First proton-proton collisions at the LHC as observed with the ALICE detector: measurement of the charged particle pseudorapidity density at $\sqrt{s}$ = 900 GeV

On 23rd November 2009, during the early commissioning of the CERN Large Hadron Collider (LHC), two counter-rotating proton bunches were circulated for the first time concurrently in the machine, at the LHC injection energy of 450 GeV per beam. Although the proton intensity was very low, with only one pilot bunch per beam, and no systematic attempt was made to optimize the collision optics, all LHC experiments reported a number of collision candidates. In the ALICE experiment, the collision region was centred very well in both the longitudinal and transverse directions and 284 events were recorded in coincidence with the two passing proton bunches. The events were immediately reconstructed and analyzed both online and offline. We have used these events to measure the pseudorapidity density of charged primary particles in the central region. In the range |$\eta$|<0.5, we obtain dNch/deta = 3.10 $\pm$ 0.13 (stat.) $\pm$ 0.22 (syst.) for all inelastic interactions, and dNch/deta = 3.51 $\pm$ 0.15 (stat.) $\pm$ 0.25 (syst.) for non-single diffractive interactions. These results are consistent with previous measurements in proton-antiproton interactions at the same centre-of-mass energy at the CERN SppS collider. They also illustrate the excellent functioning and rapid progress of the LHC accelerator, and of both the hardware and software of the ALICE experiment, in this early start-up phase.


Affiliation notes
i Also at 26 ii Also at 36 iii Now at 80 iv Now at 32 v Now at 37 vi Now at 22

Collaboration institutes 1 Introduction
The very first proton-proton collisions at Point 2 of the CERN Large Hadron Collider (LHC) [1] occurred in the afternoon of 23 rd November 2009, at a centre-of-mass energy √ s = 900 GeV, during the commissioning of the accelerator. This publication, based on 284 events recorded in the ALICE detector [2] on that day, describes a determination of the pseudorapidity density of charged primary particles 1 dN ch /dη (η ≡ − ln tan θ/2, where θ is the polar angle with respect to the beam line) in the central pseudorapidity region. The purpose of this study is to compare with previous measurements for proton-antiproton (pp) collisions at the same energy [3], and to establish a reference for comparison with forthcoming measurements at higher LHC energies.
The event sample collected with our trigger contains three different classes of inelastic interactions, i.e. collisions where new particles are produced: non-diffractive, single-diffractive, and double-diffractive 2 . Experimentally 1 Here, primary particles are defined as prompt particles produced in the collision and all decay products, except products from weak decays of strange particles such as K 0 s and Λ. 2 Inelastic pp collisions are usually divided into these classes depending on the fate of the interacting protons. If one (both) we cannot distinguish between these classes, which, however, are selected by our trigger with different efficiencies 3 .
In order to compare our data with those of other experiments, we provide the result with two different normalizations: the first one (INEL) corresponds to the sum of all inelastic interactions and corrects the trigger bias individually for all event classes, by weighting them, each with its own estimated trigger efficiency and abundance. The second normalization (non-single-diffractive or NSD) applies this correction for non-diffractive and double-diffractive processes only, while removing, on average, the singlediffractive contribution.
Multiparticle production is rather successfully described by phenomenological models with Pomeron exchange, which dominates at high energies [4,5]. These models reincoming beam particle(s) are excited into a high-mass state, the process is called single (double) diffraction; otherwise the events are classified as non-diffractive. Particles emitted in diffractive reactions are usually found at rapidities close to that of the parent proton. 3 We estimate the trigger efficiency for each class using the process-type information provided by Monte Carlo generators; the values vary by up to a factor of two between classes and are listed in Section 3. The relative abundance of each class is taken from published data (see text). late the energy dependence of the total cross section to that of the multiplicity production using a small number of parameters, and are the basis for several Monte Carlo event generators describing soft hadron collisions (see for example [6][7][8]). According to these models, it is expected that the charged-particle density increases by a factor 1.7 and 1.9 when raising the LHC centre-of-mass energy from 900 GeV to 7 and 14 TeV respectively (i.e. intermediate and nominal LHC energies). The difference in charged-particle densities between pp and pp interactions is predicted to decrease as 1/ √ s at high energies [9]. This difference was last measured at the CERN ISR to be in the range 1.5-3 % [10] at √ s = 53 GeV. Extrapolating these values to √ s = 900 GeV, one obtains a very small difference of about 0.1-0.2 %. Therefore, we will compare our measurement to existing pp data and also to different Monte Carlo models.
