betaFeedback

WorksPhysics and Astronomy

Color-Reconnection Effects on the nMPI-Nch Conditional Multiplicity Fluctuations in PYTHIA 8 pp Collisions

Abstract

Color reconnection (CR) changes both the mean and the dispersion of charged-particle multiplicity in pp collisions. We study the conditional distribution P(Nch | nMPI = m) at fixed PYTHIA interaction count for sqrt(s) = 2.76, 7, and 13 TeV, using charged particles in |eta| < 0.8 and 0.15 < pT < 20 GeV/c. CR is switched on or off while the remaining tune parameters are held fixed. Enabling CR reduces the conditional mean by a larger fraction than the standard deviation. The absolute width therefore decreases, while the coefficient of variation increases in every retained matched MPI bin. This ordering also holds in independent 13 TeV samples with Monash, A14, and CMS CP5. Comparisons with ALICE mean-transverse-momentum data and simulated activity estimators relate these conditional fluctuations to experimentally accessible final-state observables.

main.pdf

Introduction

Multiparton interactions (MPI) allow several parton–parton scatterings in a single pp collision. They help describe underlying-event activity and broad multiplicity distributions beyond a single hard scattering [1, 2]. The final charged multiplicity is also shaped by showers, beam remnants, color connections, hadronization, and decays. Its relation to the MPI count therefore depends on the full event evolution.

Color reconnection rearranges the color connections between partons before string fragmentation. In the MPI-based model used here, softer interaction systems can reconnect to harder ones, reducing the string-length measure [2]. This modifies particle production without changing the count of already generated interactions. CR is known to affect multiplicity, mean transverse momentum, and identified-particle observables [3, 4, 5].

Throughout this work, nMPIn_{\mathrm{MPI}} denotes PYTHIA’s internal interaction count. This model-dependent quantity provides a reference for studying how showers, hadronization, and decays translate partonic activity into charged-particle production. Its interpretation depends on the generator version, tune, process selection, and MPI settings; it is not directly measurable. Supplementary Material S1 specifies the implementation.

The relation between MPI activity and final-state multiplicity, including its dependence on CR, has been studied in several contexts. Weber et al. [6] examine multiplicity production and fluctuations at a given MPI count in quarkonium studies. Prasad et al. [7] compare activity classifiers with and without CR and with rope hadronization, while Prasad et al. [8] review their MPI sensitivity and selection biases. These studies motivate a joint examination of the conditional mean, absolute width, and width relative to the mean.

Here we quantify how CR changes the mean and dispersion of charged multiplicity at fixed MPI count. We evaluate the conditional mean μm\mu_m, absolute standard deviation σm\sigma_m, and coefficient of variation Cm=σm/μmC_m = \sigma_m/\mu_m together, comparing CR ON and CR OFF at the same integer m=nMPIm = n_{\mathrm{MPI}}. This comparison distinguishes a change in absolute width from a change in width relative to the local mean. We examine whether the resulting ordering persists at three collision energies and in independent tune configurations at 13 TeV, and assess its sensitivity to the stated selections.

We first use inclusive distributions to motivate conditioning on MPI count, then compare the conditional means and widths. The ALICE comparison and final-state estimators provide supporting context, followed by checks of tune and selection dependence.

Simulation, definitions, and scope

The baseline consists of five million events for each energy and CR state, with the nominal setup PYTHIA 8.316, SoftQCD:all, and Monash 2013 [9]. The CR variation uses ColourReconnection:reconnect = off, with the remaining tune parameters held fixed in the intended setup. Historical production-time seed values and effective lifetime/decay settings were not retained for the baseline samples, which limits exact generator replay. S1 and S2 describe the available production records and reproducible analysis of the retained events.

MPI and CR settings

The independent 13 TeV tune comparison uses Monash (Tune:pp = 14), central A14 NNPDF2.3 LO (Tune:pp = 21) [10], and the published CMS CP5 parameters [11, 12]. Table 1 gives the key settings. Each tune and CR state has five million generated events.

The MPI regularization scale controls the low-pTp_{\mathrm{T}} interaction rate, and its energy dependence controls how MPI activity changes with collision energy. The regularized form and energy-scaling equation are given in Supplementary Material S1, following Bierlich et al. [2].

The matter profile controls the impact-parameter dependence of MPI activity. Monash and A14 use profile 3, specified by an overlap exponent; CP5 uses a double-Gaussian profile, specified by a core radius and fraction. All three use MPI-based CR mode 0, with a range parameter controlling the reconnection probability of softer systems. The profile forms and full settings are given in Supplementary Material S1.

CP5 uses the NNPDF3.1 NNLO central PDF through LHAPDF [13], with two-loop running and αs(MZ)=0.118\alpha_{\mathrm{s}}(M_{\mathrm{Z}}) = 0.118 in all four coupling sectors. The full-precision CMS tune settings are implemented in PYTHIA 8.316, rather than the PYTHIA 8.226 environment of the original CMS production. Both CR states use the default PYTHIA decay prescription and independent recorded seeds. S1 summarizes the settings, and S2 supplies the available production records.

Particle and event definitions

The analysis multiplicity is

Nch=Nch(∣η∣<0.8, 0.15<pT<20 GeV/c).(1)N_{\mathrm{ch}} = N_{\mathrm{ch}}(\lvert\eta\rvert< 0.8,\ 0.15 < p_{\mathrm{T}} < 20\ \mathrm{GeV}/c). \tag*{(1)}

An INEL>> 0-like event contains at least one charged final-state particle in ∣η∣<1\lvert\eta\rvert< 1. Inclusive distributions use this event selection without an nMPIn_{\mathrm{MPI}} cut. The conditional response study additionally requires nMPI>2n_{\mathrm{MPI}} > 2 and compares common integer bins containing at least 500 events in each CR state.

Charged final-state particles are selected from the generator records; implementation details are given in Supplementary Material S1. No detector simulation is applied. The ALICE comparison requires Nch>0N_{\mathrm{ch}} > 0 and no MPI cut, with the primary-particle limitation discussed in Sec. 5. The estimator selections are specified in Secs. 6 and 7.

AI-assisted code development and validation

OpenAI Codex assisted code development, diagnostic calculations, plotting, language editing, and LaTeX preparation. The retained project-related session metadata identify gpt-5.5, gpt-5.6-sol, gpt-5.6-luna, gpt-6-astra, and gpt-6.1-sol. The October 2026 revision session used the recorded identifier gpt-6.1-sol through the Codex VS Code integration, client version 0.162.0-alpha.17.2. These records identify the models and client used in the documented sessions, but do not resolve immutable model snapshots or assign every historical code edit to a model.

Programming assistance covered event parsing, particle and event selections, conditional profiles, moment and uncertainty calculations, tune comparisons, and plotting scripts. Diagnostic work addressed variance conventions, covariance terms in propagated errors, empirical-bootstrap checks, and local and inclusive width normalization. All plotted physics values were calculated from event records or numerical tables using executable scripts.

Validation combined direct data reanalysis with complementary numerical checks. The central analysis processed all 60 million source event records, and a full scan of the six baseline particle samples reproduced the stored charged multiplicities, INEL>> 0-like masks, and summed transverse momenta. Conditional moments and errors were recomputed from integer joint-count tables. Absolute-width and coefficient-of-variation errors were checked with 2,000 empirical-bootstrap replicates in all 148 matched bins. For the inclusive-normalized width, the covariance with the inclusive mean was retained and the influence-function expression was checked by numerical perturbation. Source syntax and compilation checks, small validation-only generator runs, and reproduction of the numerical tables and plots from a fresh S2 extraction provided further checks. S1 and S2 give the detailed validation records and task-to-evidence map.

The validation code was also developed with AI assistance, so these checks establish numerical consistency with retained inputs rather than independent human validation. They do not recover missing historical production settings. Scientific interpretation and responsibility for the manuscript remain with the authors.

Inclusive distributions and motivation for conditioning

The inclusive multiplicity and event-mean transverse-momentum distributions show the overall CR dependence in the six Monash baseline samples (Figure 1 and Figure 2). Events satisfy the INEL>> 0-like selection without an MPI cut. The multiplicity distribution includes Nch=0N_{\mathrm{ch}} = 0: events can pass the wider event selection while containing no charged particle within the analysis cuts.

Figure 1. Inclusive charged-multiplicity distributions for the Monash baseline at s=2.76\sqrt{s} = 2.76, 7, and 13 TeV (left to right). Green solid and red dashed step histograms denote CR ON and CR OFF, respectively. Events satisfy the INEL>0> 0-like selection without an nMPIn_{\mathrm{MPI}} cut; charged particles satisfy ∣η∣<0.8|\eta| < 0.8 and 0.15<pT<200.15 < p_T < 20 GeV/c. Bin width is 5, and the ordinate is probability per unit multiplicity, normalized to unit integral. Empty bins are omitted on the logarithmic axis; statistical error bars are not displayed.

Figure 2. Distribution of event-mean transverse momentum for the same energies, particle acceptance, and CR color convention as Figure 1. The INEL>0> 0-like selection is supplemented by Nch>0N_{\mathrm{ch}} > 0, without an nMPIn_{\mathrm{MPI}} cut. Each event has equal weight. Bin width is 0.5 GeV/c, and each distribution is normalized to unit integral. Empty bins are omitted on the logarithmic axis; statistical error bars are not displayed.

For events with Nch>0N_{\mathrm{ch}} > 0, the event-mean transverse momentum is the arithmetic mean of the accepted particle transverse momenta. Each event receives equal weight. This event-level distribution complements the multiplicity-dependent ⟨pT⟩(Nch)\langle p_{\mathrm{T}}\rangle(N_{\mathrm{ch}}) profile used in the ALICE comparison. The histograms in each figure have unit integral; their bin widths are specified in the captions. The explicit event-mean definition is given in Supplementary Material S1.

CR lowers the mean charged multiplicity and raises the average event-mean transverse momentum, with larger shifts at higher energy (Table 2). It also reduces the high-multiplicity population and increases the fraction of events with large event-mean transverse momentum. The averages give equal weight to events in their respective selected samples.

These inclusive shifts mix contributions from events with different MPI counts. The multiplicity distribution can be written as

P(Nch)=∑mP(Nch∣nMPI=m)P(nMPI=m),(2)P(N_{\mathrm{ch}}) = \sum_m P(N_{\mathrm{ch}} \mid n_{\mathrm{MPI}} = m)P(n_{\mathrm{MPI}} = m), \tag*{(2)}
SettingMonashA14CMS CP5
Tune:pp (starting point)142114 + overrides
PDF familyNNPDF2.3 LO QCD+QEDNNPDF2.3 LO QCD+QEDNNPDF3.1 NNLO
MPI pT0refp_{T0}^{\mathrm{ref}} [GeV]2.282.091.41
MPI reference energy [GeV]700070007000
MPI energy exponent0.2150.2150.03344
MultipartonInteractions:bProfile332
Overlap exponent (profile 3)1.851.85–
Core radius / fraction (profile 2)––0.7634 / 0.63
ColourReconnection:mode000
ColourReconnection:range1.801.715.176
αs(MZ)\alpha_s(M_Z): hard process / running order0.130 / 10.140 / 10.118 / 2
αs(MZ)\alpha_s(M_Z): ISR / running order0.1365 / 10.127 / 10.118 / 2
αs(MZ)\alpha_s(M_Z): FSR / running order0.1365 / 10.127 / 10.118 / 2
αs(MZ)\alpha_s(M_Z): MPI / running order0.130 / 10.126 / 10.118 / 2

Table 1. Key settings for the independent 13 TeV tune comparison in PYTHIA 8.316. All use SoftQCD:all = on; only the CR physics switch changes within each tune pair. ISR and FSR denote initial- and final-state radiation; running orders 1 and 2 mean one- and two-loop running. Settings and provenance are detailed in Supplementary Materials S1 and S2.

Here the probabilities refer to the INEL>> 0-like sample without an nMPIn_{\mathrm{MPI}} cut, with Nch=0N_{\mathrm{ch}} = 0 retained. The inclusive distribution combines conditional distributions at different MPI counts, weighted by their event fractions. It therefore does not by itself reveal how CR changes multiplicity dispersion at a fixed MPI count.

To separate these contributions, we compare CR ON and CR OFF at the same integer MPI count. We require nMPI>2n_{\mathrm{MPI}} > 2 and retain common bins with at least 500 events in each sample. Within each matched bin, we measure the mean and the absolute and relative widths.

Conditional response and fluctuation measures

At each selected integer MPI count, P(Nch∣nMPI=m)P(N_{\mathrm{ch}} \mid n_{\mathrm{MPI}} = m) describes the within-bin multiplicity distribution. Its moments compare the two CR states without mixing different MPI populations [14, 15].

For the NmN_m selected events in bin mm, the conditional mean and absolute standard deviation are

μm=⟨Nch⟩m=1Nm∑i:mi=mNch,i,(3)\mu_m = \langle N_{\mathrm{ch}}\rangle_m = \frac{1}{N_m}\sum_{i:m_i=m}N_{\mathrm{ch},i}, \tag*{(3)}
σm=[1Nm∑i:mi=m(Nch,i−μm)2]1/2.(4)\sigma_m = \left[\frac{1}{N_m}\sum_{i:m_i=m}(N_{\mathrm{ch},i}-\mu_m)^2\right]^{1/2}. \tag*{(4)}

Using the divisor NmN_m gives equal weight to each event and defines the population moments of the selected sample. The standard deviation σm\sigma_m is the RMS multiplicity residual about μm\mu_m, expressed on the NchN_{\mathrm{ch}} scale.

To interpret an absolute width in relation to its conditional mean, we use the following normalized measures:

Cm=σmμm,Fm=σm2μm.(5)C_m = \frac{\sigma_m}{\mu_m}, \qquad F_m = \frac{\sigma_m^2}{\mu_m}. \tag*{(5)}

The coefficient of variation CmC_m expresses the standard deviation as a fraction of the local mean. The scaled variance FmF_m compares the variance with the mean and equals one for an ideal Poisson distribution; it is commonly used in multiplicity-fluctuation studies [16]. The relation Cm2=Fm/μmC_m^2 = F_m/\mu_m allows the two measures to have different CR orderings.

To compare the changes in mean and width, define Rμ,m=μm,ON/μm,OFFR_{\mu,m} = \mu_{m,\mathrm{ON}}/\mu_{m,\mathrm{OFF}}, Rσ,m=σm,ON/σm,OFFR_{\sigma,m} = \sigma_{m,\mathrm{ON}}/\sigma_{m,\mathrm{OFF}}, and RC,m=Cm,ON/Cm,OFFR_{C,m} = C_{m,\mathrm{ON}}/C_{m,\mathrm{OFF}}. Their algebraic relation is

s\sqrt{s} [TeV]BinsR‾μ\overline{R}_{\mu}R‾σ\overline{R}_{\sigma}R‾C\overline{R}_{C}
2.76160.7660.8951.174
7200.7410.8911.210
13230.7280.8901.232

Table 3. Common-bin-weighted ON/OFF ratios for the baseline, with wm=min⁡(Nm,ON,Nm,OFF)w_m = \min(N_{m,\mathrm{ON}}, N_{m,\mathrm{OFF}}). Ratios are evaluated bin by bin before averaging; the averages therefore need not obey the binwise identity in Eq. 6.

RC,m=Rσ,mRμ,m.(6)R_{C,m} = \frac{R_{\sigma,m}}{R_{\mu,m}}. \tag*{(6)}

When Rμ,m<Rσ,m<1R_{\mu,m} < R_{\sigma,m} < 1, CR ON has a smaller absolute width but a larger relative width: the mean decreases by a larger fraction than the standard deviation. Equation 6 expresses this balance between changes in scale and dispersion.

CR OFF shifts the conditional means to larger multiplicities and produces a steeper response. The density maps in Figure 3 retain the event spread, while Figure 4 shows the means with absolute-width bands. Supplementary Material S1 gives the correlation coefficients, fitted slopes, and sample-level width summaries. The main comparison below uses the local widths at matched MPI counts.

Figure 3. Normalized event density in the PYTHIA-specific nMPIn_{\mathrm{MPI}}–NchN_{\mathrm{ch}} plane. The color bar gives the fraction of plotted events in each bin, obtained by dividing the bin counts by their sum within each panel. The map characterizes the relation between internal MPI activity and final charged multiplicity in each simulated sample.

Figure 4. Conditional mean μm=⟨Nch⟩m\mu_m = \langle N_{\mathrm{ch}}\rangle_m at each retained integer m=nMPIm = n_{\mathrm{MPI}}. Shaded bands show μm±σm\mu_m \pm\sigma_m, where σm\sigma_m is the absolute conditional standard deviation. Only bins containing at least 500 events are shown. The absolute width is interpreted together with the normalized measures in Figure 5.

The absolute width σm\sigma_m increases with MPI count but remains smaller with CR ON (Figure 5(a)–(Figure 5c)). At a common MPI count, the widths increase with collision energy. The CR separation generally grows with activity.

Figure 5. Absolute width σm\sigma_m in panels (a)–(c), coefficient of variation CmC_m in panels (d)–(f), and scaled variance FmF_m in panels (g)–(i), all versus nMPIn_{\mathrm{MPI}}. Columns correspond to 2.76, 7, and 13 TeV from left to right. Green solid circles denote CR ON; red dashed squares denote CR OFF. Points are binwise estimates and lines only guide the eye. Vertical bars show one-standard-error statistical uncertainties propagated from the second, third, and fourth central moments; most are smaller than the markers. Only bins containing at least 500 events are shown. CR ON has a smaller absolute width but larger coefficient of variation in all 16, 20, and 23 matched bins at 2.76, 7, and 13 TeV, respectively.

The coefficient of variation CmC_m decreases with MPI count because the mean grows faster than the absolute width (Figure 5(d)–(Figure 5f)). CR ON has the larger CmC_m in every retained matched bin. Its energy dependence is weaker than that of σm\sigma_m.

