Centrality and pseudorapidity dependence of the charged-particle multiplicity density in Xe-Xe collisions at $\sqrt{s_{\rm NN}}$ = 5.44 TeV

In this Letter, the ALICE Collaboration presents the first measurements of the charged-particle multiplicity density, $\rm{d}N_{\rm{ch}}/\rm{d}\eta$, and total charged-particle multiplicity, $N_{\rm{ch}}^{\rm{tot}}$, in Xe-Xe collisions at a centre-of-mass energy per nucleon--nucleon pair of $\sqrt{s_{\rm NN}}$ = 5.44 TeV. The measurements are performed as a function of collision centrality over a wide pseudorapidity range of $-3.5<\eta<5$. The values of $\rm{d}N_{\rm{ch}}/\rm{d}\eta$ at mid-rapidity and $N_{\rm{ch}}^{\rm{tot}}$ for central collisions, normalised to the number of nucleons participating in the collision ($N_{\rm{part}}$) as a function of $\sqrt{s_{\rm NN}}$, follow the trends established in previous heavy-ion measurements. The same quantities are also found to increase as a function of $N_{\rm{part}}$, and up to the 5% most central collisions the trends are the same as the ones observed in Pb-Pb at a similar energy. For more central collisions, the Xe-Xe scaled multiplicities exceed those in Pb-Pb for a similar $N_{\rm{part}}$. The results are compared to phenomenological models and theoretical calculations based on different mechanisms for particle production in nuclear collisions. All considered models describe the data reasonably well within 15%.


Introduction
A plasma of strongly interacting quarks and gluons is formed in the hot and dense nuclear matter created in ultra-relativistic heavy-ion collisions [1,2]. The multiplicity of charged particles produced in the collisions is a key observable to characterise the properties of the matter created in these collisions, as the overall particle production is related to the initial energy density. Nuclei are extended objects and the degree of geometrical overlap between them in the collision, expressed in terms of the impact parameter (b), varies. Since b is not directly measurable, an experimental proxy of centrality is used to characterise the amount of nuclear overlap in the collisions. Typical features related to the collision centrality are the number of nucleons participating in the collision, N part , and the number of binary nucleon-nucleon collisions, N coll , among the participant nucleons. Collisions of nuclei of different sizes lead to different N part and N coll for similar relative nuclear overlap. The study of the production of charged particles with different collision systems and at various collision energies can help shed light on the role of the initial energy density and the production mechanism of final-state particles.
Previous measurements of the system-size dependence of the charged-particle pseudorapidity density (dN ch /dη) were performed at RHIC, comparing Au-Au and Cu-Cu collisions at various centre-of-mass energies [3]. The ALICE Collaboration has previously reported on dN ch /dη in Pb-Pb collisions at √ s NN = 2.76 and 5.02 TeV [4][5][6][7][8]. The dependence of the charged-particle density averaged at midrapidity (|η| < 0.5) dN ch /dη over the centre-of-mass energy shows a steeper increase in central heavyion collisions than in proton-proton (pp) and proton-nucleus (pA) collisions. The values of dN ch /dη , normalised by the number of nucleon pairs participating in the collision, increase faster than linearly with N part . No significant differences between the shapes of the N part dependence for the different collision energies were observed.
In this Letter, the ALICE Collaboration presents the first measurement of the production of charged, primary particles in Xe-Xe collisions at √ s NN = 5. 44 TeV. The size of the Xe-Xe system is intermediate between previously studied systems at the LHC, Pb-Pb [4-8] being the largest and p-Pb and pp [9, 10] the smallest. The charged-particle pseudorapidity density is presented over the interval −3.5 < η < 5 and as a function of the collision centrality. The mid-rapidity values normalised by the number of participating nucleon-nucleon pairs are also reported. The results are also compared with measurements at lower collision energies and with theoretical calculations.

Experimental setup
The data were recorded with the ALICE apparatus in 6 hours of stable data-taking with 129 Xe beams (16 bunches per beam) colliding at √ s NN = 5. 44 TeV in October 2017. The data were collected with a reduced magnetic field of 0.2 T (as compared to the nominal value of 0.5 T) in the ALICE solenoid magnet. The performance and a detailed description of ALICE can be found elsewhere [11]. In the following, the detector elements relevant to this analysis are briefly described.
