EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH (CERN)

CERN-EP-2026-216
LHCb-PAPER-2025-067
14 August 2026

Amplitude analysis of
B + → K +π +π − decays

LHCb collaboration†
arXiv:2608.12612v1 [hep-ex] 12 Aug 2026

Abstract
The branching fractions and quasi-two-body CP -violating asymmetries of interme-
diate states obtained through an amplitude analysis of the charmless three-body
decay B + → K + π + π − are reported. The analysis is based on pp collision data
√
at centre-of-mass energies s = 7 and 8 TeV recorded with the LHCb detector,
corresponding to an integrated luminosity of 3 fb−1 . The most challenging aspect of
the amplitude modelling lies in the description of the dominant K + π − and π + π −
S-wave contributions. This is achieved by three complementary approaches based
on a physically motivated analytic model built on the isobar approximation, the
K-matrix formalism, and a quasi-model-independent procedure in which overlapping
crossing partial waves are simultaneously studied. In addition, alternative sets of
results are presented, considering the π + π − final state to manifest either through
direct ω(782) decays or ρ(770)0 –ω(782) mixing. The most precise measurements
of branching fractions and CP asymmetries are obtained for the vast majority of
intermediate states, establishing firmer reference points against which to cleanly
probe model-independent physics beyond the Standard Model. The results from all
three approaches agree and provide new insight into strong dynamics and the origin
of CP -violation effects in B + → K + π + π − decays.

Submitted to Phys. Rev. D

© 2026 CERN for the benefit of the LHCb collaboration. CC BY 4.0 licence.

†
Authors are listed at the end of this paper.
ii
1 Introduction
In the Standard Model of particle physics (SM), CP violation originates from a single
irreducible complex phase in the Cabibbo–Kobayashi–Maskawa (CKM) matrix [1,2]. Thus
far, all measurements of CP violation in particle decays are consistent with this explanation.
Nevertheless, the degree of CP violation permitted in the SM is inconsistent with the
macroscopic matter-antimatter asymmetry observed in the Universe [3], motivating further
studies and searches for sources of CP violation beyond the SM.
For the manifestation of CP violation in decay, at least two interfering amplitudes with
different strong and weak phases are required. In the SM, weak phases are associated with
the complex elements of the CKM matrix and have opposite sign between charge-conjugate
processes, while strong phases are associated with hadronic final-state effects and do not
change sign under CP conjugation. In decays of b hadrons to charmless hadronic final
states, contributions from both tree-level and loop (so-called “penguin”) diagrams, which
can provide the relative weak phase that is necessary for CP violation to manifest, are
possible with comparable magnitudes.
In multibody decays, variation of strong phases across the phase space, caused by
hadronic effects, allows for a richer tapestry of CP violation effects and phenomenology
compared to two-body decays. Large CP asymmetries localised in regions of phase space
of charmless three-body B-meson decays have been observed in model-independent analy-
ses [4–8], however a description of these effects with an accurate model of the contributing
resonances has only become available for B + → π + K + K − [9] and B + → π + π + π − [10, 11]
decays.
This paper documents an analysis of the B + → K + π + π − decay amplitude in the
two-dimensional phase space known as the Dalitz plot [12, 13]. The inclusion of charge-
conjugate processes is implied, except where asymmetries are discussed. Apart from
the extraction of meaningful information on the intermediate branching fractions and
CP -violating asymmetries in this decay, it is also of paramount importance to improve the
handling of the underlying strong interaction. Focussed discussions can be found in two
companion Letters dedicated to the CP -conserving [14] and CP -violating [15] implications
facilitated by the improved understanding of strong-interaction dynamics discussed herein.
The primary quasi-two-body results reported have several uses beyond providing essen-
tial feedback for the myriad predictions on offer from several effective-theory frameworks
such as QCD factorisation [16–21], perturbative QCD [22–28], soft-collinear effective the-
ory [29,30] and SU(3) flavour symmetry [31]. These extend to model-independent tests for
new physics in loop-mediated processes via isospin sum rules [32–34] and in measurements
of the effective weak phase γ = ϕ3 ≡ arg[−Vud Vub∗ /Vcd Vcb∗ ] [35–38]. Furthermore, results for
charmonium (cc) intermediates, such as the χc0 (1P ) resonance, open the possibility [39,40]
to correct measurements of the B 0 –B 0 mixing phase β = ϕ1 ≡ arg[−Vcd Vcb∗ /Vtd Vtb∗ ], from
their related B 0 → (cc)K 0 channels arising from penguin contamination, thereby improving
the fidelity of constraints to the CKM matrix.
The present analysis is performed on data corresponding to an integrated luminosity
of 3 fb−1 collected by the LHCb experiment,
√ of which 1 fb−1 was collected in 2011 with a
−1
pp
√ collision centre-of-mass energy of s = 7 TeV and 2 fb was collected in 2012 with
s = 8 TeV. Models of the Dalitz-plot distribution are constructed in terms of intermediate
resonant and nonresonant structures. Due to their magnitude and potential importance
to the observed CP violation in B + → K + π + π − decay, particular attention is given to

1
the K + π − and π + π − S-wave contributions, which are known to consist of numerous
overlapping resonances with structures related to decay-threshold openings [41, 42]. Three
different state-of-the-art approaches to the modelling of the S-wave are used to ensure
that any inaccuracies in the description of this part of the amplitude do not impact the
interpretation of the physical quantities reported.
This paper is organised as follows: Section 2 gives a brief description of the LHCb
detector, the event reconstruction and simulation software; the signal candidate selection
procedure is described in Section 3; Section 4 describes the procedure for estimating the
signal and background yields that enter into the amplitude fit; Section 5 outlines the
formalism used for the construction of the amplitude models, as well as a description of
the mass lineshapes used to parameterise the intermediate structures; Section 6 describes
the systematic uncertainties associated with the analysis procedure; Section 7 documents
the physics parameters of interest obtained from the amplitude models and presents
comparisons between projections of the data and the fit models; these results are then
discussed in Section 8; and a summary of the work can be found in Section 9.

2 Detector and simulation

The LHCb detector [43, 44] is a single-arm forward spectrometer covering the
pseudorapidity range 2 < η < 5, designed for the study of particles containing b or
c quarks. The detector used to collect the data analysed in this paper includes a high-
precision tracking system consisting of a silicon-strip vertex detector surrounding the pp
interaction region [45], a large-area silicon-strip detector located upstream of a dipole
magnet with a bending power of about 4 T m, and three stations of silicon-strip detectors
and straw drift tubes [46] placed downstream of the magnet. The tracking system provides
a measurement of the momentum p of charged particles with relative uncertainty that
varies from 0.5% at low momentum to 1.0% at 200 GeV/c. The minimum distance of a
track to a primary vertex (PV), or impact parameter (IP), is measured with a resolution
of (15 + 29/pT ) µm, where pT is the component of the momentum transverse to the beam
(in GeV/c). Different types of charged hadrons are distinguished using information from
two ring-imaging Cherenkov detectors [47]. Photons, electrons and hadrons are identified
by a calorimeter system consisting of scintillating-pad and preshower detectors, an elec-
tromagnetic and a hadronic calorimeter. Muons are identified by a system composed of
alternating layers of iron and multiwire proportional chambers [48]. The magnetic field
deflects oppositely charged particles in opposite directions and this can lead to detection
asymmetries. Periodically reversing the magnetic field polarity throughout the data-taking
reduces this effect to a negligible level. Approximately 60% of 2011 data and 52% of 2012
data were collected in the “down” polarity configuration, and the remainder in the “up”
configuration.
The online event selection is performed by a trigger [49] which consists of a hardware
stage followed by a software stage. The hardware stage is based on information from the
calorimeter and muon systems in which events are required to contain a muon with high pT ,
or a hadron, photon or electron with high transverse energy in the calorimeters. The soft-
ware trigger requires a two- or three-track secondary vertex with significant displacement
from all primary pp interaction vertices. All charged particles with pT > 500 (300) MeV/c
are reconstructed, for data collected in 2011 (2012), in events where at least one charged

2
particle has transverse momentum pT > 1.7 (1.6) GeV/c and is inconsistent with originating
from a PV. A multivariate algorithm [50] is used for the identification of secondary vertices
consistent with b-hadron decays. Triggered data further undergo a centralised, offline
processing step to deliver physics-analysis-ready data across the entire LHCb physics
programme [51].
Simulated samples are used to investigate backgrounds from other b-hadron decays and
also to study the detection and reconstruction efficiency of the signal. In the simulation,
pp collisions are generated using Pythia [52, 53] with a specific LHCb configuration [54].
Decays of unstable particles are described by EvtGen [55], in which final-state radiation
is generated using Photos [56]. The interaction of the generated particles with the
detector and its response are implemented using the Geant4 toolkit [57, 58] as described
in Ref. [59].

3 B +-candidate selection
The selection of B + signal candidates follows well-established procedures developed for
the measurement of the relative branching fractions of B + → h+ h′+ h′− decays, where h
and h′ represent either a pion or kaon, performed on the same data sample [60]. The most
important difference is in the approach to the suppression of random associations of tracks,
commonly referred to as combinatorial background, which is based on a multivariate
boosted decision tree (BDT) classifier [61]. In the relative branching fractions measurement
of B + → h+ h′+ h′− decays, this algorithm was trained with the π + K + K − final state,
but in this analysis the training is based instead on the K + π + π − final state. The
signal training sample is taken from simulation, while the background training sample
is taken from a region of the K + π + π − mass in data where combinatorial background
dominates, 5.65 < mK + π+ π− < 6.10 GeV/c2 . The variables that enter this classifier are
B + and decay product kinematic properties, quantities based on the quality of the
reconstructed tracks and decay vertices, as well as the B + displacement from the PV.
The requirement on the output of √ the classifier is optimised to maximise the expected
approximate signal significance, Ns / Ns + Nb , where Ns is the expected signal yield within
the range 5.24 < mK + π+ π− < 5.32 GeV/c2 , corresponding to approximately ±3 standard
deviations in mass resolution around the known B + mass [62], and Nb is the corresponding
combinatorial background yield within the same region. These yield inputs to the
aforementioned figure of merit are obtained from rudimentary fits to mK + π+ π− data.
To remove the bulk of D0 decays into two oppositely charged mesons, the
following two-body mass regions are vetoed: 1.75 < mK + π− < 1.95 GeV/c2 and
1.75 < mπ+ π− < 1.90 GeV/c2 . These vetoed windows are relatively wide so as to also sup-
press backgrounds with particle misidentification. Furthermore, the requirements placed
on the particle-identification (PID) information associated with each final-state track are
more stringent than those in Ref. [60]. This suppresses further backgrounds that still arise
when any number of kaons or pions are misidentified and ensures that systematic uncertain-
ties related to such backgrounds remain negligible in the amplitude analysis. Tighter PID
requirements are imposed on the K + candidate in the region 1.95 < mK + π− < 2.00 GeV/c2
and on the π − candidate in 1.65 < mK + π− < 1.75 GeV/c2 , while tighter PID requirements
are applied to each pion candidate in 1.65 < mπ+ π− < 1.75 GeV/c2 . These requirements
reduce most misidentified D0 backgrounds to negligible levels. Finally, another veto,

