the bremsstrahlung tagged photon beam in hall b at jlab · measurement of incident photon energy in...

*Corresponding author. Permanent address: Montgomery College, Rockville, MD 20850, USA. Tel.: #1-202-319-5319; fax:#1-202-319-4448.

E-mail address: [email protected] (J.T. O'Brien)1Current address: Oregon Hearing Research Center, Portland, OR 97201-3098, USA.2Current address: Department of Physics, University of Texas at El Paso, El Paso, TX 79968, USA.3Current address: Cisco Systems, Herndon, VA 20171, USA.4Current address: Department of Physics, University of Massachusetts, Amherst, MA 01003, USA.

Nuclear Instruments and Methods in Physics Research A 440 (2000) 263}284

The bremsstrahlung tagged photon beam in Hall Bat JLab

D.I. Sober!, Hall Crannell!, Alberto Longhi!, S.K. Matthews!,1, J.T. O'Brien!,*,B.L. Berman", W.J. Briscoe", Philip L. Cole",2, J.P. Connelly",3, W.R. Dodge",

L.Y. Murphy", S.A. Philips", M.K. Dugger#, D. Lawrence#,4, B.G. Ritchie#,E.S. Smith$, James M. Lambert%, E. Anciant&, G. Audit&, T. Auger&, C. Marchand&,

M. Klusman', J. Napolitano', M.A. Khandaker), C.W. Salgado), A.J. Sarty*

!Department of Physics, The Catholic University of America, Washington, DC 20064, USA"Center for Nuclear Studies, Department of Physics, The George Washington University, Washington, DC 20052, USA

#Department of Physics and Astronomy, Arizona State University, Tempe, AZ 85287-1504, USA$The Thomas Jewerson National Accelerator Facility, Newport News, VA 23606, USA

%Department of Physics, Georgetown University, Washington, DC 20057, USA&Service de Physique Nucle&aire, DAPNIA/DSM, CEA-Saclay, France

'Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA)Department of Physics, Norfolk State University, Norfolk, VA 23504, USA

*Department of Physics, Florida State University, Tallahassee, FL 32306-3016, USA

Received 17 May 1999; accepted 17 July 1999

Abstract

We describe the design and commissioning of the photon tagging beamline installed in experimental Hall B at theThomas Je!erson National Accelerator Facility (JLab). This system can tag photon energies over a range from 20% to95% of the incident electron energy, and is capable of operation with beam energies up to 6.1 GeV. A single dipolemagnet is combined with a hodoscope containing two planar arrays of plastic scintillators to detect energy-degradedelectrons from a thin bremsstrahlung radiator. The "rst layer of 384 partially overlapping small scintillators providesphoton energy resolution, while the second layer of 61 larger scintillators provides the timing resolution necessary toform a coincidence with the corresponding nuclear interaction triggered by the tagged photon. The de"nitions of overlapchannels in the "rst counter plane and of geometric correlation between the two planes are determined using digitizedtime information from the individual counters. Auxiliary beamline devices are brie#y described, and performance resultsto date under real operating conditions are presented. The entire photon-tagging system has met or exceeded its designgoals. ( 2000 Elsevier Science B.V. All rights reserved.

0168-9002/00/$ - see front matter ( 2000 Elsevier Science B.V. All rights reserved.PII: S 0 1 6 8 - 9 0 0 2 ( 9 9 ) 0 0 7 8 4 - 6

Fig. 1. Overall geometry of tagging system. Important details referenced in the text include the shape of the magnet pole, thestraight-ahead photon path through the magnet yoke, and the relative locations of the hodoscope E- and ¹-planes. Also shows `typicalaelectron trajectories labeled according to the fraction of the incident energy that was transferred to the photon.

PACS: 29.30.Kv; 29.40.Mc; 29.70.Fm

Keywords: CLAS; Photon tagger; Photon beam; Scintillator hodoscope; Time-based logic

1. Introduction

We report the design, construction, and commis-sioning of the photon-tagging system now in use inHall B at the Thomas Je!erson National Acceler-ator Facility (JLab) for the investigation of real-photon-induced reactions. The tagger was initiallydesigned to be used in conjunction with theCEBAF (Continuous Electron Beam AcceleratorFacility) Large Acceptance Spectrometer (CLAS)[1], and has subsequently also been used in twoadditional experiments which do not make use ofCLAS. While the descriptions in this paper makefrequent reference to correlations of tagger in-formation with the CLAS detector, it is intendedthat the reader understand that all such discussionshave equivalent application to any otherdownstream detector system for photon-inducedinteractions.

The bremsstrahlung tagging technique for directmeasurement of incident photon energy in photo-nuclear interactions is well established [2}4]. The

JLab system is the "rst photon tagger in the multi-GeV energy range to combine high resolution(&10~3E

0) with a broad tagging range (20}95%

of E0).

2. Background and general description

The geometry of our system is sketched in Fig. 1,with additional, more detailed views in Figs. 2 and3. Electrons from the CEBAF accelerator strikea thin target (the `radiatora) just upstream froma magnetic spectrometer (the `taggera). The systemis based upon the electron bremsstrahlung reactionin which an electron of incident energy E

0is `decel-

erateda (scattered) by the electromagnetic "eld ofa nucleus, and in the process emits an energeticphoton (gamma ray). The energy transferred to thenucleus is negligibly small, so the reaction obeys theenergy conservation relation

Ec"E0!E

%

264 D.I. Sober et al. / Nuclear Instruments and Methods in Physics Research A 440 (2000) 263}284

Fig. 2. Views of the assembled tagger magnet; (a) is a side elevation, showing such details as yoke plate pro"les, straight-ahead vacuumpipe through yoke, coil shape and location; (b) is a central transverse section which shows the laminar construction of the yoke as well asadditional detail on relative positions of poles, coils, and vacuum walls; (c) is an isometric sketch of the entrance tip of one pole, showingpole nomenclature used in the text, and some details of pole chamfers and cross section.

where E%is the energy of the outgoing electron and

Ec is the energy of the emitted photon. Since E0

isuniquely determined by the accelerator, a measure-ment of the outgoing electron energy by a magneticspectrometer thus provides a determination of thephoton energy.

At energies above a few MeV, the outgoing elec-tron and photon emerge at very small angles rela-tive to the incident beam direction. The angulardistribution of the photons has a characteristicangle

hc"mc2/E

0

where m is the electron rest mass, while the elec-tron's characteristic angle is given by

h%"h

#Ec/E%

.

At Je!erson Lab energies ('800 MeV), both ofthese angles are of order 1 mr or smaller, so that in"rst approximation both the photon and the elec-

tron are still traveling along the original beamdirection.

Any photons which are produced in the radiatorthus continue straight ahead through the taggermagnet, toward the target farther downstream inthe experimental hall upon which photon-inducedreactions are to be studied. A collimation system tofurther de"ne the photon beam is positioned justdownstream of the tagger, centered on the photonbeamline. This system, which is indicated in Fig. 3,has two sets of interchangeable collimator inserts,with interspaced sweep magnets to clean up anycharged particle background generated in the col-limator walls.

The "eld setting of the tagger magnet is matchedto the incident beam energy so that those electronswhich do not radiate will follow a circular arc justinside the curved edge of the pole face, and will bedirected into a shielded beam dump below the #oorof the experimental hall. Electrons which do give up

D.I. Sober et al. / Nuclear Instruments and Methods in Physics Research A 440 (2000) 263}284 265

Fig. 3. Elevation view of the assembled system and gantry standing in the alcove. As a scale reference, the overall length of the taggermagnet is &6 m. Important details include the ribbed vacuum box, location of radiator, hodoscope enclosure, collimators, shielding,and beam dump entrance.

energy to generate a bremsstrahlung photon ex-perience greater curvature in the tagger "eld, andemerge from the magnet somewhere along thestraight edge of the pole gap. A scintillator hodo-scope along the #at focal plane downstream fromthis straight edge detects these energy-degradedelectrons, and thereby allows for the determinationof the energy of the radiated photon. This hodo-scope must have adequate segmentation to deliverthe desired energy resolution. Its instrumentationmust also provide timing information su$cientlyprecise to allow for correlation with a particularnuclear interaction in the downstream target towithin the 2 ns interval between beam `bucketsafrom the accelerator.