The innermost part of the tracking system of ALICE is the Silicon Pixel Detector (SPD) [12] which consists of two cylindrical layers of hybrid silicon pixel assemblies. The inner and outer SPD layers are placed at radii of 3.9 and 7.6 cm from the interaction point and cover |η| < 2 and |η| < 1.4, respectively. The Forward Multiplicity Detector (FMD) [13,14] consists of three sets of silicon strip sensors, covering the pseudorapidities −3.5 < η < −1.8 and 1.8 < η < 5. The multiplicity is obtained by measuring the total energy deposited by charged particles. The V0 detector [14,15] is used for triggering and centrality classification. It consists of two sub-detectors, V0-A and V0-C, covering the pseudorapidity regions 2.8 < η < 5.1 and −3.7 < η < −1.7, respectively. The V0 has a timing resolution better than 1 ns, allowing its fast signals to be combined in a programmable logic to reject beam-induced background events while ensuring maximum efficiency for the selection of collision events. The Zero-Degree Calorimeters (ZDCs) [16] are located at a distance of 112.5 m from the interaction point along the beam line, on either

Data sample and analysis method
The hadronic interaction rate in ALICE was about 150 (80) Hz at the beginning (end) of the data-taking. The magnetic field of 0.2 T, reduced as compared to normal Pb-Pb settings (0.5 T) increases the acceptance for low-momentum particles, thus enhancing the acceptance of the V0 system for electromagnetic (EM) interactions, which constitute a background for this analysis. In order to suppress this source of contamination, the minimum bias interaction trigger required a signal in each of the V0 sub-detectors in coincidence with a signal in each of the two neutron ZDCs. It was verified by means of a set of control triggers that such a trigger is fully efficient for hadronic interactions in the 0-90% centrality range. In addition, beam-background interactions are removed using the V0 and the ZDC timing information. The interaction probability per bunch-crossing was sufficiently small that the chance of two hadronic interactions occurring within the integration time of the involved detectors, so-called pileup events, was negligible. A total of about 1 million hadronic collisions are used in this analysis.
The classification of collisions into centrality classes uses the sum of the amplitudes of the signals in the V0-A and V0-C detectors. A model of particle production, based on a Glauber description [17,18], is fitted to the V0 amplitude distribution [19]. The number of particles in the V0 detector is calculated with a two-component model for the number of sources given by where f constrains the relative contributions of N part and N coll , coupled to a particle production model for each source parameterised by the negative binomial distribution (NBD). In the Glauber calculation, the nuclear density for 129 Xe is described by a Woods-Saxon distribution for a deformed nucleus ρ(r, ϑ ) = ρ 0 1 ( The parameter ρ 0 is the nucleon density, which provides the overall normalisation. The nuclear skin thickness is a = 0.59 ± 0.07 fm [20]. The nuclear radius R is parametrised as a function of the polar angle ϑ by R(ϑ ) = R 0 [1 + β 2 Y 20 (ϑ )], where R 0 is the average radius and the Legendre polynomial Y 20 describes the nucleus deformation for an axially symmetric case with no dependence on the azimuthal angle. For the average radius we used R 0 = 5.4 ± 0.1 fm, scaling the results for 132 Xe reported in [20] by the atomic mass number (A) dependence of the radius, namely (129/132) 1/3 [18]. The deformation parameter β 2 = 0.18 ± 0.02 is obtained by linearly interpolating the values measured for the Xe A-even isotopes from 124 to 136 [21]. In the Glauber model calculation, the orientation of the spheroid symmetry axis is randomly sampled. For √ s NN = 5.44 TeV collisions, an inelastic nucleon-nucleon cross section of 68.4 ± 0.5 mb, obtained by logarithmic interpolation of cross section measurements with respect to collision energies in pp collisions [22], is used. The NBD-Glauber fit provides a good description of the observed V0 amplitude in the region corresponding to the top 90% of the hadronic cross section, where the effects of trigger inefficiency and contamination by EM processes are negligible. The average numbers of participants N part reported in Tab. 1 are estimated from the Glauber model imposing the same cuts applied to the data on the simulated V0 response. One should note that the centrality selection based on the V0 amplitude induces a bias on the measured dN ch /dη . This leads to a dN ch /dη in the 70-80% (80-90%) centrality class about 3% (10%) lower than the value one would obtain with a centrality selection based on the impact parameter.