3
3.05 < mπ+ π− < 3.15 GeV/c2 , is imposed to remove the narrow J/ψ structure. The χc0 (1P )
resonance is retained as the only charmonium structure visible in phase space. Misidenti-
fied contributions from the dimuon and dielectron J/ψ decay channels are known to be
negligible in the K + π + π − final state [60]. Other potential contributions from decays of
broader charmonia states are included in the amplitude model.
Two-body charmless B 0 decays can inhabit the background in K + π + π − phase space
when derived from the sideband above the B + mass. Candidates containing such contribu-
tions are removed by vetoing the two-body mass ranges [5.22, 5.32] GeV/c2 . Approximately
0.2% of selected events contain multiple B + decay candidates following the aforementioned
selection procedure; all such candidates are included in the analysis. The Dalitz-plot
variables are calculated applying a kinematic mass constraint, fixing the B + candidate
mass to the known value [63] to improve resolution and to ensure that all events remain
within the Dalitz-plot boundary.

4 B +-candidate mass fit

An extended, unbinned, maximum-likelihood fit is performed to the mK + π+ π− mass
spectra to extract yields and global charge asymmetries of the B + → K + π + π − signal
and various contributing backgrounds. The fit is performed to candidates in the range
5.1 < mK + π+ π− < 6.0 GeV/c2 , and its results are used to obtain signal and background
yields in the signal region, 5.24 < mK + π+ π− < 5.32 GeV/c2 , in which the subsequent Dalitz-
plot fit is performed. All shape parameters of the probability density functions (PDFs)
comprising the fit model are shared between B + and B − candidates; the only differences
between the two final states, which are fitted simultaneously, are due to signal and
combinatorial-background yield asymmetries. At this stage, the data are also subdivided
by data-taking year, and whether the hardware trigger decision is due to hadronic
calorimeter deposits associated with the signal candidate, or due to other particles in the
pp collision, to permit correction for possible differences in efficiency between subsamples
(see Section 5.6).
The mass distribution of the B + → K + π + π − signal decay is parameterised by the sum
of an underlying broad Gaussian and a core Crystal Ball function [64], with power-law
tails on both sides of the peak in order to describe asymmetric non-Gaussian effects due
to detector resolution and final-state radiation. The parameters describing these tails are
determined from simulation relative to the peak position and width, which themselves are
free to vary in the data fit to account for small differences between simulation and data.
All remaining parameters, apart from the total yield, are obtained from a fit to simulated
decays. The global raw charge asymmetry for the signal, while free to vary in the mass
fit, is not propagated to the subsequent amplitude analysis. Rather, the fitted amplitude
model accounts for the global charge asymmetry as well as local asymmetries between the
B + and B − data.
The distribution of the combinatorial background is modelled with a falling exponential
function. Partially reconstructed backgrounds, which predominantly arise from four-body
B-meson decays where a charged hadron or neutral particle is not reconstructed, are
modelled with an ARGUS function [65] convolved with a Gaussian resolution function.
Separate components are included in the fit to account for such partially reconstructed
backgrounds from B + , B 0 and Bs0 decays. Two significant sources of peaking cross-

4
 104 104

Candidates / (9 MeV/c2)
Candidates / (9 MeV/c2)
Data LHCb LHCb
B+, B0 4-body 3 fb−1 3 fb−1
B0s 4-body
B+ B−
103 B±→ K ± π+ π− 103

102 Combinatorial 102

B± → π ± π + π −
B± → π ± K + K −
10 B±→ K ±η' (958) 10
5200 5400 5600 5800 5200 5400 5600 5800
mK +π +π − (MeV/c2) mK −π −π + (MeV/c2)

Figure 1: Mass distributions for (left) B + and (right) B − candidates reconstructed in the
K + π + π − and K − π − π + final state, respectively, for the combined 2011 and 2012 data taking
samples, along with the results of the fit. The signal region is indicated by arrows.

feed background arise from misidentified B + → π + π + π − and B + → π + K + K − decays. To

obtain an accurate model for these backgrounds when reconstructed in the B + → K + π + π −
final state, simulated decays are weighted according to their respective amplitude models
obtained by the LHCb collaboration [9–11]. Further corrections for the differences between
data and simulation in trigger, tracking and PID efficiencies are applied accounting for
dependence on kinematics to obtain more reliable shapes. These peaking-background
shapes are modelled as Crystal Ball functions, where all parameters are determined from fits
to the calibrated simulation. Furthermore, the yields of these components are constrained
to the B + → K + π + π − signal yield multiplied by the product of the relative branching
fractions of these decays and the relative overall reconstruction and selection efficiencies,
which are described in Section 5.6. Finally, a component for B + → K + η ′ (→ π + π − γ) decays
is described using simulated events, corrected to match the η ′ → π + π − γ amplitude model
from the BESIII collaboration [66] and further calibrated to represent data. Here and
throughout the paper, the symbol η ′ is used to denote the η ′ (958) meson. The shape for this
component is modelled in the same way as that of the partially reconstructed backgrounds,
while its yield is fixed relative to signal in the same way as for the misidentified peaking
backgrounds.
The mass fit results are shown in Fig. 1 with a logarithmic y-axis scale (corresponding
plots with a linear scale are provided in Appendix A), while Table 1 quotes the component
yields and phase-space-integrated background asymmetries in the B + signal region. These
values are subsequently used in the Dalitz-plot fit and to estimate the associated systematic
uncertainties outlined in Section 6.

5 Dalitz-plot fit model

The B + → K + π + π − decay amplitude can be expressed fully in terms of the mass-squared
of two pairs of the decay products m2K + π− and m2π+ π− , collectively denoted as the position
in three-body phase space, Φ3 . The Dalitz-plot distributions of the selected candidates
can be seen in Fig. 2.

5
Table 1: Component CP -integrated yields and phase-space-integrated background asymmetries in
the signal region, calculated from the results of the mass fit, where the uncertainties are statistical
and systematic, respectively. For the parameters marked as being fixed, the uncertainties derive
from either limited simulation sample sizes or from an external source as indicated by the
reference. The raw asymmetry of signal is withheld as it does not propagate to the amplitude
analysis.

Component Yield scaled to signal region Asymmetry

Signal 102 073 ± 311 ± 560 —
Combinatorial background 9 404 ± 451 ± 287 −0.022 ± 0.006 ± 0.001
B + → π+π+π− 1 358 ± 51 ± 658 0.000 ± 0.006 (fixed)
B + → π+K +K − 471 ± 39 ± 43 0.000 ± 0.008 (fixed)
B + → K +η′ 1 776 ± 200 ± 68 0.000 ± 0.011 (fixed) [67]
m2π +π − (GeV2/ c4)

m2π −π + (GeV2/ c4)

LHCb LHCb
20 3 fb−1 20 3 fb−1
B+ B−
15 15

10 10

5 5

0 0
0 5 10 15 20 2
25 0 5 10 15 20 25
m2K +π − (GeV / c4) m2K −π + (GeV2/ c4)
Figure 2: Dalitz-plot distributions for (left) B + and (right) B − candidate decays to the K + π + π −
and K − π − π + final state, respectively. Depleted regions are due to vetoes of D0 → K + π − ,
D0 → π + π − and J/ψ → π + π − decays.

5.1 Amplitude analysis formalism

In general, the total amplitude for B ± decays is taken as a coherent sum over N components,
each described by a decay amplitude Aj that is parameterised with two-body intermediate
processes [68–70],
XN
±
A (Φ3 ) ≡ a±j Aj (Φ3 ) , (1)
j

where the complex coefficients a± j represent the relative contribution of component j in the
relevant final state. For statistical stability in function minimisation, these are expressed
in the “Cartesian” parametrisation of CP violation,

a±
j ≡ (xj ± ∆xj ) + i(yj ± ∆yj ) . (2)

6
The individual amplitude Aj contains only CP -conserving dynamics, and is generically
parameterised, removing the component index j for brevity, as

A(Φ3 ) ≡ BP (Φ3 )BD (Φ3 )S(Φ3 )T (Φ3 ) , (3)

where the production and decay factors BP and BD , respectively, account for deviations
from pointlike interactions. These are taken to be the normalised Blatt–Weisskopf
penetration factors [71, 72], that depend on the effective interaction radius taken to be
rP = rD = 4.0 GeV−1 ≈ 0.8 fm, the breakup three-momentum of either decay product q
defined in the rest frame of the decaying state, and the orbital angular momentum L
between decay products. The angular correlation of the final state particles, which arises
due to total angular momentum conservation, is encoded in the spin factor S with the
Zemach tensor formalism [73, 74]. Finally, the propagator T describes the mass lineshape
of the intermediate two-body state.

5.2 Mass lineshapes

Many intermediate resonant contributions R are described by the relativistic Breit–Wigner
propagator
1
T (s) ≡ 2 , (4)
m0 + ξ(s) − s − i j gj2 ρj (s)
P

where s is the square of the two-body mass. The imaginary part is constrained by unitarity,
giving rise to the optical theorem which relates this component of the forward scattering
amplitude to the sum of probabilities of all possible transitions from the initial state to any
final state. This simplifies to the form given in Eq. (4), iterating over the j decay channels
that R may decay into with coupling gj and phase-space factor ρj . When considering
contributions only from decays into two stable particles, this further simplifies to
 2L+1
X m0 q
gj2 ρj (s) = m0 Γ(s) ≡ m0 Γ0 √ BD (q, q0 )2 , (5)
j
s q0

where q0 is the value of the break-up momentum at the resonance mass, m0 . The
energy-dependent decay width, Γ(s), is normalised to give the resonance width, Γ0 , when
evaluated at the resonance mass.
The energy-dependent dispersive correction ξ(s) is often neglected with notable excep-
tions. In particular, for contributions involving the broad ρ-resonance states, an analytic
dispersive term, known as the Gounaris–Sakurai model [75], is included to ensure unitarity
far from the pole mass. Other forms for ξ(s) appear in specific S-wave models to be
described in the following.