The scaled variance FmF_m exceeds the Poisson reference in all displayed baseline bins (Figure 5(g)–(Figure 5i)). CR ON gives larger values over most of the retained range, although endpoint uncertainties limit individual-bin interpretation. Since Cm2=Fm/μmC_m^2 = F_m/\mu_m, a falling coefficient of variation can coexist with variance above the Poisson reference.

Table 3 summarizes the ON/OFF ratios using the smaller ON/OFF event count as the common weight for each matched bin. The averaging procedure is given in Supplementary Material S1.

In all 16, 20 and 23 matched bins at 2.76, 7 and 13 TeV, respectively, Rμ,m<Rσ,m<1R_{\mu,m} < R_{\sigma,m} < 1 (Table 3). The fractional decrease in the mean is therefore larger than that in the absolute width, explaining the larger relative width with CR ON.

Comparison with data

The multiplicity dependence of mean transverse momentum provides an experimental reference for the CR response [3]. Figure 6 compares ⟨pT⟩(Nch)\langle p_{\mathrm{T}}\rangle(N_{\mathrm{ch}}) at 13 TeV with the ALICE measurement [17], using numerical values from the associated HEPData table [18]. This comparison does not condition on MPI count.

Figure 6. Mean transverse momentum versus charged-particle density. PYTHIA uses the INEL>0> 0-like selection without an nMPIn_{\mathrm{MPI}} cut. Vertical PYTHIA bars are Monte Carlo standard errors of the mean; they do not represent tune, selection, or generator-model uncertainty. ALICE values are from the associated HEPData record [17, 18].

The CR-ON curve follows the measured trend more closely over much of the displayed range. A quantitative goodnessof-fit comparison would additionally require common particle definitions and the experimental uncertainty correlations.

The stored samples do not reproduce the full ALICE primary-particle definition. The baseline arrays lack the ancestry, status and lifetime information needed to remove secondary particles according to that definition. No correction or systematic uncertainty has been assigned to this difference, and no detector response is applied. The comparison consequently indicates agreement in trend within the stated particle-definition and selection limits.

Final-state event-activity estimators in PYTHIA

Final-state activity classifiers based on multiplicity, spherocity, transverse activity, and flattenicity have been studied extensively [19, 20, 7, 8, 21]. Related approaches include MPI-sensitive regression [22] and experimental uncorrelated seeds compared with simulated MPI [23]. We examine five simulated estimators to characterize how the conditional response is reflected in different final-state regions.

The five estimators are midrapidity multiplicity, forward multiplicity, transverse-region multiplicity, the transverse-region scalar-pTp_{\mathrm{T}} sum, and forward–backward asymmetry. All use charged particles with pT>0.15p_{\mathrm{T}} > 0.15 GeV/c and no upper momentum cut. The midrapidity and forward acceptances are ∣η∣<0.8|\eta| < 0.8 and 2.5<∣η∣<5.02.5 < |\eta| < 5.0, respectively.

The transverse region is defined by π/3<∣Δϕ∣<2π/3\pi/3 < |\Delta\phi| < 2\pi/3 relative to the leading charged particle with ∣η∣<0.8|\eta| < 0.8 and pT>0.5p_{\mathrm{T}} > 0.5 GeV/c. Associated particles satisfy ∣η∣<0.8|\eta| < 0.8. Their count and scalar momentum sum define the two transverse observables.

The asymmetry is AFB=∣NF−NB∣/(NF+NB)A_{\mathrm{FB}} = |N_{\mathrm{F}} - N_{\mathrm{B}}|/(N_{\mathrm{F}} + N_{\mathrm{B}}), where NFN_{\mathrm{F}} and NBN_{\mathrm{B}} are the counts in the positive and negative forward intervals. It is evaluated only for NF+NB>0N_{\mathrm{F}} + N_{\mathrm{B}} > 0.

The baseline estimator analysis requires nMPI>2n_{\mathrm{MPI}} > 2 and positive midrapidity multiplicity without an upper momentum cut. Transverse observables additionally require the leading particle, and asymmetry requires a nonzero denominator. Correlations use the events satisfying the relevant requirements.

Figure 7 shows the event distributions, conditional means, and linear fits. Means are displayed for bins with at least 500 events; Pearson coefficients and event-level fits use all selected events, including bins below that threshold. Table 4 lists the coefficients.

Figure 7. Final-state estimators versus nMPIn_{\mathrm{MPI}} at 13 TeV for CR ON (left) and CR OFF (right). Rows show midrapidity multiplicity, forward multiplicity, transverse multiplicity, transverse scalar-pTp_T sum, and forward–backward asymmetry. Gray shading gives raw events per bin on a logarithmic scale, with common axes and color normalization within each row. These maps show unnormalized counts; the logarithmic scale retains the sparsely populated tails. Green solid and red dashed curves show the CR-ON and CR-OFF conditional means for activity bins containing at least 500 events; dotted lines show event-level least-squares fits. The full selected event ranges are retained. Each panel reports Pearson’s rr evaluated from the individual events, including activity bins below the profile-display threshold.

EstimatorCR ONCR OFF
Midrapidity NchN_{\mathrm{ch}}0.7570.880
Forward NchN_{\mathrm{ch}}0.8330.916
Transverse NchN_{\mathrm{ch}}0.6520.809
Transverse ΣpT\Sigma p_{\mathrm{T}}0.6480.762
AFBA_{\mathrm{FB}}−0.235−0.273

Table 4. Pearson association with PYTHIA nMPIn_{\mathrm{MPI}} at 13 TeV for all five estimators.

Forward multiplicity has the strongest positive linear association among these estimators under their respective selections, followed by midrapidity multiplicity and the transverse observables (Table 4). All four have rising conditional means and larger correlations with CR OFF.

Each coefficient describes the estimator under its own selection. Radiation, beam remnants, hadronization, decays, hard scattering, and acceptance also shape its response, so correlations alone give an incomplete measure of MPI resolution and experimental selection bias.

The asymmetry behaves differently: its conditional mean decreases with MPI count, with modest negative correlations in both CR states. Its curved mean response and broad distributions illustrate the limits of a single linear coefficient for describing the relation to MPI activity.

Low-, mid- and high-activity classes are defined by the one-third and two-thirds quantiles of forward multiplicity, transverse multiplicity, and transverse scalar-pTp_{\mathrm{T}} sum. Quantiles are calculated separately for each CR state, so corresponding labels need not have identical boundaries. Discrete ties can give unequal populations; boundaries and counts are supplied in Supplementary Material S1.

Figure 8 shows the mean MPI count in each activity class. Supplementary Material S1 lists the class boundaries, populations, means, and standard errors. The errors are smaller than the plotting symbols and quantify the precision of the means, rather than the event spread or model uncertainty.

Figure 8. Mean PYTHIA nMPIn_{\mathrm{MPI}} in low-, mid-, and high-activity classes at 13 TeV for three activity estimators: forward multiplicity, transverse-region multiplicity, and transverse-region scalar-pTp_{\mathrm{T}} sum. Supplementary Material S1 lists the plotted class means and their statistical uncertainties. Correlations for all five estimators, including midrapidity multiplicity and AFB, are given separately in Table 4. Standard errors of the mean are included, but are smaller than the plotting symbols and therefore not visible at this scale.

For all three estimators, mean MPI activity rises from the low- to the high-activity class in both CR states. This shows that the selected final-state observables separate populations with different average generator activity, complementing the event-level correlations.

Tune and analysis-choice sensitivity

The independent Monash, A14, and CP5 samples test the configuration dependence of the conditional result. Within each tune, disabling CR leaves the remaining physics settings unchanged (Sec. 2).

The smaller absolute width and larger coefficient of variation with CR ON persist in all common retained bins: 23 for Monash, 24 for A14, and 42 for CP5. Table 6 gives the event-weighted CmC_m and FmF_m summaries. The size of the effect varies among tunes.

The scaled variance has a different tune dependence. Its event-weighted value is larger with CR ON for Monash and A14, whereas CP5 gives nearly equal values, with CR ON slightly smaller. The common ordering of absolute width and coefficient of variation thus does not imply a common ordering of scaled variance.

The tune summaries use each sample’s event counts as bin weights and therefore differ from the common-bin ratios in Sec. 4. Both baseline and tune widths use the population divisor NmN_m, with the weighting procedure specified in S1.

The tune scan also compares forward multiplicity, transverse multiplicity, and transverse scalar-pTp_{\mathrm{T}} sum. It requires nMPI>2n_{\mathrm{MPI}} > 2 and positive central multiplicity within 0.15<pT<200.15 < p_{\mathrm{T}} < 20 GeV/cc. The estimators themselves have no upper momentum cut, and the transverse observables also require a leading particle. The additional central momentum bound and the independent samples distinguish this comparison from the baseline estimator analysis in Sec. 6.

The three estimators retain the same ranking for all tunes and have large correlations with CR OFF (Table 5). Figure 9 shows their mean responses, with statistical errors of the means for bins containing at least 500 events. Correlations use all selected events.

Figure 9. Conditional mean estimators versus nMPIn_{\mathrm{MPI}} at 13 TeV. Rows show forward multiplicity, transverse multiplicity, and transverse scalar-pTp_{\mathrm{T}} sum; columns show Monash 2013, ATLAS A14, and CMS CP5. Green solid curves with circles denote CR ON; red dashed curves with squares denote CR OFF. Error bars give statistical standard errors of the means, often smaller than the markers. Only integer nMPIn_{\mathrm{MPI}} bins containing at least 500 selected events are shown; lines guide the eye. Each tune and CR state uses five million generated events with the estimator selections described in the text. Axes are common within each row. These uncertainties describe the means, not the event-by-event spread. Each panel lists the event-level Pearson coefficients for CR ON and CR OFF, calculated from all selected events, including bins below the profile-display threshold.

The MPI selection and fitted range change the Monash slopes by several percent, while an additional positive-central-multiplicity requirement changes them by less than 0.2%. Supplementary Material S1 gives the numerical comparisons and distinguishes event-level fits from conditional-profile fits. These checks describe analysis sensitivity separately from statistical errors and do not constitute a complete model-uncertainty estimate.

Discussion and limitations

Effect of normalization on conditional widths

The relative size of conditional fluctuations depends on the reference scale. Alongside normalization by the local conditional mean, one can express the width relative to the inclusive multiplicity of each CR state. Let Gs=⟨Nch⟩sG_s = \langle N_{\mathrm{ch}}\rangle_s denote that inclusive mean for the INEL>0> 0-like sample, without an MPI cut and with Nch=0N_{\mathrm{ch}} = 0 retained. Define

Wm,s=σm,sGs,RG=GONGOFF.(7)W_{m,s} = \frac{\sigma_{m,s}}{G_s}, \qquad R_G = \frac{G_{\mathrm{ON}}}{G_{\mathrm{OFF}}}. \tag*{(7)}

The CR ON/OFF ratios then satisfy

RW,m=Rσ,mRG,RC,m=Rσ,mRμ,m.(8)R_{W,m} = \frac{R_{\sigma,m}}{R_G}, \qquad R_{C,m} = \frac{R_{\sigma,m}}{R_{\mu,m}}. \tag*{(8)}

Separate inclusive normalization gives RW,m>1R_{W,m} > 1 when RG<Rσ,m<1R_G < R_{\sigma,m} < 1, whereas normalization by the conditional mean gives RC,m>1R_{C,m} > 1 when Rμ,m<Rσ,m<1R_{\mu,m} < R_{\sigma,m} < 1. Thus the two relative widths probe different changes in scale. All comparisons here match integer mm; matching the coordinate m/⟨nMPI⟩sm/\langle n_{\mathrm{MPI}}\rangle_s would generally select different MPI counts in the two CR states. A shared fixed normalization would preserve the absolute-width ordering.

Table 7 summarizes the three width definitions across all 148 matched bins. In the 13 TeV baseline, absolute narrowing and inclusive-normalized broadening coexist in all 23 bins.

Samplemm rangeRGR_GBBRσ,m<1R_{\sigma,m} < 1RC,m>1R_{C,m} > 1RW,m>1R_{W,m} > 1
Baseline, 2.76 TeV3–180.79196916161614 (12)
Baseline, 7 TeV3–220.74993520202020 (18)
Baseline, 13 TeV3–250.72486623232323 (22)
Independent Monash, 13 TeV3–250.72887823232323 (22)
Independent A14, 13 TeV3–260.70761324242424 (23)
Independent CP5, 13 TeV3–440.61599742424242 (40)

Table 7. Absolute and relative conditional-width orderings at matched integer MPI counts. BB is the number of retained matched bins; the three count columns give bins with absolute narrowing, local-relative broadening, and inclusive-relative broadening. Parentheses in the final column give bins whose marginal normal-approximation 95% lower bound on RW,mR_{W,m} exceeds unity. The intervals are marginal rather than simultaneous.

ProxyTuneCR ONCR OFF
Forward multiplicityMonash0.8360.917
Forward multiplicityATLAS A140.8520.929
Forward multiplicityCMS CP50.8990.950
Transverse multiplicityMonash0.6620.816
Transverse multiplicityATLAS A140.6780.834
Transverse multiplicityCMS CP50.7300.874
Transverse scalar-pTp_{\mathrm{T}} sumMonash0.6560.768
Transverse scalar-pTp_{\mathrm{T}} sumATLAS A140.6710.791
Transverse scalar-pTp_{\mathrm{T}} sumCMS CP50.7050.840

Table 5. Equal-statistics 5M-event tune comparison of estimator–nMPIn_{\mathrm{MPI}} Pearson coefficients.

ConfigurationWeighted CmC_mWeighted FmF_mBins
Monash0.420 / 0.3502.808 / 2.57523 / 23
A140.414 / 0.3392.675 / 2.41624 / 24
CMS CP50.436 / 0.3582.595 / 2.59742 / 42

Table 6. 13 TeV conditional-fluctuation summaries for nMPI>2n_{\mathrm{MPI}} > 2, retaining bins with at least 500 events. Each numerical pair is CR ON / OFF. The last column lists bins with Cm,ON>Cm,OFFC_{m,\mathrm{ON}} > C_{m,\mathrm{OFF}} / all matched bins.

For example, at m=10m = 10, Rμ,m=0.663334R_{\mu,m} = 0.663334, Rσ,m=0.862401R_{\sigma,m} = 0.862401, and RG=0.724866R_G = 0.724866, giving RC,m=1.300100R_{C,m} = 1.300100 and RW,m=1.189738±0.004710R_{W,m} = 1.189738 \pm0.004710, where the uncertainty is one statistical standard error. The same central absolute-versus-inclusive-normalized ordering holds in every retained bin of the independent 13 TeV tune comparisons. At 2.76 TeV, RW,m>1R_{W,m} > 1 holds in 14 of 16 bins, whereas RC,m>1R_{C,m} > 1 holds in all 16; inclusive and local normalization are therefore not interchangeable.

Statistical errors for Wm,sW_{m,s} use empirical influence-function propagation, including covariance between the conditional standard deviation and the estimated inclusive mean. The Cm,sC_{m,s} errors include the conditional mean–variance covariance. Ratios treat the CR samples as independent, subject to the historical seed-record limitation. The variance divisor is the bin event count, as in Sec. 4. S1 gives the propagation expressions and every-bin results; S2 supplies the corresponding numerical tables and code.

This scale dependence helps interpret earlier studies of conditional fluctuations. Weber et al. [6] discuss conditional broadening alongside self-normalized MPI–multiplicity relations. The present results show how such broadening can coexist with a smaller absolute width: the inclusive mean or the local conditional mean decreases more strongly than the conditional standard deviation. Their discussion does not specify numerical conditional standard deviations or their normalization, so the width descriptions cannot be equated unambiguously. Moreover, their PYTHIA 8.230 non-diffractive samples use ALICE-defined primary charged particles in ∣η∣<1|\eta| < 1, while our samples use the definitions in § 1.2.2. The normalization comparison therefore explains a compatible pattern on the same events, without establishing direct numerical agreement between the two simulations.

Across the configurations studied here, the measured ordering Rμ,m<Rσ,m<1R_{\mu,m} < R_{\sigma,m} < 1 separates mean suppression from absolute-width suppression. It quantifies why local relative fluctuations increase despite absolute narrowing, with matched-bin uncer-

tainties and checks across energies and tunes. The persistence of this pattern characterizes the tested configurations; its generality requires further model comparisons.

The conditional correlations and the ALICE mean-pTp_{\mathrm{T}} comparison provide complementary views of the final state. CR OFF produces steeper responses and stronger associations with internal MPI activity, whereas CR ON follows the displayed ALICE mean-pTp_{\mathrm{T}} trend more closely within the limitations of Sec. 5. Correlation strength alone is therefore insufficient to judge experimental agreement. The estimator results complement established classifier studies [7, 8].

The event and particle definitions, MPI selection, bin retention, and fit range set the scope of these results. A closer experimental comparison requires the full primary-particle information and detector response.

The mapping from MPI activity to final particles depends on shower and hadronization models. PYTHIA uses string fragmentation, whereas Herwig uses a cluster model; MPI and CR are not exclusive to PYTHIA. Cross-generator comparisons [22] and separately optimized CR-OFF configurations are needed to assess the generality of the observed ordering.

Conclusions

For the tested PYTHIA configurations, CR lowers both the conditional mean multiplicity and its absolute width at fixed nMPIn_{\mathrm{MPI}}. The larger fractional decrease in the mean produces a higher coefficient of variation with CR ON in every retained matched bin at 2.76, 7, and 13 TeV.

The same absolute-versus-relative width ordering is found in independent Monash, A14, and CMS CP5 samples. Its magnitude depends on the tune and analysis choices, while the scaled variance shows a different configuration dependence.

Comparisons with other generators, separately optimized CR-OFF configurations, and samples with complete primary-particle information would establish how this conditional pattern extends beyond the present setup.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

OpenAI Codex assisted language editing, LaTeX preparation, analysis-code development, diagnostic calculations, and data-derived plotting. Section 2.3 identifies the recorded models and client version, the assisted tasks, the executable validation performed, and the limits of the historical records. The authors take responsibility for the scientific content, interpretation, and final submitted version. No AI system is listed as an author.

Data and code availability

Supplementary Material S1 gives the analysis definitions, numerical results, matched-bin uncertainties, and production-provenance limitations. Supplementary Material S2 contains the analysis and plotting code, integer conditional joint-count inputs, numerical tables, available generation configurations and production records, and the AI-assistance evidence map. Running bash RUN_ANALYSIS.sh reproduces the central conditional results, plots, and normalization comparison from the supplied inputs. The large retained event and particle files are available from the corresponding author upon reasonable request. Missing historical seed and effective lifetime/decay records limit reproducibility to reanalysis of the retained samples rather than exact historical generator replay.

Supplementary Material S1: Definitions, Calculations, and Complete Numerical Tables

Subir Sarkar; Kalyan Dey

This standalone supplement records the analysis definitions, calculation sequence, statistical estimators, uncertainty propagation, selection checks, tune comparisons, and numerical tables underlying manuscript NPB-D-26-01492R2. It is intended to make explicit details that cannot be retained in the journal article because of space and presentation constraints. It does not introduce a separate physics claim; the interpretation and limitations are the same as in the revised manuscript.

supplement.pdf

Resubmission note (21 October 2026). The approximate third configuration has been replaced by fresh CMS CP5 samples using the full-precision CMS tune card and the external NNPDF3.1 NNLO central PDF. All affected fluctuation profiles, matched-bin checks and estimator coefficients have been recalculated from five million events per CR state. Production-time settings, particle data and seeds are retained in cp5_exact/. The historical production-time seed and effective lifetime/decay records were not retained; exact historical replay is unavailable. The present resubmission draft adds checked integer joint-count inputs, a common conditional-width convention and complete matched-bin statistical uncertainties. The production-provenance limits are stated explicitly below.

Purpose and scope

The analysis asks how color reconnection (CR) changes the conditional relation between the internal PYTHIA interaction count nMPIn_{\mathrm{MPI}} and the final-state charged-particle multiplicity NchN_{\mathrm{ch}}. The two configurations are denoted CR ON and CR OFF. The intended CR variation disables ColourReconnection:reconnect while retaining the other tune parameters. Complete effective-setting equality is verified directly for corrected CP5. Historical baseline settings are supported by candidate source records rather than certified per-run dumps. No CR-OFF retuning is performed. Consequently, this is a controlled generator variation, not a comparison of two independently tuned descriptions of data.

The value nMPIn_{\mathrm{MPI}} is returned by pythia.info.nMPI(). It depends on the PYTHIA version, tune, multiparton-interaction model, regularization, matter profile, process selection, and internal counting convention. It is not an experimental observable and is not interpreted as a universal number of physical scatterings.

Generated samples and analysis selections

The historical baseline consists of six retained five-million-event samples, described in the manuscript records as PYTHIA 8.316 SoftQCD:all with Monash 2013 at s=2.76\sqrt{s}=2.76, 7 and 13 TeV in each CR state. Complete per-run version/build and effective-setting records are not certified; the retained source family and the provenance limits are documented below. Equal-statistics samples at 13 TeV are also analysed for Monash 2013, ATLAS A14, and CMS CP5.

The analysis multiplicity is

Nch=#{charged final-state rows:∣η∣<0.8,0.15<pT<20 GeV/c}.(9)N_{\mathrm{ch}} = \#\{\text{charged final-state rows}: \lvert\eta\rvert< 0.8,\quad0.15 < p_{\mathrm{T}} < 20\ \mathrm{GeV}/c\}. \tag*{(9)}

An INEL>> 0-like event contains at least one charged final-state particle in ∣η∣<1\lvert\eta\rvert< 1. The generator-level conditional study additionally requires nMPI>2n_{\mathrm{MPI}} > 2. A conditional bin is displayed only if it contains at least 500 selected events.

The stored arrays contain final-state identity, charge, and kinematics but not the complete ancestry, status, and lifetime provenance required to reproduce the exact ALICE primary-charged-particle definition. No detector simulation, tracking efficiency, momentum resolution, or smearing is applied. The ALICE-facing ⟨pT⟩(Nch)\langle p_{\mathrm{T}}\rangle(N_{\mathrm{ch}}) curve uses the INEL>> 0-like and Nch>0N_{\mathrm{ch}} > 0 requirements without any nMPIn_{\mathrm{MPI}} selection.

The tune-generation code selects charged final-state particles with isFinal() and isCharged(). The accepted multiplicity is counted after applying the analysis cuts; the implementation details are retained in S2.

Event-level quantities and bookkeeping

For each event ii:

  • mim_i is its integer PYTHIA nMPIn_{\mathrm{MPI}} value;

  • Nch,iN_{\mathrm{ch},i} is its accepted charged-particle count;

  • NmN_m is the number of selected events satisfying mi=mm_i=m;

  • N=∑mNmN=\sum_m N_m is the total number of selected events in one energy and CR sample.

The conditional distribution P(Nch∣nMPI=m)P(N_{\mathrm{ch}}\mid n_{\mathrm{MPI}}=m) contains only events having the same mm. Conditioning removes changes in the sampled nMPIn_{\mathrm{MPI}} distribution from the within-bin multiplicity dispersion.

Conditional moments

For a fixed integer mm, the conditional mean is

μm=⟨Nch⟩m=1Nm∑i:mi=mNch,i.(10)\mu_m=\langle N_{\mathrm{ch}}\rangle_m=\frac{1}{N_m}\sum_{i:m_i=m}N_{\mathrm{ch},i}. \tag*{(10)}

The residual for event ii is di=Nch,i−μmd_i=N_{\mathrm{ch},i}-\mu_m. The kkth population central moment is

μk,m=1Nm∑i:mi=mdik.(11)\mu_{k,m}=\frac{1}{N_m}\sum_{i:m_i=m}d_i^k. \tag*{(11)}

The conditional variance and absolute conditional standard deviation are

Vm=μ2,m,σm=Vm.(12)V_m=\mu_{2,m},\qquad\sigma_m=\sqrt{V_m}. \tag*{(12)}

The corrected conditional-width tables consistently use divisor NmN_m in the baseline and independent tune samples. The earlier tune tables used Nm−1N_m-1; the corrected tune widths equal those earlier widths multiplied by (Nm−1)/Nm\sqrt{(N_m-1)/N_m}. Original values were reproduced before this documented correction. The calculation characterizes the generated Monte Carlo population. The absolute width σm\sigma_m is measured on the same multiplicity scale as NchN_{\mathrm{ch}}.

Global ridge width

The global ridge width is the RMS of all event residuals after subtracting the conditional mean appropriate to each event:

σridge=[1N∑i(Nch,i−μmi)2]1/2,(13)\sigma_{\mathrm{ridge}} = \left[\frac{1}{N}\sum_i\left(N_{\mathrm{ch},i}-\mu_{m_i}\right)^2\right]^{1/2}, \tag*{(13)}
σridge2=∑mNmNσm2.(14)\sigma_{\mathrm{ridge}}^2 = \sum_m \frac{N_m}{N}\sigma_m^2. \tag*{(14)}

Thus, it is an event-weighted combination of conditional variances, not the unconditional standard deviation of the full NchN_{\mathrm{ch}} distribution. “Ridge” refers to the populated band in the nMPIn_{\mathrm{MPI}}–NchN_{\mathrm{ch}} response map and not to a two-particle angular-correlation ridge.

Normalized fluctuation measures

Since a width commonly increases with its mean, σm\sigma_m alone cannot show whether dispersion is larger relative to the local multiplicity scale. Two normalizations are used:

Cm=σmμm,Fm=σm2μm.(15)C_m = \frac{\sigma_m}{\mu_m}, \qquad F_m = \frac{\sigma_m^2}{\mu_m}. \tag*{(15)}

CmC_m is the conditional coefficient of variation. It is dimensionless and reports the standard deviation as a fraction of the conditional mean. For example, Cm=0.40C_m = 0.40 means σm=0.40μm\sigma_m = 0.40\mu_m. FmF_m is the conditional scaled variance. For an ideal Poisson multiplicity distribution, variance equals mean and Fm=1F_m = 1. A departure from one quantifies non-Poisson dispersion but does not identify its physical origin.

For a compact sample-level summary, the manuscript also gives

Cglob=σridge⟨Nch⟩,Fglob=σridge2⟨Nch⟩.(16)C_{\mathrm{glob}} = \frac{\sigma_{\mathrm{ridge}}}{\langle N_{\mathrm{ch}}\rangle}, \qquad F_{\mathrm{glob}} = \frac{\sigma_{\mathrm{ridge}}^2}{\langle N_{\mathrm{ch}}\rangle}. \tag*{(16)}

where ⟨Nch⟩=N−1∑iNch,i\langle N_{\mathrm{ch}}\rangle= N^{-1}\sum_i N_{\mathrm{ch},i}. These are not averages of CmC_m or FmF_m and must not be confused with the matched-bin ratios below.

Matched-bin CR comparison

A matched bin is the same integer m=nMPIm = n_{\mathrm{MPI}} in the CR-ON and CR-OFF samples, with at least 500 events in each. For every matched bin,

Rμ,m=μm,ONμm,OFF,Rσ,m=σm,ONσm,OFF,RC,m=Cm,ONCm,OFF,RF,m=Fm,ONFm,OFF.(17)R_{\mu,m} = \frac{\mu_{m,\mathrm{ON}}}{\mu_{m,\mathrm{OFF}}}, \qquad R_{\sigma,m} = \frac{\sigma_{m,\mathrm{ON}}}{\sigma_{m,\mathrm{OFF}}}, \qquad R_{C,m} = \frac{C_{m,\mathrm{ON}}}{C_{m,\mathrm{OFF}}}, \qquad R_{F,m} = \frac{F_{m,\mathrm{ON}}}{F_{m,\mathrm{OFF}}}. \tag*{(17)}

The common-bin weight is the smaller of the two populations,

wm=min⁡(Nm,ON,Nm,OFF),(18)w_m = \min(N_{m,\mathrm{ON}},N_{m,\mathrm{OFF}}), \tag*{(18)}

so the weight is not determined by events available in only one configuration. For X∈{μ,σ,C,F}X \in\{\mu,\sigma,C,F\}, the reported energy-level ratio is

R‾X=∑mwmRX,m∑mwm.(19)\overline{R}_X = \frac{\sum_m w_m R_{X,m}}{\sum_m w_m}. \tag*{(19)}

R‾σ<1\overline{R}_{\sigma} < 1 means a smaller absolute width with CR ON; R‾C>1\overline{R}_C > 1 means a larger width relative to the CR-ON conditional mean. It is the coexistence of these two inequalities that produces the central scale–relative-fluctuation separation.

Statistical uncertainties

For a bin containing n=Nmn = N_m events, the large-sample moment estimates used by the analysis are

Var⁡(μm)=μ2,mn,(20)\operatorname{Var}(\mu_m) = \frac{\mu_{2,m}}{n}, \tag*{(20)}
Var⁡(Vm)=μ4,m−μ2,m2n,(21)\operatorname{Var}(V_m) = \frac{\mu_{4,m} - \mu_{2,m}^{2}}{n}, \tag*{(21)}
Cov⁡(μm,Vm)=μ3,mn.(22)\operatorname{Cov}(\mu_m,V_m) = \frac{\mu_{3,m}}{n}. \tag*{(22)}

Accordingly,

SEM⁡(μm)=σmn,(23)\operatorname{SEM}(\mu_m) = \frac{\sigma_m}{\sqrt{n}}, \tag*{(23)}
SEM⁡(σm)=Var⁡(Vm)2σm.(24)\operatorname{SEM}(\sigma_m) = \frac{\sqrt{\operatorname{Var}(V_m)}}{2\sigma_m}. \tag*{(24)}

For g(μ,V)g(\mu,V), the delta-method variance is

Var⁡(g)≃(∇g)T(Var⁡(μ)Cov⁡(μ,V)Cov⁡(μ,V)Var⁡(V))∇g.(25)\operatorname{Var}(g) \simeq(\nabla g)^{T} \begin{pmatrix} \operatorname{Var}(\mu) & \operatorname{Cov}(\mu,V) \\ \operatorname{Cov}(\mu,V) & \operatorname{Var}(V) \end{pmatrix} \nabla g. \tag*{(25)}

The gradients used are

∇C=(−σμ2,12μσ),(26)\nabla C = \left(-\frac{\sigma}{\mu^{2}},\frac{1}{2\mu\sigma}\right), \tag*{(26)}
∇F=(−Vμ2,1μ).(27)\nabla F = \left(-\frac{V}{\mu^{2}},\frac{1}{\mu}\right). \tag*{(27)}

Ratio and difference errors combine ON and OFF estimates under an independent-sample assumption. Distinct fixed seeds are verified for corrected CP5; the complete historical seed histories are not available for every baseline or Monash/A14 segment. These error bars quantify finite Monte Carlo precision only. They exclude tune, generator-model, selection, particle-definition, fit-range, and detector uncertainties.

Association and fitted response

Pearson’s coefficient is

r=∑i(mi−m‾)(Nch,i−N‾ch)∑i(mi−m‾)2∑i(Nch,i−N‾ch)2,(28)r = \frac{\sum_i (m_i-\overline{m})(N_{\mathrm{ch},i}-\overline{N}_{\mathrm{ch}})} {\sqrt{\sum_i(m_i-\overline{m})^2\sum_i(N_{\mathrm{ch},i}-\overline{N}_{\mathrm{ch}})^2}}, \tag*{(28)}

and measures linear association. Spearman’s ρ\rho is Pearson’s coefficient applied to the ranks and measures monotonic ordering. The event-level fitted response is Nch=a+bnMPIN_{\mathrm{ch}} = a + b n_{\mathrm{MPI}}, where aa is the intercept and bb is the least-squares change in fitted multiplicity for a one-unit increase in nMPIn_{\mathrm{MPI}}. For fits to conditional profile points, the code uses weights Nm\sqrt{N_m}, corresponding to greater influence for more populated bins. None of these quantities establishes causality, physical superiority, or a one-to-one conversion of an additional interaction into final particles.

Measurable activity estimators

The five final-state estimators are:

  1. midrapidity multiplicity: charged particles with ∣η∣<0.8|\eta| < 0.8 and pT>0.15p_T > 0.15 GeV/cc;

  2. forward multiplicity: the sum in 2.5<η<5.02.5 < \eta< 5.0 and −5.0<η<−2.5-5.0 < \eta< -2.5, with pT>0.15p_T > 0.15 GeV/cc;

  3. transverse multiplicity: accepted central charged particles satisfying π/3<∣Δϕ∣<2π/3\pi/3 < |\Delta\phi| < 2\pi/3 relative to the leading charged particle;

s\sqrt{s}CRrrρ\rhobb
2.76 TeVON0.6350.6161.533
2.76 TeVOFF0.7890.7582.700
7 TeVON0.7140.7031.697
7 TeVOFF0.8500.8303.078
13 TeVON0.7570.7501.792
13 TeVOFF0.8800.8643.301

Table 1 (PDF p. 18). Baseline association and slope summaries for INEL>0> 0-like events with nMPI>2n_{\mathrm{MPI}} > 2. Pearson’s rr, Spearman’s ρ\rho, and the event-level fitted slope bb are defined in this section. The global width summaries are retained in the Principal numerical summaries section.

  1. transverse scalar-pTp_{\mathrm{T}} sum: the scalar pTp_{\mathrm{T}} sum of the same transverse particles;

  1. forward–backward asymmetry: AFB=∣NF−NB∣/(NF+NB)A_{\mathrm{FB}} = \lvert N_{\mathrm{F}} - N_{\mathrm{B}}\rvert/(N_{\mathrm{F}} + N_{\mathrm{B}}), evaluated only when NF+NB>0N_{\mathrm{F}} + N_{\mathrm{B}} > 0.

The baseline five-estimator sample requires nMPI>2n_{\mathrm{MPI}} > 2 and positive midrapidity multiplicity without an upper momentum cut. The transverse observables additionally require a leading particle. The separate tune-estimator sample instead uses positive central multiplicity with 0.15<pT<200.15 < p_{\mathrm{T}} < 20 GeV/cc as its event requirement; the estimators themselves retain no upper momentum cut. These separate generated samples and event definitions account for the nonidentical Monash entries in the two correlation tables.

The leading particle satisfies ∣η∣<0.8\lvert\eta\rvert< 0.8 and pT>0.5p_{\mathrm{T}} > 0.5 GeV/cc. Activity classes are low, middle, and high quantiles of the chosen estimator. They enrich different nMPIn_{\mathrm{MPI}} populations statistically; they do not measure nMPIn_{\mathrm{MPI}} event by event. All estimators also receive contributions from ISR, FSR, the hard process, beam remnants, hadronization, decays, and acceptance.

Numerical values for the activity-class figure

Table 2 gives the numerical values plotted in the manuscript activity-class figure. The low, middle, and high classes are formed from quantiles of each estimator separately for CR ON and CR OFF. Their estimator boundaries therefore need not be identical between the two CR configurations. For integer multiplicities, the lower class includes the lower boundary, the middle class is above the first boundary and includes the second, and the high class is above the second boundary. The scalar-pTp_{\mathrm{T}} boundaries and means are in GeV/cc.

s\sqrt{s} [TeV]2.76713
⟨Nch⟩\langle N_{\mathrm{ch}}\rangle6.92 / 8.748.74 / 11.6510.25 / 14.14
⟨⟨pT⟩event⟩\langle\langle p_{\mathrm{T}}\rangle_{\mathrm{event}}\rangle0.530 / 0.5020.558 / 0.5150.580 / 0.526
Multiplicity change [%]−20.8-20.8−25.0-25.0−27.5-27.5
Mean-pTp_{\mathrm{T}} change [%]5.78.310.2
Nch≥20N_{\mathrm{ch}} \ge20 [%]––15.8 / 25.3
⟨pT⟩event≥1\langle p_{\mathrm{T}}\rangle_{\mathrm{event}} \ge1 [%]––3.90 / 1.79

Table 2. Inclusive Monash results for the selections of Figs. 1 and 2. Each pair is CR ON / OFF. Mean-pTp_{\mathrm{T}} values are in GeV/cc; tail fractions are percentages.

Analysis workflow

The calculation proceeds in the following order:

  1. load the event metadata, accepted NchN_{\mathrm{ch}} array, and INEL>0> 0-like mask;

  1. apply the baseline mask INEL>0> 0-like and nMPI>2n_{\mathrm{MPI}} > 2 for the generator diagnostic;

  1. group events by integer m=nMPIm = n_{\mathrm{MPI}};

  1. retain groups with Nm≥500N_m \ge500 for profile display only;

  1. calculate μm\mu_m, central moments through fourth order, σm\sigma_m, CmC_m, FmF_m, and their statistical standard errors;

  1. calculate event-level global width, mean, correlations, and slope using all selected events, including bins below the display threshold;

  1. merge CR-ON and CR-OFF profiles on the same integer mm;

  1. calculate matched-bin ratios and their common-bin weighted averages;

Table 2: Mean PYTHIA nMPIn_{\mathrm{MPI}} in the 13-TeV activity classes used in the manuscript figure. NN is the number of events in the class, and the quoted uncertainty is SEM =σ(nMPI)/N= \sigma(n_{\mathrm{MPI})/\sqrt{N}}.

EstimatorCRClassRangeNN⟨nMPI⟩±SEM\langle n_{\mathrm{MPI}}\rangle\pm\mathrm{SEM}
Forward NchN_{\mathrm{ch}}ONlow0–325273614.1396 ± 0.0020
ONmid32–535298466.8247 ± 0.0036
ONhigh53–19250709312.1790 ± 0.0055
OFFlow0–405243453.8754 ± 0.0015
OFFmid40–735306866.6438 ± 0.0028
OFFhigh73–27351158012.5562 ± 0.0050
Transverse NchN_{\mathrm{ch}}ONlow0–36341345.2095 ± 0.0032
ONmid3–64606337.5865 ± 0.0053
ONhigh6–3743550411.5976 ± 0.0067
OFFlow0–45503324.4474 ± 0.0023
OFFmid4–95244547.0290 ± 0.0039
OFFhigh9–10346885512.3012 ± 0.0058
Transverse ∑pT\sum p_{\mathrm{T}}ONlow0–1.3085100924.9398 ± 0.0032
ONmid1.308–3.2375100896.9961 ± 0.0047
ONhigh3.237–51.16751009011.2931 ± 0.0062
OFFlow0–1.8585145474.4420 ± 0.0024
OFFmid1.858–4.3225145476.8592 ± 0.0040
OFFhigh4.322–73.43251454711.8286 ± 0.0058

Table 2 (PDF p. 19).

  1. repeat normalized profiles for Monash, A14, and CMS CP5;

  1. repeat global and fitted quantities for the selection and fit-range sensitivity variants;

  1. export CSV tables and construct the manuscript figures.

Principal numerical summaries

Table 3: Global baseline results for INEL>0> 0-like events with nMPI>2n_{\mathrm{MPI}} > 2.

s\sqrt{s}CRNN⟨Nch⟩\langle N_{\mathrm{ch}}\rangleσridge\sigma_{\mathrm{ridge}}CglobC_{\mathrm{glob}}FglobF_{\mathrm{glob}}rr
2.76 TeVON131139911.64695.05680.43422.19560.6354
2.76 TeVOFF131074415.87705.74970.36212.08220.7887
7 TeVON146869014.51055.99660.41332.47820.7138
7 TeVOFF146976720.68246.86810.33212.28070.8504
13 TeVON156611916.80366.72640.40032.69250.7569
13 TeVOFF156747024.61177.74260.31462.43570.8801

Table 3 (PDF p. 19).

Table 4: Matched-bin ratios. The final column counts bins in which the CR-ON coefficient of variation exceeds the CR-OFF value.

s\sqrt{s}rangebinsR‾σ\overline{R}_{\sigma}R‾C\overline{R}_{C}R‾F\overline{R}_{F}CON>COFFC_{\mathrm{ON}} > C_{\mathrm{OFF}}
2.76 TeV3–18160.89481.17391.048216/16
7 TeV3–22200.89121.21051.076020/20
13 TeV3–25230.88961.23221.092823/23

Table 4 (PDF p. 19).

At 13 TeV, CR ON lowers the global conditional multiplicity scale from 24.6117 to 16.8036 (31.7%) and the absolute ridge width from 7.7426 to 6.7264 (13.1%). Since the mean decreases much more strongly, the global width-to-mean ratio increases from 0.3146 to 0.4003. The matched-bin result verifies that this reversal persists across the populated nMPIn_{\mathrm{MPI}} range and is not produced by a single aggregate.

Tune summary

TuneCRnMPIn_{\mathrm{MPI}} rangeevent-weighted CmC_mevent-weighted FmF_m
MonashON3–250.42042.8083
MonashOFF3–250.34952.5755
ATLAS A14ON3–260.41392.6748
ATLAS A14OFF3–260.33852.4155
CMS CP5ON3–440.43632.5954
CMS CP5OFF3–440.35802.5974

Table 5 (PDF p. 20). Event-weighted normalized fluctuation measures at 13 TeV for retained bins with nMPI>2n_{\mathrm{MPI}} > 2.

Interpretation and limitations

The calculation supports the following precise statement: within the specified PYTHIA configurations and at the same integer nMPIn_{\mathrm{MPI}}, CR ON produces a lower conditional mean and a smaller absolute conditional width, but the width falls less strongly than the mean; hence relative dispersion is larger with CR ON. The result is stable across the tested energies and tunes, while its numerical magnitude remains model dependent.

The analysis does not establish that CR ON or CR OFF is globally superior, does not measure MPI experimentally, does not isolate the direct causal action of CR, and does not reproduce a detector-level ALICE selection. A dedicated CR-OFF retune, provenance-complete primary-particle samples, detector response, other generator models, and broader systematic variations would be required to extend those conclusions.

Inclusive distributions and estimator-profile inputs

For event ii with Nch,i>0N_{\mathrm{ch},i} > 0, the event-mean transverse momentum is

⟨pT⟩event,i=1Nch,i∑j=1Nch,ipT,ij.(29)\langle p_{\mathrm{T}}\rangle_{\mathrm{event},i} = \frac{1}{N_{\mathrm{ch},i}}\sum_{j=1}^{N_{\mathrm{ch},i}} p_{\mathrm{T},ij}. \tag*{(29)}

Here pT,ijp_{\mathrm{T},ij} is the transverse momentum of the jjth accepted charged particle.

The inclusive distributions in manuscript Figure 1 and Figure 2 use the INEL>> 0-like sample without a selection on nMPIn_{\mathrm{MPI}}. The multiplicity distribution retains zero central multiplicity. The event-mean-pTp_{\mathrm{T}} distribution additionally requires Nch>0N_{\mathrm{ch}} > 0 and assigns equal weight to each event. Its bin width is 0.5 GeV/cc, while the multiplicity-bin width is 5. The machine-readable histograms and sample summaries are supplied in Supplementary Material S2.

The five-estimator maps (manuscript Figure 7) retain all selected events and use raw events per bin with logarithmic color normalization. Conditional means are displayed for Nm≥500N_m \ge500. Pearson coefficients and ordinary least-squares fits use all selected events, rather than the conditional profile points or histogram-bin centers.

For the tune profiles (manuscript Figure 9), the standard error is sm/Nms_m/\sqrt{N_m}, where sm2=∑i(xi−xˉm)2/(Nm−1)s_m^2 = \sum_i(x_i-\bar{x}_m)^2/(N_m-1) for the estimator xx. This finite-sample convention differs from the population-moment convention used for the main conditional-width and activity-class calculations by a factor Nm/(Nm−1)\sqrt{N_m/(N_m-1)}, below 0.11% for displayed bins. The error bars quantify uncertainty in the mean, not the event spread. The twelve Monash/A14 estimator coefficients reproduce their archived values to 10−1010^{-10}. The six CP5 coefficients and their profiles are recalculated from the freshly generated CP5 events; the obsolete approximate-CP5 coefficients are excluded.

Matched-bin verification across tunes

At 13 TeV, the coefficient of variation is larger with CR ON at every common retained integer nMPIn_{\mathrm{MPI}} in each tested tune. Common-bin weights are wm=min⁡(Nm,ON,Nm,OFF)w_m = \min(N_{m,\mathrm{ON}}, N_{m,\mathrm{OFF}}), as in the energy comparison.

TunenMPIn_{\mathrm{MPI}} rangeR‾σ\overline{R}_{\sigma}R‾C\overline{R}_{C}Passing bins
Monash3–250.89021.227523/23
ATLAS A143–260.88781.250324/24
CMS CP53–440.80021.255742/42

Table 6 (PDF p. 21). Matched-bin checks in the independent tune samples. The last column counts bins with Cm,ON>Cm,OFFC_{m,\mathrm{ON}} > C_{m,\mathrm{OFF}}.

corresponding CSV is tune_matched_bin_verification.csv in Supplementary Material S2. These ratios differ from averages of CmC_m evaluated separately in each CR sample.

Complete conditional profiles

The following table reproduces every row of fresh_conditional_fluctuations_by_nmpi.csv. “SE” denotes the one-standard-error Monte Carlo uncertainty.

EnergyCRmmNmN_mμm\mu_mσm\sigma_mCmC_mFmF_mSE(μ\mu)SE(σ\sigma)SE(CC)SE(FF)
13 TeVCR OFF32751239.53734.68250.49102.29900.00890.00780.00070.0070
13 TeVCR OFF420738112.64305.48130.43352.37640.01200.01110.00080.0091
13 TeVCR OFF516656015.78926.22610.39432.45510.01530.04290.00270.0334
13 TeVCR OFF614029818.99026.78880.35752.42690.01810.01490.00070.0101
13 TeVCR OFF711975722.26957.40350.33252.46130.02140.01830.00080.0117
13 TeVCR OFF810592325.55287.94090.31082.46770.02440.02030.00080.0121
13 TeVCR OFF99357428.85458.42900.29212.46230.02760.02160.00070.0121
13 TeVCR OFF108307332.24048.95020.27762.48470.03110.02390.00070.0129
13 TeVCR OFF117339635.64109.40830.26402.48350.03470.02670.00070.0136
13 TeVCR OFF126416939.00899.80850.25142.46630.03870.03000.00070.0147
13 TeVCR OFF135497742.359310.17160.24012.44240.04340.03250.00070.0152
13 TeVCR OFF144597345.724510.55420.23082.43610.04920.03770.00080.0169
13 TeVCR OFF153738949.073010.91310.22242.42690.05640.04240.00080.0185
13 TeVCR OFF162928352.463611.30300.21542.43520.06610.05300.00100.0224
13 TeVCR OFF172242955.659811.63800.20912.43340.07770.05790.00100.0237
13 TeVCR OFF181616359.019511.97750.20292.43070.09420.06920.00110.0275
13 TeVCR OFF191139662.352412.21940.19602.39470.11450.08460.00130.0326
13 TeVCR OFF20778965.710212.64500.19242.43330.14330.10600.00160.0400
13 TeVCR OFF21516868.897312.81060.18592.38200.17820.13140.00190.0480
13 TeVCR OFF22321372.212013.28470.18402.44400.23440.16620.00230.0603
13 TeVCR OFF23199575.892213.47480.17762.39250.30170.21360.00280.0748
13 TeVCR OFF24114478.370613.35030.17032.27420.39470.27200.00350.0917
13 TeVCR OFF2561081.608214.46970.17732.56560.58590.46310.00550.1605
13 TeVCR ON32745858.07124.41290.54672.41270.00840.00790.00090.0077
13 TeVCR ON420720410.12825.06710.50032.53510.01110.01030.00090.0093
13 TeVCR ON516651012.09285.60720.46372.60000.01370.01220.00090.0103
13 TeVCR ON613954013.98816.10640.43652.66570.01630.01430.00090.0114
13 TeVCR ON712006315.87986.55790.41302.70820.01890.01620.00090.0124
13 TeVCR ON810597817.69706.97650.39422.75020.02140.01820.00090.0133
13 TeVCR ON99343219.59667.32750.37392.73990.02400.02020.00090.0141
13 TeVCR ON108299321.38627.71870.36092.78580.02680.02200.00090.0149
13 TeVCR ON117305023.15478.04450.34742.79480.02980.02450.00100.0160
13 TeVCR ON126408024.88998.30850.33382.77350.03280.02650.00100.0167
13 TeVCR ON135522126.66438.63370.32382.79550.03670.02920.00100.0179
13 TeVCR ON144594128.28818.85680.31312.77300.04130.03240.00110.0192
13 TeVCR ON153718629.99479.14660.30492.78920.04740.03820.00120.0221
13 TeVCR ON162951331.58029.29910.29452.73820.05410.04170.00120.0234
13 TeVCR ON172221533.14999.56390.28852.75920.06420.05100.00150.0281
13 TeVCR ON181635534.63189.69220.27992.71250.07580.05810.00160.0312
13 TeVCR ON191162036.26459.95500.27452.73270.09230.07220.00190.0378
13 TeVCR ON20776737.469810.11390.26992.73000.11480.08860.00220.0458
13 TeVCR ON21516438.874710.26480.26402.71040.14280.11220.00280.0570
13 TeVCR ON22311840.385210.40660.25772.68160.18640.14230.00330.0702
13 TeVCR ON23203542.212310.76120.25492.74330.23850.18250.00410.0893
13 TeVCR ON24114343.116410.54470.24462.57890.31190.24470.00540.1156

Table 1 (PDF p. 22).

EnergyCRmmNmN_mμm\mu_mσm\sigma_mCmC_mFmF_mSE(μ\mu)SE(σ\sigma)SE(CC)SE(FF)
13 TeVCR ON2564645.229110.88000.24062.61720.42810.30910.00660.1445
2.76 TeVCR OFF33149418.37574.08930.48821.99660.00730.00690.00080.0063
2.76 TeVCR OFF423449511.04684.76350.43122.05410.00980.00890.00080.0073
2.76 TeVCR OFF518465413.74205.35730.38982.08850.01250.00990.00070.0073
2.76 TeVCR OFF614899216.48805.89410.35752.10700.01530.01180.00070.0081
2.76 TeVCR OFF711947819.17726.35370.33132.10510.01840.01430.00070.0091
2.76 TeVCR OFF89431821.93156.78140.30922.09680.02210.01680.00070.0100
2.76 TeVCR OFF97089924.68237.21960.29252.11180.02710.02560.00100.0146
2.76 TeVCR OFF105187627.38187.59340.27732.10580.03330.02550.00090.0137
2.76 TeVCR OFF113541230.02067.93760.26442.09880.04220.03130.00100.0161
2.76 TeVCR OFF122333332.61428.28620.25412.10530.05420.04170.00120.0206
2.76 TeVCR OFF131444135.22068.58660.24382.09330.07150.05250.00150.0250
2.76 TeVCR OFF14839537.79368.99330.23802.14000.09820.07300.00190.0340
2.76 TeVCR OFF15478540.43629.12480.22572.05910.13190.09710.00230.0428
2.76 TeVCR OFF16249543.31469.51240.21962.08900.19040.13940.00320.0601
2.76 TeVCR OFF17121345.80879.62280.21012.02140.27630.19670.00430.0815
2.76 TeVCR OFF1855948.728110.52140.21592.27180.44500.35080.00700.1479
2.76 TeVCR ON33150947.15703.82480.53442.04410.00680.00620.00080.0060
2.76 TeVCR ON42354458.93364.34140.48602.10980.00890.00780.00080.0069
2.76 TeVCR ON518529510.61514.82620.45472.19420.01120.00960.00080.0080
2.76 TeVCR ON614969812.24405.21590.42602.22200.01350.01130.00080.0089
2.76 TeVCR ON711986513.76315.55360.40352.24100.01600.01350.00090.0101
2.76 TeVCR ON89407615.28935.86250.38342.24790.01910.01550.00090.0111
2.76 TeVCR ON97099416.73406.15960.36812.26730.02310.01860.00100.0128
2.76 TeVCR ON105079918.16786.45230.35522.29160.02860.02330.00120.0156
2.76 TeVCR ON113561619.44026.64300.34172.27000.03520.02820.00140.0182
2.76 TeVCR ON122307120.85616.93350.33242.30500.04560.03680.00170.0232
2.76 TeVCR ON131405222.25547.15600.32152.30100.06040.04790.00200.0293
2.76 TeVCR ON14823923.47717.30730.31132.27440.08050.06170.00250.0367
2.76 TeVCR ON15448824.76407.40590.29912.21480.11050.08800.00340.0505
2.76 TeVCR ON16251425.90497.52020.29032.18310.15000.11300.00410.0626
2.76 TeVCR ON17118227.17607.76360.28572.21790.22580.15240.00550.0844
2.76 TeVCR ON1856628.43827.97160.28032.23460.33510.25020.00860.1362
7 TeVCR OFF32904789.10084.44680.48862.17280.00830.00780.00080.0071
7 TeVCR OFF421750812.03465.19340.43152.24110.01110.00940.00070.0076
7 TeVCR OFF517508115.05395.84620.38842.27040.01400.01140.00070.0084
7 TeVCR OFF614529618.08816.45220.35672.30150.01690.01440.00080.0098
7 TeVCR OFF712392521.18746.99950.33042.31240.01990.02140.00100.0138
7 TeVCR OFF810675824.26057.47640.30822.30400.02290.01780.00070.0106
7 TeVCR OFF99163827.39167.96320.29072.31500.02630.02040.00070.0114
7 TeVCR OFF107776830.53508.38790.27472.30410.03010.02290.00070.0122
7 TeVCR OFF116429833.59748.79640.26182.30300.03470.02980.00090.0152
7 TeVCR OFF125081636.76019.16000.24922.28250.04060.03040.00080.0148
7 TeVCR OFF133946839.86379.53480.23922.28060.04800.03510.00090.0164
7 TeVCR OFF142941342.96949.99040.23252.32280.05830.04330.00100.0197
7 TeVCR OFF152082345.996910.16200.22092.24510.07040.05180.00110.0224
7 TeVCR OFF161396049.175610.65050.21662.30670.09010.06750.00130.0285
7 TeVCR OFF17930552.324310.98440.20992.30510.11390.08390.00160.0346
7 TeVCR OFF18573455.252911.18150.20242.26280.14770.10980.00190.0436
7 TeVCR OFF19345957.939311.35110.19592.22390.19300.14000.00240.0537
7 TeVCR OFF20193360.991711.79750.19342.28200.26830.19600.00320.0751

Table 1 (PDF p. 23).

EnergyCRmmNmN_mμm\mu_mσm\sigma_mCmC_mFmF_mSE(μ\mu)SE(σ\sigma)SE(CC)SE(FF)
7 TeVCR OFF21103563.960411.95820.18702.23570.37170.26790.00410.0983
7 TeVCR OFF2256567.437211.90070.17652.10010.50070.35060.00530.1234
7 TeVCR ON32898527.71434.16330.53972.24680.00770.00730.00080.0070
7 TeVCR ON42178039.67504.77000.49302.35170.01020.00930.00090.0083
7 TeVCR ON517481211.51555.28070.45862.42160.01260.01130.00090.0095
7 TeVCR ON614577013.34805.74730.43062.47460.01510.01300.00090.0103
7 TeVCR ON712384815.11226.15870.40752.50980.01750.01460.00090.0110
7 TeVCR ON810637916.83766.56890.39012.56270.02010.01660.00090.0121
7 TeVCR ON99167718.54886.90040.37202.56710.02280.01890.00090.0131
7 TeVCR ON107749720.22567.21720.35682.57540.02590.02100.00100.0140
7 TeVCR ON116401521.84767.47550.34222.55780.02950.02360.00100.0152
7 TeVCR ON125129623.39877.72670.33022.55150.03410.02690.00110.0168
7 TeVCR ON133945725.06218.01760.31992.56490.04040.03250.00120.0197
7 TeVCR ON142908826.37688.15270.30912.51990.04780.03820.00140.0225
7 TeVCR ON152067128.01118.41830.30052.53000.05860.04580.00150.0262
7 TeVCR ON161394129.33198.66730.29552.56110.07340.05760.00190.0325
7 TeVCR ON17924530.88578.75450.28342.48140.09100.06910.00210.0376
7 TeVCR ON18574232.19968.95990.27832.49320.11820.09510.00280.0508
7 TeVCR ON19343233.40129.21650.27592.54310.15730.12070.00340.0639
7 TeVCR ON20202734.98179.27300.26512.45810.20600.15070.00420.0775
7 TeVCR ON21104435.60739.26040.26012.40840.28660.23050.00600.1140
7 TeVCR ON2257337.93379.33430.24612.29690.38990.29220.00730.1384

Table 9.

Selection / fitCR ONCR OFF
Event fit: all selected nMPIn_{\mathrm{MPI}}1.8583.168
Event fit: nMPI>2n_{\mathrm{MPI}>2}1.7923.301
Relative change [%]−3.5-3.5+4.2+4.2
Profile fit: retained bins1.7943.301
Profile fit: up to 90th percentile1.8593.280
Change magnitude [%]3.60.6

Table 8. 13 TeV Monash slope sensitivity. Event-level fits and conditional-profile fits are different estimators and are compared only within the same row pair.

Complete selection and fit-range sensitivity table

Dashes indicate quantities not applicable to profile-only fits.

NNEnergy / CRmeanwidthCglobC_{\mathrm{glob}}rrslopeselection
33095682.76 TeV / CR ON6.92243.79180.54780.77541.6568INEL>0>0; all nMPIn_{\mathrm{MPI}}
17823132.76 TeV / CR ON9.95174.63020.46530.70031.5900INEL>0>0; nMPI>1n_{\mathrm{MPI}>1}
13113992.76 TeV / CR ON11.64695.05680.43420.63541.5326INEL>0>0; nMPI>2n_{\mathrm{MPI}>2}
13088202.76 TeV / CR ON11.66985.04930.43270.63491.5278Nch>0N_{\mathrm{ch}}>0 and nMPI>2n_{\mathrm{MPI}>2}
13109942.76 TeV / CR ON————1.5342profile bins: all with >=500>=500 events
12212662.76 TeV / CR ON————1.5979profile bins: nMPI≤90n_{\mathrm{MPI}}\mathrel{\leq}90th percentile
13022442.76 TeV / CR ON————1.5477profile bins: nMPI≤99n_{\mathrm{MPI}}\mathrel{\leq}99th percentile
33076982.76 TeV / CR OFF8.74084.26260.48770.86782.6341INEL>0>0; all nMPIn_{\mathrm{MPI}}
17826532.76 TeV / CR OFF13.21225.20960.39430.82922.6883INEL>0>0; nMPI>1n_{\mathrm{MPI}>1}
13107442.76 TeV / CR OFF15.87705.74970.36210.78872.7005INEL>0>0; nMPI>2n_{\mathrm{MPI}>2}
13094462.76 TeV / CR OFF15.89275.74550.36150.78862.6975Nch>0N_{\mathrm{ch}}>0 and nMPI>2n_{\mathrm{MPI}>2}
13102862.76 TeV / CR OFF————2.7016profile bins: all with >=500>=500 events
12196532.76 TeV / CR OFF————2.7152profile bins: nMPI≤90n_{\mathrm{MPI}}\mathrel{\leq}90th percentile
13012342.76 TeV / CR OFF————2.7046profile bins: nMPI≤99n_{\mathrm{MPI}}\mathrel{\leq}99th percentile
33374787 TeV / CR ON8.73764.61980.52870.81651.7849INEL>0>0; all nMPIn_{\mathrm{MPI}}
19004477 TeV / CR ON12.48775.51960.44200.76361.7422INEL>0>0; nMPI>1n_{\mathrm{MPI}>1}
14686907 TeV / CR ON14.51055.99660.41330.71381.6974INEL>0>0; nMPI>2n_{\mathrm{MPI}>2}
14666847 TeV / CR ON14.53035.99220.41240.71351.6944Nch>0N_{\mathrm{ch}}>0 and nMPI>2n_{\mathrm{MPI}>2}
14681697 TeV / CR ON————1.6990profile bins: all with >=500>=500 events
13429497 TeV / CR ON————1.7669profile bins: nMPI≤90n_{\mathrm{MPI}}\mathrel{\leq}90th percentile
14553517 TeV / CR ON————1.7156profile bins: nMPI≤99n_{\mathrm{MPI}}\mathrel{\leq}99th percentile
33376037 TeV / CR OFF11.65115.31400.45610.89832.9679INEL>0>0; all nMPIn_{\mathrm{MPI}}
19016707 TeV / CR OFF17.40636.26870.36010.87763.0562INEL>0>0; nMPI>1n_{\mathrm{MPI}>1}
14697677 TeV / CR OFF20.68246.86810.33210.85043.0781INEL>0>0; nMPI>2n_{\mathrm{MPI}>2}
14687487 TeV / CR OFF20.69676.86560.33170.85033.0763Nch>0N_{\mathrm{ch}}>0 and nMPI>2n_{\mathrm{MPI}>2}
14692617 TeV / CR OFF————3.0785profile bins: all with >=500>=500 events
13435667 TeV / CR OFF————3.0670profile bins: nMPI≤90n_{\mathrm{MPI}}\mathrel{\leq}90th percentile
14565357 TeV / CR OFF————3.0796profile bins: nMPI≤99n_{\mathrm{MPI}}\mathrel{\leq}99th percentile
335507513 TeV / CR ON10.24815.29350.51650.83771.8576INEL>0>0; all nMPIn_{\mathrm{MPI}}
197515713 TeV / CR ON14.53916.21450.42740.79751.8294INEL>0>0; nMPI>1n_{\mathrm{MPI}>1}

Table 10.

NNEnergy / CRmeanwidthCglobC_{\mathrm{glob}}rrslopeselection
156611913 TeV / CR ON16.80366.72640.40030.75691.7924INEL>0; nMPI>2n_{\mathrm{MPI}}>2
156430013 TeV / CR ON16.82326.72320.39960.75671.7901Nch>0 and nMPI>2n_{\mathrm{MPI}}>2
156535913 TeV / CR ON————1.7945profile bins: all with ≥500\geq500 events
142859713 TeV / CR ON————1.8591profile bins: nMPI≤90n_{\mathrm{MPI}}\leq90th percentile
155325313 TeV / CR ON————1.8088profile bins: nMPI≤99n_{\mathrm{MPI}}\leq99th percentile
335537413 TeV / CR OFF14.13796.19700.43830.91223.1680INEL>0; all nMPIn_{\mathrm{MPI}}
197661113 TeV / CR OFF20.87057.09860.34010.90073.2755INEL>0; nMPI>1n_{\mathrm{MPI}}>1
156747013 TeV / CR OFF24.61177.74260.31460.88013.3014INEL>0; nMPI>2n_{\mathrm{MPI}}>2
156661113 TeV / CR OFF24.62527.74090.31430.88003.3001Nch>0 and nMPI>2n_{\mathrm{MPI}}>2
156678313 TeV / CR OFF————3.3010profile bins: all with ≥500\geq500 events
143020413 TeV / CR OFF————3.2800profile bins: nMPI≤90n_{\mathrm{MPI}}\leq90th percentile
155465313 TeV / CR OFF————3.2996profile bins: nMPI≤99n_{\mathrm{MPI}}\leq99th percentile

Table 12.

Complete tune-normalized profiles

This table reproduces every row of fresh_tune_normalized_fluctuation_profiles.csv, with at least two events per conditional bin. Singleton bins remain in the comprehensive machine-readable results but are excluded here because they cannot constrain a width uncertainty. The manuscript summary applies Nm≥500N_m \ge500 and m>2m > 2.

mmμm\mu_mσm\sigma_mNmN_msampleCmC_mFmF_m
00.36360.481022ATLAS A14 CR OFF1.32290.6364
14.74145.00761362354ATLAS A14 CR OFF1.05615.2888
26.75163.7603397605ATLAS A14 CR OFF0.55692.0943
39.91154.7845268435ATLAS A14 CR OFF0.48272.3096
413.14715.5973201978ATLAS A14 CR OFF0.42572.3830
516.46856.3446162575ATLAS A14 CR OFF0.38532.4443
619.85406.9385137401ATLAS A14 CR OFF0.34952.4248
723.34697.5743118609ATLAS A14 CR OFF0.32442.4573
826.79758.1323104075ATLAS A14 CR OFF0.30352.4679
930.40198.723892729ATLAS A14 CR OFF0.28702.5033
1033.98589.157183047ATLAS A14 CR OFF0.26942.4673
1137.63659.597574497ATLAS A14 CR OFF0.25502.4474
1241.257110.037966661ATLAS A14 CR OFF0.24332.4422
1344.935510.471659199ATLAS A14 CR OFF0.23302.4403
1448.528510.823551527ATLAS A14 CR OFF0.22302.4140
1552.267211.394743068ATLAS A14 CR OFF0.21802.4842
1655.752311.630635555ATLAS A14 CR OFF0.20862.4263
1759.263912.020828682ATLAS A14 CR OFF0.20282.4382
1862.958812.255622079ATLAS A14 CR OFF0.19472.3857
1966.560812.672616656ATLAS A14 CR OFF0.19042.4128
2070.296012.907412014ATLAS A14 CR OFF0.18362.3700
2173.840712.95418124ATLAS A14 CR OFF0.17542.2726
2277.313213.45705674ATLAS A14 CR OFF0.17412.3423
2380.575713.88393670ATLAS A14 CR OFF0.17232.3923
2484.364414.02872294ATLAS A14 CR OFF0.16632.3328
2587.475614.01621352ATLAS A14 CR OFF0.16022.2458
2692.594914.9668738ATLAS A14 CR OFF0.16162.4192
2794.862213.4305479ATLAS A14 CR OFF0.14161.9015
2898.021714.0274230ATLAS A14 CR OFF0.14312.0074
29101.015015.9019133ATLAS A14 CR OFF0.15742.5033
30105.938515.203165ATLAS A14 CR OFF0.14352.1818
31109.156316.965232ATLAS A14 CR OFF0.15542.6368
32108.941213.832917ATLAS A14 CR OFF0.12701.7564
33114.166724.68076ATLAS A14 CR OFF0.21625.3355
34117.600014.94795ATLAS A14 CR OFF0.12711.9000
37130.50009.50002ATLAS A14 CR OFF0.07280.6916
00.41180.492217ATLAS A14 CR ON1.19520.5882
14.30733.80041364304ATLAS A14 CR ON0.88233.3532
26.04623.6895398860ATLAS A14 CR ON0.61022.2515
38.34834.4901267771ATLAS A14 CR ON0.53782.4150
410.51025.1448202108ATLAS A14 CR ON0.48952.5185
512.57145.7271162705ATLAS A14 CR ON0.45562.6091
614.49226.2164137101ATLAS A14 CR ON0.42892.6665
mmμm\mu_mσm\sigma_mNmN_msampleCmC_mFmF_m
716.50056.7134118606ATLAS A14 CR ON0.40692.7314
818.37437.0844104537ATLAS A14 CR ON0.38562.7315
920.34227.581692727ATLAS A14 CR ON0.37272.8257
1022.27947.917583128ATLAS A14 CR ON0.35542.8137
1124.17008.345875304ATLAS A14 CR ON0.34532.8818
1226.12738.658966843ATLAS A14 CR ON0.33142.8697
1327.92588.945958653ATLAS A14 CR ON0.32032.8658
1429.74129.241350886ATLAS A14 CR ON0.31072.8715
1531.54839.459743074ATLAS A14 CR ON0.29982.8365
1633.17949.656535155ATLAS A14 CR ON0.29102.8104
1734.95279.941028322ATLAS A14 CR ON0.28442.8274
1836.527710.182721746ATLAS A14 CR ON0.27882.8386
1938.371910.283316517ATLAS A14 CR ON0.26802.7558
2039.957710.615711879ATLAS A14 CR ON0.26572.8203
2141.359010.86308134ATLAS A14 CR ON0.26272.8532
2242.796910.72615509ATLAS A14 CR ON0.25062.6883
2344.567111.27633449ATLAS A14 CR ON0.25302.8531
2445.609911.18232253ATLAS A14 CR ON0.24522.7416
2547.901811.65691415ATLAS A14 CR ON0.24332.8367
2648.557710.8293832ATLAS A14 CR ON0.22302.4151
2750.991211.8177457ATLAS A14 CR ON0.23182.7389
2853.214012.4844285ATLAS A14 CR ON0.23462.9289
2954.479711.6952148ATLAS A14 CR ON0.21472.5106
3055.475411.641761ATLAS A14 CR ON0.20992.4430
3153.055611.484236ATLAS A14 CR ON0.21652.4858
3257.153812.954613ATLAS A14 CR ON0.22672.9363
3359.166711.581812ATLAS A14 CR ON0.19572.2671
3466.25009.67924ATLAS A14 CR ON0.14611.4142
01.45271.21669516CMS CP5 CR OFF0.83751.0189
14.35434.95831283769CMS CP5 CR OFF1.13875.6460
25.42153.3545331346CMS CP5 CR OFF0.61872.0755
37.63344.2164221216CMS CP5 CR OFF0.55242.3289
49.88754.8934164980CMS CP5 CR OFF0.49492.4218
512.18565.5074131473CMS CP5 CR OFF0.45202.4892
614.54286.0583109599CMS CP5 CR OFF0.41662.5238
716.90266.576293355CMS CP5 CR OFF0.38912.5585
819.21457.063682007CMS CP5 CR OFF0.36762.5967
921.61477.529073221CMS CP5 CR OFF0.34832.6226
1023.95707.971566635CMS CP5 CR OFF0.33272.6525
1126.40918.369460507CMS CP5 CR OFF0.31692.6524
1228.88488.813955854CMS CP5 CR OFF0.30512.6894
1331.25509.258852000CMS CP5 CR OFF0.29622.7428
1433.81119.554248566CMS CP5 CR OFF0.28262.6998
1536.26779.897745886CMS CP5 CR OFF0.27292.7011
1638.752010.264643688CMS CP5 CR OFF0.26492.7189
1741.172710.605841252CMS CP5 CR OFF0.25762.7320
1843.722210.996039223CMS CP5 CR OFF0.25152.7655
1946.148511.331437254CMS CP5 CR OFF0.24512.7725
2048.694311.591335461CMS CP5 CR OFF0.23802.7592
2151.294111.924133911CMS CP5 CR OFF0.23252.7719
2253.828012.245032033CMS CP5 CR OFF0.22752.7855

Table 1 (PDF p. 27).

mmμm\mu_mσm\sigma_mNmN_msampleCmC_mFmF_m
2356.317912.550630082CMS CP5 CR OFF0.22292.7969
2458.664312.658528027CMS CP5 CR OFF0.21582.7314
2561.226313.030126270CMS CP5 CR OFF0.21282.7730
2663.567913.332924286CMS CP5 CR OFF0.20972.7965
2765.933913.534022149CMS CP5 CR OFF0.20532.7781
2868.448913.863520242CMS CP5 CR OFF0.20252.8079
2970.917113.955217742CMS CP5 CR OFF0.19682.7461
3072.913014.122015830CMS CP5 CR OFF0.19372.7352
3175.306714.410713596CMS CP5 CR OFF0.19142.7576
3277.363614.533211769CMS CP5 CR OFF0.18792.7301
3380.008714.76999755CMS CP5 CR OFF0.18462.7266
3482.142214.85548361CMS CP5 CR OFF0.18092.6866
3584.285615.11146813CMS CP5 CR OFF0.17932.7093
3686.769615.10235472CMS CP5 CR OFF0.17412.6286
3789.037315.68314419CMS CP5 CR OFF0.17612.7624
3890.986315.68523362CMS CP5 CR OFF0.17242.7040
3992.762315.86372684CMS CP5 CR OFF0.17102.7129
4094.921716.36982004CMS CP5 CR OFF0.17252.8231
4196.315115.99631466CMS CP5 CR OFF0.16612.6567
42100.375917.31881056CMS CP5 CR OFF0.17252.9882
43101.324515.9759826CMS CP5 CR OFF0.15772.5189
44103.331616.9147570CMS CP5 CR OFF0.16372.7688
45104.771617.3394429CMS CP5 CR OFF0.16552.8696
46107.537516.2885253CMS CP5 CR OFF0.15152.4672
47109.835016.7961206CMS CP5 CR OFF0.15292.5685
48114.864916.7861111CMS CP5 CR OFF0.14612.4531
49115.322616.544693CMS CP5 CR OFF0.14352.3735
50120.298215.841357CMS CP5 CR OFF0.13172.0860
51116.794915.352639CMS CP5 CR OFF0.13142.0181
52121.357116.782714CMS CP5 CR OFF0.13832.3209
53113.058818.193317CMS CP5 CR OFF0.16092.9276
54132.857125.75957CMS CP5 CR OFF0.19394.9945
55138.000012.35583CMS CP5 CR OFF0.08951.1063
56126.333315.41283CMS CP5 CR OFF0.12201.8804
01.47261.23049612CMS CP5 CR ON0.83551.0280
13.84993.48001283524CMS CP5 CR ON0.90393.1456
24.75383.1354332004CMS CP5 CR ON0.65962.0680
36.29283.7647221434CMS CP5 CR ON0.59832.2522
47.67914.2504165316CMS CP5 CR ON0.55352.3526
58.95834.6672131559CMS CP5 CR ON0.52102.4316
610.16475.0466109656CMS CP5 CR ON0.49652.5055
711.38255.381293807CMS CP5 CR ON0.47282.5440
812.51955.697182318CMS CP5 CR ON0.45512.5925
913.71586.022473114CMS CP5 CR ON0.43912.6444
1014.78826.263566808CMS CP5 CR ON0.42352.6529
1115.98166.536560409CMS CP5 CR ON0.40902.6734
1217.11696.834756014CMS CP5 CR ON0.39932.7291
1318.22117.073252413CMS CP5 CR ON0.38822.7457
1419.43387.336148787CMS CP5 CR ON0.37752.7693
1520.54757.548045808CMS CP5 CR ON0.36732.7727
1621.73527.835743782CMS CP5 CR ON0.36052.8248
mmμm\mu_mσm\sigma_mNmN_msampleCmC_mFmF_m
1722.84648.057941494CMS CP5 CR ON0.35272.8420
1824.02448.243639251CMS CP5 CR ON0.34312.8286
1925.13228.483837319CMS CP5 CR ON0.33762.8638
2026.30128.607035271CMS CP5 CR ON0.32722.8166
2127.43528.877233631CMS CP5 CR ON0.32362.8724
2228.41358.947131919CMS CP5 CR ON0.31492.8174
2329.56949.093429574CMS CP5 CR ON0.30752.7965
2430.73549.357828327CMS CP5 CR ON0.30452.8491
2531.78279.534026261CMS CP5 CR ON0.30002.8600
2632.77799.619524236CMS CP5 CR ON0.29352.8231
2733.82979.645822216CMS CP5 CR ON0.28512.7503
2834.78269.834720090CMS CP5 CR ON0.28272.7807
2935.88549.986017962CMS CP5 CR ON0.27832.7788
3036.72859.975615847CMS CP5 CR ON0.27162.7094
3137.719310.097713799CMS CP5 CR ON0.26772.7032
3238.597910.302111735CMS CP5 CR ON0.26692.7497
3339.476110.33049800CMS CP5 CR ON0.26172.7033
3440.318910.51668329CMS CP5 CR ON0.26082.7431
3541.239910.53186828CMS CP5 CR ON0.25542.6896
3642.085810.57905410CMS CP5 CR ON0.25142.6592
3742.923410.68664306CMS CP5 CR ON0.24902.6606
3843.772910.94113439CMS CP5 CR ON0.25002.7348
3944.221510.68642641CMS CP5 CR ON0.24172.5824
4045.667610.84631880CMS CP5 CR ON0.23752.5761
4146.505910.44701431CMS CP5 CR ON0.22462.3468
4247.353111.34601031CMS CP5 CR ON0.23962.7185
4347.334611.1628795CMS CP5 CR ON0.23582.6325
4448.304011.0955546CMS CP5 CR ON0.22972.5486
4549.089011.2416382CMS CP5 CR ON0.22902.5744
4649.389110.9993257CMS CP5 CR ON0.22272.4496
4751.398812.3418173CMS CP5 CR ON0.24012.9635
4850.900010.4326120CMS CP5 CR ON0.20502.1383
4952.000011.647276CMS CP5 CR ON0.22402.6088
5054.244911.821049CMS CP5 CR ON0.21792.5760
5155.192314.390226CMS CP5 CR ON0.26073.7519
5258.550012.698320CMS CP5 CR ON0.21692.7540
5352.44449.713818CMS CP5 CR ON0.18521.7992
5460.363613.532911CMS CP5 CR ON0.22423.0340
5550.80007.88425CMS CP5 CR ON0.15521.2236
00.66670.471415Monash CR OFF0.70710.3333
14.82834.97111388633Monash CR OFF1.02965.1181
27.08603.9887410363Monash CR OFF0.56292.2453
310.36005.1010275713Monash CR OFF0.49242.5116
413.70815.8981207062Monash CR OFF0.43032.5378
517.14496.6448167031Monash CR OFF0.38762.5753
620.61037.3024139716Monash CR OFF0.35432.5873
724.12337.9099120237Monash CR OFF0.32792.5936
827.72488.5502105743Monash CR OFF0.30842.6368
931.33649.067893594Monash CR OFF0.28942.6240
1034.90479.537082401Monash CR OFF0.27322.6058
1138.600610.003373523Monash CR OFF0.25912.5924
mmμm\mu_mσm\sigma_mNmN_msampleCmC_mFmF_m
1242.390310.496564069Monash CR OFF0.24762.5991
1346.034910.992855159Monash CR OFF0.23882.6250
1449.622411.368545900Monash CR OFF0.22912.6045
1553.284711.779237199Monash CR OFF0.22112.6039
1656.951612.104429363Monash CR OFF0.21252.5726
1760.425512.487422257Monash CR OFF0.20672.5806
1864.043712.942416398Monash CR OFF0.20212.6155
1967.680513.073911677Monash CR OFF0.19322.5255
2071.440413.59347693Monash CR OFF0.19032.5865
2175.065414.33755017Monash CR OFF0.19102.7385
2278.658113.97313182Monash CR OFF0.17762.4822
2381.565714.22961909Monash CR OFF0.17452.4824
2485.869914.18631268Monash CR OFF0.16522.3437
2589.311315.0059652Monash CR OFF0.16802.5213
2690.705715.0222384Monash CR OFF0.16562.4879
2795.833315.5103204Monash CR OFF0.16182.5103
2898.372517.0765102Monash CR OFF0.17362.9643
29107.816317.609949Monash CR OFF0.16332.8763
30110.480014.269225Monash CR OFF0.12921.8430
31108.538521.695913Monash CR OFF0.19994.3368
32112.50008.64584Monash CR OFF0.07690.6644
34129.000018.00002Monash CR OFF0.13952.5116
00.21740.412523Monash CR ON1.89740.7826
14.41963.86221388156Monash CR ON0.87393.3751
26.35173.9013409118Monash CR ON0.61422.3962
38.77374.7442275233Monash CR ON0.54072.5653
411.02275.4301206553Monash CR ON0.49262.6751
513.18726.0363167209Monash CR ON0.45772.7631
615.23236.5691140141Monash CR ON0.43132.8330
717.30857.0746120188Monash CR ON0.40872.8916
819.37937.5169105786Monash CR ON0.38792.9156
921.32117.898993406Monash CR ON0.37052.9264
1023.37478.321883218Monash CR ON0.35602.9627
1125.38528.658873376Monash CR ON0.34112.9535
1227.21519.030764322Monash CR ON0.33182.9966
1329.11889.279055759Monash CR ON0.31872.9568
1430.98579.579445849Monash CR ON0.30922.9615
1532.79189.800337436Monash CR ON0.29892.9289
1634.531110.056729539Monash CR ON0.29122.9289
1736.216610.258622488Monash CR ON0.28332.9058
1837.994110.637816298Monash CR ON0.28002.9784
1939.484910.689011523Monash CR ON0.27072.8936
2040.927910.75927743Monash CR ON0.26292.8284
2142.556411.19185189Monash CR ON0.26302.9433
2243.995011.19933172Monash CR ON0.25462.8509
2346.002511.70031970Monash CR ON0.25432.9758
2447.634111.40461104Monash CR ON0.23942.7305
2548.872011.4472633Monash CR ON0.23422.6812
2649.369911.3347346Monash CR ON0.22962.6023
2751.989412.0510189Monash CR ON0.23182.7934
2855.623912.9395117Monash CR ON0.23263.0101
mmμm\mu_mσm\sigma_mNmN_msampleCmC_mFmF_m
2953.333312.650945Monash CR ON0.23723.0008
3055.312514.171832Monash CR ON0.25623.6310
3155.090911.106511Monash CR ON0.20162.2391
3247.400010.19025Monash CR ON0.21502.1907
3358.00009.00002Monash CR ON0.15521.3966

Table 2 (PDF p. 29).

Files and provenance

The principal calculation is archived at submission_tools/reproduce_conditional_results.py. Current generation sources are in generation/; original CP5 workflow programs are retained as historical provenance in archived_analysis_code/cp5_historical_workflow/. Production-time settings and particle data are in cp5_exact/production_settings/; sixteen independent seeds and available logs are in production_records/. Complete event-summary files remain externally retained and identified by source hashes; lossless joint-count inputs for the conditional analysis are supplied in reproduction_data/. Both tune analyses use the same fresh CP5 samples. The external PDF is NNPDF31_nnlo_as_0118 member 0, accessed with LHAPDF 6.5.4 in PYTHIA 8.316.

The exported numerical files are in new_analysis_editor_response/. The archive also contains historical acceptance, ALICE comparison, activity-estimator, tune-generation, tune-analysis, and output-assembly programs. Historical workflows are provenance records rather than supported execution instructions; current portable commands are listed in the S2 README. Large generated event arrays are available from the corresponding author upon reasonable request and are not embedded in this PDF. Supplementary Material S2 also includes the inclusive and estimator plotting programs, their numerical inputs, and a script for reproducing the revised estimator figures without the large event samples. The Supplementary Material S1 and manuscript use identical definitions; the explicit primary-particle limitation remains unresolved by the available records.

SettingMonashA14CMS CP5
Tune:pp (starting point)142114 + overrides
PDF:pSet1313LHAPDF6 (member 0)
PDF familyNNPDF2.3 LO QCD+QEDNNPDF2.3 LO QCD+QEDNNPDF3.1 NNLO
MPI pT0refp_{\mathrm{T0}}^{\mathrm{ref}} [GeV]2.282.091.41
MPI reference energy [GeV]700070007000
MPI energy exponent0.2150.2150.03344
MultipartonInteractions:bProfile332
Overlap exponent (profile 3)1.851.85–
Core radius / fraction (profile 2)––0.7634 / 0.63
ColourReconnection:mode000
ColourReconnection:range1.801.715.176
αs(MZ)\alpha_{\mathrm{s}}(M_{\mathrm{Z}}): hard process / running order0.130 / 10.140 / 10.118 / 2
αs(MZ)\alpha_{\mathrm{s}}(M_{\mathrm{Z}}): ISR / running order0.1365 / 10.127 / 10.118 / 2
αs(MZ)\alpha_{\mathrm{s}}(M_{\mathrm{Z}}): FSR / running order0.1365 / 10.127 / 10.118 / 2
αs(MZ)\alpha_{\mathrm{s}}(M_{\mathrm{Z}}): MPI / running order0.130 / 10.126 / 10.118 / 2
SpaceShower:rapidityOrderononon
SigmaDiffractive:mode111
SigmaTotal:zeroAXBononoff
SigmaTotal:mode110
SigmaTotal:sigmaTot [mb]100100100.309
SigmaTotal:sigmaEl [mb]252521.89

Table 11. Settings of the independently generated 13 TeV configuration scan in PYTHIA 8.316. Monash and A14 values are reconstructed from archived source; CP5 values are saved during the fresh production. All configurations use SoftQCD:all = on; only the physics setting ColourReconnection:reconnect changes between ON and OFF; distinct fixed seeds are documented for the initial Monash/A14 commands and all corrected CP5 segments; complete historical top-up seed linkage is not certified. The baseline is associated with a Monash source family, but no complete per-run production configuration is certified. Candidate sources include lifetime and species decay overrides, and actual historical seeds are unavailable in retained records; these candidates must not be presented as verified effective settings.

Complete tune settings and production details

The manuscript retains the key tune settings and a short explanation of their physical roles. This section records the complete comparison table and implementation details moved out of the main prose. They support reproducibility without changing the analysis.

The CP5 PDF is LHAPDF6:NNPDF31_nnlo_as_0118, member 0, accessed with LHAPDF 6.5.4. The four coupling sectors use two-loop running with αs(MZ)=0.118\alpha_{\mathrm{s}}(M_{\mathrm{Z}}) = 0.118. Central diffraction is retained with SigmaTotal:zeroAXB = off. The full-precision parameters are pT0ref=1.41p_{\mathrm{T0}}^{\mathrm{ref}} = 1.41 GeV, energy exponent 0.03344, core radius 0.7634, core fraction 0.63 and CR range 5.176. The total-cross-section settings are SigmaTotal:mode = 0, σtot=100.309\sigma_{\mathrm{tot}} = 100.309 mb and σel=21.89\sigma_{\mathrm{el}} = 21.89 mb. The published settings are implemented in PYTHIA 8.316; the original CMS PYTHIA 8.226 environment is not reproduced.

The MPI regularization and energy scaling follow the PYTHIA 8.3 guide by Bierlich et al. (cited in the main manuscript):

dσdpT2∝αs2(pT2+pT02)(pT2+pT02)2,pT0=pT0ref(sEref)ϵ.(30)\frac{\mathrm{d}\sigma}{\mathrm{d}p_{\mathrm{T}}^{2}} \propto\frac{\alpha_{\mathrm{s}}^{2}(p_{\mathrm{T}}^{2}+p_{\mathrm{T0}}^{2})}{(p_{\mathrm{T}}^{2}+p_{\mathrm{T0}}^{2})^{2}}, \qquad p_{\mathrm{T0}} = p_{\mathrm{T0}}^{\mathrm{ref}}\left(\frac{\sqrt{s}}{E_{\mathrm{ref}}}\right)^{\epsilon}. \tag*{(30)}

The reference scale regulates the low-pTp_{\mathrm{T}} interaction rate, while ϵ\epsilon controls its energy dependence.

For profile 3, the overlap has the form O(b)∝exp⁡[−(b/b0)p]O(b) \propto\exp[-(b/b_{0})^{p}], with p=expPowp = \texttt{expPow}. Profile 2 is a double-Gaussian matter distribution whose narrow component is set by coreRadius and coreFraction. In CR mode 0, a softer system of scale pTp_{\mathrm{T}} reconnects with probability pT,rec2/(pT,rec2+pT2)p_{\mathrm{T,rec}}^{2}/(p_{\mathrm{T,rec}}^{2}+p_{\mathrm{T}}^{2}), where pT,rec=RpT0p_{\mathrm{T,rec}} = R p_{\mathrm{T0}} and R=ColourReconnection:rangeR = \texttt{ColourReconnection:range}. The m0m_{0} and time-dilation parameters of mode 1 do not enter these mode-0 comparisons.

The CP5 pair uses the same default PYTHIA decay prescription without particle-specific overrides. Eight independent fixed seeds contribute five million events to each CR state: four verified logged prefixes from an interrupted production plus four new independent-seed runs with unchanged physics settings. Recovered prefixes lack final end-of-run warning summaries; checkpoint validation and complete top-up logs are retained. Per-run counts, seeds, settings, particle data and source records are supplied with the analysis material. The same CP5 events support both fluctuation and estimator analyses. Monash and A14 settings are reconstructed from archived source; the historical six-baseline decay settings and seeds remain subject to production-record confirmation.

Production provenance and independent reproducibility

The twelve central analysis samples are distinguished in S2’s production_records/sample_provenance_matrix.csv. All six original baseline samples retain five million metadata records and their particle chunks. Candidate generation sources include ParticleDecays:limitTau0 = on, tau0Max = 10.0, and decay overrides for species 333, 310, 3122, 3312 and 3334 and the specified antiparticles. The source-to-run link, production executable, complete effective settings, actual time-derived seeds and full historical logs have not been recovered as auditable per-run records. These candidates are supplied as unverified evidence, rather than substituted for missing production records. Thus exact historical generator replay is not established. The retained-event selection and conditional numerical analysis can nevertheless be independently checked, and the supplied joint counts reproduce the conditional moments and uncertainties without event-order information. This distinction does not establish that a newly generated default-decay sample is equivalent to the historical baseline.

Monash and A14 tune configurations are reconstructed from archived source; available shell commands and logs are supplied, but complete seed-to-segment linkage for all historical top-ups is not certified. Corrected CP5 has complete per-segment settings and particle dumps, sixteen fixed seeds, checked event-file hashes and available checkpoint/final logs. Interrupted prefixes lack final warning summaries and are identified as such. The retained CMS card, PDF metadata and central PDF grid accompany the production records. Fresh short generator validation runs are kept only in the author check directory; they are not production samples and do not establish missing historical provenance.

To reproduce the conditional results from the supplied integer joint counts, run pythonsubmission_tools/reproduce_conditional_results.py from S2. The outputs include central moments through fourth order, errors for every conditional mean/width/coefficient of variation/scaled variance, and every retained ON/OFF matched-bin comparison. Histogram aggregation preserves all information needed for these one-event conditional statistics, but does not preserve particle ancestry, generator replay, event ordering or time correlations.

Every retained matched bin: ratios and statistical uncertainties

The following tables use the common population-moment convention and empirical moments through fourth order. The quoted errors are marginal one-standard-error delta-method estimates; CR samples are treated as independent. N+N_{+} and N−N_{-} denote ON and OFF counts. All 148 central-value orderings satisfy Rμ<Rσ<1R_{\mu}<R_{\sigma}<1, hence RC>1R_{C}>1. The two signed differences, σ−−σ+\sigma_{-}-\sigma_{+} and C+−C−C_{+}-C_{-}, also pass a Bonferroni normal-approximation diagnostic across the 296 directional comparisons, with family error target 0.05. This is an asymptotic diagnostic, not an exact finite-sample guarantee. The machine-readable file additionally supplies conditional means, widths, coefficients of variation, their errors, ratio confidence intervals and central moments through fourth order.

Baseline: 13 TeV

mmN+N_{+}N−N_{-}RμR_{\mu}SERσR_{\sigma}SERCR_{C}SE
32745852751230.8462810.0011860.9424150.0023111.1135960.002428
42072042073810.8010950.0011650.9244340.0026511.1539630.003006
51665101665600.7658940.0011420.9006050.0065151.1758870.008271
61395401402980.7365940.0011110.8994770.0028771.2211300.003604
71200631197570.7130760.0010920.8857760.0030991.2421910.004038
81059781059230.6925690.0010680.8785520.0032081.2685410.004333
993432935740.6791530.0010540.8693230.0032711.2800100.004516
mmN+N_{+}N−N_{-}RμR_{\mu}SERσR_{\sigma}SERCR_{C}SE
1082993830730.6633340.0010480.8624010.0033711.3001000.004795
1173050733960.6496650.0010480.8550420.0035571.3161270.005172
1264080641690.6380570.0010530.8470720.0037451.3275800.005580
1355221549770.6294790.0010810.8488110.0039511.3484330.005991
1445941459730.6186630.0011230.8391720.0042851.3564270.006602
1537186373890.6112260.0011950.8381260.0047821.3712220.007479
1629513292830.6019460.0012800.8227110.0053401.3667530.008525
1722215224290.5955810.0014210.8217770.0059941.3797900.009663
1816355161630.5867860.0015890.8092000.0067391.3790380.011070
1911620113960.5816060.0018260.8146840.0081671.4007480.013478
20776777890.5702280.0021440.7998380.0096991.4026630.016354
21516451680.5642420.0025350.8012730.0120071.4200880.020594
22311832130.5592590.0031550.7833530.0145191.4006980.025011
23203519950.5562140.0038430.7986150.0185411.4358060.032189
24114311440.5501600.0048490.7898470.0243901.4356680.043155
256466100.5542230.0065840.7519160.0321761.3567050.056256

Table 13.

Baseline: 2.76 TeV

mmN+N_{+}N−N_{-}RμR_{\mu}SERσR_{\sigma}SERCR_{C}SE
33150943149410.8544920.0011020.9353190.0021901.0945910.002335
42354452344950.8087000.0010840.9113990.0023641.1269930.002704
51852951846540.7724550.0010760.9008550.0024491.1662230.002938
61496981489920.7426020.0010680.8849450.0026161.1916810.003294
71198651194780.7176810.0010830.8740830.0028971.2179280.003791
894076943180.6971380.0011190.8645060.0031301.2400770.004258
970994708990.6779770.0011970.8531730.0039721.2584100.005594
1050799518760.6634980.0013210.8497300.0041971.2806820.005989
1135616354120.6475630.0014840.8368970.0048521.2923800.007154
1223071233330.6394780.0017580.8367500.0061181.3084890.009152
1314052144410.6318860.0021400.8334000.0075551.3189090.011477
14823983950.6211940.0026720.8125260.0095171.3080080.014841
15448847850.6124230.0033860.8116250.0129511.3252690.020390
16251424950.5980640.0043480.7905650.0165931.3218730.026778
17118212130.5932490.0060910.8067930.0228641.3599580.038136
185665590.5836090.0087000.7576600.0346941.2982310.057917

Table 14.

Baseline: 7 TeV

mmN+N_{+}N−N_{-}RμR_{\mu}SERσR_{\sigma}SERCR_{C}SE
32898522904780.8476560.0011460.9362430.0023261.1045080.002471
42178032175080.8039330.0011290.9184730.0024431.1424740.002765
51748121750810.7649530.0010990.9032640.0026201.1808090.003148
61457701452960.7379430.0010810.8907520.0028341.2070750.003567
71238481219250.7132620.0010630.8798690.0034061.2335850.004520
81063791067580.6940330.0010570.8786120.0030551.2659520.004138
991677916380.6771710.0010560.8665470.0032461.2796570.004520
1077497777680.6623740.0010710.8604370.0034281.2990190.004900
1164015642980.6502770.0011060.8498360.0039291.3068830.005757
1251296508160.6365240.0011650.8435310.0040581.3252150.006094
1339457394680.6286950.0012640.8408790.0046001.3374980.006992
1429088294130.6138510.0013890.8160530.0052101.3293980.008154
1520671202230.6089770.0015780.8284110.0061761.3603320.009722
1613941139600.5964720.0018500.8137920.0074691.3643420.012009
17924593050.5902730.0021630.7971320.0087591.3504450.014387
18574257340.5827680.0026470.8013130.0115881.3750130.019153
19343234590.5764870.0033260.8119400.0146061.4084290.024410
20202719330.5735490.0042160.7860160.0182641.3704430.031419
21104410350.5567080.0055270.7744010.0259371.3910360.044277
225735650.5625040.0071330.7843530.0337151.3943950.058748

Table 15.

Tune: ATLAS A14

mmN+N_{+}N−N_{-}RμR_{\mu}SERσR_{\sigma}SERCR_{C}SE
32677712684350.8422830.0011760.9384610.0032391.1141880.003591
mmN+N_{+}N−N_{-}RμR_{\mu}SERσR_{\sigma}SERCR_{C}SE
42021082019780.7994340.0011540.9191670.0025351.1497730.002890
51627051625750.7633590.0011290.9026800.0043181.1825110.005390
61371011374010.7299390.0010900.8959260.0028831.2273980.003664
71186061186090.7067530.0010680.8863320.0030001.2540900.003964
81045371040750.6856730.0010410.8711430.0035431.2704930.004893
992727927290.6691120.0010340.8690660.0056431.2988350.008139
1083128830470.6555500.0010140.8646310.0044171.3189400.006473
1175304744970.6421970.0010060.8695800.0035901.3540710.005303
1266843666610.6332810.0010080.8626240.0036811.3621510.005538
1358653591990.6214640.0010150.8542970.0039181.3746530.006006
1450886515270.6128610.0010370.8538160.0041481.3931650.006467
1543074430680.6035970.0010780.8301820.0050421.3753910.008049
1635155355550.5951210.0011340.8302640.0054681.3951180.008870
1728322286820.5897800.0012220.8269860.0052651.4021940.008580
1821746220790.5801840.0013340.8308610.0060311.4320650.010004
1916517166560.5764950.0014730.8114600.0067561.4075760.011359
2011879120140.5684210.0016810.8224500.0081001.4469030.013783
21813481240.5601110.0019620.8385810.0099721.4971690.017199
22550956740.5535520.0022650.7970620.0114771.4399040.020014
23344936700.5531080.0028550.8121830.0144281.4683980.025375
24225322940.5406290.0033650.7971010.0168701.4743960.030410
25141513520.5476020.0042710.8316680.0227681.5187460.040871
268327380.5244100.0051160.7235530.0268671.3797450.049986

Table 16.

Tune: CMS CP5

mmN+N_{+}N−N_{-}RμR_{\mu}SERσR_{\sigma}SERCR_{C}SE
32214342212160.8243730.0014270.8928770.0027611.0830980.002926
41653161649800.7766430.0014190.8685850.0028041.1183840.003177
51315591314730.7351550.0013980.8474360.0029991.1527300.003634
61096561095990.6989460.0013680.8329920.0031791.1917830.004089
793807933550.6734170.0013480.8182820.0033471.2151210.004525
882318820070.6515670.0013300.8065470.0033741.2378570.004732
973114732210.6345610.0013150.7998980.0036201.2605540.005214
1066808666350.6172830.0012870.7857320.0042431.2728870.006443
1160409605070.6051560.0012740.7809980.0038631.2905720.005932
1256014558540.5925900.0012590.7754510.0038111.3085790.005981
1352413520000.5829830.0012450.7639390.0111211.3103970.018629
1448787485660.5747760.0012280.7678350.0044341.3358870.007272
1545808458860.5665510.0012110.7626050.0039981.3460490.006629
1643782436880.5608800.0012000.7633720.0041101.3610240.006883
1741494412520.5548930.0011910.7597660.0042221.3692110.007157
1839251392230.5494780.0011800.7496880.0050221.3643640.008701
1937319372540.5445940.0011760.7500200.0049101.3772110.008573
2035271354610.5401290.0011630.7425340.0043941.3747350.007726
2133631339110.5348610.0011600.7447540.0044871.3919040.007971
2231919320330.5278560.0011470.7306730.0045421.3842270.008201
2329574300820.5250440.0011560.7245420.0045441.3799650.008286
2428327280270.5239200.0011640.7392500.0047781.4109990.008746
2526261262700.5191030.0011780.7316930.0049411.4095330.009100
2624236242860.5156370.0011940.7214800.0050481.3992020.009372
2722216221490.5130860.0012100.7127080.0050381.3890630.009505
2820090202420.5081540.0012450.7093950.0054231.3960250.010251
2917962177420.5060190.0012890.7155780.0056721.4141310.010840
3015847158300.5037300.0013350.7063870.0060611.4023120.011596
3113799135960.5008750.0014070.7007080.0064031.3989680.012350
3211735117690.4989150.0015020.7088670.0069861.4208160.013560
33980097550.4933980.0015970.6994220.0075001.4175610.014749
34832983610.4908420.0017060.7079280.0081311.4422710.016099
35682868130.4892870.0018480.6969450.0087241.4244080.017463
36541054720.4850290.0020120.7004930.0103061.4442280.020622
37430644190.4820830.0022310.6814070.0106981.4134640.021630
38343933620.4810930.0025000.6975460.0125241.4499180.025384
39264126840.4767190.0027390.6736390.0137631.4130740.027993
40188020040.4811080.0032220.6625840.0157161.3772040.031513
41143114660.4828520.0035510.6530880.0172471.3525640.035127
42103110560.4717570.0043210.6551240.0205661.3886890.042379
437958260.4671590.0046730.6987270.0278141.4956950.057086
445465700.4674660.0056030.6559660.0300441.4032380.061936

Table 17.

Tune: Monash

mmN+N_{+}N−N_{-}RμR_{\mu}SERσR_{\sigma}SERCR_{C}SE
32752332757130.8468840.0011800.9300550.0120771.0982090.014042
42065532070620.8041050.0011570.9206560.0048901.1449460.005840
51672091670310.7691640.0011280.9084320.0029201.1810640.003485
61401411397160.7390630.0011030.8995880.0028821.2172000.003591
71201881202370.7175010.0010840.8943950.0030711.2465410.003963
81057861057430.6989890.0010650.8791450.0031891.2577390.004253
993406935940.6803930.0010460.8710960.0033221.2802840.004575
1083218824010.6696710.0010440.8725830.0034191.3030030.004812
1173376735230.6576390.0010400.8655930.0037311.3162130.005366
1264322640690.6420130.0010490.8603470.0037721.3400770.005575
1355759551590.6325380.0010690.8440940.0041341.3344550.006206
1445849459000.6244310.0011220.8426270.0046381.3494320.007098
1537436371990.6154080.0011840.8319980.0046931.3519450.007285
1629539293630.6063230.0012730.8308340.0052901.3702820.008333
1722488222570.5993600.0014040.8215120.0059121.3706490.009486
1816298163980.5932530.0016030.8219290.0069831.3854620.011319
1911523116770.5834010.0018030.8175850.0082261.4014120.013579
20774376930.5728960.0021150.7915030.0094551.3815820.015937
21518950170.5669240.0025730.7805970.0187171.3768980.031946
22317231820.5593190.0030810.8014890.0150541.4329730.026086
23197019090.5639930.0039390.8222460.0195711.4578990.033906
24110412680.5547240.0047540.8039160.0253711.4492190.044238
256336520.5472100.0062380.7628430.0312521.3940590.055158

Table 18.

Declaration on historical production records and conditional new production

For the historical baseline samples, production-time records of the actual random-seed values and effective lifetime/decay settings, including ParticleDecays:limitTau0 and ParticleDecays:tau0Max, were not retained. Candidate generation sources survive, but their correspondence to each historical run cannot be certified. Complete reproduction of the original event-generation chain, including exact replay of the historical samples, therefore cannot currently be achieved from the retained records. Here the lifetime setting concerns the decay prescription, not specifically tau-lepton decays. Reanalysis of the retained data and reproduction of the supplied conditional numerical results remain possible; neither establishes the missing historical generator settings. The separate, stronger records for corrected CP5 and the remaining Monash/A14 top-up linkage gap are documented above.

If the editor or referees consider new production essential for assessment, we are prepared to generate fresh samples with explicitly documented physics and decay settings, independent recorded seeds, complete settings and particle dumps, and production logs, and to repeat the affected analyses and update the manuscript accordingly. Such production would provide new documented samples rather than recover the unknown history of the existing samples. Full replacement production is deferred and has not been started.

Effect of normalization on conditional widths

For CR state ss, define Gs=⟨Nch⟩sG_s = \langle N_{\mathrm{ch}}\rangle_s over the complete INEL>> 0-like selected sample, including Nch=0N_{\mathrm{ch}} = 0 and all MPI counts. At fixed integer mm, Cm,s=σm,s/μm,sC_{m,s} = \sigma_{m,s}/\mu_{m,s} and Wm,s=σm,s/GsW_{m,s} = \sigma_{m,s}/G_s. Consequently RC,m=Rσ,m/Rμ,mR_{C,m} = R_{\sigma,m}/R_{\mu,m} and RW,m=Rσ,m/RGR_{W,m} = R_{\sigma,m}/R_{G}, with RG=GON/GOFFR_G = G_{\mathrm{ON}}/G_{\mathrm{OFF}}. A common fixed normalization preserves width ordering; separate CR-state means can change it. The self-normalized MPI coordinates are also supplied, but integer mm, rather than the normalized coordinate, is matched here.

KindSampleBBRGR_GRσ<1R_{\sigma} < 1RC>1R_C > 1RW>1R_W > 1
baseline13 TeV230.724866232323
baseline2.76 TeV160.791969161614
baseline7 TeV200.749935202020
tuneATLAS A14240.707613242424
tuneCMS CP5420.615997424242
tuneMonash230.728878232323

Table 19.

Here BB counts retained matched bins. In the 13 TeV baseline and all three independent tune comparisons, absolute narrowing and inclusive-relative broadening coexist in every retained bin. Across the six comparison groups, inclusive-relative broadening holds centrally in 146 of 148 bins, whereas local-relative broadening holds in all 148. The associated marginal 95% lower bounds for RWR_W exceed one in 137 bins; this is not a simultaneous confidence statement. For a conditional standard deviation and an inclusive mean estimated in the same sample, their covariance must be retained. Let pm=P(M=m)p_m = P(M = m), vm=σm2v_m = \sigma_m^2, and X=NchX = N_{\rm ch}. The empirical influence functions used are

ψσm(X,M)=1M=mpm(X−μm)2−vm2σm,(31)\psi_{\sigma_m}(X,M)=\frac{\mathbf{1}_{M=m}}{p_m}\frac{(X-\mu_m)^2-v_m}{2\sigma_m}, \tag*{(31)}
ψWm(X,M)=ψσm(X,M)G−σmG2(X−G).(32)\psi_{W_m}(X,M)=\frac{\psi_{\sigma_m}(X,M)}{G}-\frac{\sigma_m}{G^2}(X-G). \tag*{(32)}

The estimated variance of W^m\widehat{W}_m is the empirical mean of ψWm2\psi_{W_m}^2 divided by the total selected sample size. ON/OFF ratio errors assume independent CR samples; historical seed gaps limit independent certification of that assumption. The influence function passed a numerical perturbation check. All normalization ratio identities were checked in all 148 matched bins. Weber et al., arXiv:1811.07744v2, Sections 2–3 and Figure 2, already discuss conditional fluctuations alongside self-normalized relations. Their conditional-width discussion does not specify a numerical standard-deviation estimator or supply a conditional-width table with uncertainties. The present test establishes normalization-based compatibility on our retained events, not an exact matched-configuration reproduction or an unambiguous identification of their width metric. Their calculation uses PYTHIA 8.230, non-diffractive collisions, and ALICE-defined primary charged particles at ∣η∣<1|\eta| < 1. Our selections and historical primary-particle limitations remain distinct. The added empirical information is separate mean and absolute-width suppression with uncertainty, not the algebraic ratio identities or discovery of conditional fluctuations. Run pythonconditional_width_analysis/code/compare_normalizations.py from the S2 root, or use bashRUN_ANALYSIS.sh. The inputs are the complete integer joint counts and the reproduced central matched-bin table. The following tables give RWR_W and one-standard-error uncertainties for every retained matched bin; conditional_width_analysis/results/matched_normalization_comparison.csv additionally gives RCR_C, RσR_\sigma and marginal confidence bounds.

Baseline: 13 TeV

mmRσR_\sigmaSERCR_CSERWR_WSE
30.9424150.0023111.1135960.0024281.3001230.003297
40.9244340.0026511.1539630.0030061.2753180.003746
50.9006050.0065151.1758870.0082711.2424430.009015
60.8994770.0028771.2211300.0036041.2408880.004046
70.8857760.0030991.2421910.0040381.2219860.004343
80.8785520.0032081.2685410.0043331.2120200.004490
90.8693230.0032711.2800100.0045161.1992880.004574
100.8624010.0033711.3001000.0047951.1897380.004710
110.8550420.0035571.3161270.0051721.1795860.004962
120.8470720.0037451.3275800.0055801.1685910.005218
130.8488110.0039511.3484330.0059911.1709900.005501
140.8391720.0042851.3564270.0066021.1576920.005955
150.8381260.0047821.3712220.0074791.1562490.006636
160.8227110.0053401.3667530.0085251.1349830.007401
170.8217770.0059941.3797900.0096631.1336950.008299
180.8092000.0067391.3790380.0110701.1163440.009323
190.8146840.0081671.4007480.0134781.1239090.011288
200.7998380.0096991.4026630.0163541.1034280.013398
210.8012730.0120071.4200880.0205941.1054090.016579
220.7833530.0145191.4006980.0250111.0806860.020041
230.7986150.0185411.4358060.0321891.1017410.025588
mmRσR_{\sigma}SERCR_CSERWR_WSE
240.7898470.0243901.4356680.0431551.0896450.033655
250.7519160.0321761.3567050.0562561.0373180.044393

Table 20.

Baseline: 2.76 TeV

mmRσR_{\sigma}SERCR_CSERWR_WSE
30.9353190.0021901.0945910.0023351.1810050.002834
40.9113990.0023641.1269930.0027041.1508020.003044
50.9008550.0024491.1662230.0029381.1374880.003146
60.8849450.0026161.1916810.0032941.1173990.003353
70.8740830.0028971.2179280.0037911.1036840.003700
80.8645060.0031301.2400770.0042581.0915900.003992
90.8531730.0039721.2584100.0055941.0772820.005043
100.8497300.0041971.2806820.0059891.0729330.005326
110.8368970.0048521.2923800.0071541.0567300.006150
120.8367500.0061181.3084890.0091521.0565440.007743
130.8334000.0075551.3189090.0114771.0523140.009554
140.8125260.0095171.3080080.0148411.0259570.012029
150.8116250.0129511.3252690.0203901.0248200.016361
160.7905650.0165931.3218730.0267780.9982280.020957
170.8067930.0228641.3599580.0381361.0187190.028876
180.7576600.0346941.2982310.0579170.9566790.043809

Table 21.

Baseline: 7 TeV

mmRσR_{\sigma}SERCR_CSERWR_WSE
30.9362430.0023261.1045080.0024711.2484310.003190
40.9184730.0024431.1424740.0027651.2247360.003338
50.9032640.0026201.1808090.0031481.2044550.003565
60.8907520.0028341.2070750.0035671.1877720.003842
70.8798690.0034061.2335850.0045201.1732600.004592
80.8786120.0030551.2659520.0041381.1715830.004130
90.8665470.0032461.2796570.0045201.1554950.004380
100.8604370.0034281.2990190.0049001.1473470.004620
110.8498360.0039291.3068830.0057571.1332120.005280
120.8435310.0040581.3252150.0060941.1248050.005451
130.8408790.0046001.3374980.0069921.1212680.006169
140.8160530.0052101.3293980.0081541.0881640.006976
150.8284110.0061761.3603320.0097221.1046430.008260
160.8137920.0074691.3643420.0120091.0851490.009979
170.7971320.0087591.3504450.0143871.0629340.011697
180.8013130.0115881.3750130.0191531.0685090.015464
190.8119400.0146061.4084290.0244101.0826800.019487
200.7860160.0182641.3704430.0314191.0481120.024363
210.7744010.0259371.3910360.0442771.0326240.034589
220.7843530.0337151.3943950.0587481.0458940.044962

Table 22.

Tune: ATLAS A14 Tune: CMS CP5

mmRσR_{\sigma}SERCR_CSERWR_WSE
30.9384610.0032391.1141880.0035911.3262350.004661
40.9191670.0025351.1497730.0028901.2989680.003686
50.9026800.0043181.1825110.0053901.2756690.006156
60.8959260.0028831.2273980.0036641.2661240.004160
70.8863320.0030001.2540900.0039641.2525650.004319
80.8711430.0035431.2704930.0048931.2311010.005070
90.8690660.0056431.2988350.0081391.2281650.008009
100.8646310.0044171.3189400.0064731.2218980.006290
110.8695800.0035901.3540710.0053031.2288910.005136
120.8626240.0036811.3621510.0055381.2190620.005263
130.8542970.0039181.3746530.0060061.2072950.005592
140.8538160.0041481.3931650.0064671.2066140.005914
150.8301820.0050421.3753910.0080491.1732150.007164
160.8302640.0054681.3951180.0088701.1733300.007763
170.8269860.0052651.4021940.0085801.1686980.007480
180.8308610.0060311.4320650.0100041.1741730.008557
190.8114600.0067561.4075760.0113591.1467560.009578
200.8224500.0081001.4469030.0137831.1622880.011472
210.8385810.0099721.4971690.0171991.1850840.014114
220.7970620.0114771.4399040.0200141.1264090.016235
230.8121830.0144281.4683980.0253751.1477790.020403
240.7971010.0168701.4743960.0304101.1264650.023853
250.8316680.0227681.5187460.0408711.1753140.032186
260.7235530.0268671.3797450.0499861.0225260.037974

Table 23.

mmRσR_{\sigma}SERCR_CSERWR_WSE
30.8928770.0027611.0830980.0029261.4494830.004614
40.8685850.0028041.1183840.0031771.4100490.004673
50.8474360.0029991.1527300.0036341.3757150.004976
60.8329920.0031791.1917830.0040891.3522670.005258
70.8182820.0033471.2151210.0045251.3283880.005522
80.8065470.0033741.2378570.0047321.3093370.005563
90.7998980.0036201.2605540.0052141.2985440.005952
100.7857320.0042431.2728870.0064431.2755460.006951
110.7809980.0038631.2905720.0059321.2678610.006339
120.7754510.0038111.3085790.0059811.2588560.006255
130.7639390.0111211.3103970.0186291.2401670.018069
140.7678350.0044341.3358870.0072721.2464920.007254
150.7626050.0039981.3460490.0066291.2380020.006554
160.7633720.0041101.3610240.0068831.2392470.006733
170.7597660.0042221.3692110.0071571.2333930.006911
180.7496880.0050221.3643640.0087011.2170330.008198
190.7500200.0049101.3772110.0085731.2175720.008018
200.7425340.0043941.3747350.0077261.2054200.007186
210.7444750.0044871.3919040.0079711.2085700.007337
220.7306730.0045421.3842270.0082011.1861650.007424
230.7245420.0045441.3799650.0082861.1762100.007427
240.7392500.0047781.4109990.0087461.2000880.007806
250.7316930.0049411.4095330.0091001.1878190.008067
260.7214800.0050481.3992020.0093721.1712410.008239

Table 24.

mmRσR_{\sigma}SERCR_CSERWR_WSE
270.7127080.0050381.3890630.0095051.1570010.008223
280.7093950.0054231.3960250.0102511.1516220.008843
290.7155780.0056721.4141310.0108401.1616580.009249
300.7063870.0060611.4023120.0115961.1467380.009875
310.7007080.0064031.3989680.0123501.1375200.010428
320.7088670.0069861.4208160.0135601.1507640.011372
330.6994220.0075001.4175610.0147491.1354310.012204
340.7079280.0081311.4422710.0160991.1492400.013227
350.6969450.0087241.4244080.0174631.1314100.014189
360.7004930.0103061.4442280.0206221.1371700.016752
370.6814070.0106981.4134640.0216301.1061860.017386
380.6975460.0125241.4499180.0253841.1323850.020349
390.6736390.0137631.4130740.0279931.0935750.022358
400.6625840.0157161.3772040.0315131.0756290.025525
410.6530880.0172471.3525640.0351271.0602140.028011
420.6551240.0205661.3886890.0423791.0635180.033396
430.6987270.0278141.4956950.0570861.1343030.045160
440.6559660.0300441.4032380.0619361.0648860.048778

Table 25.

Tune: Monash

mmRσR_{\sigma}SERCR_CSERWR_WSE
30.9300550.0120771.0982090.0140421.2760100.016579
40.9206560.0048901.1449460.0058401.2631150.006753
50.9084320.0029201.1810640.0034851.2463430.004083
60.8995880.0028821.2172000.0035911.2342090.004030
70.8943950.0030711.2465410.0039631.2270840.004283
80.8791450.0031891.2577390.0042531.2061630.004440
90.8710960.0033221.2802840.0045751.1951190.004618
100.8725830.0034191.3030030.0048121.1971600.004751
110.8655930.0037311.3162130.0053661.1875690.005172
120.8603470.0037721.3400770.0055751.1803720.005228
130.8440940.0041341.3344550.0062061.1580740.005716
140.8426270.0046381.3494320.0070981.1560610.006402
150.8319980.0046931.3519450.0072851.1414780.006478
160.8308340.0052901.3702820.0083331.1398800.007292
170.8215120.0059121.3706490.0094861.1270920.008142
180.8219290.0069831.3854620.0113191.1276630.009607
190.8175850.0082261.4014120.0135791.1217040.011307
200.7915030.0094551.3815820.0159371.0859200.012990
210.7805970.0187171.3768980.0319461.0709570.025685
220.8014890.0150541.4329730.0260861.0996210.020665
230.8222460.0195711.4578990.0339061.1280980.026860
240.8039160.0253711.4492190.0442381.1029500.034815
250.7628430.0312521.3940590.0551581.0465990.042881

Table 26.

AI assistance and validation record

The manuscript Methods subsection on AI-assisted code development and validation identifies OpenAI Codex, the assisted programming, diagnostic, plotting and editorial tasks, and the executable validation performed. Selected retained project-related metadata record gpt-5.5, gpt-5.6-sol, gpt-5.6-luna, gpt-6-astra and gpt-6.1-sol; the current revision session records gpt-6.1-sol through the Codex VS Code integration, client 0.162.0-alpha.17.2. These are recorded identifiers, not independently verified immutable snapshots or a complete model-to-artifact history. S2 supplies ai_disclosure/LOCAL_MODEL_RECORDS.json and ai_disclosure/TASK_VALIDATION_MAP.csv. Private conversation bodies are not included. The available numerical evidence includes the 60-million-event central reanalysis, full six-baseline particle scan, fourth-moment and empirical-bootstrap errors, normalization influence-function perturbation check, source checks, validation-only generator tests, and fresh-archive numerical/plot reproduction. Automated checks were themselves developed with AI assistance and are not independent human validation. They verify numerical consistency with retained inputs and do not restore missing historical generation records. Scientific interpretation and responsibility remain with the authors.

References

  1. [1]T. Sjöstrand, M. van Zijl, A multiple-interaction model for the event structure in hadron collisions, Phys. Rev. D 36 (1987) 2019–2041. doi: 10.1103/PhysRevD.36.2019.
  2. [2]C. Bierlich, et al., A comprehensive guide to the physics and usage of PYTHIA 8.3, SciPost Phys. Codeb. 2022 (2022) 8. arXiv:2203.11601, doi:10.21468/SciPostPhysCodeb.8.DOI
  3. [3]B. B. Abelev, et al., Multiplicity dependence of the average transverse momentum in pp, p-Pb, and Pb-Pb collisions at the LHC, Phys. Lett. B 727 (2013) 371–380. arXiv:1307.1094, doi:10.1016/j.physletb.2013.10.054.
  4. [4]A. Ortiz Velasquez, P. Christiansen, E. Cuautle Flores, I. A. Maldonado Cervantes, G. Paić, Color reconnection and flowlike patterns in pp collisions, Phys. Rev. Lett. 111 (2013) 042001. arXiv:1303.6326, doi:10.1103/PhysRevLett.111.042001.DOI
  5. [5]A. Tumasyan, et al., Cms pythia 8 colour reconnection tunes based on underlying-event data, Eur. Phys. J. C 83 (2023) 587. doi:10.1140/epjc/s10052-023-11630-8.DOI
  6. [6]S. G. Weber, A. Dubla, A. Andronic, A. Morsch, Elucidating the multiplicity dependence of J/ψ production in proton–proton collisions with PYTHIA8, Eur. Phys. J. C 79 (1) (2019) 36. arXiv:1811.07744, doi:10.1140/epjc/s10052-018-6531-4.DOI
  7. [7]S. Prasad, B. Sahoo, S. Tripathy, N. Mallik, R. Sahoo, Probing strangeness with event topology classifiers in pp collisions at energies available at the CERN Large Hadron Collider with the rope hadronization mechanism in PYTHIA, Phys. Rev. C 111 (4) (2025) 044902. arXiv:2409.05454, doi:10.1103/PhysRevC.111.044902.DOI
  8. [8]S. Prasad, S. Tripathy, B. Sahoo, R. Sahoo, Event topology classifiers at the Large Hadron Collider, Phys. Rept. 1181 (2026) 1–75. arXiv:2506.03782, doi:10.1016/j.physrep.2026.04.001.DOI
  9. [9]P. Skands, C. Carrazza, J. Rojo, Tuning pythia 8.1: The monash 2013 tune, Eur. Phys. J. C 74 (2014) 3024. arXiv:1404.5630, doi:10.1140/epjc/s10052-014-3024-y.DOI
  10. [10]ATLAS Collaboration, Atlas run 1 pythia8 tunes, Tech. Rep. ATL-PHYS-PUB-2014-021, CERN (2014). URL https://cds.cern.ch/record/1966419
  11. [11]A. M. Sirunyan, et al., Extraction and validation of a new set of cms pythia8 tunes from underlying-event measurements, Eur. Phys. J. C 80 (2020) 4. arXiv:1903.12179, doi:10.1140/epjc/s10052-019-7499-4.DOI
  12. [12]CMS Collaboration, CP5 settings in CMSSW, branch CMSSW_10_6_X, https://github.com/cms-sw/cmssw/blob/CMSSW_10_6_X/Configuration/Generator/python/MCTunes2017/PythiaCP5Settings_cfi.py, accessed 10 October 2026; source and checksum archived with the analysis (2026).
  13. [13]A. Buckley, et al., LHAPDF6: parton density access in the LHC precision era, Eur. Phys. J. C 75 (3) (2015) 132. arXiv:1412.7420, doi:10.1140/epjc/s10052-015-3318-8.DOI
  14. [14]G. Cowan, Statistical Data Analysis, Oxford University Press, Oxford, 1998.
  15. [15]M. I. Gorenstein, M. Gaździcki, Strongly intensive quantities, Phys. Rev. C 84 (2011) 014904. arXiv:1101.4865, doi:10.1103/PhysRevC.84.014904.DOI
  16. [16]S. Acharya, et al., Charged-particle multiplicity fluctuations in pp–pb collisions at √sNN = 2.76 TeV, Eur. Phys. J. C 81 (2021) 1012. arXiv:2105.05745, doi:10.1140/epjc/s10052-021-09784-4.DOI
  17. [17]S. Acharya, et al., Charged-particle production as a function of multiplicity and transverse sphericity in pp collisions at √s = 5.02 and 13 TeV, Eur. Phys. J. C 79 (2019) 857. arXiv:1905.07208, doi:10.1140/epjc/s10052-019-7350-y.DOI
  18. [18]ALICE Collaboration, HepData record for average transverse momentum as a function of event multiplicity in pp collisions at 13 tev, table 7, HEPData, associated with arXiv:1905.07208 (2019). doi:10.17182/hepdata.91996.v2/t7.
  19. [19]A. Ortiz, G. Paić, E. Cuautle, Mid-rapidity charged hadron transverse sphericity in pp collisions simulated with Pythia, Nucl. Phys. A 941 (2015) 78–86. arXiv:1503.03129, doi:10.1016/j.nuclphysa.2015.05.010.DOI
  20. [20]T. Martin, P. Skands, S. Farrington, Probing Collective Effects in Hadronisation with the Extremes of the Underlying Event, Eur. Phys. J. C 76 (5) (2016) 299. arXiv:1603.05298, doi:10.1140/epjc/s10052-016-4135-4.DOI
  21. [21]J. E. M. Méndez, A. Ortiz, Investigating the most active pp collisions (top 0.1%) using the tools developed by experiments at the LHC, J. Phys. G 52 (9) (2025) 095001. arXiv:2506.06588, doi:10.1088/1361-6471/ae02ac.DOI
  22. [22]A. Ortiz, A. Paz, J. D. Romo, S. Tripathy, E. A. Zepeda, I. Bautista, Multiparton interactions in pp collisions from machine learning-based regression, Phys. Rev. D 102 (7) (2020) 076014. arXiv:2004.03800, doi:10.1103/PhysRevD.102.076014.DOI
  23. [23]B. Abelev, et al., Multiplicity dependence of two-particle azimuthal correlations in pp collisions at the LHC, JHEP 09 (2013) 049. arXiv:1307.1249, doi:10.1007/JHEP09(2013)049.DOI

Paper details

vlnc:W_zdgxg9qw
Version
v2 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Codex

View statement
Looking for related papers…
Contents