This article is organized as follows: Section 2 describes the experimental conditions during data taking; the main features of the ALICE detector subsystems used for this analysis are decribed in Section 3; Section 4 is dedicated to the event selection and data analysis; the results are discussed in Section 5 and Section 6 contains the conclusion.

LHC and the run conditions
The LHC, built at CERN in the circular tunnel of 27 km circumference previously used by the Large Electron-Positron collider (LEP), will provide the highest energy ever explored with particle accelerators. It is designed to collide two counter-rotating beams of protons or heavy ions. The nominal centre-of-mass energy for proton-proton collisions is 14 TeV. However, collisions can be obtained down to √ s = 900 GeV, which corresponds to the beam injection energy.
The results from the first proton-proton collisions presented here were obtained during the early commissioning phase of the LHC, when two proton bunches were circulating for the first time concurrently in the machine. The bunches used were the so-called "pilot bunches": low intensity bunches used during machine commissioning, with a few 10 9 protons per bunch. The two beams were brought into nominal position for collisions without a specific attempt to maximize the interaction rate. The nominal r.m.s. size of LHC beams at injection energy is about 300 µm in the transverse direction and 10.5 cm in the longitudinal (z-axis) direction. However, at this early stage, the beam parameters can deviate from these nominal values; they were not measured for the fill used in this analysis. For the previous fill, for which the longitudinal size was measured, it was found to be shorter, with an r.m.s. of about 8 cm. Assuming Gaussian beam profiles, the luminous region should be smaller than the beam size by a factor of √ 2 in all directions. Shortly after circulating beams were established, the ALICE data aquisition system [11] started collecting events with a trigger based on the Silicon Pixel Detector (SPD), requiring two or more hits in the SPD in coincidence with the passage of the two colliding bunches as inferred from beam pickup detectors. As a precaution, only a small subset of the detector subsystems, including the silicon tracking detectors and the scintillator trigger counters, was turned on, in order to assess the beam conditions provided by the LHC.
The trigger rate was measured just before collisions with the same trigger conditions. Without beams we measured a rate of 3 × 10 −4 Hz (in coincidence with one bunch crossing interval per orbit). In coincidence with the passage of the bunch of one circulating beam the rate was 0.006 Hz. As soon as the second beam was injected in the accelerator, the event rate increased significantly, to 0.11 Hz. The first event that was analyzed and displayed in the counting room by the offline reconstruction software AliRoot [12] running in online mode is shown in Fig. 1. This marked symbolically the keenly anticipated start of the physics exploitation of the ALICE experiment 4 . The online reconstruction software implemented in the High-Level Trigger (HLT) computer farm [13] also analyzed the events in real time and calculated the vertex position of the collected events, shown in Fig. 2. The distributions are very narrow in the transverse plane (sub-millimetre, including contributions from detector resolution and residual misalignment), of about the expected size in the longitudinal direction and well positioned with respect to the nominal centre of the ALICE detector. This provided immediate evidence that a substantial fraction of the events corresponded to collisions between the protons of the two counter-rotating beams.
After 43 minutes, the two beams were dumped in order to proceed with the LHC commissioning programme. In total, 284 events were triggered and recorded during this short, but important, first run of the ALICE experiment with colliding beams.
3 The ALICE experiment ALICE, designed as the dedicated heavy-ion experiment at the LHC, also has excellent performance for protonproton interactions [14]. The experiment consists of a large number of detector subsystems [2] inside a solenoidal magnet (B = 0.5 T). The magnet was off during this run.
During the several months of running with cosmic rays in 2008 and 2009, all of the ALICE detector subsystems were extensively commissioned, calibrated and used for data taking [15]. Data were collected for an initial alignment of the parts of the detector that had sufficient exposure to the mostly vertical cosmic ray flux. Data were also taken during various LHC injection tests to perform timing measurements and other calibrations.