5.3 S-wave models

The K + π − and π + π − S-wave (L = 0) components of the B + → K + π + π − amplitude are
both large in magnitude and contain many overlapping resonances. In addition, the
kinematic ranges of these components include several relevant decay channel thresholds,
where increasing two-body mass opens additional decay channels that at lower mass are
kinematically inaccessible, which can modulate the intensity. This analysis includes three

7
distinct treatments of the S-wave components in B + → K + π + π − decays in an attempt to
better understand its behaviour. The use of multiple approaches also increases confidence
that parameters reported for the non-S-wave contributions are robust, and provides
additional information for further study.
As such, three sets of results are presented here, corresponding to the cases where
the K + π − and π + π − S-waves are described by: (i) a coherent sum of specific two-body
contributions (this model is referred to subsequently as “Isobar”); (ii) substitution of the
Isobar S-wave in π + π − with a monolithic, two-body unitarity-preserving model informed
by historical scattering data (“K-matrix”); and (iii) a quasi-model-independent binned
approach (“QMI”). All approaches contain the same contributions from higher-order
partial waves, where L > 0.

5.3.1 Isobar model

In addition to the K0∗ (1950)0 , f0 (1500) and f0 (1710) states described by the Breit–Wigner
function, the Isobar model S-wave amplitude includes dedicated lineshapes tuned to various
phenomena. The low K + π − mass region is modelled with the so-called Generalised LASS
(GLASS) lineshape [76], an extension to the unitarity-preserving LASS function [77,78] that
breaks two-body unitarity in scattering by allowing the complex coupling of the K0∗ (1430)0
resonance to vary relative to an underlying elastic πK–πK scattering contribution. This is
justified as additional phase motion may be induced by contributions from Kη (′) coupled
channels, from the I = 3/2 amplitude, from interaction between the companion π + and
the K + π − S-wave, and from penguin amplitudes. The first three of these have been
considered as potential sources of unitarity violation in the Kπ S-wave obtained from
D+ → K − π + π + decays [79]; the fourth is an additional possibility in B + → K + π + π −
decays.
As a function of the K + π − mass squared, s, the GLASS lineshape is given by [76]

AGLASS (s) ≡ ARes + AER = aGLASS sin δRes eiδRes e2iδER + aER sin δER eiδER ,


(6)

with
mK0∗ Γ(s)
 
δRes (s) = ϕNonres + arctan ,
m2K0∗ − s
 
1 rq
δER (s) = ϕER + arccot + , (7)
aq 2

where aGLASS and aER are free complex couplings of the model determined separately for
each B-meson charge. In the resonance-sensitive term ARes , mK0∗ is the K0∗ (1430)0 mass
and the usual energy-dependent width Γ(s) contains the K0∗ (1430)0 resonance width. In
the effective-range-like term AER , the parameters a and r, multiplying the K + π − breakup
momentum q, represent the scattering length and effective interaction length, respectively.
The parameters ϕNonres and ϕER are offsets to the scattering strong phases δRes and δER ,
respectively. In the δRes term, ϕNonres sets the strength of nonresonant entanglement; this
term approaches purely resonant behaviour when ϕNonres tends to zero. All parameters of
the GLASS lineshape are determined from data in this analysis.
For the broad scalar f0 (500) resonance at low π + π − mass, the modified Breit–Wigner
model from Bugg is used [80]. Besides f0 (500) → ππ decays, it includes contributions

8
from the f0 (500) → KK, ηη and 4π decay modes as well as dispersive effects in the elastic
channel. The baseline Isobar model has parameters fixed to the values from fit (iii) of
Ref. [80]. The Flatté function [81] is used to parameterise the f0 (980) resonance, whose
pole is close to the KK threshold. The couplings to the ππ and KK channels governing
the total width and the f0 (980) resonance mass are left as free parameters of the model.
Concerning the f0 (1370) state, where the physical lineshape is not clear at this time, a
simple T -matrix pole expression is used,
1
TPole (s) = , (8)
s0 − s
√
where the pole position, s0 = m0 −iΓ0 /2, is determined from dispersive phenomenological
studies with meson-meson scattering data [82]. The pole extracted from ππ → KK data
is taken, since that from ππ scattering in the elastic channel is less robust due to the need
to separate more contributions with different isospin. Moreover, trials with the pole from
the elastic channel give a much worse −2 log L value, as defined in Section 5.9, in the fit
to data.
For the lineshapes discussed above, only the direct production of the final state under-
going no further interaction is accounted for in the isobaric sum of S-wave resonant states.
However, rescattering effects need also be considered for a more complete description of
final-state interactions. Assuming the additional particle produced in the B decay does
not interact with the two-meson intermediate state, Watson’s theorem guarantees the
phase motion in decay will be identical to that of two-meson scattering [83]. Precise phe-
nomenological amplitude parameterisations based on ππ–ππ scattering data are available,
from which the parameters of the dispersively constrained (“CFD”) model are fixed in
this analysis [41].
This ππ–ππ scattering amplitude, |η00 (s)| exp(iδ00 (s)), contains a noninteractive term,
which is already represented in production by the S-wave isobaric sum, and therefore must
first be subtracted. An additional phenomenological form factor is typically included to
mimic a mild s-dependence of the scattering terms for the initial-state two-meson pair.
It comes from the partonic decay amplitude that produces the three-meson final state,
in which the relative momentum between the pair of mesons is distributed among the
quarks in the momentum loop within the microscopic amplitude, e.g. the tree and penguin
diagrams, and probes the internal structure of the mesons involved in the initial and final
states [84, 85].
The overall ππ–ππ rescattering amplitude is then given by
0
|η 0 (s)|eiδ0 (s) − 1
Tππ–ππ (s) ≡ 0 , (9)
1 + s/Λ2π

where Λπ would be the only free parameter of the model. Studies with this rescattering
model, however, indicate that Λπ , which is typically taken to be of order 1 GeV [9–11],
tends to infinity, indicating that the partonic form factor is not essential to the model.
Therefore, the denominator of the expression in Eq. (9) is set to unity.
While the concept of ππ–KK rescattering was originally developed within the context
of two-body interactions, for three-body decays rescattering means that a pair of mesons
produced in one channel will appear in the final state of a coupled channel. Theoretically
motivated amplitude models have been derived from scattering data with constraints from

9
πK–πK and ππ–KK dispersion relations [42], from which the parameters referred to as
the “CFDB ” set are taken for this analysis. The alternative constrained (“CFDC ”) set
includes spurious experimental data that are in conflict with Watson’s theorem near the
KK threshold, and hence is not considered in this analysis.
This ππ–KK scattering amplitude is presented as unnormalised, so in order
p to facilitate
comparison with the data, it is multiplied by the phase-space factor 4 qπ qK /s, where
q is the usual breakup momenta for the indicated two-meson final state.1 The overall
ππ–KK rescattering amplitude is then given by [42]
0
p
4 qπ (s)qK (s) |g00 (s)|eiϕ0 (s)
Tππ–KK (s) ≡ i √ , (10)
s 1 + s/Λ2K
where the partonic form factor that modulates the unknown production magnitude is
similarly applied as in Eq. (9), within which ΛK would be the only free parameter of the
model. However, studies with this rescattering model indicate the partonic form factor
can also be discarded from the overall model, and hence the denominator of the right-most
term in Eq. (10) is set to unity.

5.3.2 K-matrix model

The coherent sum of resonant contributions modelled with Breit–Wigner lineshapes can be
used to describe the dynamics of three-body decays when the quasi-two-body resonances
are relatively narrow and isolated. However, when there are broad, overlapping resonances
(with the same isospin and spin-parity quantum numbers) or structures that are near
open decay channels, this model does not satisfy S-matrix unitarity, thereby violating the
conservation of quantum mechanical probability current.
Assuming that the dynamics is dominated by two-body processes (i.e. that the
S-wave does not interact with other decay products in the final state), then two-body
unitarity is naturally conserved within the K-matrix approach [87]. This approach was
originally developed for two-body scattering [88] and the study of resonances in nuclear
reactions [89, 90], but was extended to describe resonance production and n-body decays
in a more general way [91]. In this analysis, the K-matrix approach is applied exclusively
to the π + π − S-wave, while the K + π − S-wave retains the lineshape model of the Isobar
approach.
The K-matrix is parameterised as a sum over the n coupled channels,
u ∈ ππ, 4π, KK, ηη, ηη ′ , as


n
X
Tu (s) ≡ [(Iˆ − iK̂(s)ρ̂(s))−1 ]uv P̂v (s) , (11)
v=1

where the diagonal matrix ρ̂ accounts for the amount of phase space available to each
channel. The scattering matrix, K̂, is taken as the combination of the sum over the f0 (500),
f0 (980), f0 (1370), f0 (1500) and f0 (1710) bare mass poles together with nonresonant, or
“slowly varying”, parts and is fixed from a global analysis of ππ scattering data [76, 92].
Differences in the initial-state environment between particle scattering and decay are
1
This analysis is based on a preliminary ππ-KK √ scattering model adapted from Ref. [86] in which the
definition of q was incorrect by a factor of 2/ s, with s given in GeV. The difference with the correct
version shown in Eq. (10) is included as a model systematic uncertainty.

10
accounted for by splicing the P̂ production vector with the scattering propagator. The
parameterisation of P̂ is similar in structure to K̂ with the exception that the channel
couplings are complex and different for B + and B − decays to allow for CP violation.
These couplings are free parameters of the model. However, there is no sensitivity to the
parameter governing the shape of the slowly varying part in production, sprod
0 , which is a
part of P̂ . Therefore, this parameter is fixed from an external measurement [93].

