Monitored drift tubes in ATLAS

Abstract The precision chamber system of the ATLAS muon spectrometer is described. Particular emphasis is put on the mechanical aspects and on the choice of the operation point of the detector.


The ATLAS Muon Spectrometer
In the ATLAS detector [1] muons will be measured in two completely independent systems, the Inner Tracker and the Muon Spectrometer. The inner tracker has a radius of 1.1 m and is surrounded by a superconducting coil which generates a solenoidal eld of 2 T. It is instrumented with pixel detectors, silicon strips and straw tubes with transition radiation detection capability. In this region, muons momenta are measured in a light material but high occupancy environment. The momentum measurement i s competitive u p t o p T 50 GeV=c.
In the muon spectrometer the measurement is made after the calorimeters and then is limited at low p T by m ultiple scattering and energy losses. The ATLAS muon spectrometer is based on three 8-coil air-core toroid magnets, one to cover the barrel region (jj < 1:2) and two in the forward regions (1:2 j j 2 : 7) , which generate an approximately toroidal eld over a radial distance of the order of 5 m with an average eld of 2.6 and 6 Tm in the barrel and endcaps respectively. The muon spectrometer is designed for stand-alone measurement capability.
The muon spectrometer instrumentation consists of precision chambers for coordinate measurements in the bending plane: Monitored Drift Tubes (MDT) [2] and Cathode Strip Chambers (CSC) [3] and trigger chambers: Resistive Plate chambers (RPC) [4] in the barrel region and Thin Gap Chambers (TGC) [5] in the forward regions; the trigger chambers provide bunch crossing identication and measurement of the second coordinate in the non bending plane.

The concept of Monitored Drift Tubes
The design consideration with the strongest impact on the design of the AT-LAS muon spectrometer is the requirement for stand-alone measurement capability. This safeguards against any unanticipated backgrounds and ensures a good discovery potential in the presence of unexpected event topologies due to unknown physics at the TeV scale. To accomplish a good momentum measurement e v en at low momenta, this requirement leads to the choice of an air-core magnet system, which h o w ever puts high demands on the muon chamber system due to unavoidable eld inhomogeneities and the relatively low elds in such systems.
The open structure of the barrel toroid allows to reconstruct the muon momenta from a measurement of the sagitta in three muon stations in front, in the center and behind the magnetic eld in the range jj < 1:4. In the forward the muon momenta are obtained from a point-angle measurement with one point in front and two point behind the endcap toroid cryostats.
To a c hieve the required momentum resolution at high momenta ( the sagitta for a 1 T eV muon is about 500 m) t w o requirements have t o b e fullled: the muon chambers have to provide a position measurement accuracy of about 50 m per station and the alignment of the three stations with respect to each other has to be known to about 30 m.
To a c hieve the rst goal a system of high pressure (3 bar absolute) proportional drift tubes -the Monitored Drift Tubes -has been chosen. An MDT chamber consists of six layers of drift tubes, arranged such that three layers of closely-packed tubes (multilayers) are mounted on each side of a support structure. The two m ultilayers and the support structure form a chamber, see Fig.1.
To fully exploit the drift tubes space resolution wires have to be positioned to better than 10 m inside a tube and better than 20 m inside a chamber using high precision assembly methods.
An in-plane alignment system allows to track sag/distortion in each c hamber, see Fig.1. The comb plates of the support structure accurately dene the positions of the multilayers. They are equipped with RASNIK alignment systems [6]. These systems measure the position, in three coordinates, of the middle component (the lens) with respect to the center point b e t w een the outer components (an illuminated coded mask and a CCD with video readout respectively). There are two systems along the chamber edges parallel to the tubes, and two diagonal systems.
It will not be possible to stabilize the geometry of the chambers and their positions in the spectrometer on the scale of a measurement accuracy of 50 m due to movements induced by v ariations of temperature or magnetic eld. Therefore chamber alignment i s a c hieved o-line using corrections obtained from the measurement of the path of light-rays by light sensitive detectors (See Sect.6).
The chambers layout which has a direct impact on the spectrometer resolution has been optimized with respect to geometrical acceptance, access to the electronics, alignment facilities, eld inhomogeneities and chamber con-struction. Figure 2 shows the MDT system layout. MDT chambers cover almost all the rapidity range but a very small region at high rapidity where the expected rates are higher than 100 Hz=cm 2 . In this region, to reduce occupancy, CSC chambers are used because of their ner granularity. The MDT system consists of 880 chambers which c o v ers an area of 5500 m 2 ; the total number of tubes is 300000, for a total tube and wire length of 1075 Km.