Electrons must be transported to the detectors invacuum, in order to minimize the e!ects of multiplescattering on the achievable energy resolution. As itwas considered impractical to locate such a largedetector system inside a vacuum chamber, it wasnecessary to have a thin exit window located asclose as practical to the detector plane. The vacuumwindow covers an aperture of 9.6 m]20 cm, and iscomposed of aluminized Mylar backed by Kevlar

cloth, with a combined average thickness of&6]10~4 radiation lengths. Construction and in-stallation details of this thin vacuum window havealready been presented in an earlier publication[5].

In the sections which follow, we discuss the op-tical design of the magnet, structural details of themagnet, vacuum, and hodoscope assemblies, andthe details of our electronics readout scheme. We pro-vide brief descriptions of various auxiliary beam-line devices, and we report on the performance ofthe system during its early running period.

3. The magnet

3.1. Optical design

The desire for the largest practical dynamicrange } photon energies from &20% to &95% ofthe initial electron energy } required an open-yokegeometry that would permit such a large rangeof orbit radii to reach the detector array. A goalto achieve energy resolution of &10~3E

0placed


requirements on "eld uniformity, and arguedstrongly in favor of a design with a relatively simplefocal surface. Finally, a strong interest in the possi-bility of future expansion to use electron/photonout-of-plane angular correlations for identi"cationof photon linear polarization demanded magnetoptics that would preserve the transverse angleinformation for the electron, and would optimallymap this distribution onto the focal surface.

The optical design of the tagger magnet was doneat Catholic University [6]. The "nal design isa modi"ed Elbek spectrometer [7] } a single, uni-form-"eld dipole with a total de#ection angle of303, the smallest angle that would allow a sensibleplacement of the beam dump. A magnet with a lar-ger de#ection angle would have been superior fromthe standpoint of optics, but much heavier andmore expensive to build.

At the time that this design was "nalized, theapproved upper limit for the CEBAF acceleratorwas 4 GeV, a beam energy that is very comfortablywithin the capabilities of this magnet. The as-builtdesign, with the present power supply, will in factallow operation at up to &6 GeV with somewhatdegraded "eld uniformity. Planning for possiblefuture upgrades of CEBAF toward operation atthat energy, and perhaps beyond, has indeed beensubsequently authorized.

The de"ning characteristic of an Elbek magnet isa uniform magnetic "eld con"gured so that the exit"eld boundary is a straight edge which projectsback to the point at which the incident particleenters the "eld. Outgoing trajectories are parallel,and the bend-plane images lie along a plane outsidethe "eld. A particularly simple pole face geometryfor such a magnet would be a segment of a circle,with particles both entering and leaving across thestraight (chord) edge.

This basic shape was adapted, by trimming o!the sharp points at either end, to produce the polepro"le shown in Fig. 1. This adjustment modi"edthe optics in three signi"cant ways:

a. The projection of the long exit edge now inter-cepts the incident trajectory somewhat upstreamfrom its entrance into the "eld. This e!ectivelyrotates the edge slightly from a true Elbek con-"guration, and increases the path length inside

the "eld for the lower-energy electrons. This hasthe advantage of moving the low-momentumend of the image plane farther away from the"eld boundary, providing more space for instal-lation of detectors. With this pole shape the exittrajectories are no longer parallel, but the imagesurface is still essentially a plane.

b. The incident electron beam now crosses normalto the local "eld boundary.

c. The full-energy exit beam also crosses normal tothe local "eld boundary, so that the full-energytrajectories into the beam dump are more nearlyparallel. Since this shortens the total path lengthof such trajectories inside the "eld, the de#ectionangle for full-energy electrons is now a few de-grees smaller than would be the case along theunmodi"ed Elbek shape.

The spectrometer magnetic "eld is defocusing in thetransverse plane, so that a beam spot on the radi-ator will be imaged to a line in the detector plane.This property is potentially useful in seeking toselect polarization by angular correlation as men-tioned above, since transverse position at the imageplane is proportional to transverse angle at theradiator. This e!ect could be enhanced by displac-ing the radiator upstream from its nominal positionof 50 cm from the "eld entry edge, which can bedone without seriously degrading the energy res-olution. However, there are no longer any plans toinstrument this particular polarization option atthe time of this writing.

The shaping of the pole edges and the dimen-sions of the return yoke were designed using thetwo-dimensional magnetic "eld program POIS-SON [8]. Because of the elongated shape of themagnet, it was reasonable to approximate the mag-net in two dimensions by using pole cross sectionsat the midpoint and at three other positions be-tween the center and the end. At each of thesepositions, the cross section of the return yoke wasadjusted until the saturation curves at the fourpositions were equal and provided satisfactory"elds up to 1.75 T (the "eld value required to trans-port a 6 GeV beam along the full-energy orbit tothe beam dump). The curvature of the designedreturn yoke was determined by a "t to these fourcalculated values.


POISSON calculations were also used in design-ing the cross-section pro"le of the magnet pole (thehatched surface in Fig. 2c). A two-step linear con-tour was used as a starting con"guration, and theshape was adjusted slightly to minimize saturatione!ects and to make the e!ective "eld boundarycoincide with the edge of the pole root. The innerchamfer did not extend all the way to the ends ofthe long exit edge (see Fig. 2c), where the pole widthwas only &15 cm, so as not to interfere with the"eld uniformity at the entry edge and the full-energy exit edge of the magnet.

3.2. Magnet and vacuum specixcations

Structural and assembly design for the magnet,vacuum system, and related mechanical compo-nents was developed at Catholic University andre"ned by the Je!erson Lab engineering sta!. Twoviews of the assembled magnet are shown in Fig. 2.It is topologically a C-magnet, with the yoke com-pletely open along the straight edge to allow freepassage for degraded electrons along this entirelength.

The magnet has a full-energy radius of curvatureof 11.80 m and de#ection angle of 303. It is 6.06 min length along the open chord, has a gap width of5.7 cm, and weighs &68 000 kg. Pole width is&0.5 m at the midpoint, tapering to 0.16 m at theends. A typical magnetic "eld in the gap is 1.13 Tfor a beam energy of 4 GeV.

The gap width of 5.7 cm was determined by thecriterion that the lowest-energy electrons at an inci-dent energy of 800 MeV (the lowest anticipatedbeam energy at JLab) should be detected with 90%e$ciency. A narrower gap would have unaccep-tably encroached upon the production cone forthese electrons. The pole faces are themselves in-corporated as the side walls of the vacuum contain-ment, primarily because any separate vacuumchamber with walls thick enough to be self-sup-porting would have required an enlargement of thegap (with associated upscaling of other design para-meters) in order to preserve this clearance.

The magnet yoke is built up in a laminar fashion(see Fig. 2b) from seven separate steel plates thatwere cut and machined by outside vendors andwere then assembled on site at JLab. They are

made from AISI 1006 steel, which is a very lowcarbon steel (0.06%) with essentially no other alloycontent. The three innermost layers, which form anintegral assemblage with the pole tip, coil, andvacuum structures, are precision machined on allmating faces to control for the ultimate uniformityof the magnet gap. The major components in thiscentral group are bolted together, to allow for polegap alignment, vacuum welding, and coil installa-tion. The two outermost yoke layers on either side,which were torch-cut from rolled plates by thesupplier with almost no "nish machining, were sim-ply tack-welded to the central assembly aroundtheir perimeters at the completion of the assemblyprocess.

The mid-plane layer is of double thickness(36 cm, as opposed to 18 cm for the others), and is,in fact, built up from two rolled plates that werewelded together by the manufacturer, with muchmore careful surface preparation to minimize en-closed voids than was needed for the outermostlayers. This doubled layer was then cut and ma-chined in two sections in order to providea straight-through slot (shown in Figs. 1 and 2)toward the primary target downstream in the ex-perimental hall. This slot provides passage for thephoton beam during tagger operations, or for elec-tron beams directly into the hall when the taggersystem is not in use.

As noted above, the magnet pole tips themselvesconstitute the lateral walls of the pole-gap vacuumvessel. Stainless-steel vacuum walls were weldeddirectly to the pole root around its entire perimeter.The curved wall along the high-energy edge of thepoles incorporates a length of 6.35 cm diameterbeam pipe to carry the vacuum path outside theyoke through the straight-ahead slot. The remain-ing width of this slot to either side of this pipe was"lled with solid slabs of magnet iron, tack welded inplace during the assembly.