5.3.3 Quasi-model-independent analysis

The QMI approach simply describes each of the Kπ and ππ S-waves by binning in mass and
allowing a CP -violating complex parameter for each bin with no other structure considered.
This set of parameters, varied in the fit, thus represents the average amplitude in each
bin. This approach exploits the distinctive flat angular structure of the S-wave amplitude
to disentangle this component from other contributions to the phase space, assuming
the higher-order waves to be well modelled by the isobar approximation. Building upon
the technical implementation of the B + → π + π + π − amplitude analysis from the LHCb
collaboration [10, 11], which in turn developed methods pioneered by previous works such
as Ref. [79], both the overlapping K + π − and π + π − S-wave contributions are modelled.
This marks the first attempt with such an approach in which multiple partial waves
produced by particle decay are simultaneously modelled. The binning is largely ad hoc,
except that bin edges are tuned to align with the kinematic openings of coupled channels,
since cusps in the amplitude are always expected at these thresholds from scattering theory.
In K + π − the relevant thresholds are for the K3π, Kη, Kη ′ , Kηc and Ds D channels, while
in π + π − the 4π, KK, ηη, ηη ′ , ηηc , DD and Ds+ Ds− /η ′ ηc thresholds are considered. Each
S-wave is assigned 50 bins. In K + π − 35 of these bins are allocated below the open charm
threshold, while in π + π − this number is set to 40. Overlapping crossing binned S-waves
contain an inherent mathematical ambiguity in the expression of their sum that is not
present when a singular partial wave is modelled [94]. For fit convergence, it is therefore
essential to fix a value for each component of the couplings, i.e. x, y, ∆x and ∆y, in one
QMI bin. This bin need not necessarily be the same for each coupling component. As
far as the −2 log L value (see Eq. (19) below) is concerned, the value in each of the fixed
bins may take on any finite number. However, these should be as close as possible to
their true physical values in the interest of obtaining meaningful separation between the
crossing partial S-waves. For this purpose, one bin in each coupling component is fixed
from the measured Isobar amplitude, by means of a one-dimensional χ2 scan, ultimately
selecting the bin that leads to the closest agreement between the Isobar prediction and the
QMI S-wave fit results. Due to the potential for bias in the derived one-dimensional QMI
S-wave amplitudes, it is also necessary to include an additional systematic uncertainty on
these projections, accounting for the full spread of the amplitude value in the fixed bins
coming from the Isobar result. The two-dimensional sum of the K + π − and π + π − QMI
S-waves is unaffected by this effect.

5.4 Measurement quantities

The primary outputs of the Dalitz-plot fit are the complex coefficients a±j defined in
Eq. (1). However, since these are convention-dependent, they have limited physical
meaning apart from their phase values, arg(a±
j ), which are reported. It is often useful

11
to compare CP -averaged fit fractions for each intermediate component, j, defined as
Fj = Fj− + Fj+ , where
R ±
|aj Aj (Φ3 )|2 dΦ3
Fj± ≡ R P − , (12)
(| j aj Aj (Φ3 )|2 + | j a+
P 2
j Aj (Φ3 )| ) dΦ3

for charge-separated B ± decays. These fit fractions alone will not sum to unity if there is
net constructive or destructive interference. Such effects are described by the interference
fit fractions, given by
2Re[a± ±∗ ∗
R
± j ak Aj (Φ3 )Ak (Φ3 )] dΦ3
Ij < k ≡ R P − , (13)
(| j aj Aj (Φ3 )|2 + | j a+
P 2
j Aj (Φ3 )| ) dΦ3

where the sum over all fit and interference fractions is unity by definition. Another
important physical quantity is the quasi-two-body parameter of CP violation in decay
associated with a particular intermediate contribution,
Fj− − Fj+
AjCP ≡ − . (14)
Fj + Fj+

The violation of CP originating from long-distance interference effects cannot be

quantified through an asymmetry parameter defined analogously to Eq. (14) by replacing
F ± with the corresponding I ± terms of Eq. (13), as the denominators could become null
or negative. Instead, this effect is described within the quasi-two-body asymmetry of the
combined components,
− +
Fjk − Fjk Fj Fk
Ajk
CP ≡ − + = AjCP + AkCP + 2Ajk
I , (15)
Fjk + Fjk Fjk Fjk

where Ajk − + − +
I = (Ijk − Ijk )/Fjk . The compound fit fractions with two indices, Fjk and Fjk ,
are determined using Eq. (12), with coherent sums of the j and k components instead
− +
comprising the numerator, and Fjk = Fjk + Fjk . In the decomposition of Eq. (15), the
first two terms arise from quasi-two-body CP asymmetry in their respective components,
while the third term, AjkI , quantifies CP asymmetry in their interference.

5.5 Square Dalitz plot

Since resonances tend to populate the edges of the conventional Dalitz plot in charmless
B decays, it is useful to define the so-called “square” Dalitz plot [95], which provides
improved resolution in these critical regions when using uniform binning, for example
when modelling experimental effects. Furthermore, the mapping to a square space aligns
the constant bin boundaries to the kinematic phase-space boundaries. The square Dalitz
plot is defined in terms of the variables m′ and θ′
mπ+ π− − mmin
 
′ 1 π+ π− ′ 1
m ≡ arccos 2 max − 1 , θ ≡ θπ + π − , (16)
π mπ+ π− − mmin
+
π π − π

where mmax min

π + π − = mB + − mK + and mπ + π − = 2mπ + represent the kinematic limits of the
π π mass permitted in the B → K + π + π − decay, and θπ+ π− is the angle between the K +
+ − +

12
and π − three-momentum vectors in the π + π − rest frame. Transition between phase-space
bases, dΦ3 → |det J|dm′ dθ′ , is facilitated by the Jacobian of the transformation,
′ ′
|det J| = 2π 2 |⃗pπ− ||⃗pK + |mπ+ π− (mmax min
π + π − − mπ + π − ) sin(πm ) sin(πθ ) , (17)

where the three-momenta p⃗ are also evaluated in the π + π − rest frame.

All efficiencies and backgrounds described in Sections 5.6 and 5.8 are determined
as functions of the square Dalitz-plot variables, modelled as histograms with a uniform
26 × 25 binning scheme in m′ and θ′ . These histograms are determined separately for
B + and B − decays and further smoothed by a two-dimensional cubic spline to mitigate
effects of discontinuity at the bin edges, with bins abutting kinematic boundaries reflected
to ensure good behaviour at the edge of the phase space. The D0 and J/ψ vetoes are
not included in these maps, as they cover very narrow regions of phase space where the
subsequent binning and smoothing of the histograms would otherwise lead to biased
results.

5.6 Efficiency model

The efficiency of selecting a signal decay is parameterised in the two-dimensional square
Dalitz plot and determined separately for B + and B − decays. Nonuniformities in the
efficiency of selecting a signal decay arise as a result of the detector geometry, trigger
and reconstruction algorithms, particle identification selections, production and detection
asymmetries, and other background rejection requirements such as that imposed by the
BDT classifier to discriminate against combinatorial background. The efficiency map is
first obtained using simulated decays, however differences between data and simulation
arising in the aforementioned effects are each calibrated with data control samples. These
corrections are applied to the baseline model except for those related to the combinatorial
background suppression BDT, which are considered in the systematic uncertainties.
The effect of an asymmetry between the production rates of B − and B + mesons is
indistinguishable in this analysis from a global detection efficiency asymmetry. Therefore,
the B + production asymmetry, as measured within the LHCb acceptance [96], is taken
into account by introducing a global asymmetry of approximately −0.6% into the efficiency
maps. This is obtained as an average of the measured production asymmetries, weighted
by the relative integrated luminosity obtained in 2011 and 2012.
The hardware-trigger efficiency correction is calculated using pions from D0 → K − π +
decays, arising from promptly produced D∗+ → D0 (→ K − π + )π + decays, and affects two
disjoint subsets of the selected candidates: those where the trigger requirements were
satisfied by hadronic calorimeter deposits as a result of the signal decay and those where
the requirements were satisfied only by deposits from the rest of the event. In the first
case, the probability to satisfy the trigger requirements is calculated using calibration
data as a function of the transverse energy of each final-state particle of a given species,
the dipole-magnet polarity, and the hadronic calorimeter region. In the second subset,
a smaller correction is applied following the same procedure in order to account for the
requirement that these tracks did not fire the hardware trigger. These corrections are
combined according to the relative abundance of each category in data.
Prompt J/ψ → µ+ µ− decays provide a calibration of the tracking efficiency [97], by
comparing the performance of the default track-finding algorithm to that obtained when
dropping information from at least one subdetector. This calibration is determined as a

13
Figure 3: Square Dalitz-plot distributions for the (left) B + and (right) B − signal efficiency
models, ϵ(m′ , θ′ ), smoothed using a two-dimensional cubic spline. Depleted regions that traverse
the plots arise where more stringent PID requirements are applied near the D0 meson.

function of the track momentum and pseudorapidity, and of the multiplicity of the event,
and is assumed to factorise with respect to the final-state tracks so that the efficiency
for each track is multiplied to form the overall efficiency. The particle identification
efficiency, also assumed to factorise, is calculated from calibration data corresponding to
the D∗+ → D0 (→ K − π + )π + decay, where pions and kaons can be identified without the
use of the LHCb particle identification system [98]. The particle identification efficiencies
for the background-subtracted pions and kaons are parameterised in terms of their total
and transverse momentum, and the number of tracks in the event. Finally, with the
approach devised for the b-hadron production asymmetry measurements [96], residual
detection asymmetries not already accounted for are determined from prompt charm
control samples in bins of momentum.
The overall efficiency, as a function of square Dalitz-plot position, can be seen in Fig. 3
for B + and B − decays separately. Here, and in similar plots, depleted regions associated
to the more stringent PID requirements placed around the D0 meson are visible.

5.7 Resolution effects

Due to narrow π + π − mass structures, particularly the ω(782) and χc0 (1P ) resonances,
the signal PDF is convolved with a one-dimensional Gaussian function representing the
detector resolution, R. Since there are no structures in the mK + π− spectrum that are
narrow compared to the detector resolution, only a one-dimensional smearing is required.
The resolution itself is taken from simulated decays depending on the mass square sπ+ π− .
The signal PDF for B + or B − decays is then given by

± ϵ± (m′ , θ′ )|A± (Φ3 )|2 ⊗ R(sπ+ π− )

PSig (Φ3 ) =Z , (18)
(ϵ− (m′ , θ′ )|A− (Φ3 )|2 + ϵ+ (m′ , θ′ )|A+ (Φ3 )|2 ) dΦ3

where ϵ± represents the Dalitz-plot dependent efficiency for the B ± decay. Furthermore,
the convolution is not applied to the denominator as this would amount to a trivial
additive constant in the −2 log L value (see Eq. (19) below) due to Fubini’s theorem.

14
Figure 4: Square Dalitz-plot distributions for the (left) B + and (right) B − combinatorial
background models, scaled to represent their respective yields in the signal region.

Figure 5: Square Dalitz-plot distributions for the misidentified B + → π + π + π − background

models in (left) B + and (right) B − samples, scaled to represent their respective yields in the
signal region.