For all the collisions in the 0-90% centrality range the coordinates of the primary interaction point can be reconstructed with good accuracy by correlating hits in the two SPD layers. The measurement of the charged-particle multiplicity density at mid-rapidity uses information from the SPD. The acceptance of the SPD for charged particles spans different pseudorapidity regions depending on the position of the interaction point along the beam line, z. For example, for collisions with the vertex located within |z| < 7 cm a maximum acceptance of |η| < 1.5 can be reached, with approximately constant acceptance for |η| < 0.5. To extend the pseudorapidity coverage up to |η| < 2, all collisions with a primary vertex located within |z| < 20 cm have been considered.
Following the method developed earlier [4-8], tracklets (short track segments) are formed using the position of the primary vertex and all possible combinations of hits between the two SPD layers. The primary charged-particle multiplicity density dN ch /dη is obtained from the number of tracklets that pass the quality selection criteria, after correcting for detector acceptance, reconstruction and selection efficiencies and contamination from combinatorial background and secondary charged particles. This selection allows primary charged-particle detection down to a momentum of 30 MeV/c. The corrections are estimated using a detailed simulation based on events generated with the HIJING event generator [23] with particle transport in ALICE performed by GEANT3 [24]. The decay products of long-lived decaying particles like K 0 S , Λ,Λ and other strange hadrons are classified as secondary particles [25] and the contamination from these particles is subtracted from data. It is known that HIJING underestimates the relative production rate of strange particles in high-energy heavy-ion collisions. For this reason, the simulation has been reweighed to reproduce the relative particle abundances observed in the data which are about 30% (50%) higher than HIJING in the most central (peripheral) collisions. The reweighing factors have been derived from an estimate of K 0 S , Λ andΛ relative production in the data, obtained via invariant mass reconstruction and compared to HIJING.
The deposited energy signal in the FMD is used to measure the charged-particle pseudorapidity density in the forward regions (−3.5 < η < −1.8 and 1.8 < η < 5), following the method described elsewhere [5]. The energy loss is measured in the 51,200 Si strip sensors of the detector and a statistical approach is used to calculate the inclusive number of charged particles. A data-driven correction derived from previous studies [6] corrects for the background of secondary particles, which are abundant in the forward regions.

Systematic uncertainties
The systematic uncertainties on N part are obtained by varying the parameters of the Glauber model independently within their estimated uncertainties and repeating the NBD-Glauber fit. The uncertainty due to the centrality determination is estimated by changing the value of V0 amplitude that corresponds to the top 90% of the hadronic cross section by ±0.5%. This results in an uncertainty on dN ch /dη of 0.1% to 4.8% from central to peripheral collisions. An additional 4% uncertainty assigned to the most peripheral class, arising from the remaining contamination from EM processes, was estimated by studying the energy deposition in the ZDCs [26].
For the tracklet analysis at mid-rapidity the relative systematic uncertainty on the measurement of the charged-particle multiplicity in peripheral (central) events arises from the following sources: tracklet selection 0.1% (0.8%), calculated by varying the tracklet quality cut up to 4 times the nominal value; combinatorial background subtraction 0.5% (2.0%), estimated from simulations and cross-checked using an alternative method where artificial SPD clusters are added to the data and the number of corresponding artificial reconstructed tracklets are used for background subtraction; particle composition 0.2% (0.2%), estimated by changing the relative abundances of protons, pions and kaons by ±30% in the simulation; contamination by weak decays 0.3% (0.3%), estimated by changing the reweighting factors; extrapolation to zero transverse momentum 0.6% (0.6%), obtained from the variation of the estimated yield of particles at low transverse momentum by a factor of two in the simulation; variations in detector accep- tance and efficiency 1% (1%), evaluated by carrying out the analysis for different slices of the z-position of the interaction vertex and with subsamples in azimuth. At forward rapidities, the uncertainties related to the measurement of multiplicity arise from the following sources: the data-driven correction for secondary particles [6] 6.1%; the merging algorithms of signals from Si strips to a single particle 1%; variation in rejection threshold for calculation of the charged-particle multiplicity per event +1% −2% ; particle composition 2%, estimated in the same way as in the tracklet analysis.
The systematic uncertainties from centrality selection and electromagnetic interactions affect the overall normalisation of the results. The total systematic uncertainty, obtained by adding in quadrature all contributions, amounts to 6.4% (2%) for peripheral (central) in |η| < 2, to 6.9% for η > 3.5 and to 6.4% elsewhere in the forward region, and is partially correlated over η and between different centrality classes.