The pole tips, which are also made from AISI1006 steel, required substantially more precisionmachine work than any other component. The "n-ished pieces are 15 cm thick, and are both taperedand beveled for optimum "eld-edge shaping as dis-cussed in Section 3.1 above. A pre-assembly ofpoles, spine, and inner yoke plates was initiallybolted together in order to allow for precise


measurement of the spacing throughout the polegap, and thus for computation of appropriate shimthicknesses between pole and yoke at each boltlocation. The components were then separated andthe entire system was re-assembled, this timeincluding the shims, coils, and welded vacuumelements.

The coils are wound from copper conductor hav-ing a 2.54]1.83 cm2 rectangular section witha central hole 0.953 cm diameter for coolant #ow.There are 20 turns on each pole, wrapped withepoxy glass tape for electrical isolation, and ar-ranged in four layers of "ve turns. Although electri-cally in series, the separate coolant loops for thefour layers are connected in parallel to minimizethe pressure drop in the cooling water circuit. Atfull "eld (1.75 T, 6.1 GeV), the coils carry a currentof 2400 A with a power dissipation of 150 kW.

The focal surface of the magnet, as discussedabove, is nearly straight and can be approximatedas a plane without a!ecting the resolution at theorder of 10~3E

0. With the radiator foil at the

nominal position, the focal plane lies at an angle of22.43 below the incident beam line, or equivalently7.43 below the magnet pole chord. Over thedynamic range of photon energies from 20% to95% of E

0, this focal plane is more than 9 m in

length. The geometry of this arrangement is in-dicated in Fig. 1.

In order to bring vacuum all the way out to thefocal plane, while introducing no obstacle to anydegraded electron trajectory over this entire range,we settled upon a welded structure of 0.635 cmthick stainless-steel plate (AISI 304), entirely sup-ported by external ribs, as may be seen in Fig. 3. Inthe upstream zone (toward the left on the "gure),these cantilever loads were shed by welding thesupport ribs directly to the outer yoke plates of themagnet. Farther downstream, where the focal planeextends well beyond the magnet, it was necessary towrap the external rib structure all the way aroundthe closed side of the vacuum box, and to sti!en itagainst buckling failure by means of cross ribs andedge #anges.

The focal plane lies &4 cm beyond, and parallelto, the exit face of the vacuum box. This side of thebox is a rectangular #ange 1.27 cm thick, of ex-ternal dimensions 9.86 m]91.4 cm, with a central

rectangular aperture 961]20 cm2. This 20 cm di-mension exactly matches the interior dimension ofthe ribbed portion of the box, which in turn waschosen to match the largest transverse angles of thelowest-energy electrons accepted through the polegap. The #ange is drilled for two bolt patterns } oneat the outer perimeter to mate with the light-tightenclosure that supports the detector hodoscope,and a second at the perimeter of the central aper-ture to accommodate the thin-membrane vacuumwindow. Its external face is polished over the zonewithin 15 cm of the lip of the aperture, to providean appropriate O-ring sealing surface for the win-dow frame.

3.3. Field mapping

Extensive mapping of the magnetic "eld, usingthe `Ziptracka [9] apparatus on loan from Fer-milab, was performed prior to installation of thetagger magnet in its operating position. In order toallow the e!ective use of the overhead crane formanipulation of the various heavy components, themagnet was actually assembled lying on its side onthe #oor of the main experimental hall } i.e. with itssymmetry plane horizontal rather than vertical.Mapping in this orientation allowed much moree$cient use of the Ziptrack apparatus. Some hard-ware and software modi"cations were needed toextend the probe into our comparatively narrowgap, and to allow measurements to be madethroughout the crescent-shaped pole gap withoutdanger of collision with the already installed vac-uum wall along the curved rear edge of the poles.

The Ziptrack device provides simultaneous map-ping of three orthogonal components of the mag-netic "eld with automated probe positioning ona prede"nable matrix of points along all three Car-tesian axes. The available range of travel duringa scan is &1.5 m vertically, and on one horizontalaxis, but can be as large as 13 m along the remain-ing horizontal axis, a feature which made it wellsuited to our magnet dimensions.

Excitation curves ("eld vs. current) were mappedat several key points within the "eld over the fullrange of the power supply. The Hall-probe sensorson the Ziptrack unit were regularly calibratedagainst both an NMR probe and a high-precision


Table 1Tagger magnet excitation data

Current B.!9

Incident energy(A) (T) (GeV)

150 0.130 0.449300 0.257 0.900600 0.510 1.795800 0.677 2.388

1000 0.844 2.9801400 1.174 4.1441800 1.475 5.1892000 1.588 5.5632200 1.682 5.8672400 1.761 6.120

Fig. 4. Field map contour plots for the tagger magnet at threedi!erent excitations. Contours are drawn at 99.5%, 99%, 97%,96%, 95%, 90%, 70%, 50%, 30% and 10% of the maximum"eld value.

Hall probe, using appropriate mechanical jigs toensure repetitive centering of each probe at exactlythe same location in the "eld.

Once the excitation data had been established,and hysteresis and degaussing e!ects were under-stood, a series of extended maps over the entire polesurface, both in and out of the mid-plane, wasundertaken using a grid spacing of 2.54 cm parallelto the long pole edge and 1.27 cm in the transversedirection. Data were taken from at least 2.54 cmbeyond the full-energy orbit in the pole gap toa distance of 30 cm (approximately "ve gap widths)outside the pole edge.

Because of vacuum box structure at the two endsof the magnet, extending the Ziptrack measurementgrid beyond the pole boundary at the entrance andfull-energy exit edges of the magnet was not pos-sible. Measurements along these two axes weretaken manually, using a high-precision Hall probewhich was moved along a calibrated arm anchoredto alignment holes in the magnet. These measure-ments extended from 50 cm outside the "eldboundary to 25 cm inside the "eld, and so extendedwell into the region of the Ziptrack measurements,with which they were fully consistent.

Field maps were measured for 10 di!erent mag-net currents, corresponding to the "elds requiredfor incident electron energies from 0.45 to6.12 GeV. The "eld maps were stored in the form oftwo-dimensional arrays suitable for input to theray-tracing program SNAKE [10]. The additionaldata points for the entry and full-energy exit re-gions were not incorporated into the central Zip-track grid, but instead were used to create separate`"eld boxesa, a convenient separation for theSNAKE calculation.

When the magnet had been rotated to its "nalvertical-bending orientation and installed in itsgantry, a partial re-measurement of the "eld wasundertaken using a precision Hall probe and a col-lection of improvised alignment devices insertedinto the pole gap from beneath. The re-measuredpoints were consistent with the original map towithin 0.2%.

Table 1 shows the basic parameters of the 10 "eldsettings. The incident energy was calculated by raytracing with the requirement that the full-energyelectron beam be de#ected by exactly 303. The

resulting outgoing rays for the 10 "eld settingscoincided to within 1.0 mm. The e!ective "eldboundary along the straight edge (calculated byintegrating the "eld from the maximum to the endof the fringe "eld and dividing by B

.!9) agreed with

the POISSON calculations to within 2 mm.Fig. 4 shows some sample contour plots. The

contours at the curved boundary show some spuri-ous structure due to the rectangular grid of themap. The right-angle jogs in the contour along thestraight edge are due to the termination of the poleedge chamfer near the two ends. As noted above,this chamfer was not extended all the way to the


Fig. 5. A scale drawing of a short section of hodoscope with a few typical electron trajectories superposed. Shows the `Venetian Blindageometry and indicates the general relationship between the E- and ¹-planes of scintillators.

ends of the chord so as not to a!ect the uniformityof the "eld at the entry and full-energy exit edges.

The uniformity of the "eld was within 1% overmost of the pole area for the settings between 0.51and 1.48 T, corresponding to incident electron en-ergies of approximately 1.8 and 5.2 GeV. At thehigher excitations, a clear fallo! of the "eld near thepole edges was observed, while at very low excita-tions a small "eld enhancement appeared near eachend. Neither of these features had any signi"cante!ect on the calculated energy resolution, and thecalculated energy calibration of the focal plane (asa fraction of E

0) varied by less than 0.2% between

"eld settings below 1.17 T, increasing to a max-imum shift of 1.0% at 1.76 T.

Ray-tracing calculations with SNAKE, using themeasured "eld maps, were used to calculate theenergy boundaries of the individual hodoscopescintillators, and the permissible coincidence com-binations between the "rst and second layers ofcounters. The calculations agreed exactly withcounter correlations subsequently observed in thedata (see Section 7.2.3 below).