5.8 Background model

The dominant source of background in the signal region is combinatorial in nature,
modelled using upper-sideband data from the region 5.65 < mK + π+ π− < 6.10 GeV/c2 .
The corresponding combinatorial background distributions can be seen in Fig. 4. In the
Dalitz-plot fit, the charge asymmetry in the combinatorial background yield is fixed to
that obtained in the B + mass fit described in Section 4.
Sources of peaking background arise from misidentified decays reconstructed under
the B + → K + π + π − decay hypothesis. These backgrounds, including B + → π + π + π − and
B + → π + K + K − decays, are modelled using simulated events with corrections to account
for differences between data and simulation applied in the same way as for the signal
efficiency model. To account for the phase-space distributions of these backgrounds, the
events are further weighted according to their respective amplitude models obtained by
the LHCb collaboration [9–11]. The corresponding distributions can be seen in Figs. 5
and 6.
A final category of partially reconstructed peaking background originates from
B + → K + η ′ (→ π + π − γ) decays. These are treated in the same way as the misidenti-
fied peaking backgrounds except that the corrected simulation is reweighted by the
η ′ → π + π − γ amplitude model from the BESIII collaboration [66]. The background distri-

15
Figure 6: Square Dalitz-plot distributions for the misidentified B + → π + K + K − background
models in (left) B + and (right) B − samples, scaled to represent their respective yields in the
signal region.

Figure 7: Square Dalitz-plot distributions for the partially reconstructed B + → K + η ′ (π + π − γ)

background models in (left) B + and (right) B − , scaled to represent their respective yields in
the signal region.

butions are shown in Fig. 7. All background models transformed back to the traditional
Dalitz plot can be found in Appendix B.

5.9 Fit procedure

Of the three approaches to the S-wave, the Isobar and K-matrix fits are performed using
the —Laura++— Dalitz-plot fitter package v3r5p3 [99], which interfaces to the MINUIT
function minimisation algorithm [100, 101]. In contrast, the QMI approach relies on the
—Mint2— [102] amplitude-analysis interface to Minuit2 [101]. The fundamental difference
between these amplitude-analysis software packages is in the handling of the normalisation.
The former approximates the definite integral by employing a Gaussian quadrature
approach, while the latter invokes a stochastic sampling technique. Additionally, due
to the size of its parameter space, the QMI approach greatly benefits from the use of
GPU-accelerated solutions.
In all cases, the K ∗ (892)0 component is set to be the “reference” amplitude. In practice,
this means that the average magnitude of the B + and B − coefficients for this component
is set to unity, i.e. x = 1 in Eq. (2), while the ∆x parameter is left free to vary to allow
for CP violation. Since there is no sensitivity to the phase difference between the B + and

16
Table 2: Components of the K + π − and π + π − S-waves in each of the three approaches considered
in the analysis, together with their fixed lineshape parameters.

Component Approach Mass lineshape, T (s) Fixed parameters

(K + π − )Res Isobar/K-matrix GLASS [76, 78] —
(K + π − )ER
K0∗ (1950)0 Isobar/K-matrix Breit–Wigner m0 = 1.945 ± 0.022 GeV/c2 [63]
Γ0 = 0.201 ± 0.086 GeV [63]
f0 (500) Isobar Bugg fit (iii) of Ref. [80]
f0 (980) Isobar Flatté [81] —
f0 (1370) Isobar T-matrix pole m0 = 1.38 ± 0.07 GeV/c2 [82]
Γ0 /2 = 0.22 ± 0.08 GeV [82]
f0 (1500) Isobar Breit–Wigner m0 = 1.506 ± 0.006 GeV/c2 [63]
Γ0 = 0.112 ± 0.009 GeV [63]
f0 (1710) Isobar Breit–Wigner m0 = 1.704 ± 0.012 GeV/c2 [63]
Γ0 = 0.123 ± 0.018 GeV [63]
ππ–ππ Isobar Rescattering CFD of Ref. [41]
ππ–KK Isobar Rescattering CFDB of Ref. [42]
π + π − S-wave K-matrix K-matrix [87–91] Refs. [76, 92, 93]
K + π − and QMI Constant per bin —
π + π − S-waves

B − amplitudes, the imaginary part of the K ∗ (892)0 component is set to zero for both B +
and B − (y = ∆y = 0), which means that the coefficients of all other contributions to the
model are measured relative to this component.
The extended likelihood function that is optimised is of the form
P "
NCand X
#
e− j Nj Y i
L= Nj Pj (Φ3 ) , (19)
N! i=1 j

where Nj is the yield for the candidate category j (fixed from Table 1), NCand is the total
number of candidates in the signal region, and Pj is the probability density function for
candidates in category j in terms of the Dalitz-plot coordinates, Φi3 . The B + and B −
data are modelled simultaneously and the optimal values of the fitted parameters are
found by minimising twice the negative log-likelihood, −2 log L.
Contrary to the majority of amplitude analyses, there is no formal model building
procedure conducted to optimise which resonant contributions to include per se. This is
because every resonance that is considered as established in Ref. [63], and which plausibly
decays to either K + π − or π + π − in the charmless region, is found to be individually
significant considering statistical uncertainties only, based on the reduction in −2 log L for
the additional degrees of freedom. Since Dalitz-plot analyses are susceptible to multiple
solutions, depending upon the initial parameter values the results may correspond to
a local, rather than global, minimum of the −2 log L function. To attempt to find the
global minimum, a large number of fits are performed where the initial values of the
complex coefficients are randomised. The fit result with the smallest −2 log L value out

17
Table 3: Resonances common to each S-wave approach, together with their default lineshapes
and fixed lineshape parameters.

Resonance Spin Mass lineshape, T (s) Fixed parameters [63]

χc0 (1P ) 0 Breit–Wigner m0 = 3.41471 ± 0.00030 GeV/c2
Γ0 = 0.0105 ± 0.0008 GeV
K ∗ (892)0 1 Breit–Wigner m0 = 0.89555 ± 0.00020 GeV/c2
Γ0 = 0.0473 ± 0.0005 GeV
K ∗ (1410)0 1 Breit–Wigner m0 = 1.414 ± 0.015 GeV/c2
Γ0 = 0.232 ± 0.021 GeV
K ∗ (1680)0 1 Breit–Wigner m0 = 1.718 ± 0.018 GeV/c2
Γ0 = 0.322 ± 0.110 GeV
†
ρ(770)0 1 Gounaris–Sakurai [75] m0 = 0.7690 ± 0.0009 GeV/c2
Γ0 = 0.1509 ± 0.0017 GeV
ω(782) 1 Breit–Wigner m0 = 0.78266 ± 0.00013 GeV/c2
Γ0 = 0.00868 ± 0.00013 GeV
ρ(1450)0 1 Gounaris–Sakurai [75] m0 = 1.465 ± 0.025 GeV/c2
Γ0 = 0.400 ± 0.060 GeV
ρ(1700)0 1 Gounaris–Sakurai [75] m0 = 1.720 ± 0.020 GeV/c2
Γ0 = 0.250 ± 0.100 GeV
K2∗ (1430)0 2 Breit–Wigner m0 = 1.4324 ± 0.0013 GeV/c2
Γ0 = 0.109 ± 0.005 GeV
f2 (1270) 2 Breit–Wigner m0 = 1.2755 ± 0.0008 GeV/c2
Γ0 = 0.1859 ± 0.0028 GeV
f2′ (1525) 2 Breit–Wigner m0 = 1.5174 ± 0.0025 GeV/c2
Γ0 = 0.0869 ± 0.0023 GeV
K3∗ (1780)0 3 Breit–Wigner m0 = 1.779 ± 0.008 GeV/c2
Γ0 = 0.161 ± 0.017 GeV
ρ3 (1690)0 3 Breit–Wigner m0 = 1.686 ± 0.004 GeV/c2
Γ0 = 0.186 ± 0.014 GeV
†
The “neutral only, other reactions” values from Ref. [63] are used.

of this ensemble is then taken to be the baseline result for each S-wave method. No
secondary solutions are found within 25 units of −2 log L of this baseline. A summary of
the components in each of the three models considered in this analysis is given in Tables 2
and 3.

6 Systematic uncertainties
Sources of systematic uncertainty are separated into two major categories: those that arise
from experimental effects and those from the inherent lack of knowledge of the amplitude
models. The sources are described below, with the numerical systematic uncertainties for
the CP -averaged fit fractions, quasi-two-body CP asymmetries, and relative phases for B +

18
and B − decays summarised in Appendix C. Throughout this section, labels in brackets
are assigned to facilitate matching with each source listed in the systematic-uncertainty
tables.
Beginning with the experimental systematic uncertainties, the uncertainties on the
signal yield and the background yields and asymmetries, given in Table 1, comprise
a statistical component as well as systematic effects due to the B-candidate mass fit
procedure. Apart from varying parameters of the baseline model within one standard
deviation, such experimentally motivated models are themselves changed in order to
mitigate imperfections in the mK + π+ π− model. The uncertainty arising from assumptions
regarding the signal parameterisation is found by releasing the underlying Gaussian
of its restrictions relative to the core Crystal Ball function. Similarly, the model for
the combinatorial background is replaced with a second-order polynomial. The relative
fractions between the remaining peaking backgrounds are varied within two standard
deviations as they are derived from simulation, while their charge asymmetries are varied
within their world-average expectations [67]. The combined statistical and systematic
uncertainties on the signal and background yields and asymmetries are then propagated to
the amplitude analysis, where those variations causing the largest upward and downward
deviations with respect to the baseline yield and charge asymmetry values are taken to
form a covariance matrix between these parameters. Values sampled from this matrix are
then fixed in the amplitude-analysis fit to assign the systematic uncertainty relating to
the three-body B-candidate mass fit (“B mass”).
To account for the statistical uncertainty on the efficiency description, an ensemble of
efficiency maps is created by sampling bin-by-bin from the baseline description, according
to uncorrelated Gaussian distributions with means corresponding to the calibrated central
value of the efficiency in each bin, and widths corresponding to the uncertainty derived from
data. Sources of these systematic uncertainties include the B + production asymmetry [96],
tracking [97], particle identification [98] and residual detection asymmetries [96]. To
account for extraneous potential biases in the method used to correct the hardware trigger
efficiency, an alternative method using B 0 → J/ψ(→ µ+ µ− )K + π − decays in the region
of the K ∗ (892)0 resonance is employed. Corrections to the simulation are determined
from events first triggered by muons from the J/ψ decay, with the trigger decision then
recomputed for the kaon and pion tracks [103, 104]. Additionally, to account for potential
variation of the efficiency within a square Dalitz-plot bin, the efficiency map is constructed
using a finer binning scheme. The standard deviation of the distribution of resulting Dalitz-
plot fit parameters obtained with these variations are then taken to be the associated
systematic uncertainty (“Eff”).
As the detector resolution in m2π+ π− convolved into the signal model is determined
from simulation, this candidate-dependent resolution is increased by a global scale factor
determined from data. This is found from a one-dimensional fit to mπ+ π− in the immediate
vicinity of the prominent χc0 (1P ) contribution, which is clearly visible Fig. 2. Adopting
a Voigtian function for the signal and a polynomial for the remainder, the difference
from the baseline fit parameters with the scaled resolution is taken as their associated
systematic uncertainties (“Resol”).
The uncertainties on the combinatorial and peaking background distributions are
propagated to the Dalitz-plot fit results with an ensemble procedure similar to that for
the efficiency maps. To account for a scenario in which the combinatorial background
phase-space structure varies with the B-candidate mass, an alternative combinatorial