Results
Figure 1 presents the charged-particle multiplicity density dN ch /dη as a function of pseudorapidity for 12 centrality classes. The measurement is obtained from the SPD at mid-rapidity, FMD in forwardrapidities, and combined in regions of overlap (1.8 < |η| < 2) between the two detectors by taking the weighted average using the non-shared uncertainties as weights. The data are symmetrised around η = 0, averaging positive and negative η results wherever possible, and extended into the non-measured region −5 < η < −3.5 by reflecting the 3.5 < η < 5 values around η = 0. Averaged values (left and right) agree within the uncertainties. Assuming that the charged-particle rapidity density dN ch /dy has Gaussian shape and using an effective Jacobian, the measured dN ch /dη is fitted with this ansatz and a width of σ = 4.4 ± 0.1 is found, consistent with the value obtained in Pb-Pb at √ s NN = 5.02 TeV [8].
The multiplicity density averaged over |η| < 0.5 in different centrality classes is shown in Tab. 1. The total charged-particle multiplicity N tot ch is determined from the data in the measured region and from  Fig. 1: Charged-particle pseudorapidity density for 12 centrality classes over a broad η range in Xe-Xe collisions at √ s NN = 5.44 TeV. Boxes around the points reflect the total systematic uncertainties, while the filled squares on the right reflect the normalisation uncertainty from the centrality determination. Statistical errors are negligible.
The reflection (open circles) of the 3.5 < η < 5 values around η = 0 is also shown. The lines correspond to fits to a gaussian distribution in rapidity multiplied by an effective Jacobian of transformation from η to y.
extrapolations, up to η = ±y beam , in the unmeasured region. Three different functions are used to extrapolate the data points: the difference of two Gaussian distributions centred at η = 0; a Woods-Saxon-like distribution in rapidity as proposed by PHOBOS [27]; and a trapezoidal form. The trapezoid ansatz in the forward unmeasured regions corresponds to a linear extrapolation up to η = ±y beam with the starting point constrained by the measurements. A colour-glass-condensate ansatz [28] yields similar extrapolated values beyond the measured region. A Gaussian dN ch /dy in rapidity results in a distribution in pseudorapidity which is very similar to the difference of two Gaussians centered at η = 0. The central value in the unmeasured regions (−8.6 < η < −3.5 and 5 < η < 8.6) is taken as the average between the trapezoidal function (which gives the lowest N tot ch ) and the difference of two Gaussian distributions centred at η = 0 (which gives the highest N tot ch ). The contribution from the extrapolated region is less than 30% of N tot ch . The systematic uncertainty of the extrapolated N tot ch is calculated as the quadratic sum of contributions from the systematic uncertainty of the data and a conservative contribution obtained by comparing the results from the different fit functions. It amounts to about 9% (3%) of N tot ch in peripheral (central) events. In order to compare bulk particle production at different energies and in different collision systems, the average charged-particle multiplicity density dN ch /dη at mid-rapidity is divided by the average number of participating nucleon pairs, N part /2. This allows one to compare nuclear collisions to pp and pp collisions. The N part values are calculated within the Glauber model.
Figure 2 (top) shows the mid-rapidity charged-particle multiplicity normalised by the number of nucleon pairs participating in the collision, 2 N part dN ch /dη , in pp, pp, p(d)A and in central heavy-ion collisions as a function of the centre-of-mass energy. The lines represent fits to lower energy results. The Xe-Xe result is in agreement within the uncertainties with the trend established from previous heavy-ion measurements, which shows a stronger rise as a function of √ s NN than for pp and p-Pb collisions. Figure  2 (bottom) shows the total charged-particle multiplicity per participant nucleon pair 2 N part N tot ch , which follows the trend for central heavy-ion collisions. Figure 3 shows the centrality dependence of the mid-dN ch /dη in Xe-Xe at √ s NN = 5.44TeV ALICE Collaboration by the number of participant pairs shows a slight increase as a function of the number of participants in Fig. 3 (bottom) , similar to that of the midrapidity results, albeit with larger experimental uncertainties.
The study of the centrality dependence of particle multiplicity for different collision systems provides a variable number of nucleon-nucleon collisions at equal number of participating nucleons and therefore may provide further information to clarify the measured deviation from N part scaling. The scaling of the charged-particle multiplicity by the number of participant nucleons was studied in detail and a deviation from N part -scaling was observed at RHIC energies [3, 32, 40-42]. Such a deviation was first interpreted as being due to an increasing contribution of hard processes for central events, but no conclusive evidence was found to support this interpretation.  functions were fitted to the Pb-Pb data at √ s NN = 5.02 TeV [7]. For the Xe-Xe data only the absolute normalisation was adjusted. The values of the parameters are also consistent with those obtained at SPS and RHIC energies [32,43]. While no unique physics conclusion can be drawn from such fits, this suggests that geometrical arguments may be sufficient to provide a good description of particle production across different colliding systems and beam energies.
Describing particle production in relativistic heavy-ion collisions as a superposition of emission from a thermal core and hard scatterings in a corona [44], one can classify the participating nucleons into those that scatter only once (N corona part ) and those that scatter multiple times (N core part ). The multiplicity can then be fitted with the sum of those contributions, dN ch /dη pp N corona part + dN ch /dη core N core part , where dN ch /dη pp is the multiplicity measured in inelastic pp collisions [9] and dN ch /dη core is the contribution to the charged-particle multiplicity from the core of the fireball, which is fitted to the data. Figure 4 also shows dN ch /dη per participant quark N q-part calculated with a Glauber model using effective wounded constituent quarks [46] [45], as a function of N part , as was done for Pb-Pb collisions at √ s NN = 5.02 TeV [47]. In the implementation of the quark-Glauber model the partonic degrees of freedom (3 or 5) are located around the nucleon centres [45]. The effective inelastic scattering cross section for collisions of constituent quarks is set to 20.38 mb and 9.76 mb, for N q = 3 and N q = 5, respec-  [28,48,49,[51][52][53][54][55][56][57][58][59][60][61][62][63][64][65][66]. The bottom panel shows the ratio of the models to the data. The shaded band around the points reflects the correlated systematic uncertainties.
tively, adjusted to reproduce the 68.4 mb nucleon-nucleon inelastic cross section at 5.44 TeV. N q-part has been divided by the average value in pp collisions µ = N q-part , which is 3.5 (4.3) for N q = 3 (N q = 5). Comparing the behaviour of dN ch /dη in terms of the dependence on the number of nucleon or quark participants in the collision, one concludes that N q-part scaling describes the data better than N part scaling as previously observed [42,47] except the 0-10% centrality range in Xe-Xe collisions where a clear scaling violation is observed. Figure 5 shows a comparison of the Xe-Xe data to calculations from theoretical models at mid-rapidity. HIJING 2.1 [48,49] combines perturbative QCD processes with soft interactions, and includes a strong impact parameter dependence of parton shadowing [50].
For Xe-Xe data at √ s NN = 5.44 TeVit uses a large gluon shadowing parameter of 0.28 to limit the multiplicity per participant. With this choice, the same as in Pb-Pb collisions at √ s NN = 5.02 TeV, the multiplicities at mid-rapidity and the centrality dependence in the most central collisions are reproduced. AMPT [52,53] is a model which implements hydrodynamical evolution of an initial state produced by dN ch /dη in Xe-Xe at √ s NN = 5.44TeV ALICE Collaboration HIJING. It includes spatial coalescence of quarks to hadrons, followed by hadronic scattering. AMPT describes both the shape and the overall magnitude of the mid-rapidity data. PYTHIA/Angantyr [54] extends the nucleon-nucleon model of PYTHIA 8.230 [55] to the case of heavy-ion collisions, essentially performing individual nucleon-nucleon collisions at the parton level, while the resulting Lund-strings are hadronised as an ensemble. It is interesting to note that this model agrees reasonably well with the data even though it was developed as an extension of a generator for nucleon-nucleon collisions. EPOS LHC [51] is a parton model based on the Gribov-Regge theory, designed for minimum bias hadronic interactions, which incorporates collective effects treated via a flow parameterisation and a separation of the initial state into core-corona parts. The shape of the centrality dependence is reproduced fairly well at intermediate centralities, however, the model underestimates the absolute values of the multiplicity, as was the case in Pb-Pb collisions at √ s NN = 5.02 TeV [7]. The Duke global calibrated model is based on a Bayesian Statistics analysis using T R ENTo initial conditions for high-energy nuclear collisions [67,68]. The subsequent transport dynamics is then simulated using the iEBE-VISHNU event-by-event simulations for relativistic heavy-ion collisions which uses a hybrid approach based on (2 + 1)-dimensional viscous hydrodynamics coupled to a hadronic cascade model [69]. The Duke global calibrated model can reproduce the shape of the mid-rapidity distribution, but overestimates slightly the overall magnitude.