Collisions take place at the centre of the ALICE detector, inside a beryllium vacuum beam pipe (3 cm in radius and 800 µm thick). The tracking system in the AL-ICE central barrel covers the full azimuthal range in the   pseudorapidity window |η| < 0.9. It has been designed to cope with the highest charged-particle densities expected in central Pb-Pb collisions. The following four detector subsystems were active during data taking and were used in this analysis.
-The Silicon Pixel Detector (SPD) consists of two cylindrical layers with radii of 3.9 and 7.6 cm and has about 9.8 million pixels of size 50 × 425 µm 2 . It covers the pseudorapidity ranges |η| < 2 and |η| < 1.4 for the inner and outer layers respectively, for particles originating at the centre of the detector. The effective ηacceptance is larger due to the longitudinal spread of the position of the interaction vertex. The detector is read out by custom-designed ASICs bump-bonded directly on silicon ladders. Each chip contains 8192 channels and also provides a fast trigger signal if at least one of its pixels is hit. The trigger signals from all 1200 chips are then combined in a programmable logic unit which provides a level-0 trigger signal to the central trigger processor. The total thickness of the SPD amounts to about 2.3 % of a radiation length. About 83 % of the channels were operational for particle detection and 77 % of the chips were used in the trigger logic. The SPD was aligned using cosmic-ray tracks collected during 2008 [16], and the residual misalignment was estimated to be below 10 µm for the modules well covered by mostly vertical tracks. The modules on the sides are likely to be affected by larger residual misalignment. The time resolution of this detector is better than 1 ns. Its response is recorded in a time window of ±25 nsec around the nominal beam crossing time. For events collected in this run, the arrival times of particles at the detector relative to this "time zero" is shown in Fig. 3. Note that in general several particles are registered for each event. Particles hitting one of the detectors before the beam crossing have negative arrival times and are typically due to interactions taking place outside the central region of ALICE.
More details about the ALICE experiment and its detector subsystems can be found in [2]. The trigger used to record the events for the present analysis is defined by requiring at least two hit chips in the SPD, in coincidence with the signals from the two beam pick-up counters indicating the presence of two passing proton bunches. The efficiency of this trigger as well as all other corrections have been studied using two different Monte Carlo generators, PYTHIA 6.4.14 [17] tune D6T [18] and PHOJET [8], for INEL and NSD interactions. The trigger efficiencies for non-diffractive, singlediffractive, and double-diffractive events were evaluated separately, and found to be 98-99 %, 48-58 %, and 53-76 % respectively. The ranges are determined by the two event generators. These event classes were combined for the corrections using the fractions measured by UA5 [19]: non-diffractive 0.767 ± 0.059; single-diffractive 0.153± ±0.031; double-diffractive 0.08 ± 0.05. The resulting efficiencies were found to be 87-91 % for the INEL normalization and 94-97 % for the NSD normalization, again depending on the event generator used.
The results presented in the following sections are those obtained with PYTHIA. The difference between results corrected with PYTHIA and PHOJET is used in the estimate of the systematic uncertainty.

Data analysis
The data sample used in the present analysis consists of 284 events recorded without magnetic field. The results presented here are based on the analysis of the SPD data. However, information from the SDD, SSD and VZERO was used to crosscheck the identification and removal of background events.
In the SPD analysis, the position of the interaction vertex is reconstructed [20] by correlating hits in the two silicon-pixel layers to obtain tracklets. The achieved resolution depends on the track multiplicity and for this specific vertex reconstruction is approximately 0.1-0.3 mm in the longitudinal direction and 0.2-0.5 mm in the transverse direction. For events with only one charged track, the vertex position is determined by intersecting the SPD tracklet with the mean beam axis determined from the vertex positions of other events in the sample. A vertex was reconstructed in 94 % of the selected events. The distribution of the vertex position in the longitudinal direction (z-axis) is shown in Fig. 4. For events originating from the centre of the detector, the vertex-reconstruction efficiency was estimated, using Monte Carlo simulations, to be 84 % for INEL interactions and 92 % for NSD collisions. These efficiencies decrease for larger |z|-values of the vertex in low-multiplicity events; therefore, only events with vertices within |z| < 10 cm were used. This allows for an accurate charged-particle density measurement in the pseudorapidity range |η| < 1.6 using both SPD layers.