19
map much closer to the signal region 5400 < mK + π+ π− < 5650 MeV/c2 is produced, where
contamination from B + decays is still minimal. As peaking-background models derive from
simulation, the systematic effects considered for the efficiency maps are also accounted for
here. Furthermore, the amplitude-model weights in phase space are varied within their
known uncertainties [9–11, 66]. The standard deviation in the variation of the subsequent
Dalitz-plot fit results is taken to be the systematic uncertainty due to these effects (“Bkg”)
and generally constitutes the dominant systematic uncertainty in this analysis.
Systematic uncertainties related to possible intrinsic fit bias are investigated using an
ensemble of pseudoexperiments based on the baseline fit result (“Fit bias”). Differences
between the input and fitted values from the ensemble for the fit parameters are generally
found to be small. However, for the fits involving the QMI approach, an additional bias
arises from the intrinsic limitations in the capacity of the approach to reproduce the
underlying analytic S-wave. The cause of such a bias is primarily due to the approximation
of the analytic lineshape by a constant amplitude in each bin; this is by far the dominant
systematic uncertainty in this approach. Its evaluation involves reusing the ensemble of
pseudoexperiments generated for estimating the Isobar fit bias, fitting them with the QMI
model, and determining the difference between the obtained and true bin-averaged values
of the S-wave amplitude (“QMI bias”). Systematic uncertainties are assigned as the sum
in quadrature of the difference between the input and output values and the uncertainty
on the mean of the output value determined from a fit to the ensemble.
The remaining sources of uncertainty are endemic to the amplitude model and are
presented as a separate major category because their improvement typically lies outside
the realm of experiment. Lack of knowledge on the radius parameter of the ad hoc
Blatt–Weisskopf barrier factors is accounted for by modifying its value between 3 and
5 GeV−1 using a flat prior, with uncertainties calculated from an ensemble (“Barrier”).
The systematic uncertainty due to fixed partial-wave shape parameters (“Lineshape”)
such as masses and widths is assigned with an ensemble technique, where their values for the
resonances common to each S-wave approach are fluctuated according to the uncertainties
listed in the Particle Data Group tables [63]. Specific to the Isobar approach, the S-wave
lineshape variations are made according to the uncertainties listed in Table 2, whereas
for the K-matrix an alternative fit is performed releasing the slowly-varying production
parameter, sprod
0 . Use of the corrected ππ–KK rescattering lineshape described in Eq. (10)
is also included in this category for the Isobar model. Systematic uncertainties due to the
fixed values of the parameters reported in Tables 2 and 3 are negligible, and the impact of
updated measurements of these parameters since Ref. [63] was published is also neglected.
Finally, potential additional resonant contributions to the amplitude are considered
through the inclusion of the K4∗ (2045)0 and f2 (1430) states into the model, with the
systematic impact of each summed in quadrature (“Extra”).
In general, the largest sources of systematic uncertainty for the Isobar and K-matrix
approaches are due to the combinatorial-background distribution across phase space. For
the CP -averaged fit fractions this source dominates the uncertainty, while for the CP
asymmetries, which are more robust against such effects, the statistical and systematic
uncertainties are somewhat comparable. For the relative phases between components, the
background and efficiency models comprise the dominant systematic uncertainty, while
the impact of additional resonances is the largest contributor to the model uncertainty.
The total systematic uncertainty combines the effects of each category in quadrature as
they are considered to be uncorrelated.

20
Table 4: Properties of the global −2 log L minima for each of the three S-wave approaches with
respect to the K-matrix result.

Isobar K-matrix QMI

∆(−2 log L) −24 0 −291
∆(ndf) −9 0 +338

7 Results
Information on the relative global minima of the −2 log L function between the baseline
results in each of the three S-wave approaches can be found in Table 4. Naturally,
the QMI approach gives the best absolute minimum, however the additional number of
free parameters by which this is achieved compared to the other approaches indicates
comparable fit quality and hence that the Isobar and K-matrix S-wave models are fairly
reliable. It is also interesting to note the comparison between the Isobar and K-matrix
minima. The improvement in −2 log L for fewer degrees of freedom supports the use of
more physically motivated dispersive S-wave lineshapes, compared to what has been done
previously in similar analyses, as this is the first incidence of the K-matrix being surpassed
in this fit-quality metric by an isobaric approach in which all lineshapes have physical
meaning.
The goodness of fit is assessed by comparing the fit model with the data in square
Dalitz-plot bins that can be seen in Appendix D and determining an associated χ2 value.
A binning common to B + and B − decays is chosen through an adaptive procedure that
requires an approximately constant number of candidates from the combined B + and
B − data samples in each bin. For various values of the required number of candidates
per bin, the ratio of the χ2 (in the range 5000–5800 for 3600 bins) to the number of bins
is approximately 1.6 accounting for statistical uncertainties only. Given the impact of
the systematic uncertainties on the results discussed in Section 6, the agreement of the
fit models with the data is reasonable. Smaller χ2 values tend to be obtained for the
S-wave models with larger numbers of free parameters, such that all three approaches
have comparable goodness-of-fit overall. The exception is for B + decays, where the χ2
metric favours the K-matrix model over the Isobar. The distribution in the square Dalitz
plot of bins that contribute significantly to the χ2 does not reveal any clear source of
mismodelling.
Due to the number of numerical results involved in this analysis, it is not practical to
tabulate every measured value and uncertainty necessary to replicate a complete expression
of the amplitudes. Therefore, these are recorded by electronic means as discussed in the
Supplemental Material [105]. In particular, these contain additional tables of interference
fractions and their CP -violating asymmetries as defined in Eq. (15) that are not further
discussed here in the text. Instead, the focus in the main body of this paper is placed
on the measured quasi-two-body quantities and relative phases, while the free lineshape
parameters appear in Appendices E and F.

21
F (%)

F (%)
10
10
1

1 10−1

LHCb 10−2 LHCb

3 fb−1 3 fb−1
10−1

ω (782)

f 2(1270)

f 0(500)
f 0(980)
f 0(1370)
f 0(1500)
f 0(1710)
f 2'(1525)
(K +π −)Res
K (892)0

K (1410)0

K (1680)0

K 2(1430)0

K 3(1780)0

K 0(1950)0

ρ (770)0

ρ (1450)0
ρ (1700)0

ρ3(1690)0
(K +π −)ER

π π -K K
π π -π π
*

*
*

*
Figure 8: Comparison between Isobar (green), K-matrix (magenta) and QMI (blue) fit fractions
for states decaying to (left) K + π − and (right) π + π − . Solid bands show the statistical spread,
with progressively lighter shading indicating the combined statistical and systematic uncertainties,
followed by the total uncertainty including that from the model.

7.1 Fit fractions

The CP -averaged fit fractions are given in Table 5 for all three S-wave approaches and
are visually represented in Fig. 8 to ease comparison. In all cases, statistical uncertainties
are calculated using 68% confidence intervals obtained from the results of fits performed
to data sets sampled from the nominal fit models. Generally, the statistical uncertainty
is lowest for the model with the fewest parameters (Isobar), and highest for the model
with the largest number of parameters (QMI). For the prominent K ∗ (892)0 and ρ(770)0
vector states, and the K2∗ (1430)0 and f2 (1270) tensor states, the systematic uncertainties
overwhelm the statistical error, whereas for the remaining contributions, the tendency
is for these to be at comparable levels. Another trend is that the model uncertainty
dominates if the associated state has low statistical significance. Including systematic and
amplitude-model sources of uncertainty, several intermediate states are observed for the
first time as discussed in a companion Letter [14].
While the total fit fractions given in Table 5 may appear to be in disagreement between
the S-wave approaches, this quantity can be deceiving. As composite objects, the internal
interference-fraction contributions between terms comprising the K-matrix and QMI
S-waves are included in their total fit fractions, while in the Isobar approach, these are
absent. To facilitate a meaningful comparison between approaches, the total S-wave is
taken as a single amplitude in each approach. Given in Table 6, the S-wave sum and total
fit fractions for the B + and B − amplitudes are found to be in excellent agreement between
approaches. When split into individual overall K + π − and π + π − S-waves, the Isobar and
K-matrix results are not in agreement, whereas the QMI results are in agreement with
both, showing a slight preference for the Isobar results.

7.2 CP asymmetries
Quasi-two-body CP asymmetries associated to each component are shown in Table 7
and Fig. 9, also demonstrating good agreement between the S-wave approaches. Several
significant measurements of CP violation can be noted, however an in-depth discussion is
left to another companion Letter [15]. As far as the uncertainties are concerned, many
of the points discussed for the fit fractions remain relevant. However, because of the

22
Table 5: CP -averaged fit fractions for each approach, where the first uncertainty is statistical,
the second the experimental systematic and the third is the model systematic.