4. The hodoscope

4.1. Overview and design considerations

The tagger instrumentation must provide twokey pieces of information for each detected electron} momentum data of su$cient precision to allowdetermination of the photon energy to the desiredresolution, and timing data adequate to permit the

forming of reliable coincidences with any sub-sequent events triggered by interaction of thatphoton in downstream targets. The "rst of theserequires a high degree of segmentation, and hencedemands scintillators that are quite small in thedispersive dimension, even on a focal plane that ismore than 9 meters in length. The latter calls foroutput pulse shapes which will allow for precisetiming, and hence argues for scintillators which arethick enough to ensure su$cient light for this pur-pose.

The focal-plane hodoscope, a section of which issketched in Fig. 5, consists of two separate planes ofscintillator detectors. Each detector element isoriented with its working surface normal to thelocal beam trajectory, the so-called `Venetianblinda geometry. An important bene"t of this ge-ometry is that the counters subtend a fairly smallangular acceptance for detection of secondary par-ticles that do not come directly from the radiator,especially those generated by slightly degradedelectrons which strike the #ange and walls of thevacuum chamber adjacent to the full-energy exitport. The two separate planes provide for addi-tional background suppression capability throughthe use of appropriate logic circuitry to establishgeometric constraints on the trajectories of accep-ted particles.

The "rst detector plane is called the E-plane,where E denotes energy. This plane, used only formomentum de"nition, lies along the magnet focalsurface and contains 384 narrow scintillators (the`E-countersa). By using an overlapping design,this level of segmentation provides momentum


Fig. 6. Color photograph of the detector through an opened service panel. Shows details of box construction, relative locations ofdetector planes, both sets of phototubes, "ber guides, and several E-counters seen in re#ection from the vacuum window.

resolution of &1]10~3, a compromise chosen toreduce cost and complexity while still more thanadequate to meet the needs of experiments current-ly proposed. The intrinsic resolution of the magnetitself is &2]10~4, and upgrades to this capabilityof the detector thus remain possible if needed in thefuture.

The second detector plane (called `¹a for timing)lies 20 cm behind the E-plane (downstream), asmeasured along the normal to both planes. Thisplane contains 61 scintillators of considerably lar-ger dimensions which are instrumented to providethe timing precision needed for the coincidencewith CLAS.

Electron trajectories cross the focal plane atangles ranging from 9.53 at the high electron mo-mentum (low photon energy) end to 253 at thelow-momentum end. Hence the linear distancealong the trajectory from plane-to-plane rangesfrom &40 to &120 cm. The electronic instrumen-

tation of these two detector planes is described inthe following sections.

The entire detector hodoscope resides insidea light-tight enclosure suspended beneath the tag-ger vacuum box. Fig. 6 shows the downstreamregion of the box opened for service access. The sidewalls of this enclosure comprise a series of light-weight removable panels made from opaque blackplastic, sti!ened by expanded metal mesh (whichalso provides electromagnetic shielding). All panelsare interlocked with the high-voltage supplies, bothfor personnel protection and to prevent phototubesbeing accidentally energized while exposed to am-bient room light. A forced ventilation system, usinga small fan unit originally designed for a photo-graphic darkroom, protects against heat buildupwhen the system is powered.

Structurally, the enclosure is a welded trussworkof aluminum angle members, with su$cient stabil-ity to support the full weight of the detectors,


phototubes, and cable segments. In normal opera-tion, this entire load is suspended from a pin-located bolt pattern on the tagger vacuum box.A set of external jacks allows the entire assembly tobe withdrawn and then repositioned as an integralunit when necessary to service or replace the vac-uum window. The rear (downward-facing) wall hasa thin panel in the midplane region to minimize thegeneration of backscattered background, #ankedby solid panels containing all of the signal andhigh-voltage feedthroughs. The mounting rails forthe two detector planes are supported on postsrising from this rear surface.

4.2. The E-plane

The E-plane is aligned along a #at surface essen-tially coincident with the optical focal plane of themagnet. These E-counters record de#ected electronpositions at the focal surface, and thus the momen-ta of those degraded electrons. The array is com-posed of 384 plastic scintillators 20 cm long and4 mm thick, and with widths (along the dispersiondirection) that range from 6 to 18 mm in order tosubtend approximately constant momentum inter-vals of 0.003E

0. Each counter optically overlaps its

adjacent neighbors by one-third of their respectivewidths, thus creating 767 separate photon energybins through appropriate recording of coincidencesand anti-coincidences, and providing an energy res-olution of 0.001E

0.

No pulse amplitude data beyond `hit/no-hitainformation is required from these small scintil-lators. We were thus able to realize signi"cant costsavings in the instrumentation of this large set ofcounters, while still retaining adequate time resolu-tion to support the E}¹ coincidence logic discussedin Section 5.3 below. High voltage for the E-plane isbussed to groups of four tubes, for a signi"cantadditional savings in power supply and controlcosts.

For optimum resolution, the E-plane must belocated close to the exit #ange on the tagger magnetvacuum box, in order to minimize the e!ects ofmultiple scattering as electrons pass through thethin exit window. Since the E-plane must lie alongthe focal surface of the magnet, the vacuum vesselwas designed to bring the exit window as close as

practical to this location. The scintillator strips aremounted within 4 cm of the vacuum window frame,supported from below on a system of aluminumrails. This clearance allows for the installation ofa layer of metal plates between the E-plane and thevacuum box for personnel protection when per-forming service on the detector system while thethin window remains under vacuum load. Thesmall red and white features visible at the top centerof Fig. 6 are individual E-counters, actually seenhere in re#ection from the thin vacuum window.

Each scintillator is wrapped in one layer of whitepaper to increase light collection e$ciency and toreduce any cross talk between adjacent channels.Optical "ber light guides, seen arcing through theupper section of Fig. 6, carry the signal from oneend of each strip to a series of Lucite cylinders,&2.5 cm]2.5 cm, located well back in the de-tector enclosure, entirely behind the second de-tector plane and its phototubes. (These cylindersare hidden behind the central aluminum beam inthe photograph.) The "bers pass entirely throughan axial hole drilled in each cylinder and are gluedin place for mechanical support. The exit face ofeach "ber/cylinder assembly is polished for opti-mum optical contact with the photomultiplier tube.

Su$cient light is transmitted via the optical"bers to permit reliable readout with low-costphotomultiplier tubes, clearly visible in the lowerportion of the photo. The tubes are held in positionagainst the "ber/cylinder assemblies by a spring-loaded arrangement which allows quick access forreplacement and repair. The scintillator stripsthemselves were cut to speci"cation and polishedby the manufacturer (Bicron). All other compo-nents in this scintillator/light-guide arrangementwere cut, polished, wrapped and assembled locallyat Catholic and George Washington Universities.

4.3. The T-plane

To properly associate a tagged electron with theaccelerator microstructure (one micropulse each2 ns into each of the three experimental halls), andthereby with any related downstream event, re-quires a timing resolution of p"300 ps or better.The ¹-plane provides this information. This planecontains 61 counters, 2 cm thick, read out using


Fig. 7. Schematic of the tagger overall logic design.

phototubes attached by solid light guides at bothends (transverse to the momentum axis) of eachscintillator. These units were shipped by the manu-facturer (Bicron) with scintillator and lightguidealready integrated. The row of green circuit boardsseen in the upper section of Fig. 6 marks the loca-tion of the ¹-plane tube bases.

The ¹-counter scintillators are organized in twogroups, with the "rst group of 19 narrower countersspanning the photon energy range from 75% to95% of the incident electron energy, and the re-maining group of 42 counters spanning the rangefrom 20% to 75%. The ¹-counter widths (along thedispersion direction) are varied to compensate forthe 1/Ec behavior of the bremsstrahlung cross sec-tion, so that the counting rate in each detectorremains approximately the same within eachgroup, with the rate in the "rst group approxim-ately 1

3of that in the second group. The dis-

continuity in the width progression (visible inFig. 1) permits operation at considerably highertagged-photon rates for some experiments whichare only interested in the highest-energy photons.

¹-counter width acceptances overlap their nearestneighbors by a few mm, just su$cient to ensure thatthere are no inter-counter gaps through which anelectron could escape undetected.

¹-counter lengths (transverse to the momentumdirection) are also varied from 20 cm at the highelectron momentum end to 9 cm at the low-mo-mentum end, tracking the energy dependence of thebremsstrahlung characteristic angle distribution, inorder to help suppress potential background fromambient radiation. These lengths were chosen to belarge enough to accept &98% of electrons fromthe radiator at E

0"0.8 GeV.

5. Electronics/readout scheme

The basic logic diagram for the Tagger elec-tronics is shown in Fig. 7. The signature of anyevent is a set of times from both the E- and ¹-planes, measured with FASTBUS TDCs, whichare recorded upon receipt of a valid tagger/CLASevent trigger. These times are subsequently used to


determine which counters were hit at the correcttime relative to a CLAS trigger to represent a validevent. Appropriate coincidences and counting ratescan then all be determined afterward through theuse of software.

5.1. T-counter electronics

Each of the 122 T-counter photomultiplier sig-nals is fed to one input of a fast (200 MHz) con-stant-fraction discriminator (Phillips 715). Thesediscriminators provide output signals for the tag-ger's own timing system, and also for the photonevent logic circuitry that is used to establish a validtagger photon event.

For the timing system, the output of each fastdiscriminator is sent to one of four 64-input, 12-bit,multi-hit, FASTBUS TDCs (LRS 1872A), operat-ing in the 50 ps/channel resolution mode. TheseTDCs are located near the master trigger elec-tronics for the CLAS detector, in racks on thedownstream side of the experimental hall. They areoperated in common-start mode, with the discrimi-nator signals providing the stops for the individual¹-channels. Start signals come either from theCLAS event logic, or from one of the beam linephoton monitoring devices (see below), dependingon whether the event is part of a data collectionperiod or a beam monitoring run. The cables forthe discriminator output signals must traversehalf-way around the circumference of Hall B toreach these racks, and this provides a delay of380 ns which is su$cient to get them in time withthe CLAS electronics.

For the photon event logic, remaining outputsignals from the ¹-counter discriminators are usedto establish a coincidence between the two photo-multipliers which view each individual ¹-planescintillator. Signals from the left and right ends ofeach ¹-counter are fed to a fast AND logic unit(Phillips 754). One output of each AND gate isscaled using a 100 MHz scaler (CAEN V560E } notshown in the "gure), while another output is usedas input to the Master OR.

The Master OR logic is a simple cascade of theoutputs of the AND gates into 4-input fast ORgates (Phillips 754). While Fig. 7 shows all 61 ¹-counters included in the logic, the system is easy to

recon"gure to use any multiple of 4 ¹-counters asthe Master OR generator. Pick-o!s from di!erentlevels in the OR gate chain allow for fast identi"ca-tion of the photon energy in the broad energyrange covered by signals to that particular OR gate.In fact, the Master OR system has already beenreplicated for the "rst 19 ¹-counters, correspond-ing to the highest photon-energy tagging range(75}95% of the incident electron energy), in orderto provide a totally separate signal for an indepen-dent experiment running parasitically with theCLAS detector.

The Master OR output provides the signal thata photon has been tagged for the CLAS "rst-leveltrigger. It is important that the Master OR signalalways arrive at the trigger logic at the same timerelative to a photon leaving the radiator, indepen-dent of which of the ¹-counters was struck. Thisrelative timing synchronization has been achievedby cutting the cables for each of the 122 ¹-countersignals to the correct length to assure this condi-tion.

The shape of the bremsstrahlung spectrum gen-erates many more tagged lower-energy photonsthan higher-energy photons, especially when tag-ging over a large energy range. For the Hall B tag-ger, the rate in the lowest-energy region of ourhodoscope is 4.5 times the rate in an equal widthinterval in the highest-energy region. Especiallywhen operating with a fairly unrestricted CLAStrigger, this places many more low photon energyevents into the data stream, which contributes tothe rapidity at which storage media become "lled.Moreover, since in many experiments one is prim-arily interested in events triggered by high-energyphotons, this surplus of low-energy events in thetrigger could prolong the total experimental run-ning time necessary to generate acceptable statisticsin the desired energy range.

To ameliorate this problem somewhat, a systemwas developed for vetoing the output of the ¹-counter left}right coincidence gates, which allowsfor the selective prescaling of various segments ofthe tagging spectrum that is used to generate thephoton event trigger. A software controlled VMEoutput register generates the veto signals, with ap-propriate level conditioning. Control and monitoringsoftware for this system was developed at Arizona


State University. By adjusting the rate and dura-tion of the veto gates, zones of ¹-counters can beseparately removed from the trigger for di!erenttime intervals. The veto gates are typically o! for33 ms, and then on for a time which is an integralmultiple of the o! time.

A typical veto scheme would divide the hodo-scope into four equal zones by photon energy, withthe lowest-energy zone being live one-eighth of thetime, the next zone one-quarter, the next one-half,and the highest photon energy range operated withno veto. This particular pattern results in a morenearly uniform event rate across the entire photonspectrum (with, in fact, a moderate enhancementfor the highest-energy events). The prescale factorsare adjustable by software control to meet theneeds of particular experiments.

5.2. E-counter electronics

Modularized, front-end E-plane electronics forthis apparatus were speci"cally designed and builtat Arizona State University. The output of each ofthe E-counter photomultiplier tubes is fed to a sig-nal-ampli"er, discriminator, multiplexer and logicmodule (ADML) The ampli"er discriminator inthis module is a Phillips 6816, 16-channel,time-over-threshold unit. Eight ADML units areinstalled in a single custom crate with a dedicatedcontroller module, so three such crates are requiredfor the full complement of 24 modules servicing the384 photomultiplier inputs from the E-plane.

The control module in each crate is used tomultiplex the photomultiplier signals so that theanalog output from any one of the 128 E-counterscan be viewed at a single output jack on the frontpanel. Selection can be made locally via a hand-held controller, or remotely via a computer con-trolled switch, for multiplexing onto a single line forviewing in the experimental hall or in the countingroom.

The output from each E-counter discriminator isfed to one input of a 96-input, 12-bit, multi-hit,FASTBUS TDC (LRS 1877), running in the com-mon-stop mode with a timing resolution of500 ps/channel. The stop signal is supplied by theCLAS event trigger. The relative time of an event inthe E-counter is used to determine which of the

E-counters were hit during the time bucket of thisparticular trigger. Because of the 1

3geometric over-

lap of the E-plane scintillators, software-deter-mined coincidences between adjacent countersallow subdivision of the 384 E-counter signals into767 energy channels.

5.3. E}T coincidence logic

The timing information from the E- and ¹-counters is used during data analysis to establishthe hit patterns and to make tight timing coincid-ences between the counters. Timing windows of lessthan 0.5 ns for the ¹-counters and 3 ns for theE-counters allow software rejection of most back-ground events.

All of this "nal coincidence logic takes place insoftware after the full CLAS event has been re-corded. It was therefore decided to provide anoption for some reduction of background eventsprior to generation of any CLAS trigger signal, byestablishing a hardware coincidence between each¹-counter and the group of E-counters that mightbe paired with a legitimate electron trajectorythrough that ¹-counter.

Establishing appropriate E}¹ coincidences re-quires dealing with a very complex matrix of over-lapping coverage. Each ¹-counter is geometricallyassociated with a group of E-counters and, becauseof the angular divergence of the electrons leavingthe tagger magnet, many E-counters can be legit-imately associated with adjacent ¹-counters. Toprovide a completely correct logic signal for each ofthe 61 ¹-counters would require a matrix logiccircuit with 384 inputs.

In order to simplify the coincidence logic process,the ADML units were built to provide logic ORECL outputs for each contiguous set of eight E-counters. The coincidence logic matrix (E}¹ LogicUnit), built in-house for this purpose, then mapsthese 48 logic signals from the ADMLs onto theappropriate ¹-counter signals. Using the eight-foldlogic from the E-counter array does allow some`falsea coincidences between E and ¹ pairs that,while relatively close together, are not geomet-rically correlated with any allowed trajectory. Thisadds only slightly to the chance background whilegreatly simplifying the logic matrix.


Each output from the E}¹ logic unit is fed to oneof the inputs of the AND gate used for each ¹-counter. The output gate width of the E}¹ logicmodule is deliberately set to be considerably longerthan the outputs from the ¹-plane discriminators,so that the timing of the event is still entirely deter-mined by the ¹-counter signal.

If desired, the E}¹ coincidence requirement canbe removed from the trigger logic by setting all theoutputs of the E}¹ logic unit to the &&on'' state. Thiscan be done by remote control.

6. Other beamline components

6.1. Radiator foils

The bremsstrahlung radiator target ladder isbased on a modi"cation of a HARP design [11}13]used throughout the CEBAF accelerator as an elec-tron beam position monitor. This modi"cationprovides for a longer travel stroke (15.25 cm) whichallows the mounting of a standard wire scanner andfour radiator targets on the same ladder assembly.Thus, the position of the electron beam can beprecisely determined at the exact location wherethe radiators are to be inserted.

The scanner incorporates a pair of wires, ortho-gonal to each other but at 453 to the direction ofscanning motion, in order to allow two-axis recon-struction of the beam pro"le. There is no directreadout of any electrical signal from the scanningwires used in the tagger beamline. Rather, the wirefunctions as a very localized radiator whichproduces a strong signal of slightly degradedelectrons in a set of counters surrounding thebeam dump vacuum line. This signal is highlysensitive to the wire position within the incidentbeam.

Radiators of high atomic number were used inorder to minimize the e!ects from electron}electronbremsstrahlung in the spectrum. Several di!erentthicknesses of radiator targets, ranging from1]10~6 to 1]10~4 radiation lengths, have beenprepared and mounted at Florida State University.The foils are prepared through deposition ontostandard 2.54]7.62 cm2 microscope slides treated

with a release agent consisting of a detergent(Neodol 25-3S from Shell Chemical Company, orSteol CS-460 from Stepan Company) and/or NaCl.The thinner radiators are combination "lms of goldon a backing of carbon made in one of twomethods: (1) carbon deposition by carbon-arc dis-charge, followed by gold deposition with an elec-tron beam, or (2) carbon and gold both depositedby electron beam in one vacuum cycle. The thickerradiators are free-standing gold "lms.

Once deposited onto the slide, the "lms arescored to size and #oated in distilled water. Theradiator holding frame is then lifted from beneaththe "lm for mounting. The radiator targets are then`slackeneda by drawing air across one side of the"lm to create a pressure di!erential; this slackeningimproves the hardiness of the "lms for a longerlifetime. These "lms then need no special storageenvironment and can be stored under ambient con-ditions. Speci"cations for the radiators which havebeen produced are given in Table 2.

6.2. Beam dump and shielding

The tagger beam dump, designed at CatholicUniversity, must stop all of the full-energy elec-trons, and must contain the secondary particlescreated by stopping these electrons, during opera-tion of the tagger system. The shielding around thedump must accomplish two purposes; it must pre-vent the ground water located outside the dumptunnel from becoming excessively radioactive, andit must shield the CLAS detector system from sec-ondary radiation.

In both of these considerations, it is the neutronsproduced inside the dump that are the most pen-etrating and therefore the most di$cult to shield.Ground water activation comes principally fromfast neutron spallation reactions. In addition, manylower-energy neutrons are produced from photo-nuclear excitation of the giant resonance. The ma-jor source of background for the CLAS comes fromneutrons which exit back through the beam trans-port pipe that carries the electron beam to thedump. Details of the modeling and calculation ofthese various e!ects [14], and of the anticipatedbeam pro"le in the dump region [15], are availablein CEBAF internal publications.


Table 2Radiator composition

Radiation length Composition

1]10~6 4 lg/cm2 Au, on 15 lg/cm2 C2]10~6 10 lg/cm2 Au, on 15 lg/cm2 C5]10~6 30 lg/cm2 Au, on 15 lg/cm2 C1]10~5 62 lg/cm2 Au, on 15 lg/cm2 C2]10~5 129 lg/cm2 Au free standing5]10~5 323 lg/cm2 Au free standing1]10~4 646 lg/cm2 Au free standing

The design plans for the dump had to take intoaccount the beam power to be dissipated. In thenormal, thin-radiator mode of operation this poweris quite low, typically a few hundred watts. How-ever, plans had been put forward for producingpolarized photons by certain techniques whichwould require a considerably more intense electronbeam. The dump was therefore designed from thebeginning to be capable of handling a power levelof &10 kW, since it would be prohibitively expen-sive to retro"t to this capability once any lower-power installation had been completed. This powerlevel is believed to be near the maximum that canbe handled with a relatively inexpensive, forced gas#ow cooling system for the dump.

The original dump tunnel in the civil construc-tion plans for Hall B is a cylindrical, corrugatediron pipe, 3.05 m in diameter and approximately12 m long, sloping down beneath the #oor at anangle of 303 as indicated in Fig. 3. At its maximumdepth the center of the tunnel is 6.1 m below the#oor, and almost directly under the CLAS target.The tunnel is surrounded by bedrock, concrete,and construction back"ll, but the exact proportionsand locations of these materials are uncertain.Nevertheless, since most of Hall B is locatedbelow the water table, ground water certainly existsin the environment immediately outside the dumptunnel.

The dump itself is a laminar array in the shape ofa 30 cm diameter cylinder. The "rst section struckby the beam is composed of "ve discs of graphite(carbon) 10 cm thick and separated by &1 cm airgaps. The rear section is composed of 10 sheets oftungsten 7 mm thick, also with air gaps between.

This array sits within an airtight steel enclosure of62 cm diameter, with appropriate interior ba%ing,and with 18.2 cm diameter pipe connections up to#oor level, to provide for easy future implementa-tion of forced air #ow if upgrade to multi-kilowattcapacity is ever required. The tunnel is entirely"lled with iron for 1.5 m downstream of the pri-mary dump, and the central dump enclosure issurrounded by iron &1.2 m thick (radial direction)for an additional 1.3 m (axial).

An inner pipe of diameter &1 m runs from theface of this iron shielding up to #oor level, concen-tric with the original tunnel walls. The spacebetween the two cylinders is back"lled with gravel,and sealed at #oor level with poured concrete toa depth of &1 m. The remaining opening is thenplugged with another steel can, &1.5 m tall, withpenetrations for a central beam pipe &25 cm dia-meter and for the two air#ow pipes previouslymentioned. The air#ow pipes pass through the canat an angle strongly o!set from the axis of thetunnel in order to provide some shadow shielding.The remaining volume of this can is "lled withsand. The beam line vacuum extends from the tag-ger magnet to a stainless steel window just belowthis can, at which point there is &1 m of sandand/or concrete in direct line between this windowand the CLAS.

The beam dump and its associated shielding arevery e!ective. With no radiator, the measuredbackground from the dump while receiving 400 Wof power is less than 5 kHz across the entiretagger detector system. Most of these events couldprobably be eliminated with timing and coincid-ence checks. Even without any such cuts, thisbackground is less than 10~3 of the typical taggingrate.

A much more serious source of backgroundcomes from minimally degraded electrons insidethe tagger which strike the beam pipe walls, or thetagger vacuum box in the vicinity of the full-energyexit #ange, and never make it into the dump tunnel.This background in the tagger hodoscope can bekept well in hand by appropriate timing and coin-cidence cuts (see Section 7.2.3 below). However,additional lead shielding has been installed in thisarea, as indicated in Fig. 3, to shadow the CLASfrom these components.


Fig. 8. Sketch showing the general layout of downstream #ux monitoring devices (not to scale!)

6.3. Flux monitoring devices

Several devices located downstream of the CLAStarget, shown in Fig. 8, are used to measure andmonitor photon #ux. Brief descriptions of the hard-ware are included in this section. Details of themonitoring algorithm are discussed below in Sec-tion 7.2.2.

The most reliable and direct measurement is ac-complished with a set of four lead-glass blocks,mounted in a 2]2 square array, called the TotalAbsorption Shower Counter (TASC). This device,constructed and tested by Norfolk State University,is usable only at low rates and is remotely retractedfrom the photon beam during data runs. The TASCis the primary absolute measure of the photon #ux,with close to 100% detection e$ciency, and thus,when compared to the sum of all tags in the taggerfocal plane, it yields the photon tagging ratio, dis-cussed in more detail in Section 7.2.2.

The lead glass used in the TASC is of type SF3. Itis 55% PbO and 45% SiO

2, with radiation length

of 2.36 cm and index of refraction 1.67. The dimen-sions of each block are 10]10]40 cm3, for ane!ective thickness of almost 17 radiation lengths.Each block is read out at one end by a 7.6 cm,Philips XP4312B phototube.

Since the TASC detector cannot handle photonbeam rates higher than &105 Hz without notice-able degradation, the device is cross calibrated withtwo other devices } the Pair Spectrometer (PS) andthe Pair Counter (PC) } which are able to functionat much higher beam intensity.