Using the reconstructed vertex as the origin, we calculate the differences in azimuthal (∆ϕ, bending plane) and polar (∆θ, non-bending direction) angles of pairs of hits with one hit in each SPD layer. These tracklets [21] are selected by a cut on the sum of the squares of ∆ϕ and ∆θ, each normalized to its estimated resolution (80 mrad and 25 mrad, respectively). When more than one hit in a layer matches a hit in the other layer, only the hit combination with the smallest angular difference is used. This occurs in only 2 % of the matched hits.
The number of primary charged particles is estimated by counting the number of tracklets. This number was corrected for: trigger inefficiency; detector and reconstruction inefficiencies; contamination by decay products of long-lived particles (K 0 s , Λ, etc.), gamma conversions and secondary interactions.
The corrections are determined as a function of the zposition of the primary vertex, and on the pseudorapidity of the tracklet. For the analyzed sample the average correction factor for tracklets is about 1.5.
The beam-gas and beam-halo background events were removed by a cut on the ratio between the number of tracklets and the total number of hits in the tracking system (SPD, SDD, and SSD); this ratio is smaller for background events (as measured in the previous fills triggering on the bunch passage from one side) than for collisions [22]. In addition, the timing information from the VZERO detector was used for background rejection by removing events with negative arrival time (see Fig. 3). The event quality and event classification was crosschecked by a visual scan of the whole event sample. In total 29 events (i.e. about 10 %) were rejected as beam induced background, which is consistent with the rate expected from previous fills. The remaining background was estimated from the vertex distribution and found to be negligible. The contamination from coincidence with a cosmic event was estimated to be one event in the full sample. Indeed, two cosmic events were identified by scanning, both without reconstructed vertex. Particular attention has been paid to events having zero or one charged tracklets in the SPD acceptance. The vertex-finding efficiency for events with one charged particle in the acceptance is about 80 %. The number of zerotrack events has been estimated by Monte Carlo calculations. The total number of collisions used for the normalization was calculated from the number of events selected for the analysis, corrected for the vertex-reconstruction inefficiency. In order to obtain the normalization for INEL and NSD events, we further corrected the number of selected events for the trigger efficiency for these two event classes. In addition, for NSD events, we subtract the singlediffractive contribution. These corrections, as well as those for the vertex finding efficiency, depend on the event charged-particle multiplicity, see Fig. 5. The dependence of the event-finding efficiency (combining event selection and vertex finding) on multiplicity was calculated for different interaction types using our detector simulation, and is above 98 % for events with at least two charged particles. The averaged combined corrections for the vertex reconstruction efficiency and the selection efficiency is 20 % for INEL interactions and much smaller for NSD interactions, due to the cancelation of some contributions.
The various corrections mentioned above were calculated using the full GEANT 3 [23] simulation of the AL-ICE detector as included in the offline framework Ali-Root. In order to estimate the systematic uncertainties, the above analysis was repeated by: applying different cuts for the tracklet definition (varying the angle cut-off by ±50 %); varying by ±10 % the density of the material in the tracking system, thus changing the material budget; using the non-aligned geometry; varying by ±30 % the composition of the produced particle types with respect to the yields suggested by the event generators; varying the particle yield below 100 MeV/c by ±30 %;  [19] and using two different models for diffraction kinematics (PYTHIA and PHOJET).
An additional source of systematic error comes from the limited statistics used so far to determine the efficiencies of the SPD detector modules. In test beams, the SPD efficiency in active areas was measured to be higher than 99.8 %. This was crosschecked in-situ with cosmic data, but only over a limited area and with limited statistics. At this stage, we have assigned a conservative value of 4 % to this uncertainty. The triggering efficiency of the SPD was estimated from the data itself, using the trigger information recorded in the data stream for events with more than one tracklet, and found to be very close to 100 %, with an error of about 2 % (due to the limited statistics).