Fit fraction (%)

Component Isobar K-matrix QMI
K ∗ (892)0 12.167 ± 0.128 ± 0.494 ± 0.157 12.071 ± 0.118 ± 0.397 ± 0.047 12.669 ± 0.223 ± 0.304 ± 0.286
K ∗ (1410)0 0.700 ± 0.092 ± 0.149 ± 0.053 0.517 ± 0.088 ± 0.169 ± 0.037 0.891 ± 0.166 ± 0.343 ± 0.050
K ∗ (1680)0 0.219 ± 0.060 ± 0.069 ± 0.029 0.223 ± 0.065 ± 0.077 ± 0.024 0.322 ± 0.111 ± 0.111 ± 0.018
ρ(770)0 7.834 ± 0.192 ± 0.579 ± 0.301 8.122 ± 0.194 ± 0.543 ± 0.101 7.326 ± 0.259 ± 0.543 ± 0.158
ω(782) 0.202 ± 0.030 ± 0.033 ± 0.009 0.219 ± 0.035 ± 0.037 ± 0.002 0.221 ± 0.032 ± 0.027 ± 0.004
ρ(1450)0 1.213 ± 0.212 ± 0.123 ± 1.016 0.881 ± 0.153 ± 0.095 ± 0.081 1.908 ± 0.409 ± 0.460 ± 0.102
ρ(1700)0 0.340 ± 0.088 ± 0.076 ± 0.372 0.317 ± 0.073 ± 0.070 ± 0.035 0.351 ± 0.142 ± 0.223 ± 0.009
K2∗ (1430)0 2.252 ± 0.125 ± 0.259 ± 0.157 2.434 ± 0.117 ± 0.265 ± 0.083 2.220 ± 0.119 ± 0.130 ± 0.229
f2 (1270) 2.152 ± 0.110 ± 0.242 ± 0.136 2.145 ± 0.094 ± 0.224 ± 0.087 2.442 ± 0.174 ± 0.337 ± 0.227
f2′ (1525) 0.056 ± 0.019 ± 0.050 ± 0.015 0.047 ± 0.019 ± 0.030 ± 0.011 0.053 ± 0.021 ± 0.023 ± 0.019
K3∗ (1780)0 0.222 ± 0.039 ± 0.036 ± 0.024 0.274 ± 0.042 ± 0.071 ± 0.029 0.420 ± 0.084 ± 0.145 ± 0.158
ρ3 (1690)0 0.408 ± 0.047 ± 0.056 ± 0.076 0.439 ± 0.042 ± 0.050 ± 0.029 0.258 ± 0.043 ± 0.038 ± 0.085
χc0 (1P ) 2.000 ± 0.065 ± 0.093 ± 0.028 1.927 ± 0.052 ± 0.079 ± 0.007 1.823 ± 0.077 ± 0.146 ± 0.014
(K + π − )Res 13.076 ± 0.633 ± 1.293 ± 0.958 20.610 ± 0.765 ± 2.280 ± 0.582 —
(K + π − )ER 26.852 ± 1.281 ± 0.664 ± 1.031 17.857 ± 0.460 ± 1.491 ± 0.304 —
K0∗ (1950)0 0.793 ± 0.092 ± 0.076 ± 0.110 0.842 ± 0.079 ± 0.058 ± 0.093 —
f0 (500) 1.830 ± 0.214 ± 0.205 ± 0.245 — —
f0 (980) 16.582 ± 0.440 ± 0.467 ± 0.429 — —
f0 (1370) 16.786 ± 0.951 ± 0.615 ± 1.566 — —
f0 (1500) 1.857 ± 0.150 ± 0.110 ± 0.264 — —
f0 (1710) 0.714 ± 0.094 ± 0.052 ± 0.586 — —
ππ–ππ 5.365 ± 0.833 ± 1.023 ± 1.467 — —
ππ–KK 0.227 ± 0.101 ± 0.121 ± 0.656 — —
π + π − S-wave — 30.901 ± 0.303 ± 1.018 ± 0.124 —
K + π − and — — 67.547 ± 0.606 ± 0.570 ± 0.291
π + π − S-waves
Total 113.847 ± 2.593 ± 3.381 ± 1.548 99.825 ± 0.749 ± 2.244 ± 0.858 98.451 ± 0.431 ± 0.545 ± 0.158

Table 6: Sum of fit fractions in percent deriving from the indicated overall S-waves instead
of individual components comprising the S-wave in each of the three approaches, and their
impact on the total fit fractions. The first uncertainty is statistical, the second the experimental
systematic and the third is the model systematic. The fourth uncertainty arises from the
approach adopted to lift the mathematical ambiguity inherent to the one-dimensional QMI
S-waves.

Fit fraction (%)

Component Isobar K-matrix QMI
K + π − S-wave 63.606 ± 2.019 ± 3.490 ± 3.264 52.010 ± 0.489 ± 1.524 ± 0.298 66.196 ± 3.341 ± 8.468 ± 2.099 ± 2.267
π + π − S-wave 38.800 ± 1.201 ± 1.733 ± 4.405 30.901 ± 0.303 ± 1.018 ± 0.124 36.905 ± 2.231 ± 6.692 ± 3.512 ± 2.991
K + π − and 68.248 ± 0.387 ± 0.252 ± 0.747 68.475 ± 0.329 ± 0.389 ± 0.190 67.547 ± 0.606 ± 0.570 ± 0.291
π + π − S-waves
Total 98.433 ± 0.301 ± 0.316 ± 0.221 98.090 ± 0.377 ± 0.294 ± 0.169 98.451 ± 0.431 ± 0.545 ± 0.158

propensity for partial cancellations in asymmetry parameters, the systematic uncertainties

tend to be at the same level or smaller than the statistical errors, leaving space for future
improvements.
As with the naive total fit fractions from Table 5, the total CP asymmetries associated

23
Table 7: Quasi-two-body CP asymmetries for each approach, where the first uncertainty is
statistical, the second the experimental systematic and the third is the model systematic.

Quasi-two-body CP asymmetry
Component Isobar K-matrix QMI
K ∗ (892)0 −0.038 ± 0.010 ± 0.018 ± 0.002 −0.035 ± 0.012 ± 0.017 ± 0.001 −0.027 ± 0.014 ± 0.022 ± 0.002
K ∗ (1410)0 −0.116 ± 0.106 ± 0.113 ± 0.113 +0.024 ± 0.150 ± 0.048 ± 0.040 +0.447 ± 0.163 ± 0.102 ± 0.045
K ∗ (1680)0 −0.412 ± 0.220 ± 0.681 ± 0.221 −0.340 ± 0.206 ± 0.539 ± 0.217 +0.549 ± 0.277 ± 0.165 ± 0.077
ρ(770)0 +0.280 ± 0.022 ± 0.016 ± 0.024 +0.271 ± 0.022 ± 0.014 ± 0.003 +0.359 ± 0.037 ± 0.047 ± 0.009
ω(782) +0.011 ± 0.158 ± 0.048 ± 0.043 −0.023 ± 0.131 ± 0.061 ± 0.015 −0.004 ± 0.139 ± 0.014 ± 0.004
ρ(1450)0 −0.005 ± 0.180 ± 0.131 ± 0.433 +0.081 ± 0.178 ± 0.143 ± 0.062 −0.273 ± 0.197 ± 0.332 ± 0.068
ρ(1700)0 −0.711 ± 0.189 ± 0.113 ± 0.975 −0.750 ± 0.153 ± 0.111 ± 0.051 −0.735 ± 0.240 ± 0.408 ± 0.058
K2∗ (1430)0 −0.084 ± 0.052 ± 0.144 ± 0.073 −0.119 ± 0.046 ± 0.155 ± 0.044 −0.176 ± 0.066 ± 0.102 ± 0.007
f2 (1270) −0.354 ± 0.047 ± 0.036 ± 0.066 −0.402 ± 0.051 ± 0.036 ± 0.020 −0.279 ± 0.067 ± 0.115 ± 0.006
f2′ (1525) −0.827 ± 0.175 ± 0.206 ± 0.157 −0.757 ± 0.167 ± 0.224 ± 0.169 −0.764 ± 0.263 ± 0.419 ± 0.128
K3∗ (1780)0 −0.936 ± 0.080 ± 0.639 ± 0.372 −0.934 ± 0.069 ± 0.632 ± 0.195 −0.971 ± 0.036 ± 0.165 ± 0.004
ρ3 (1690)0 +0.603 ± 0.090 ± 0.103 ± 0.064 +0.477 ± 0.100 ± 0.067 ± 0.069 +0.698 ± 0.135 ± 0.136 ± 0.091
χc0 (1P ) +0.043 ± 0.028 ± 0.013 ± 0.005 +0.027 ± 0.028 ± 0.011 ± 0.004 +0.012 ± 0.038 ± 0.020 ± 0.008
(K + π − )Res −0.094 ± 0.020 ± 0.038 ± 0.027 −0.121 ± 0.015 ± 0.048 ± 0.013 —
(K + π − )ER +0.023 ± 0.022 ± 0.033 ± 0.042 +0.092 ± 0.021 ± 0.048 ± 0.009 —
K0∗ (1950)0 +0.050 ± 0.097 ± 0.042 ± 0.041 +0.014 ± 0.081 ± 0.031 ± 0.014 —
f0 (500) −0.547 ± 0.091 ± 0.058 ± 0.157 — —
f0 (980) −0.054 ± 0.017 ± 0.011 ± 0.010 — —
f0 (1370) −0.075 ± 0.054 ± 0.045 ± 0.062 — —
f0 (1500) −0.046 ± 0.069 ± 0.030 ± 0.077 — —
f0 (1710) −0.474 ± 0.099 ± 0.064 ± 0.664 — —
ππ–ππ −0.079 ± 0.049 ± 0.039 ± 0.057 — —
ππ–KK +0.371 ± 0.272 ± 0.309 ± 0.213 — —
π + π − S-wave — +0.052 ± 0.033 ± 0.108 ± 0.014 —
K + π − and — — +0.033 ± 0.010 ± 0.018 ± 0.002
π + π − S-waves
Total −0.036 ± 0.010 ± 0.008 ± 0.012 +0.011 ± 0.004 ± 0.010 ± 0.003 +0.030 ± 0.007 ± 0.010 ± 0.003

Table 8: Quasi-two-body CP asymmetries deriving from the indicated overall S-waves instead of
individual components comprising the S-wave in each of the three approaches, and their impact
on the total CP asymmetry. The first uncertainty is statistical, the second the experimental
systematic and the third is the model systematic. The fourth uncertainty arises from the
approach adopted to lift the mathematical ambiguity inherent to the one-dimensional QMI
S-waves.

Quasi-two-body CP asymmetry
Component Isobar K-matrix QMI
K + π − S-wave −0.016 ± 0.011 ± 0.014 ± 0.019 −0.025 ± 0.008 ± 0.016 ± 0.003 −0.035 ± 0.077 ± 0.049 ± 0.010 ± 0.061
π + π − S-wave +0.080 ± 0.012 ± 0.007 ± 0.060 +0.052 ± 0.033 ± 0.108 ± 0.014 +0.022 ± 0.072 ± 0.032 ± 0.023 ± 0.038
K + π − and +0.023 ± 0.006 ± 0.013 ± 0.025 +0.024 ± 0.006 ± 0.014 ± 0.003 +0.033 ± 0.010 ± 0.018 ± 0.002
π + π − S-waves
Total +0.021 ± 0.004 ± 0.010 ± 0.016 +0.020 ± 0.004 ± 0.010 ± 0.001 +0.030 ± 0.007 ± 0.010 ± 0.003

to the S-wave components in Table 7 also show an apparent discrepancy when calculated
from these fit fractions. When considering the S-wave either as a whole or as individual
overall K + π − and π + π − contributions, their CP asymmetries and that of the total model
are found to be in excellent agreement between approaches, as shown in Table 8.