The pair spectrometer consists of a large aper-ture dipole magnet and eight scintillator counters }paired four on each side of the beam line. A photonimpinging on a thin converter produces e`e~ pairs,which are diverged by the magnetic "eld and thendetected in the symmetrically positioned counters.This device is designed to operate at the highphoton beam #ux (5x107 tagged photons/s) of theHall B facility, and can provide rough energyinformation on the photon beam.

The housing for the PS converters is a weldedbox of steel. It has #anges which permit connectionto the Hall B vacuum system to support vacuuminto the main Hall B dump during electron beamoperations. The converters are thin foils of alumi-num, 20 cm square. They are held by a three "ngerspoke holder, also aluminum, which allows for re-mote insertion of the converter of choice understepper motor control. Thicknesses of 1% and 2%(radiation lengths) are typically installed, alongwith a blank slot to allow for free passage of the


beam during electron running. All parts for thevacuum box, converters, and spoke holder were cutand machined in-house at Rensselaer PolytechnicInstitute.

The positioning of the scintillator detectors wasdesigned to cover as much of the energy spectrumas possible while overlapping with the Tagger'stagging range. Of the eight scintillators, six wereplaced with their long dimension vertically, perpen-dicular to the dispersion plane of the PS magnet.The lowest-energy scintillators were placed withtheir longest dimension in the dispersion plane inorder to fully cover the low end of the Taggerspectrum. The placement of these detector elementscan be seen in Fig. 8.

The pair counter, a joint project of Saclay andGeorge Washington, is a much simpler device, int-ended as a backup intensity monitor to be usedwith the full beam. It is reliable and easy to main-tain and it can remain in the beamline, locatedbetween the PS and the TASC, at all times. A thinconverter produces electron}positron pairs thatare detected in coincidence between a singlefront scintillator and a rear layer of four, margin-ally overlapping, scintillators positioned quadrant-wise around the beam centroid. This segmentationof the second layer, together with lateral motionunder stepper motor control, provides rough in-formation on the lateral position stability of thephoton beam, and also on the beam pro"le. A vetoscintillator is located upstream of the converter toeliminate charged particle background comingfrom the CLAS. The e$ciency of the Pair Counteris roughly 1.5% which, at a beam intensity of5]107 photons/s, yields a detection rate of&750 kHz.

The link between the measure of photon #ux atlow rates and that at nominal data taking beamrates is established by the cross calibration betweenthe TASC and the Pair Spectrometer. The latterdevice is always in the photon beam just upstreamof the TASC. A comparison is made between thesedevices at lower rates, and the TASC is then re-moved from the beamline. As the beam intensity isincreased, the ratio of the sum of the individual¹-counter scalers to the PS has been monitoredand found to be linear. At high beam rates, the ratioof the `Master ORa (i.e. the single scaler which

counts the logical OR of all ¹-counters) to PC andPS is not linear with beam intensity because ofscaler dead time, so individual ¹-counter scalersmust be summed o!-line to obtain the correctphoton-beam intensity.

7. Performance to date

7.1. Early calibration with beam

The "rst beam through the tagger system wasdelivered in December 1996. Between May 1997and May 1998 the Hall B Photon Tagger wasexercised through several testing and commission-ing sessions. Of prime concern during this periodwas the optimization of the timing characteristicsof the tagging system. This required the precisetime alignment of tagger signals relative to eachother, and also relative to the beam RF structure.It was also important to minimize the widths ofthe time-of-#ight spectra, which dependupon the pulse characteristics of the individualPMTs as well as upon the various geometric factorslisted above.

Electron beams from 0.9 to 4.0 GeV were usedduring the commissioning period. These valueswere determined primarily by the needs of earlyexperiments sharing the beam, and not by consid-eration of tagger parameters. Most tagger commis-sioning was done without photon beam collimationin order to be able to measure the full bremsstrah-lung beam during photon transmission e$ciencymeasurements.

Relative timing among the ¹-counters, as well asthe E}¹ coincidence timing, was veri"ed during theearly test runs. ¹-counter cables were adjusted inlength to bring ¹-counter relative time di!erencesdown to the nanosecond level, so that timing coin-cidence gate widths could be kept small in order toreduce random background. Timing characteristicsare particularly important for the ¹-counters, sincethe CLAS event trigger is constructed from theMaster OR of these counters. It was also importantto obtain reasonably good relative timing amongE-counters so that the hardware E}¹ coincidencelogic signals could be formed.


Fig. 9. Plot of the `tagging ratioa (de"ned in Section 7.2.2) forindividual ¹-counters vs. fraction of incident energy.

7.2. Current status

7.2.1. Tagger ezciencyThe absolute e$ciency of the tagger detector

system has not been measured. Because the taggersignal is in the event trigger, tagger e$ciency can-cels in the ratio of detected to tagger events. Thusthe knowledge of the absolute tagger e$ciency isnot necessary to determine cross sections. How-ever, we believe the absolute e$ciency of the taggeris quite high. The pulses produced by a minimumionizing particle passing through the 2-cm thick¹-counter scintillators provide a spectrum ofevents that are well separated from the back-ground. The ¹-counter discriminators are operatedwith the threshold set at one-half the value neededto pass essentially all these minimum ionizingpulses, insuring a very high overall absolute e$-ciency for the tagger.

7.2.2. Tagging ratioThe fraction of photons that have been tagged

which actually hit the CLAS target is called theTagging Ratio. This ratio is monitored by insertingthe downstream total absorption shower counter(TASC) into the beam. The output of the TASC hasbeen tuned so that all tagged pulses are well abovethe TASC discriminator threshold, insuring thatthe TASC e$ciency is essentially 100%.

At low photon #ux, a tagged photon is de"ned bya coincidence, ¹

i)TASC, between a signal in the

TASC and a signal from any of the ¹icounters in

the focal plane. The TDC information for each¹iand E

jcounter allows determination of the num-

ber Nc(Ej) of tagged photons for each energy chan-

nel. The rate of the ¹i)TASC coincidences and the

rate ¹3!8i

in the tagger are both measured byscalers. The tagging ratio R is de"ned for each¹-channel as being the ratio of these two numbers

Ri"(¹

i)TASC)/¹3!8

i.

If there were no losses of photons, and no spuriousbackground in the tagger, this ratio would be unity.

First estimates of this ratio were made during anearly calibration run. Data were obtained usinga small auxiliary lead-glass TASC positioned justafter an exit window installed directly onto thetagger's straight-through beam pipe, approxim-

ately 8 m from the radiator. Except for low ¹-counter numbers, corresponding to the highestphoton energy, the ratio was approximately 98%.The measured ratio cannot be unity since the tag-ger has "nite acceptance for M+ller scattered elec-trons, where no photon is created. Such eventscontribute to the number of tags (Nc ) withoutgenerating any event in the TASC, and this e!ectwas indeed seen in this early calibration data asa fallo! in the measured ratio for the highestphoton energies (corresponding to low ¹-counternumbers).

Fig. 9 displays the tagging ratio as a function of¹-counter number from a later measurement usingthe permanent downstream TASC detector,located &50 m from the radiator. It shows a mea-sured tagging ratio of about 78%, which is typicalfor photon energies over most of the upper opera-tional range } consistent with the expected addi-tional losses due to target, air, and beam pipewindows between the radiator and the permanentTASC. A plastic bag "lled with helium is installedduring photon beam operations between the CLAStarget and the downstream devices to help reducesuch losses (GEANT simulation indicates a correc-tion of &3%). The collimation system, which is


Fig. 10. Cluster plot of ¹-counter hits vs. E-counter number.(Size of each plotted box represents number of events. Thatdetail is not easily seen at this scale, but is also not critical to theconclusions being drawn.) Shows the correlation between theE and ¹ hodoscope planes. E-counters are sized to representa constant momentum bite, so horizontal scale is linear inphoton energy.

designed to limit the beam to a diameter of 33 mmat the CLAS target, accepts a cone of half-angle0.74 mr. This corresponds to 2.4h

cat 1.6 GeV inci-

dent energy, and reduces the overall ratio to about75% at this lower end of the operational range. Thetagging ratio depends strongly on the beam set-upthrough the accelerator and switchyard, and vari-ations of &5% have been detected. Thus, it mustbe regularly monitored.

The Pair Spectrometer e$ciency ePS

is measuredat low intensity by comparison with the TASC

ePS"(¹

i)PS)/(¹

i)TASC)

where ¹i)TASC and ¹

i)PS stand for the number of

coincidences between a ¹-counter and the TASCor the PS, respectively. These numbers are recordedon scalers. This measurement is done at low inten-sity during a `normalization runa. At high inten-sity, i.e during a `production runa, the TASC isremoved from the beam line and only the numberof ¹

i)PS coincidences (true and accidental) are

recorded. The number of photons during a produc-tion run, N130$c (E

j), is then simply deduced from the

number of photons measured during a normaliz-ation run, N/03.c (E

j), as

N130$c (Ej)"N/03.c (E

j)g

where

g"(¹i)PS)130$/(¹

i)PS)/03..

The ratio g is a constant proportional to thenumber of incoming electrons, and takes into ac-count any variation of the tagging ratio. It is ex-pected to be independent of ¹-counter number(and hence is not subscripted) as long as there is novariation within the electronics between thenormalization and the production runs. Anormalization run is performed every day duringan experiment to ensure that e

PSdoes not vary.

7.2.3. E- and T-counter correlationThe dual planes of detectors, the E- and ¹-

counters, are arranged such that electrons comingfrom the radiator can pass through a restricted setof allowed pairs of scintillators. The arrival timedata can be scanned for allowed pairs and fordisallowed pairs coming from background and ran-dom noise events. The allowed pairs of detectors

were determined from a calculation of possible elec-tron orbits using the magnetic "eld maps and themeasured positions of the detectors. A display ofthe number of detected events in the ¹-counters asa function of E-counter number is shown in Fig. 10.The agreement between the calculated and ob-served hit pattern is excellent, indicating that wehave good knowledge of the magnetic "eld and thedetector placement. In this "gure the x-axis, repres-enting E-counter number, is linear in electron en-ergy. The y-axis, the ¹-counter number, is notlinear in energy. The ridge of maximum number ofevents shows the discontinuity in slope caused bythe "ner gradation of the "rst 19 ¹-counters, fol-lowed by the curving section caused by the pro-gressive narrowing of the remaining ¹-counters tomaintain constant counting rate in a rising brem-sstrahlung spectrum.

Any events which are not in the vicinity of themaximum ridge are due to background. One can


Fig. 11. Tagger timing spectrum, showing the timing resolutionof the tagger ¹-plane. The 2 ns interval between beam `bucketsainto Hall B is indicated. The small intervening peaks, located at2/3 of this 2 ns spacing, represent spill from `bucketsa beingdelivered to one of the two adjacent Experimental Halls.

see that the overall background is quite small, andthat it is more signi"cant near the lower photonenergy portion of the spectrum. The additionalbackground in this region is thought to be due tountagged bremsstrahlung electrons that have lostenough energy to strike the vacuum pipe or #angeat the exit of the tagger vacuum chamber, locatedjust above the low photon energy end of the de-tector hodoscope. Background from these eventswill clearly be coming from the wrong direction andwill not strike properly correlated detectors. It willtherefore be straightforward to eliminate theseevents from the analysis.

7.2.4. Event timingThe time of arrival of an event is determined very

precisely from a combination of the time signalsfrom each end of the ¹-counter. With the measure-ment of both times, the time jitter from the approx-imately 1.4 ns variation in the light transit time,caused by the lateral spread in the electrons, can beeliminated. The sum of the times from the left andright ends of the scintillators includes the total lightdrift time, independent of the starting point in thescintillator, and thus is a constant.

We have timed these signals against a signalreceived from the RF drive of the accelerator.Fig. 11 shows the timing spectrum thus generated.The timing resolution obtained is approximately

110 ps for 1 p, greatly exceeding the design goal of300 ps for this quantity. This is by far the tightesttime resolution in the system, and allows us touniquely identify the RF beam bucket responsiblefor initiating an event, and essentially eliminatesthe uncertainty in start time used for time-of-#ightmeasurements in CLAS. The "gure clearly showsthe 2 ns microstructure of the CEBAF beam. Inter-estingly, the "gure also shows that with the acceler-ator tune employed for these measurements there isa small spill from the million times more intenseinterspersed beam being sent to another experi-mental hall } the small peak located at two-thirdsof the spacing (1.33 ns) between the peaks producedby the main beam entering Hall B.

7.3. Conclusion

The JLab photon-tagging system has satis"ed itsdesign objectives. It meets, and frequently exceeds,the capabilities required in order to achieve thephysics goals of any real- photon experiment so farapproved for Hall B at Je!erson Lab. At the time ofthis writing, we are nearing the end of the "rst yearof data collection with this facility. Early analyseslook very encouraging, and presentation of prelimi-nary physics results at professional meetings hasalready begun.

Acknowledgements

The list of people who could be thanked at theend of a project of more than "ve years durationbecomes all but unmanageable, but the contribu-tions of the following individuals in particular aregratefully acknowledged: Addison Lake, JohnO'Meara and John Robb of the JLab engineeringsta! for their invaluable participation in all stagesof the design and assembly process; BernhardMecking and Arne Freyberger of JLab, and DaveDoughty of Christopher Newport University, formany helpful suggestions and contributionsthroughout the extended lifetime of this project;Paul Branch of CUA for his machinist skills duringthe prototyping phase; Irwin Price of GWU foressential "nancial support to one of us (WJB);Randall Stevens and Aaron Vanderpool of ASUfor technical assistance in fabricating and testing


tagger electronics; R. DiBari, K. Knowles and R.Thomas of RPI for technical contributions to thePair Spectrometer; Scott Teige of Indiana U. forproviding the lead glass blocks for the TASC, andPete Hemler of JLab for designing its supportstructure; Powell Barber of FSU for his expertise inmanufacturing the thin radiator targets, and RobinChappell and the FSU machine shop for fabricat-ing the HARPs; Doug Tilles and the entire JLabHall B crew of riggers and welders for cheerful andreliable day to day support during installation andcalibration; Students Henrik Ayvazian, ZoranBucalo, Juan Carlos Sanabria, Bryan Carnahan,Catalina Cetina, Katherine Keilty, Aziz Sha",Michele Schottenbauer, and Michael Wood, andtechnician Shane Briscoe, for their participation inconstruction and testing. This project was sup-ported in part by the National Science Foundation(ASU, CUA, FSU, RPI) and by the Department ofEnergy (GWU, JLab).

References

[1] V.D. Burkert, B.A. Mecking, in: B. Frois, I. Sick (Eds.),Modern Topics in Electron Scattering, World Scienti"c,Singapore, 1991.

[2] J.D. Kellie, I. Anthony, S.J. Hall, I.J.D. MacGregor, A.McPherson, P.J. Thorley, S.L. Wan, F. Zettl, Nucl. Instr.and Meth. A 241 (1985) 154.

[3] I. Anthony, J.D. Kellie, S.J. Hall, G.J. Miller, J. Ahrens,Nucl. Instr. and Meth. A 301 (1991) 230.

[4] P. Detemple, K.H. Altho!, G. Anton, J. Arends, A. Bock,M. Breuer, K. BuK chler, G. NoK ldeke, M. Serwazi, W.Schneider, D. Urban, B. Zucht, Nucl. Instr. and Meth.A 321 (1992) 479 and works cited therein.

[5] S.K. Matthews, Hall Crannell, J.T. O'Brien, D.I. Sober,Nucl. Instr. and Meth. A 421 (1999) 23.

[6] D.I. Sober, A Guide to the Optics of the Tagged PhotonMagnet, CLAS Note 91-012.

[7] J. Borggreen, B. Elbek, L.P. Nielson, Nucl. Instr. andMeth. 24 (1963) 1.

[8] Poisson/Super"sh Reference Manual, Los Alamos Accel-erator Code Group, LA-UR-87-126, 1987.

[9] ZIPTRAK Operations Manual, Fermilab Survey Group,private communication.

[10] Pascal Vernin, private communication.[11] C. Yan et al., Nucl. Instr. and Meth. A 365 (1995)

261.[12] R. Abbott, B. Bowling, E. Woodworth, HARP User's

Manual, CEBAF Tech Note TN92-050, 1992.[13] J.E. Feldl, private communication.[14] Hall Crannell, Considerations for the Design of the CLAS

Tagger Beam Dump, CLAS Note 93-023.[15] D. I. Sober, Tagger Electron Beam Dump Considerations,

CLAS Note 92-013.


the bremsstrahlung tagged photon beam in hall b at jlab · measurement of incident photon energy in...

Documents