These contributions to the systematic uncertainty on the charged particle pseudorapidity density are summarized in Table 1. Our conclusion is that the total systematic uncertainty on the pseudorapidity density is less than ±7.2 % for INEL collisions and ±7.1 % for NSD collisions. The largest contribution comes from uncertainties in cross sections of diffractive processes and their kinematic simulation.
More details about this analysis, corrections, and the evaluation of the systematic uncertainties can be found in [24]. Figure 6 shows the charged primary particle pseudorapidity density distributions obtained for INEL and NSD interactions in the range |η| < 1.6. The pseudorapidity density obtained in the central region |η| < 0.5 for INEL interactions is 3.10±0.13(stat.)±0.22(syst.) and for NSD interac- Table 2. Comparison of charged primary particle pseudorapidity densities at central pseudorapidity (|η| < 0.5) for inelastic (INEL) and non-single diffractive (NSD) collisions measured by the ALICE detector in pp interactions and by UA5 in pp interactions [3] at a centre-of-mass energy of 900 GeV. For ALICE, the first error is statistical and the second is systematic; no systematic error is quoted by UA5. The experimental data are also compared to the predictions for pp collisions from different models. For PYTHIA the tune versions are given in parentheses. The correspondence is as follows: D6T is tune (109); ATLAS CSC is tune (306); Perugia-0 is tune (320). tions is 3.51±0.15(stat.)±0. 25(syst.). Also shown in Fig. 6 are the previous measurements of proton-antiproton interactions from the UA5 experiment [3]. Our results obtained for proton-proton interactions are consistent with those for proton-antiproton interactions, as expected from the fact that the predicted difference (0.1-0.2 %) is well below measurement uncertainties. The measurements at central pseudorapidity (|η| < 0.5) are summarized in Table 2 together with model predictions obtained with QGSM, PHOJET and three different PYTHIA tunes. PYTHIA 6.4.14, tune D6T, and PHOJET yield respectively the lowest and highest charged particle densities. Therefore, these two have been used for the evaluation of our systematic errors. PYTHIA 6.4.20, tunes ATLAS CSC and Perugia-0, are candidates for use by the LHC experiments at higher LHC energies and are shown for comparison. Figure 7 shows the centre-of-mass energy dependence of the pseudorapidity density in the central region (|η| < 0.5). The data points are obtained in the |η| < 0.5 range from this experiment and from references [3,10,[28][29][30][31], and are corrected for differences in pseudorapidity range where necessary, fitting the pseudorapidity distribution around η = 0. As noted above, there is good agreement between pp and pp data at the same energy. The dashed and solid lines (for INEL and NSD interactions respectively) are obtained by fitting the density of charged particles in the central pseudorapidity rapidity region with a power-law dependence on energy.

Experiment
Using this parametrization, the extrapolation to the nominal LHC energy of √ s = 14 TeV yields dN ch /dη = 5.5 and dN ch /dη = 5.9 for INEL and NSD interactions respectively.

Conclusion
Proton-proton collisions observed with the ALICE detector in the early phase of the LHC commissioning have been used to measure the pseudorapidity density of charged primary particles at √ s = 900 GeV. In the central pseudorapidity region (|η| < 0.5), we obtain dN ch /dη = 3.10 ± 0.13(stat.) ± 0.22(syst.) for all inelastic and dN ch /dη = 3.51 ± 0.15(stat.) ± 0.25(syst.) for non-single diffractive proton-proton interactions. The results are consistent with earlier measurements of primary charged-particle production in proton-antiproton interactions at the same energy. They are also compared with model calculations.
These results have been obtained with a small sample of events during the early commissioning of the LHC. They demonstrate that the LHC and its experiments have finally entered the phase of physics exploitation, within days of starting up the accelerator complex in November 2009.

Acknowledgements
The ALICE collaboration would like to thank all its engineers and technicians for their invaluable contributions to the construction of the experiment. We would like to thank and congratulate the CERN accelerator teams for the outstanding performance of the LHC complex at start up, and for providing us with the collisions used for this paper on such a short notice! The ALICE collaboration acknowledges the following funding agencies for their support in building and running the AL-ICE detector: