected to particle identification criteria in order to reject background from tex2html_wrap_inline557 (tex2html_wrap_inline559) and tex2html_wrap_inline561 (tex2html_wrap_inline563) decays. To reject electrons, we require that there be no track-associated Cherenkov hit and that tracks with p>2 GeV/c (<2 GeV/c) have tex2html_wrap_inline573 (<0.52). The low momentum track on the right side of the detector was required to deposit <0.66tex2html_wrap_inline579tex2html_wrap_inline521 in the front PbG blocks. To reject muons which passed the MHO veto in the trigger, we cut events with a hit in the MRG which was consistent with the projection of a track and which corresponded to at least 65% of the expected range of a muon with that track's momentum.

Lambda candidates were selected by requiring that tex2html_wrap_inline583 be less than 4 times the mass resolution of tex2html_wrap_inline501 decays (0.55 MeV/tex2html_wrap_inline463). The data was then divided into two streams: a normalization stream consisting of tex2html_wrap_inline455's which project back to the production target, and a signal stream consisting of tex2html_wrap_inline455's which do not. The former were selected by requiring that the square of the collinearity angle tex2html_wrap_inline593 be less than 1.5 mradtex2html_wrap_inline595, where tex2html_wrap_inline593 is the angle between the reconstructed tex2html_wrap_inline455 momentum vector and a line connecting the production target with the decay vertex. This sample contains negligible background. The signal sample was selected by requiring that tex2html_wrap_inline511>145 MeV/c, where tex2html_wrap_inline511 is the tex2html_wrap_inline455 momentum transverse to the line connecting the production target with the decay vertex. This cut value was chosen to eliminate tex2html_wrap_inline611 decays, which have a kinematic endpoint of 135 MeV/c. The tex2html_wrap_inline511 distribution of tex2html_wrap_inline455's from two-body tex2html_wrap_inline449 decays exhibit an approximate Jacobian peak (not exact because the vertex is the tex2html_wrap_inline455's) with an endpoint which depends upon tex2html_wrap_inline471. A large fraction of high-tex2html_wrap_inline511 tex2html_wrap_inline455's were found to project back to a collimator located just upstream of the decay tank. We thus required that the point in our beamline to which a tex2html_wrap_inline455 projects back be located downstream of this collimator: tex2html_wrap_inline631 m.

A signal region for H candidates was defined by the criteria tex2html_wrap_inline635 MeV/c and tex2html_wrap_inline639, where tex2html_wrap_inline641 is the distance in proper lifetimes between the decay vertex and the nearest material (beamline element) to which the momentum vector projects back. The tex2html_wrap_inline511 cut rejects tex2html_wrap_inline645 decays which survive the CER, PbG, MHO, and MRG vetoes due to detector inefficiency, while the tex2html_wrap_inline641 cut rejects tex2html_wrap_inline455's which originate from collimators, flanges, and other beamline elements. All cuts were determined without looking at events in the signal region, in order that our final limit on H's be unbiased. After fixing cuts we looked in the signal region and observed two events. The estimated background is 0.15 events from tex2html_wrap_inline455's originating from beamline elements, and <0.21 events from tex2html_wrap_inline645 decays (all tex2html_wrap_inline559 as the tex2html_wrap_inline511 is too high for tex2html_wrap_inline563). The former is estimated by studying the tex2html_wrap_inline641 distribution of tex2html_wrap_inline455's originating from a ``hot'' flange located immediately upstream of 9.65 m. The latter is estimated by first counting the number of final events cut because the low-momentum track had tex2html_wrap_inline669 (these are electrons); this is then multiplied by the ratio of the number of electrons passing PbG analysis cuts to the number having tex2html_wrap_inline669, as determined from a sample of tex2html_wrap_inline559 decays. The tex2html_wrap_inline641 vs. tex2html_wrap_inline511 plot for the final high-tex2html_wrap_inline511 tex2html_wrap_inline455 sample is shown in Fig. 2. In this figure the Cherenkov veto for the freon counter is not imposed. A band of tex2html_wrap_inline563 decays is visible at tex2html_wrap_inline685 MeV/c which results from the tex2html_wrap_inline689 MeV/c cut and the tex2html_wrap_inline693 cut; this latter cut constrains tex2html_wrap_inline511 from above. tex2html_wrap_inline455's which originate from beamline elements are visible at low tex2html_wrap_inline641. For the freon subset, when we require that there be no signal in the left Cherenkov counter, all but two tex2html_wrap_inline563 decays are eliminated while all tex2html_wrap_inline455's at low tex2html_wrap_inline641 remain.