Saturation-inspired models (rcBK-MC [56,57], KLN [28,[58][59][60], ASW [61], IP-Glasma [62,63] and EKRT [64][65][66]) rely on perturbative QCD and an energy-dependent saturation scale, which limits the number of produced partons, and in turn the number of produced particles. This results in a factorisation of the energy and centrality dependence of particle production or, in other words, in the invariance of the centrality growth, as observed in the experimental data [70]. The rcBK-MC model limits the centrality growth using the rc-BK equation. It provides a good description of the mid-rapidity data, both of the shape and the highest multiplicity reached in central collisions. The ASW prediction overestimates the data, while it was very accurate in Pb-Pb at √ s NN = 5.02 TeV. The KLN model does not describe the shape well and, although it agrees with the value measured for most central collisions, it is significantly above the centrality dependence of the data. The IP-Glasma model naturally produces initial energy fluctuations computed within the Color Glass Condensate framework combining an impact parameter dependent saturation model. It uses a gluon multiplicity scaled to describe hadron multiplicities measured in Pb-Pb collisions at √ s NN = 5.02 TeV [7]. The centrality dependence is stronger than that observed in mid-rapidity data over(under)-predicting the data in central (peripheral) collisions. The EKRT model for heavy-ion collisions uses perturbative QCD with a conjecture of gluon saturation to suppress soft parton production. The saturation scale is also dependent on the local product of thickness functions, implying a geometrical scaling. The space-time evolution of the system is then described with viscous fluid dynamics event-by-event. The normalisation is fixed by exploiting the 0-5% most central multiplicity measurement in Pb-Pb collisions at √ s NN = 2.76 TeV [71]. As for Pb-Pb collisions at √ s NN = 5.02 TeV, the EKRT model can describe both the shape and the overall magnitude of multiplicity on centrality.
In Fig. 6, the models are compared to the pseudorapidity dependence of the dN ch /dη for the top 5% central collisions. HIJING 2.1 reproduces the pseudorapidity dependence at mid-rapidity well, but overestimates the data at forward rapidity, due to the large value of the shadowing parameter used. AMPT and PYTHIA/Angantyr describe the data fairly well, with a slight overestimate at forward rapidities. EPOS LHC reproduces the shape well, but under-predicts the multiplicity overall. The rcBK-MC is restricted to |η| < 2.5 since its formalism can only be used for rapidities far from the fragmentation regions. It shows a narrower distribution than what is seen in data. KLN agrees with the data at mid-rapidity, but not at forward rapidity, where it under-predicts the data. For IP-Glasma the rapidity dependence is provided by the IP-Sat model [72] and it is converted to pseudorapidity using an effective mass of 0.2 GeV/c 2 . The shape is wider than that of the data.

Conclusions
The measurements of the charged-particle multiplicity density and its centrality dependence in Xe-Xe collisions at √ s NN = 5.44 TeVhave been presented over the pseudorapidity range −3.5 < η < 5 using the full acceptance of the ALICE detector. For the 5% most central collisions, the average charged-particle pseudorapidity density at mid-rapidity (|η| < 0.5) is 1167 ± 26 and the total number of charged particles is 13230 ± 370. Scaled by the number of participant pairs, these are found to follow the same power-law dependence with energy established in previous heavy-ion measurements.
The centrality dependences of 2 N part dN ch /dη and 2 N part N tot ch are very similar to those previously measured in Pb-Pb collisions at similar or lower energies up to the 10% most central Xe-Xe collisions, where the Xe-Xe results are larger than the Pb-Pb results at a similar number of participating nucleons. Similar conclusions were drawn at RHIC from the comparison of the data for Cu-Cu and Au-Au collisions [73]. While measurements of particle production in large and medium-sized colliding systems such as Xe-Xe are abundant and become even more precise, the underlying mechanism to describe the increase with dN ch /dη in Xe-Xe at √ s NN = 5.44TeV ALICE Collaboration energy and centrality is still not completely understood. Deeper insight of the system size dependence of particle production may come from the study of light-nuclei collisions, still not much explored at high energy, which could bridge the gap between the trends observed in pp and pA collisions and those of the mid-sized Xe-Xe and the large Pb-Pb systems.