WorksPhysics and Astronomy
Improved nonflow control for multiparticle correlators and system-size-dependent evolution from OO to Pb–Pb collisions at the LHC
Abstract
Multiparticle correlations of event-plane angles, known as symmetry plane correlators (SPCs), and correlations of anisotropic flow amplitudes of different harmonics, known as symmetric cumulants (SCs), help constrain the initial conditions and the transport properties of the system formed in heavy-ion collisions. Such multiparticle correlations are robust against nonflow biases in heavy-ion collisions; however, in small systems the nonflow contribution is sizeable and can distort SPCs and SCs at low multiplicity, mimicking a flow-like signal. Evidence of nuclear-geometry-induced collectivity in light-ion collisions, such as OO and Ne–Ne, has been established at the LHC. Measurements of SPCs and SCs in these systems, and their comparison to heavy-ion collisions, can therefore clarify the system-size dependence of these correlations and the strength of nonlinear mode coupling. In this work, an improved nonflow subtraction framework based on multiplicity reweighting (\czero) and -estimator-based (\cone) scaling is validated for selected SPCs and SCs using a controlled toy model and the Pythia~8~Angantyr model, where collective flow is absent by construction. We find that \cone provides the more reliable nonflow control and apply it to predict the SPCs and normalized SCs in OO collisions at ~TeV with the ampt string-melting model. The results are compared with published Pb–Pb measurements from ALICE and from ampt as a function of midrapidity charged-particle multiplicity , placing light- and heavy-ion systems on a common system-size axis.
Introduction
Azimuthal anisotropy is a key signature of the strongly interacting matter created in ultrarelativistic heavy-ion collisions [1–3]. In a noncentral collision, the nuclear overlap is anisotropic in the transverse plane, arising from the almond-shaped collision geometry or from the event-by-event fluctuations of nucleon (and sub-nucleonic) positions. Hydrodynamic evolution converts this initial spatial anisotropy into momentum-space anisotropies, characterized by the harmonic amplitudes and the associated symmetry planes . Correlations among the and among the then provide complementary constraints on the initial state and on the transport properties of the medium.
Symmetric cumulants (SCs) and their normalized form (NSCs) measure the correlations of event-by-event fluctuations of anisotropic flow harmonic amplitudes, as studied extensively in Pb–Pb collisions [5, 6] [4]. They isolate genuine correlations between and beyond trivial factorization and are sensitive to both initial-state fluctuations and the nonlinear hydrodynamic response [7]. In parallel, correlations among symmetry planes, known as symmetry plane correlators (SPCs), encode how event-plane angles of different-order harmonics align. Early measurements using the scalar-product method [8] were shown to be biased by simultaneous correlations among flow amplitudes [9, 10]. The Gaussian estimator (GE) removes that bias and enables a cleaner extraction of SPCs and has been studied in Pb–Pb collisions at and 5.02 TeV, including higher-order combinations involving up to five planes by ALICE [10, 11]. Hydrodynamic models describe the overall trends but often overestimate correlators that are particularly sensitive to nonlinear mode coupling [10, 11].
A natural next step is to extend SPC and SC measurements from heavy ions to light-ion systems. Collisions of and at the LHC bridge the multiplicity gap between small ( and ) and large (Pb–Pb) systems. Recent measurements have already established finite elliptic and triangular flow, and , and strong evidence of nuclear-geometry-driven collectivity in OO and Ne–Ne collisions [12, 13, 14]. Those single-harmonic observables, however, largely reflect the overall strength of the medium in converting initial-state spatial anisotropy and energy fluctuations into the final-state momentum space anisotropy. The study of SPCs and SCs goes further as they probe the correlations among different symmetry planes and among different flow amplitudes. Together, these observables are therefore more sensitive than alone to nonlinear mode coupling and to finer details of the initial geometry, including features of the nuclear structure [15, 16]. A system-size scan of SPCs and SCs can further reveal whether the same hydrodynamic response that describes Pb–Pb persists in light ions, and how initial-state geometry shapes multiparticle correlations from small to large systems [17].
In heavy-ion collisions, multiparticle correlators are typically dominated by long-range collective flow, while1 short-range few-body correlations—commonly called nonflow—are comparatively small when the analysis is restricted to the soft sector (typically GeV/), where jet-driven contributions are suppressed [18] [1]. Nonflow arises from sources unrelated to the global collision geometry: jet fragmentation, resonance decays, quantum statistics (HBT), and other local particle clusters [19] [1]. Although multiparticle cumulants and subevent methods can suppress nonflow contributions, residual contamination may remain, particularly at relatively low multiplicities, motivating further development and systematic validation of alternative approaches [18, 20, 21, 22, 23, 24]. Because these contributions involve only a limited number of correlated particles, their relative weight grows as the multiplicity decreases, scaling roughly as for an -particle correlator in the independent-source picture [19, 20]. Consequently, in small and intermediate systems—where is much lower than in mid-central Pb–Pb—nonflow can dominate SPCs and SCs at low multiplicity and mimic a flow-like signal. Reliable interpretation of these observables in OO and Ne–Ne therefore requires an explicit nonflow estimate and subtraction, beyond the partial suppression offered by rapidity gaps or subevent methods alone [19].
In this paper, improved nonflow subtraction methods are tested for SPCs and SCs in OO collisions at TeV before being applied to the predictions from A Multi-Phase Transport (AMPT) string-melting model. Section II introduces the observables and the nonflow subtraction framework based on multiplicity reweighting () and -estimator scalings (). Section III validates the procedure on a controlled toy model with known flow and nonflow, and in OO collisions at TeV from PYTHIA 8 Angantyr model, where collective flow is absent by construction. Section IV applies the prescription that best suppresses the nonflow to AMPT string-melting OO collisions at the same energy and compares the results with published Pb–Pb measurements from ALICE and the AMPT model. Section V summarizes the findings. Details of the toy-model generator, the AMPT model, the scaling factors, and the remaining higher-order SPCs and NSCs are given in the appendices.
Observables and nonflow subtraction
The nonflow subtraction methods used in this work are introduced in this section, together with the definitions of the prime observables, SPCs and normalized SCs. The choice of the nonflow estimator for each correlator and the domain of applicability are also mentioned.
Low-multiplicity anchored subtraction
In the independent-source picture, nonflow in an -particle correlator comes from local clusters whose number grows with the multiplicity [19]. Of the distinct -tuples in an event, only are drawn from one correlated source, so the nonflow contribution scales as
The first relation is the multiplicity baseline (). The second is the dipole refinement (): a correlator that contains follows the same multiplicity dependence and absorbs the extra dependence induced by multiplicity selection. Such correlators are dominated by nonflow, because the genuine dipolar flow changes sign with and nearly cancels after integration [19].
The nonflow template is the same correlator in the low-multiplicity (LM) class of the same data set, taken here as the 70–100% multiplicity class. Collective flow in that class, if any, is neglected, which is justified when the target multiplicity lies well above the LM multiplicity [19]. The subtracted correlators are given as, scaling:
scaling:
Here, is a nonflow estimator of the same particle order as the target and contains at least one power of . The multiplicity in Eq. (2) is fixed by the -tuple average of [19],
where the sum runs over events in the class and is the number of distinct -tuples. The same average in the LM class defines . For , and reduces to . It satisfies . High-multiplicity events dominate that average inside a broad class, so exceeds the mean multiplicity. Using in place of under subtracts the nonflow, and the bias is largest in the wide LM class.
Domain of applicability, assumptions, and advantages over other methods. The nonflow subtraction used in this work relies on the independent-source picture [19]. Here, the scaling of Eq. (1) is the leading expectation and not an exact rule. This scaling assumes that the measured correlator is a sum of genuine flow and nonflow, that the shape of the nonflow correlation does not change with multiplicity, and that the LM class has no genuine collective flow. It is important to note that any flow left in the LM template is subtracted along with the nonflow, which biases the final result [19]. Thus, the method is applicable only when the multiplicity of the target class is much larger than that of the LM class, i.e., , and at low ( GeV/), where the nonflow is dominated by dipolar correlations. At high , or when a single jet dominates the event, the scaling breaks down [19]. In this study, the observables are integrated over GeV/, which goes beyond this limit. However, the closure test with Pythia 8 Angantyr in the same range shows that the subtraction removes most of the nonflow, leaving a residual at the 10% level or below for most of the observables (Table II). Moreover, the results are shown for the 0–60% classes, which are well separated from the 70–100% LM class.
The scaling further assumes that the -integrated dipolar flow is negligibly small. The -based estimator can therefore be used to test the assumed scaling of nonflow. It also accounts for deviations associated with multiplicity selection, which are not captured by the pure multiplicity scaling of [19]. Moreover, for the SPCs, only the numerator is corrected, which assumes that the nonflow in the denominator is small. Reference [19] has validated this framework only with the Heavy Ion Jet Interaction Generator (HIJING), which has no flow, for , , and , and found that the best estimator differs from one observable to another. Therefore, before applying it to the SPCs and normalized SCs involving higher harmonics, we validate the method in Sec. III using Pythia 8 and a toy model, where the latter also tests whether a known flow signal is recovered. In contrast to the subevent and pseudorapidity-gap methods, the template subtraction uses the full detector acceptance, thus retaining the statistical precision. Further, it can be applied in the same way to both SPCs and SCs using either LM class of the same data set or minimum-bias collisions as the reference.
Symmetry plane correlators
The -particle azimuthal correlator is given as
where is the azimuthal angle of the -th particle, and the average runs over all distinct sets of particles in one event and then over the event sample [7, 11]. The integers are the signed coefficients of the positive harmonics . They satisfy , and the particle order is . Equation (5) factorizes into the flow vectors
| Observable | Numerator | Estimator |
| Symmetry plane correlators (SPC) | ||
| Normalized symmetric cumulants (NSC) | ||
| NSC(4, 2) | SC(4, 2) | SC(1, 2) |
| NSC(3, 2) | SC(3, 2) | SC(1, 2) |
| NSC(4, 3) | SC(4, 3) | SC(1, 3) |
| NSC(5, 2) | SC(5, 2) | SC(1, 2) |
| NSC(5, 3) | SC(5, 3) | SC(1, 3) |
Table 1. Estimators for the subtraction of each symmetry plane correlator and normalized symmetric cumulant. The subtraction uses the same numerators with the multiplicity factor of Eq. (2). Here, , , and .
, as
where is the -th order symmetry plane angle. The index counts only the distinct harmonics, so .
The numerator used in the symmetry plane correlator is the real part of the measured correlator in Eq. (6),
The denominator is the -particle correlator in which every harmonic is paired with its complex conjugate, such that the event-plane angles are dropped,
Using the Gaussian estimator (GE), the SPC is estimated by [9, 11]
Because the magnitudes remain correlated inside , this estimator avoids the bias of the scalar-product form, whose denominator factorizes into single-harmonic moments assuming that event-by-event fluctuations of flow amplitudes are mutually independent. Both and are obtained from the recursive algorithm of the generic framework [7]. The recursion builds each correlator from -vectors in one pass over the event and removes autocorrelations by construction. This also reduces the computational cost significantly. The measured numerator in Eq. (7) still contains nonflow. That contribution is subtracted with Eqs. (2) and (3); the denominator is left unchanged, and the estimator in Eq. (9) is rebuilt from the subtracted numerator. In the scaling, the nonflow estimator is an estimator of the same particle order as the numerator itself: and for the three-particle correlators such as , , and , and for the four-particle correlators such as (Table I).
Normalized symmetric cumulants
The symmetric cumulant of two flow harmonics and is defined as a four-particle cumulant [7]
where and are the two- and four-particle correlators of Eq. (5), with events weighted by the number of distinct pairs and quadruplets, respectively. In the absence of nonflow contributions, the SC measures the correlation between the event-by-event fluctuations of and through the covariance term of and in Eq. (10). A positive (negative) value of indicates that the two harmonics are correlated (anticorrelated).
To reduce the dependence on the magnitudes of the flow harmonics, the SC is normalized as [5] [4]
where estimates . All correlators are calculated using the -vector formalism [7, 21]. The calculations are performed for the harmonic pairs , , , , and . Results for and are presented in the main text, while those for the remaining combinations are given in Appendix D.
Since the two terms of Eq. (10) carry different event weights, averaging over a broad multiplicity class can produce a spurious SC even without genuine four-particle correlations. Following Ref. [19], is therefore calculated in unit- bins and averaged with the four-particle weights
where is the number of distinct quadruplets, as in Eq. (4). The two-particle cumulants used to calculate are evaluated over the full multiplicity class.
Nonflow subtraction using the template method
Equations (2) and (3) are applied to the four-particle cumulant, . For the scaling,
with the four-particle effective multiplicities from Eq. (4). For the scaling, the estimator is the symmetric cumulant between the first harmonic and the lower harmonic of the target, ,
Thus, is used for , and , while is used for and (Table I). is obtained by replacing with in Eq. (11).
Kinematics and multiplicity selection
In this study, the SPCs and NSCs are calculated using primary charged particles [25] with at midrapidity, , similar to the acceptance of the ALICE detector at the LHC. For the multiplicity percentile estimation, the acceptance regions of the forward-timing (FT0) detectors of ALICE in Run 3 [26], known as the FT0M multiplicity, which is the sum of the amplitudes measured with FT0A () and FT0C (), are used. This selection ensures a minimal autocorrelation among the tracks, since the tracks used for observables at midrapidity are different from the ones used for the estimation of the event multiplicity at forward rapidity. The multiplicity percentiles are then divided into nine classes with edges at 0, 5, 10, 20, 30, 40, 50, 60, 70, and 100%, where the LM template belongs to the 70–100% multiplicity class.
Nonflow closure
In this section, the nonflow subtraction closure is demonstrated first on a toy model and then on PYTHIA 8 to ensure the effectiveness of the methods in removing the nonflow as well as to estimate the nonflow residuals for each observable considered in this study.
Toy-model closure
The nonflow subtraction methods described in Sec. II are validated on event samples generated using a toy model. The parameters of the toy model are presented in Appendix Appendix A. This closure is crucial since the flow and nonflow pieces in the toy model are known separately. Three samples with an equal number of events are generated: one with only flow, one with only nonflow, and one with both (mixed). All three samples use the nine multiplicity classes as stated in Sec. II D.
In each event a fraction of the tracks, or at least six, is reserved for nonflow clusters, and the remaining tracks form the bulk. This fraction is held fixed so that the nonflow contribution stays finite at high multiplicity. The same split is used in every sample, so at a given multiplicity the three samples have the same number of tracks in the reserved groups and the same number in the bulk. This is called occupancy-matching. Each such reserved group of tracks forms a cluster of six tracks in the nonflow and mixed samples: six particles at fixed relative azimuths and a random overall orientation. This adds the known nonflow correlation. However, in the flow sample, for these reserved groups, the six azimuths are drawn independently for each cluster, so there is no nonflow correlation. The remaining tracks in the event, forming the bulk, are emitted from one common plane in the flow and mixed samples, with and fixed . In the nonflow sample, the bulk is drawn uniformly in azimuth, i.e., with no flow modulation. To remove any correlation between the nonflow cluster and the bulk, each cluster is rotated at random and is not aligned with the common plane, so particles from a cluster are uncorrelated with particles from the plane. The correlator measured in the mixed sample is then the sum of the correlator of the flow sample and the correlator of the nonflow sample. In the mixed sample, the 70–100% class is filled differently from the other eight classes. The clusters are generated as in the other classes, while the remaining tracks have uniform azimuths and are not attached to a common plane. The correlator of this class therefore contains the clusters and no flow. Equations (2) and (3) use it as . For the sake of simplicity, the values of are made multiplicity independent, and the relative azimuths of the tracks inside a cluster are given in Appendix A.
Each sample is passed through the same recursive SPC analysis framework independently. Equations (2) and (3) are applied only to the mixed sample, using its own LM bin. The flow and nonflow samples are the truth against which that result is compared. Figure 1 shows the closure for the numerators and in the 0–60% multiplicity classes. In the upper panels, the mixed points follow an approximate sum of correlations from the flow truth and the nonflow truth. The gradual increase of correlations towards low-multiplicity events in the mixed sample arises purely from nonflow in this toy model. After and scalings are applied to the mixed sample, the nonflow-subtracted correlations move closer to the flow truth. The lower panel is the ratio of each subtracted correlator to the flow truth, which shows that the scaling works better than the scaling, as the latter seems to oversubtract nonflow towards the peripheral bins. The
Figure 1. Toy-model closure examples for the numerators (top) and (bottom). Upper panels: mixed sample, flow truth, nonflow truth, and the mixed sample after and subtraction. Lower panels: subtracted correlator divided by the flow truth.
nonflow sample shows the contribution that the subtraction should remove.
Pythia 8 as a nonflow reference
To evaluate the performance of the proposed nonflow subtraction methods, the PYTHIA 8 ANGANTYR model [27, 28, 29] is used to generate approximately 165 million minimum-bias OO collisions at TeV. The nucleon–nucleon cross section uses the built-in parameters (2.15, 18.42, 0.33), with no refit [29]. Particles with mm are left undecayed, following the ALICE primary-particle selection [25]. In this configuration, there are no collective flow components in the PYTHIA 8 ANGANTYR model. The azimuthal correlations come from sources such as jet production and fragmentation, multiparton interactions, resonance decays, and momentum conservation, which make the sample a nonflow-only reference.
In PYTHIA events, the standard multiparticle cumulants are dominated by dijets [18]. This has been used to study nonflow in symmetric cumulants [23], to test two-particle factorization and low-multiplicity template methods [30, 31, 32, 33], and to study the correlation of with the mean transverse momentum [34]. A multiplicity-scaled subtraction of multiparticle cumulants has been tested with HIJING in OO collisions at the same energy [19]. In this study, the and subtractions are applied to the SPCs and NSCs computed from the Pythia events. In the absence of collective flow, a complete nonflow subtraction should produce values consistent with zero for the SPCs and NSCs, and any remaining signal is a residual bias of the method.
Figure 2 shows the multiplicity-percentile dependence of , , NSC(4, 2), and NSC(3, 2) in OO collisions at TeV from Pythia 8 Angantyr, without and with - and -scaling-based nonflow subtractions. Even without collective flow, Pythia 8 Angantyr produces finite correlations for both SPCs and NSCs. The NSCs grow towards low multiplicity, whereas the SPCs show little multiplicity dependence. This is the nonflow reference the methods need to subtract. When the nonflow subtraction is applied, both scalings drive the SPCs close to zero within uncertainties. For the NSCs, scaling performs better, giving values consistent with zero, while the scaling oversubtracts and leaves a negative residual.
Figure 2. Multiplicity-percentile dependence of multiparticle correlators in OO collisions at TeV from Pythia 8 Angantyr showing the unsubtracted, , and -scaled correlations. (a) SPCs and . (b) NSC(4, 2) and NSC(3, 2). Dashed lines show constant fits where applicable. Ratio panels show the scaled result divided by the unsubtracted one.
| Correlator | residual [%] | residual [%] |
| Symmetry plane correlators (SPC) | ||
| Normalized symmetric cumulants (NSC) | ||
| NSC(4, 2) | ||
| NSC(3, 2) | ||
| NSC(4, 3) | ||
| NSC(5, 2) | ||
| NSC(5, 3) |
Table 2. Residual nonflow in Pythia 8 Angantyr OO collisions at TeV after and subtraction. Each entry is , where is a constant (pol0) fit to the multiplicity-percentile dependence for 0–60% (residual fraction as in [19]). Uncertainties are from the fits. A residual consistent with zero indicates successful nonflow removal in this flow-free sample.
To quantify the remaining nonflow bias, Table II reports the residual as the ratio of constant (pol0) fits to the subtracted and unsubtracted observables over the 0–60% multiplicity-percentile range. Both methods remove most of the nonflow, and for most entries the residual is at the 10% level or below. For the SPCs, the two methods give similar results, with providing slightly better overall nonflow control. One exception is the four-particle SPC , for which neither method works well: and leave residuals of and , respectively. For the NSCs, the difference between the methods is more pronounced. For NSC(4, 2), NSC(3, 2), and NSC(5, 3), the residual is consistent with zero, while oversubtracts by about 20%. For NSC(4, 3) and NSC(5, 2), performs better, whereas leaves a positive residual of around 20%. For completeness, the remaining SPC and NSC observables are collected in Appendix § 0.2.4.
It is important to note that, for the SCs and NSCs, nonflow can also be suppressed using the subevent method, where the particles of the correlator are taken from two or more separate pseudorapidity regions (subevents), with or without a pseudorapidity gap between them [18, 23]. We have checked this approach in the present setup and found that it reduces the statistical precision, which gets worse with an increase in the number of subevents and in the size of the pseudorapidity gap. Thus, the subevent methods are not included in the present study.
Results and discussion
The AMPT model provides a realistic testing ground for nonflow-control methods, as its final-state correlations reflect the interplay of collective flow and nonflow effects (see Appendix B for a brief model description). The collective-flow signals emerge primarily from the response of the partonic medium to initial-state geometry and fluctuations [35], while nonflow contributions can arise from minijets, fragmentation, and resonance decays, which can be smeared and modified by hadronization and hadronic rescattering [24, 36, 37, 38]. Although it is not trivial to uniquely disentangle the flow and nonflow components in the model, their coexistence allows the robustness and limitations of the proposed nonflow subtraction methods to be systematically checked.
Figure 3 shows the multiplicity-percentile dependence of , , NSC(4, 2), and NSC(3, 2) in OO collisions at TeV from the AMPT string-melting model [35], before and after and subtraction. The model produces finite, multiplicity-dependent correlations for both SPCs and NSCs. For multiplicity percentiles , the correlators rise abruptly towards more peripheral classes, indicating growing nonflow contamination at low multiplicity. Both and scalings remove this peripheral enhancement and restore a smoother multiplicity dependence in the low-multiplicity region. In more central (high-multiplicity) collisions the subtraction changes the values only mildly: the scaled points remain close to the unsubtracted AMPT results, consistent with a weaker relative nonflow contribution where the multiplicity is large.
Figure 3. Multiplicity-percentile dependence of multiparticle correlators in OO collisions at TeV from the AMPT model showing the unsubtracted, , and -scaled correlations. (a) SPCs and . (b) NSC(4, 2) and NSC(3, 2). Ratio panels show the scaled result divided by the unsubtracted one.
Motivated by the toy-model closure and the PYTHIA 8 null test, where provides the more reliable nonflow control—particularly for NSC(4, 2) and NSC(3, 2)—and by the AMPT results, where and perform similarly with slightly preferred, we recommend scaling as the best choice for studying , , NSC(4, 2), and NSC(3, 2) in OO collisions. Figure 4 compares the -scaled SPCs and NSCs in OO collisions at TeV from AMPT with Pb–Pb collisions at TeV from ALICE [6, 11] and also from AMPT [39]. To place both collision systems on an equal footing, the multiplicity-percentile bins are mapped to the corresponding mean charged-particle multiplicity at midrapidity, () [40]. This comparison is preferable because the relation between collision centrality and impact parameter—and thus the connection to the collision geometry—is weaker in light-ion collisions than in heavy-ion collisions. Plotting versus furthermore allows the system-size dependence of these correlators to be examined across OO and Pb–Pb in a common variable [17].
Figure 4. Nonflow-subtracted (-scaled) multiparticle correlators measured as a function of charged-particle multiplicity at midrapidity in OO collisions at TeV from the AMPT model. Results are compared with Pb–Pb collisions at TeV from ALICE [6, 11] and AMPT [39]. (a) SPCs and . (b) NSC(4, 2) and NSC(3, 2).
In Figure 4(a), both and in AMPT OO increase from low- to high-multiplicity events. The correlator, , rises to a broad maximum near and then decreases mildly towards the most central OO bin, whereas the three-harmonic correlator, , continues to increase up to the highest OO multiplicity (). In the overlap region around –250, the OO values reach a magnitude comparable to peripheral Pb–Pb. Extending into the Pb–Pb domain, both correlators attain their global maxima in mid-central collisions (roughly –800) and then decrease towards the most central, highest-multiplicity Pb–Pb events. AMPT underpredicts the Pb–Pb correlators relative to ALICE, while reproducing the qualitative multiplicity dependence. This suggests that OO data may follow a similar trend to AMPT OO, although the absolute magnitudes could differ.
For NSC(4, 2), the correlator in AMPT OO rises steeply with multiplicity, peaks near , and already declines towards the highest OO multiplicities. Across the OO–Pb–Pb overlap it connects smoothly in magnitude to peripheral Pb–Pb, after which the Pb–Pb values fall steadily towards zero in the most central collisions. AMPT describes the Pb–Pb data well at high multiplicity and underpredicts them somewhat in more peripheral (lower-multiplicity) Pb–Pb bins. For NSC(3, 2), the OO correlator remains close to zero from low to mid multiplicity and becomes only modestly negative () in the highest-multiplicity OO bin. In Pb–Pb, by contrast, the correlator starts more strongly negative () in peripheral collisions and approaches zero towards the most central events. AMPT again reproduces the qualitative Pb–Pb trend but underpredicts the magnitude of the anticorrelation. Therefore, the NSCs in OO data may also follow a similar trend to AMPT OO, although the magnitudes could differ.
The Pb–Pb trends are driven by how initial geometry and hydrodynamic response change with system size. ALICE finds that and related plane correlators are weakest in central collisions, where the overlap is nearly isotropic, and strongest in mid-central collisions, where the almond-shaped geometry and the nonlinear medium response that couples to (and to and ) are most pronounced, before easing again towards the periphery [11]. Likewise, NSC(4, 2) is positive because event-by-event and fluctuations are correlated through both the initial – geometry and the nonlinear response (hence NSC(4, 2) is also sensitive to ), while NSC(3, 2) is negative because and inherit the anticorrelation of and , making NSC(3, 2) largely an initial-state probe [5, 6] [4]. Mapped versus , the AMPT OO results sit on the rising, smaller-system side of this picture: as multiplicity grows within OO, the overlap develops a clearer anisotropic geometry and nonlinear couplings strengthen, so the SPCs and NSC(4, 2) increase towards the peripheral Pb–Pb values near , whereas NSC(3, 2) remains near zero until the system is large enough for a Pb–Pb-like – anticorrelation to emerge. Continuing into Pb–Pb, the correlators then follow the familiar mid-central maximum (SPCs) or the fall of NSC(4, 2) towards zero and the approach of NSC(3, 2) towards zero in the most central, nearly isotropic collisions.
Summary and outlook
The multiplicity-reweighted () and -based-estimator () nonflow subtraction methods are studied for SPCs and NSCs in OO collisions at TeV, and the preferred prescription is used to predict these correlators in the AMPT model and to compare them with results from Pb–Pb collisions. The important findings are summarized below.
In a toy model with known flow and nonflow components, both scalings remove the nonflow rise towards low multiplicity; tracks the flow truth more closely, while tends to oversubtract the collective-flow-like correlations in peripheral bins.
In the PYTHIA 8 Angantyr model, used as a nonflow-only reference, both methods drive most SPCs close to zero (residuals typically ). For NSC(4, 2) and NSC(3, 2), leaves residuals consistent with zero, whereas oversubtracts by about 20%.
In the AMPT model, and both remove the peripheral nonflow enhancement and leave the high-multiplicity points nearly unchanged in OO collisions; following the closure tests, the scaling is recommended for the main SPCs and NSCs studied in this work.
When studied as a function of , the -scaled , , and NSC(4, 2) in OO collisions from AMPT rise with multiplicity and approach peripheral Pb–Pb collisions near . On the other hand, NSC(3, 2) in OO collisions stays near zero and only attains a modest negative value towards high multiplicity. Results from Pb–Pb collisions then show the familiar mid-central SPC maximum and the decline (approaching zero) of NSC(4, 2) [NSC(3, 2)] towards central collisions, consistent with evolving initial geometry and nonlinear hydrodynamic response.
These predictions from the AMPT model, after subtraction and versus , provide a concrete baseline for forthcoming SPC and NSC measurements in pO, OO, and Ne–Ne collisions at the LHC. Comparing data to this baseline and extending the same nonflow framework to additional correlators and nuclear geometries will provide necessary constraints on models incorporating initial-state effects, transport properties, and medium evolution, and on the onset of nonlinear mode coupling from small to large systems.
Acknowledgments
The authors would like to thank Dr. Dong Jo Kim, University of Jyväskylä, Finland, for fruitful discussions. NM is supported by the Academy of Finland through the Center of Excellence in Quark Matter (Grant No. 346328). SP gratefully acknowledges support from the Hungarian National Research, Development and Innovation Office (NKFIH) under Contract Nos. NKFIH NKKP ADVANCED-25 K153456, 2025-1.1.5-NEMZ_KI-2025-00005, 2024-1.2.5-TET-2024-00022, FuSe COST action CA241011. The authors also acknowledge the Wigner Scientific Computing Laboratory (WSCLAB) in Budapest, Hungary for allocating computational resources for this study.
Parameters of the toy model
The analysis uses the same nine multiplicity percentile bins as the main analysis, and the corresponding multiplicity bin edges are,
ordered from low to high multiplicity. The interval from 0 to 25 is the most peripheral class, 70–100%, and the interval from 238 to 3000 (the last edge is an upper bound, not a typical multiplicity) is the most central class, 0–5%. The event multiplicity () is drawn randomly from a Gaussian distribution with mean and standard deviation for each event class. From central to peripheral bins, the nine means are
Draws that fall outside the class, or below , are rejected. The requirement leaves room for one group of six tracks (for the nonflow cluster) and twelve tracks for the bulk formation which carry the flow modulation.
The number of groups () is found with
and is reduced if the bulk would otherwise contain fewer than 12 tracks. For example, if the event has 100 particles, , so the reserved clusters are made with tracks, and 82 tracks are left for the bulk.
A particle emitted from the common plane with plane angle is drawn from
where is chosen uniformly in , and is the azimuthal angle of the particle. The flow coefficients are kept the same in every class,
For the nonflow clusters of six particles each, the six relative azimuths are
in radians. Track is placed at , with a new angle , drawn uniformly, for every cluster. A track that belongs neither to the plane nor to a cluster is given an azimuth drawn uniformly on . In the LM class, this uniform azimuth is used for the bulk of every sample (no flow is assumed in the LM class), while the mixed and nonflow samples still fill their groups with clusters. For every track, whatever its azimuth, is uniform on , is uniform on GeV/, and the charge is random. None of these depends on , so the only azimuthal correlations are the common plane and the clusters.
AMPT model
A Multi-Phase Transport (AMPT) model is a Monte Carlo transport framework developed to describe the evolution of relativistic nuclear collisions through the initial-state, partonic, hadronization, and hadronic stages [35]. The initial conditions are generated using the Heavy Ion Jet Interaction Generator (HIJING) [41], while partonic scatterings are described by Zhang’s Parton Cascade (ZPC) model [42]. Hadronization proceeds through Lund string fragmentation in the default version of AMPT, whereas, in the string-melting version, the excited strings are converted into partons that subsequently form hadrons through quark coalescence [35, 43, 44]. The produced hadrons can undergo further interactions described by the A Relativistic Transport (ART) model [45]. Here, the AMPT settings used for OO collisions are similar to those in Ref. [46], while for Pb–Pb collisions, the same settings as in Ref. [39] are used. This framework enables investigations of how initial-state fluctuations, partonic interactions, hadronization, and hadronic rescattering influence final-state particle production and azimuthal correlations. AMPT has consequently been used to study anisotropic flow and multiparticle correlations, providing a means of investigating the interplay between initial-state geometry and final-state dynamics [24, 35, 47]. Studies based on AMPT in OO collisions have investigated anisotropic flow, normalized symmetric cumulants, two-particle azimuthal correlations, and global observables for different oxygen nuclear configurations, including Woods–Saxon, harmonic-oscillator, and some exotic -clustered geometries [15, 16, 46, 48]. These studies demonstrate the relevance of the AMPT model for predicting collective-flow-related observables in light-ion collisions and motivate further investigations of correlations between different flow harmonics such as the SPCs and NSCs [5, 47] [4].
Scaling factors
The nonflow subtractions of Eqs. (2) and (3) are fixed by the scaling factors
where is the effective multiplicity of Eq. (4) and is the nonflow estimator of the same particle order as the target correlator (Table I). Both factors are defined to equal unity in the LM reference class (70–100%) and to fall below one towards higher multiplicity, where the relative nonflow weight is smaller. Figure 5 shows these scaling factors as a function of multiplicity percentiles in OO collisions at TeV from Pythia 8 (top) and the AMPT model (bottom). For , the three-particle estimators and are used for the three-particle SPC numerators, and the four-particle estimator is used for . For
Figure 5. Scaling factors used for the numerators in various SPCs versus multiplicity percentiles in OO collisions at TeV from Pythia 8 (top) and the AMPT model (bottom). The gray band marks the LM class.
, and correspond to three- and four-particle correlators, respectively. The gray band marks the LM class. The factors track the multiplicity trend of the -based estimators and can differ from the pure multiplicity powers ; that difference is what allows to absorb residual multiplicity-selection effects beyond the independent-source baseline.
Additional Pythia SPCs and NSCs
Figure 6 shows the remaining SPCs and NSCs in Pythia 8 Angantyr OO collisions after and subtraction. As for the main-text observables, a successful subtraction should drive the correlators to zero in this flow-free sample. Both scalings bring and close to zero, whereas retains a sizeable residual with either method. For the NSCs, performs well for NSC(5, 3), while is preferred for NSC(4, 3) and NSC(5, 2); the corresponding residual fractions are listed in Table II.
Figure 6. Multiplicity percentile dependence of SPCs and NSCs in OO collisions at TeV from Pythia 8 Angantyr model: no subtraction, , and . (a) Additional SPCs , , and . (b) Additional NSCs: NSC(4, 3), NSC(5, 2), and NSC(5, 3). Dashed lines show constant fits where applicable; ratio panels are scaled/no-subtraction.
References
- [5]S. Acharya et al. (ALICE), Systematic studies of correlations between different order flow harmonics in Pb-Pb collisions at √sNN = 2.76 TeV, Phys. Rev. C 97, 024906 (2018), arXiv:1709.01127 [nucl-ex].
- [6]S. Acharya et al. (ALICE), Measurements of mixed harmonic cumulants in Pb–Pb collisions at √sNN = 5.02 TeV, Phys. Lett. B 818, 136354 (2021), arXiv:2102.12180 [nucl-ex].
- [7]A. Bilandzic, C. H. Christensen, K. Gulbrand-sen, A. Hansen, and Y. Zhou, Generic framework for anisotropic flow analyses with multiparticle azimuthal correlations, Phys. Rev. C 89, 064904 (2014), arXiv:1312.3572 [nucl-ex].DOI
- [8]G. Aad et al. (ATLAS), Measurement of event-plane correlations in √sNN = 2.76 TeV lead-lead collisions with the ATLAS detector, Phys. Rev. C 90, 024905 (2014), arXiv:1403.0489 [hep-ex].
- [9]A. Bilandzic, M. Lesch, and S. F. Taghavi, New estimator for symmetry plane correlations in anisotropic flow analyses, Phys. Rev. C 102, 024910 (2020), arXiv:2004.01066 [nucl-ex].DOI
- [10]S. Acharya et al. (ALICE), Symmetry plane correlations in Pb-Pb collisions at √sNN = 2.76 TeV, Eur. Phys. J. C 83, 576 (2023), arXiv:2302.01234 [nucl-ex].
- [11]S. Acharya et al. (ALICE), Higher-order symmetry plane correlations in Pb-Pb collisions at sNN=5.02TeV, Phys. Rev. C 111, 064913 (2025), arXiv:2409.04238 [nucl-ex].
- [12]I. J. Abualrob et al. (ALICE), Evidence of Nuclear Geometry-Driven Anisotropic Flow in O+O and Ne+Ne Collisions at sNN=5.36 TeV, Phys. Rev. Lett. 137, 082301 (2026), arXiv:2509.06428 [nucl-ex].DOI
- [13]G. Aad et al. (ATLAS), Measurement of the azimuthal anisotropy of charged particles in √sNN = 5.36 TeV 16O+16O and 20Ne+20Ne collisions with the ATLAS detector, Phys. Rev. C 113, 045205 (2026), arXiv:2509.05171 [nucl-ex].
- [14]A. Hayrapetyan et al. (CMS), Observation of Long-Range Collective Flow in O+O and Ne+Ne Collisions and Implications for Nuclear Structure Studies, Phys. Rev. Lett. 137, 082302 (2026), arXiv:2510.02580 [nucl-ex].DOI
- [15]D. Behera, S. Prasad, N. Mallick, and R. Sahoo, Effects of clustered nuclear geometry on the anisotropic flow in O-O collisions at the LHC within a multiphase transport model framework, Phys. Rev. D 108, 054022 (2023), arXiv:2304.10879 [hep-ph].DOI
- [16]S. Prasad and R. Sahoo, Nuclear geometry driven symmetry plane correlations in OO and Ne–Ne collisions at the Large Hadron Collider (2026), arXiv:2605.00866 [hep-ph].arxiv.org/abs/2605.00866
- [17]S. Acharya et al. (ALICE), Investigations of Anisotropic Flow Using Multiparticle Azimuthal Correlations in pp, p-Pb, Xe-Xe, and Pb-Pb Collisions at the LHC, Phys. Rev. Lett. 123, 142301 (2019), arXiv:1903.01790 [nucl-ex].
- [18]J. Jia, M. Zhou, and A. Trzupek, Revealing long-range multiparticle collectivity in small collision systems via subevent cumulants, Phys. Rev. C 96, 034906 (2017), arXiv:1701.03830 [nucl-th].DOI
- [19]Z. Wang, J. Jia, J. Chen, S. Huang, C. Zhang, and Z. Yan, Nonflow Subtraction Beyond Two-Particle Correlations, Rept. Prog. Phys. 89, 108601 (2026), arXiv:2606.10258 [nucl-th].DOI
- [20]N. Borghini, P. M. Dinh, and J.-Y. Ollitrault, Flow analysis from multiparticle azimuthal correlations, Phys. Rev. C 64, 054901 (2001), arXiv:nucl-th/0105040.DOI
- [21]A. Bilandzic, R. Snellings, and S. Voloshin, Flow analysis with cumulants: Direct calculations, Phys. Rev. C 83, 044913 (2011), arXiv:1010.0233 [nucl-ex].DOI
- [22]M. Aaboud et al. (ATLAS), Measurement of long-range multiparticle azimuthal correlations with the subevent cumulant method in pp and p + Pb collisions with the ATLAS detector at the CERN Large Hadron Collider, Phys. Rev. C 97, 024904 (2018), arXiv:1708.03559 [hep-ex].
- [23]P. Huo, K. Gajdošová, J. Jia, and Y. Zhou, Importance of non-flow in mixed-harmonic multi-particle correlations in small collision systems, Phys. Lett. B 777, 201 (2018), arXiv:1710.07567 [nucl-ex].DOI
- [24]M.-W. Nie, P. Huo, J. Jia, and G.-L. Ma, Multiparticle azimuthal cumulants in p+Pb collisions from a multi-phase transport model, Phys. Rev. C 98, 034903 (2018), arXiv:1802.00374 [hep-ph].DOI
- [25]ALICE Collaboration, The ALICE definition of primary particles, Tech. Rep. ALICE-PUBLIC-2017-005 (CERN, 2017).
- [26]S. Acharya et al. (ALICE), ALICE upgrades during the LHC Long Shutdown 2, JINST 19 (05), P05062, arXiv:2302.01238 [physics.ins-det].
- [27]T. Sjöstrand, S. Ask, J. R. Christiansen, R. Corke, N. Desai, P. Ilten, S. Mrenna, S. Prestel, C. O. Rasmussen, and P. Z. Skands, An introduction to PYTHIA 8.2, Comput. Phys. Commun. 191, 159 (2015), arXiv:1410.3012 [hep-ph].
- [28]C. Bierlich et al., A comprehensive guide to the physics and usage of PYTHIA 8.3, SciPost Phys. Codeb. 2022, 8 (2022), arXiv:2203.11601 [hep-ph].arxiv.org/abs/2203.11601
- [29]C. Bierlich, G. Gustafson, L. Lönnblad, and H. Shah, The Angantyr model for Heavy-Ion Collisions in PYTHIA8, JHEP 10, 134, arXiv:1806.10820 [hep-ph].DOI
- [30]D. Kikola, L. Yi, S. Esumi, F. Wang, and W. Xie, Non-flow ”factorization” and a novel method to disentangle anisotropic flow and nonflow, Phys. Rev. C 86, 014901 (2012), arXiv:1110.4809 [nucl-ex].DOI
- [31]S. H. Lim, Q. Hu, R. Belmont, K. K. Hill, J. L. Nagle, and D. V. Perepelitsa, Examination of flow and nonflow factorization methods in small collision systems, Phys. Rev. C 100, 024908 (2019), arXiv:1902.11290 [nucl-th].DOI
- [32]J. L. Nagle, R. Belmont, S. H. Lim, and B. Seidlitz, Checking nonflow assumptions and results via PHENIX published correlations in p+p, p+Au, d+Au, and 3He+Au at √sNN=200 GeV, Phys. Rev. C 105, 024906 (2022), arXiv:2107.07287 [nucl-th].DOI
- [33]S. Ji, M. Virta, T. Kallio, S. H. Lim, and D. J. Kim, Toward unbiased flow measurements in pp collisions at the CERN Large Hadron Collider, Phys. Rev. C 108, 034909 (2023), arXiv:2303.05806 [hep-ph].
- [34]C. Zhang, A. Behera, S. Bhatta, and J. Jia, Non-flow effects in correlation between harmonic flow and transverse momentum in nuclear collisions, Phys. Lett. B 822, 136702 (2021), arXiv:2102.05200 [nucl-th].DOI
- [35]Z.-W. Lin, C. M. Ko, B.-A. Li, B. Zhang, and S. Pal, A Multi-phase transport model for relativistic heavy ion collisions, Phys. Rev. C 72, 064901 (2005), arXiv:nucl-th/0411110.DOI
- [36]G.-L. Ma and X.-N. Wang, Jets, Mach cone, hot spots, ridges, harmonic flow, dihadron and γ-hadron correlation in high-energy heavy-ion collisions, Phys. Rev. Lett. 106, 162301 (2011), arXiv:1011.5249 [nucl-th].
- [37]L. Xu, L. Yi, D. Kikola, J. Konzer, F. Wang, and W. Xie, Model-independent decomposition of flow and nonflow in relativistic heavy-ion collisions, Phys. Rev. C 86, 024910 (2012), arXiv:1204.2815 [nucl-ex].DOI
- [38]J. Jia and S. Mohapatra, Disentangling flow and non-flow correlations via Bayesian unfolding of the event-by-event distributions of harmonic coefficients in ultrarelativistic heavy-ion collisions, Phys. Rev. C 88, 014907 (2013), arXiv:1304.1471 [nucl-ex].DOI
- [39]S. Tripathy, S. Prasad, and R. Sahoo, Event-shape dependence of symmetry plane correlations using the Gaussian estimator in Pb-Pb collisions at the LHC using a multi-phase transport model, Phys. Rev. D 112, 114012 (2025), arXiv:2504.09275 [nucl-ex].DOI
- [40]J. Adam et al. (ALICE), Centrality Dependence of the Charged-Particle Multiplicity Density at Midrapidity in Pb-Pb Collisions at √sNN = 5.02 TeV, Phys. Rev. Lett. 116, 222302 (2016), arXiv:1512.06104 [nucl-ex].DOI
- [41]X.-N. Wang and M. Gyulassy, HIJING: A Monte Carlo model for multiple jet production in pp, pA and AA collisions, Phys. Rev. D 44, 3501 (1991).pubmed.ncbi.nlm.nih.gov/10013812
- [42]B. Zhang, ZPC 1 0.1: A Parton cascade for ultrarelativistic heavy ion collisions, Comput. Phys. Commun. 109, 193 (1998), arXiv:nucl-th/9709009.
- [43]B. Andersson, G. Gustafson, G. Ingelman, and T. Sjöstrand, Parton Fragmentation and String Dynamics, Phys. Rept. 97, 31 (1983).
- [44]Z.-w. Lin and C. M. Ko, Partonic effects on the elliptic flow at RHIC, Phys. Rev. C 65, 034904 (2002), arXiv:nucl-th/0108039.
- [45]B.-A. Li and C. M. Ko, Formation of superdense hadronic matter in high-energy heavy ion collisions, Phys. Rev. C 52, 2037 (1995), arXiv:nucl-th/9505016.DOI
- [46]D. Behera, N. Mallick, S. Tripathy, S. Prasad, A. N. Mishra, and R. Sahoo, Predictions on global properties in O+O collisions at the Large Hadron Collider using a multi-phase transport model, Eur. Phys. J. A 58, 175 (2022), arXiv:2110.04016 [hep-ph].DOI
- [47]M. Nasim, Systematic study of symmetric cumulants at √sNN = 200 GeV in Au+Au collision using transport approach, Phys. Rev. C 95, 034905 (2017), arXiv:1612.01066 [nucl-ex].DOI
- [48]X.-L. Zhao, Z.-W. Lin, Y. Zhou, C. Zhang, and G.-L. Ma, Nuclear cluster structure effect in 16O+16O collisions at the top RHIC energy, Phys. Lett. B 874, 140254 (2026), arXiv:2404.09780 [nucl-th].DOI