Figure 2: tex2html_wrap_inline641 vs. tex2html_wrap_inline511 for the high-tex2html_wrap_inline511 tex2html_wrap_inline455 sample. The signal region is denoted by dashed lines. The band of events from tex2html_wrap_inline511=145 to tex2html_wrap_inline511tex2html_wrap_inline461150 MeV/c are tex2html_wrap_inline563 decays; the leftmost edge is due to a tex2html_wrap_inline511 cut, while the rightmost edge is due to a lower cut on tex2html_wrap_inline727.

Figure 3: tex2html_wrap_inline729 vs. tex2html_wrap_inline511 for the final high-tex2html_wrap_inline511 tex2html_wrap_inline455 sample. The two events in the signal region are circled. The cluster of events at tex2html_wrap_inline685 GeV/c, tex2html_wrap_inline741 GeV/tex2html_wrap_inline463 are consistent with Monte Carlo simulated tex2html_wrap_inline563 decays. There is a third event which is well-separated from the tex2html_wrap_inline563 decays and which lies just outside the signal region; the 3 separated events are consistent with tex2html_wrap_inline449 decay if tex2html_wrap_inline751 GeV/tex2html_wrap_inline463.

Also visible in Fig. 2 are our two candidates, which have tex2html_wrap_inline511 of 187 and 191 MeV/c and tex2html_wrap_inline641 of 6.7 and 9.4. The tex2html_wrap_inline511 values correspond to a Jacobian peak from tex2html_wrap_inline449 decay if tex2html_wrap_inline751 GeV/tex2html_wrap_inline463. The probability for a tex2html_wrap_inline563 decay to have such high tex2html_wrap_inline511 is extremely small, as it is kinematically forbidden for a tex2html_wrap_inline563 decay to have both tex2html_wrap_inline693 and tex2html_wrap_inline777 MeV/c (Fig. 3). The probability for a tex2html_wrap_inline559 decay to look like these events is also very small, as the PbG response for the electron candidate tracks is very uncharacteristic of electrons: tex2html_wrap_inline783 = 0.44 and 0.27, and for both events tex2html_wrap_inline785 = 0 (Fig. 4). This response is typical of pions from tex2html_wrap_inline501 decay. To investigate background from neutrons in the beam interacting with residual gas molecules in the decay tank, we recorded and analyzed a sample of data equivalent to 1% of the total sample with the decay tank vacuum spoiled by a factor tex2html_wrap_inline789. This sample yielded one event in the signal region, implying a background level in the rest of the data of 0.04 events. We also studied potential background from tex2html_wrap_inline611 decays where the tex2html_wrap_inline793 originates from a beamline element; from Monte Carlo simulation and the number of tex2html_wrap_inline455's observed originating from beamline elements, we estimate a background of less than 0.10 events. The total background estimate from known sources is less than 0.50 events. The probability of 0.50 events fluctuating up to two or more events is 0.090; if such a fluctuation occurred, it is remarkable that the tex2html_wrap_inline511 of the events is so similar.

Figure 4: tex2html_wrap_inline799 (PbG) vs. tex2html_wrap_inline783 for: a) the low momentum track of tex2html_wrap_inline455's from the final high-tex2html_wrap_inline511 sample with only the PbG cuts relaxed, and b) low momentum electrons from tex2html_wrap_inline559 decay. In (a), the tracks from the two events in the signal region are circled. There are 4.7 times as many events in (b) as in (a).

Figure 5: 90% C.L. upper limits on the H production cross section as a function of tex2html_wrap_inline471 or tex2html_wrap_inline813 (see Ref. [3]). The dashed contour corresponds to an H lifetime half that given on the top scale.