The Drift Tubes
The basic detection cell of the MDT's is a drift tube of 3 cm outer diameter and 400 m wall thickness made with an hard aluminum alloy of the Al Mn type (ALUMAN 100) . The tubes are produced by extrusion and their thickness is then reduced by drawing. The use of this technique allows to satisfy tight tolerances on the outer diameter (+0/-20 m), wall thickness ( 20 m), eccentricity ( < 15 m). The tubes are closed at both ends by endplugs which also provide the xation of the W-Re wire of 50 m diameter and its connection to preampliers and the high voltage supply as well as the gas inlets. They also contain the gas seals of each tube. The wire is stretched in the tubes to the 60% of the elastic limit. The method for wire xation is not decided yet -under consideration are gluing or crimping of the wire.
The wire is located in the center of the tube by disk-shaped precision wire locators. The wire locators x the wire position at both ends of the tubes, right in front of the endplugs with a precision of 10 m. In the long chambers (tube length > 4m), an additional wire locator is used for reasons of electrostatic stability in the middle of the tube. The support structure of the chamber makes the tube gravitational sag match the wire sag within 100 m (see Sect.4). The accurate position of the wire respect to the outer tube surface is guaranteed by the tight tolerances on the tube specications.
The goal is to fabricate individual tubes and test them for gas tightness and verify electrical properties before their assembly into multilayers. Given the large number of tubes, semi-automatic assembly of the individual tubes with wire, wire locators and endplugs is foreseen. Pilot production of drift tubes has shown a yield of good tubes higher than 99% with a measured leak rate lower than 10 5 torr l s 1 corresponding to a pressure drop of < 1 % per tube every 10 days. 4 Chamber assembly Tested drift tubes are assembled into units consisting of three layers of closelypacked tubes (multylayers). Essential to the construction of the multilayers is the use of precision jigging to establish and maintain, during assembly and while the adhesive cures, the position (20m) of the tubes. At present, prototypes are being constructed using two somewhat dierent techniques for the assembly of multilayers. In the rst approach each l a y er of tubes is precisely controlled. In a rst step tubes in a layer are glued together, then two l a y ers are glued and nally the third layer is added on top. The jigging controls each tube to very high accuracy at those points where the wires are located, namely the wire supports. The second approach puts more emphasis on the jigging for assembling of complete multilayers. Only the top and bottom layers are controlled by jigging and the middle layer is positioned by the tubes. In this case the precision is again guaranteed by the strict tolerances of the tubes.
A nal decision on the tube stacking method will be taken on the grounds of full scale prototypes evaluation consisting on precise measurements of the wire positions using X-ray tomography techniques.
An MDT chamber is built from two m ultilayers mounted to either side of a support structure as shown in Fig.1. Considerations of pattern recognition performance, autocalibration 2 , and angular resolution in a single chamber have also guided the decision to separate the two m ultilayers by a distance of 200-400 mm.
The support structure uses a minimum of material and provides precision positioning of the two m ultilayers and components of the alignment systems (see Sect. 6). The stiness of the beams is chosen to permit the desired gravitational sag of the chamber to match the wire sag. The assembly of spacer and the multylayers is again done using high precision jigs.

The choice of the operating point
The choice of the operating point for the MDT chambers is driven by few basic requirements: excellent position resolution (< 100 m), suciently high lifetime to survive 1 0 y ears of operation at LHC, and suciently low occupancy of the chambers to allow for ecient pattern recognition in the high background conditions expected at LHC.
The expected deposited charge per unit of wire length accumulated after 10 years of LHC operation can be computed from: Q = r d < N e > G e t 0 : 067C=cm where: r = 100Hz is the maximum expected counting rate in the MDTs, < N e > 700e the average number of primary electrons, G = 2 10 4 the gas gain (the choice of the gas gain will be justied later), e the charge of the electron and t = 1 0 8 s the running time of the experiment. In the nal setup the integrated deposited charge will be higher due to the higher primary ionization produced by neutrons and photons, which dominate the background and counting rates. Background calculations indicate that these particles have, averaged over the energy spectrum, a primary ionization yield which is higher by a factor 2 than the one for minimum ionizing ones. This gives an integrated deposited charge of Q = 0 : 13C=cm.
Streamer signals have t o b e a v oided as already small fractions of streamer pulses increase the integrated deposited charge signicantly. If the charge in a streamer pulse is 100 times higher than in the proportional avalanche a fraction of 1% streamer would double the integrated charge. Finally a safety factor of 5 is applied to account for uncertainties in the background calculations. We therefore require that the MDT have to be designed to survive Q = 0 : 67C=cm. Ultra-clean tests made at CERN show gain variations of the order of 10 20% per C/cm.
From the previous discussion on the chamber aging it is clear that one has to keep the lowest possible gain compatible with a good performance. The lowest possible gain is dened by the interplay of noise, discriminator threshold and shaping time. We require a discriminator threshold ve times higher than the electronic noise. For a gas gain of 2 10 4 at an electronic noise of N = 5000 equivalent electrons, the threshold can be set to be around the 25th electron for a shaping time of 15 ns. As a direct consequence of this choice the single wire position resolution averaged over the drift radius is expected to be of about 80m. A decrease of the discriminator threshold to the 12th electron, possible with a gas gain of at least 4 10 4 , w ould improve the average resolution by about 30%.
Once the operating parameters xed, one has to nd a suitable gas mixture. To fulll the ATLAS requirements, the lling gas has to be non ammable for safety requirements, the maximum drift time has to be less than 500 ns to minimize the occupancy. T o improve both resolution and autocalibration, the Lorenz angle should be kept below 1 0 o and, despite the 1=r E-eld dependence, the r-t relation should be as linear as possible. Finally the streamer rate should be < 1% to limit aging eects.
At present the best gas candidate is Ar : N 2 : CH 4 (91 : 4 : 5) which fullls all the above requirements. The small percentage of N 2 improves the linearity of the space-time relation as can be seen from Fig.3 where the rt relation for the preferred gas is shown as calculated by the GARFIELD package [7]. Fig.4 shows the single wire space resolution calculated using GARFIELD. Shown are also the dierent contributions to the resolution. The main source of uncertainty is, as expected, due to clustersize uctuations. Test beam measurements on small MDT prototypes are in good agreement with both simulation results. 6 The Global Alignment system As discussed in Sec.2 to achieve a sagitta measurement at the level of 50 m one needs a global alignment system. In the case of sti chambers the corrections to a false sagitta of a particle from chamber displacements can be corrected for using the apparent sagitta of projective (pointing to the vertex) light-rays, see Fig. 5. The accuracy of these corrections is limited by initial chamber displacements in the order of a centimeter and several milliradians and the vertex spread of several centimeters in the direction of the beam. Alignment errors of better than 30m are expected after correction, if the position of the track through the chambers is known to about 2 cm. Axial alignment r a ys, parallel to the beam direction, reduce the dependence on the initial chamber positioning error. In-plane alignment measurements within each c hamber are used to correct for deviations from stiness.
The positions of projective alignment t o w ers with respect to each other and to other subdetectors, necessary for the determination of the invariant mass of multiparticle nal states, needs to be known with a precision in the centimeter range. This is achieved by an external survey of the detector. Tower-to-tower alignment as necessary for the reconstruction of lowmomentum muons traversing more than one alignment t o w ers is achieved by proximity sensors, measuring displacements between adjacent c hambers. These also allow to combine more than one chamber per layer to form a projective t o w er, thus reducing the number of required light paths and adding redundancy to the alignment system. In the endcaps, the cryostat of the endcap toroids would obstruct any light path, thus making complete projective alignment impossible. Therefore a dierent solution is adopted for this region: precise carbon-ber bars are projectively aligned with only 8 alignment r a ys and the chamber alignment is done with respect to the bars.

Spectrometer performance
Assuming a single tube space resolution of 80 m, one obtains a multilayer resolution of 50 m and a chamber space resolution of 35 m. Adding to this a 30 m contribution due to the global alignment, the design 50 m sagitta resolution is obtained. On these grounds a full simulation of the spectrometer gives the results shown in Fig.6. At p T = 100 GeV=c, the structures seen in the resolution curve are due to regions where only a two station measurement is possible and to multiple scattering in the magnet struts. At the highest momenta the multiple scattering eects become negligible.
Two full scale prototypes have been already tested on a beam at CERN and more prototypes are expected to be ready by the end of 1996. Three of these prototypes will form a full size tower to be exposed to cosmic rays for a demonstration of the alignment s c heme. The series production is expected to start around 1999 to end in late 2003 in time for installation in the experimental area.

Conclusions
The ATLAS MDT system is a very ambitious muon detector aiming for a precision of 50 m in a huge system: 45 m long, 22 m in diameter and covering an area of 5500 m 2 . It has excellent stand-alone capability and almost hermetic coverage up to jj = 2 : 7. Quite advanced prototype studies show that the challenge can be met, allowing to explore the exciting physics scenario opened by LHC.

Longitudinal beam
In-plane alignment Multilayer Cross plate