A 90% C.L. upper limit on the H production cross section can be expressed in terms of the inclusive tex2html_wrap_inline455 production cross section as follows:
where tex2html_wrap_inline821 is the number of tex2html_wrap_inline455's originating from the target, tex2html_wrap_inline825 and tex2html_wrap_inline827 are geometric acceptances for tex2html_wrap_inline455's originating from the target and from H decays, respectively, tex2html_wrap_inline833 and tex2html_wrap_inline835 are branching ratios, tex2html_wrap_inline837 is the inclusive tex2html_wrap_inline455 production cross section, and tex2html_wrap_inline841 is the factor which multiplies the single-event sensitivity to give the value of tex2html_wrap_inline843 which has a 10% chance of producing tex2html_wrap_inline8452 detected events. Here we conservatively assume no background and take tex2html_wrap_inline847. The acceptance tex2html_wrap_inline827 accounts for the fact that tex2html_wrap_inline455's from H's must project back to a restricted region of the beamline. Since tex2html_wrap_inline501 decays are common to both signal and normalization channels, all trigger and detection efficiencies divide out of Eq. (1).

The acceptances tex2html_wrap_inline857 and tex2html_wrap_inline859 were determined from Monte Carlo simulation using several different estimates of the production momentum spectra. For the H simulation, a central production spectrum was used with a broad peak at tex2html_wrap_inline863. A spectrum corresponding to a tex2html_wrap_inline481 coalescence model for H production [7] resulted in a limit on tex2html_wrap_inline843 about 50% lower. We quote here the more conservative limit resulting from the central production spectrum. The inclusive tex2html_wrap_inline455 production spectrum was taken from a measurement by Abe et al.[8]; comparison with our data shows very good agreement.

The acceptance tex2html_wrap_inline859 also depends crucially on H lifetime. Here we assume the relationship between tex2html_wrap_inline471 and tex2html_wrap_inline813, tex2html_wrap_inline881, and tex2html_wrap_inline883 calculated by Donoghue et al.[3], and obtain 90% C.L. upper limits on tex2html_wrap_inline885 as a function of tex2html_wrap_inline471. Our acceptance is maximum for tex2html_wrap_inline889 ns and becomes small for tex2html_wrap_inline891 ns due to the tex2html_wrap_inline631 m cut. Our limits for tex2html_wrap_inline885 are plotted in Fig. 5. For tex2html_wrap_inline897 GeV/tex2html_wrap_inline463, Jaffe's original prediction,
From Abe et al.[8], tex2html_wrap_inline901 mb/sr, so tex2html_wrap_inline903 tex2html_wrap_inline905b/sr. For tex2html_wrap_inline907 GeV/tex2html_wrap_inline463, consistent with the observed tex2html_wrap_inline455 tex2html_wrap_inline511, the acceptance is lower and the two candidate events correspond to a differential cross section of tex2html_wrap_inline915 tex2html_wrap_inline905b/sr. The authors of Ref. [3] note that tex2html_wrap_inline813 may be shorter than their predicted value by up to a factor of two; this would increase our acceptance for tex2html_wrap_inline921 GeV/tex2html_wrap_inline463 and decrease our acceptance for tex2html_wrap_inline471 greater than this value. The resultant 90% C.L. upper limits are plotted as the dashed contour in Fig. 5. If we assume that the invariant cross section tex2html_wrap_inline927 has the form tex2html_wrap_inline929, then our limit (2) corresponds to tex2html_wrap_inline931 nb for a wide range of parameters B and C.

There are few theoretical predictions of the H production cross section. Cousins and Klein[7] predict a differential cross section of tex2html_wrap_inline939100 tex2html_wrap_inline905b/sr for p-Cu interactions at our targeting angle based on a tex2html_wrap_inline481 coalescence model. Cole et al.[9] considers tex2html_wrap_inline481 and tex2html_wrap_inline949 coalescence and predicts tex2html_wrap_inline951 for p-Cu collisions at AGS energies; taking the inelastic cross section tex2html_wrap_inline955 to be tex2html_wrap_inline461780 mb[10] gives tex2html_wrap_inline959 tex2html_wrap_inline905b. Rotondo[11] considers only tex2html_wrap_inline949 coalescence and predicts a total cross section at Fermilab energies of 1.2 tex2html_wrap_inline905b.

We are indebted to the E791 and E871 collaborations, which built or supported most of the apparatus used here. We thank V.L.Fitch for much encouragement, and S. Black, K. Schenk, and N. Mar for their help in various stages of this work. We gratefully acknowledge the strong support of BNL, in particular R. Brown, A. Pendzick, the AGS staff, and the C.C.D. We also thank the SLAC computing division and the Princeton C.I.T., where all the data was reconstructed. This work was supported in part by the U.S. Department of Energy, the National Science Foundation, and the R.A. Welch Foundation.

Brent Ware
Sun Dec 21 13:13:28 PST 1997