24
ACP 1 1

ACP
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
−0.2 −0.2
−0.4 −0.4
−0.6 LHCb −0.6 LHCb
−0.8 3 fb−1 −0.8 3 fb−1
−1 −1

ω (782)

f 2(1270)

f 0(500)
f 0(980)
f 0(1370)
f 0(1500)
f 0(1710)
f 2'(1525)
(K +π −)Res
K (892)0

K (1410)0

K (1680)0

K 2(1430)0

K 3(1780)0

K 0(1950)0

ρ (770)0

ρ (1450)0
ρ (1700)0

ρ3(1690)0
(K +π −)ER

π π -K K
π π -π π
*

*
*

*
Figure 9: Comparison between Isobar (green), K-matrix (magenta) and QMI (blue) ACP for
states decaying to (left) K + π − and (right) π + π − . Solid bands show the statistical spread, with
progressively lighter shading indicating the combined statistical and systematic uncertainties,
followed by the total uncertainty including that from the model.

7.3 Relative phases

For completeness, the relative phases for each contribution in B + (B − ) decays are reported
in Table 9 (10) and Fig. 10, showing broad agreement between the S-wave approaches.
The results for these amplitude-level quantities tend to be dominated by the systematic
uncertainty.

25
Table 9: Phase comparison for B + decay amplitudes between the three S-wave approaches
where the first uncertainty is statistical, the second systematic and the third from the model.

Phase of B + decay amplitude (◦ )

Component Isobar K-matrix QMI
∗ 0
K (892) 0 (fixed) 0 (fixed) 0 (fixed)
K ∗ (1410)0 −47.4 ± 4.1 ± 8.4 ± 4.5 −49.0 ± 4.4 ± 8.6 ± 1.5 −33.6 ± 8.8 ± 13.1 ± 1.5
K ∗ (1680)0 −175.5 ± 8.2 ± 19.2 ± 16.0 +179.3 ± 8.5 ± 17.2 ± 3.8 −120.7 ± 29.3 ± 60.8 ± 17.7
ρ(770)0 +0.1 ± 3.3 ± 2.4 ± 1.5 −7.2 ± 3.2 ± 5.4 ± 1.1 −4.3 ± 6.4 ± 9.0 ± 1.3
ω(782) +104.3 ± 7.2 ± 2.8 ± 1.4 +100.8 ± 7.4 ± 3.7 ± 1.5 +103.8 ± 9.0 ± 5.9 ± 2.1
ρ(1450)0 +114.5 ± 4.8 ± 7.8 ± 2.4 +128.5 ± 5.7 ± 5.8 ± 2.3 +128.8 ± 5.9 ± 8.9 ± 5.5
ρ(1700)0 +42.1 ± 9.3 ± 6.1 ± 4.7 +62.2 ± 8.3 ± 3.8 ± 2.8 +62.1 ± 13.6 ± 3.8 ± 2.3
K2∗ (1430)0 +96.1 ± 1.8 ± 2.4 ± 1.9 +99.2 ± 1.5 ± 2.7 ± 0.6 +102.3 ± 5.8 ± 3.5 ± 2.0
f2 (1270) +137.8 ± 2.2 ± 2.8 ± 2.6 +138.1 ± 2.3 ± 2.8 ± 2.5 +137.0 ± 4.5 ± 4.0 ± 3.6
f2′ (1525) +19.5 ± 9.8 ± 17.8 ± 6.4 +26.7 ± 9.3 ± 14.6 ± 2.7 +42.7 ± 11.6 ± 15.2 ± 1.6
∗ 0
K3 (1780) −110.4 ± 5.4 ± 10.2 ± 3.7 −104.4 ± 5.1 ± 8.6 ± 2.5 −106.0 ± 7.5 ± 16.5 ± 3.6
ρ3 (1690)0 −66.4 ± 10.3 ± 12.2 ± 10.1 −78.3 ± 7.7 ± 10.7 ± 10.2 −64.9 ± 14.3 ± 11.3 ± 14.9
χc0 (1P ) +115.6 ± 3.8 ± 1.6 ± 1.6 +116.1 ± 3.1 ± 2.0 ± 0.3 +119.3 ± 5.3 ± 2.6 ± 0.7
(K + π − )Res +119.7 ± 2.0 ± 5.5 ± 3.8 +85.5 ± 1.6 ± 8.7 ± 1.3 —
(K + π − )ER −64.9 ± 1.9 ± 2.8 ± 3.7 −76.9 ± 1.5 ± 5.3 ± 0.5 —
K0∗ (1950)0 −16.7 ± 5.2 ± 6.7 ± 5.2 −14.2 ± 4.6 ± 6.0 ± 4.8 —
f0 (500) −73.6 ± 3.3 ± 6.8 ± 3.8 — —
f0 (980) +62.1 ± 2.8 ± 3.4 ± 2.2 — —
f0 (1370) +154.9 ± 2.3 ± 5.5 ± 2.2 — —
f0 (1500) −39.6 ± 3.7 ± 4.1 ± 2.1 — —
f0 (1710) +30.8 ± 5.6 ± 6.8 ± 4.9 — —
ππ–ππ +116.0 ± 3.9 ± 7.9 ± 4.2 — —
ππ–KK +40.4 ± 19.0 ± 17.5 ± 25.3 — —

26
Table 10: Phase comparison for B − decay amplitudes between the three S-wave approaches
where the first uncertainty is statistical, the second systematic and the third from the model.

Phase of B − decay amplitude (◦ )

Component Isobar K-matrix QMI
∗ 0
K (892) 0 (fixed) 0 (fixed) 0 (fixed)
K ∗ (1410)0 −54.4 ± 4.0 ± 4.4 ± 3.7 −54.4 ± 4.9 ± 1.8 ± 0.8 −39.8 ± 5.1 ± 5.2 ± 0.8
K ∗ (1680)0 +123.5 ± 13.5 ± 35.1 ± 37.7 +123.4 ± 11.1 ± 41.2 ± 11.8 −164.5 ± 9.0 ± 14.0 ± 1.5
ρ(770)0 −40.2 ± 2.8 ± 3.3 ± 8.0 −47.5 ± 2.7 ± 4.4 ± 2.0 −38.0 ± 4.8 ± 5.9 ± 0.6
ω(782) +82.4 ± 6.5 ± 4.0 ± 8.4 +74.2 ± 6.9 ± 4.6 ± 2.0 +85.9 ± 7.6 ± 4.4 ± 0.8
ρ(1450)0 +77.4 ± 5.0 ± 3.4 ± 14.4 +80.3 ± 5.8 ± 3.2 ± 2.6 +109.7 ± 5.9 ± 5.1 ± 1.9
ρ(1700)0 +49.1 ± 25.0 ± 16.3 ± 44.9 +75.2 ± 25.7 ± 10.0 ± 9.8 +83.8 ± 25.4 ± 29.8 ± 7.8
K2∗ (1430)0 +85.9 ± 1.6 ± 2.6 ± 0.5 +87.4 ± 1.6 ± 2.8 ± 1.0 +100.8 ± 3.5 ± 2.3 ± 0.4
f2 (1270) +169.6 ± 3.0 ± 2.8 ± 4.4 +167.3 ± 3.5 ± 3.4 ± 2.0 −176.5 ± 4.4 ± 4.0 ± 1.6
f2′ (1525) −112.5 ± 31.0 ± 30.6 ± 55.0 −95.6 ± 30.5 ± 30.1 ± 8.8 −135.0 ± 38.8 ± 23.0 ± 30.0
K3∗ (1780)0 −131.1 ± 35.8 ± 13.7 ± 13.9 −115.1 ± 33.7 ± 16.5 ± 4.1 −99.5 ± 34.2 ± 10.8 ± 17.1
ρ3 (1690)0 −111.5 ± 4.8 ± 5.7 ± 27.9 −105.4 ± 4.6 ± 5.5 ± 1.3 −84.1 ± 8.3 ± 5.3 ± 1.3
χc0 (1P ) +115.2 ± 3.2 ± 2.0 ± 0.9 +116.0 ± 3.0 ± 1.7 ± 0.2 +126.7 ± 5.0 ± 2.1 ± 0.2
+ −
(K π )Res +112.4 ± 2.3 ± 2.9 ± 1.8 +77.2 ± 1.7 ± 6.3 ± 1.6 —
(K + π − )ER −60.7 ± 1.9 ± 1.8 ± 2.6 −75.9 ± 1.3 ± 3.6 ± 0.6 —
K0∗ (1950)0 −19.7 ± 4.5 ± 6.4 ± 8.6 −18.8 ± 4.1 ± 7.6 ± 6.7 —
f0 (500) −111.9 ± 5.6 ± 6.8 ± 11.7 — —
f0 (980) +65.2 ± 3.1 ± 2.2 ± 4.8 — —
f0 (1370) +148.8 ± 2.4 ± 3.8 ± 2.9 — —
f0 (1500) −48.1 ± 3.4 ± 2.3 ± 6.4 — —
f0 (1710) +66.3 ± 7.4 ± 7.2 ± 39.1 — —
ππ–ππ +133.5 ± 5.1 ± 3.7 ± 9.2 — —
ππ–KK −31.1 ± 9.3 ± 5.8 ± 15.3 — —

27
B+ phase (°)

B+ phase (°)
150 150
100 100
50 50
0 0
−50 −50
−100 −100
LHCb LHCb
−150 3 fb−1 −150 3 fb−1

ω (782)

f 2(1270)

f 0(500)
f 0(980)
f 0(1370)
f 0(1500)
f 0(1710)
f 2'(1525)
(K +π −)Res
K (1410)0

K (1680)0

K 2(1430)0

K 3(1780)0

K 0(1950)0

ρ (770)0

ρ (1450)0
ρ (1700)0

ρ3(1690)0
(K +π −)ER

π π -K K
π π -π π
*

*
*

*
B− phase (°)

B− phase (°)

150 150
100 100
50 50
0 0
−50 −50
−100 −100
LHCb LHCb
−150 3 fb−1 −150 3 fb−1
ω (782)

f 2(1270)

f 0(500)
f 0(980)
f 0(1370)
f 0(1500)
f 0(1710)
f 2'(1525)
(K +π −)Res
K (1410)0

K (1680)0

K 2(1430)0

K 3(1780)0

K 0(1950)0

ρ (770)0

ρ3(1690)0
(K +π −)ER

π π -K K
π π -π π
ρ (1450)
ρ (1700)
*

*
*

Figure 10: Comparison between Isobar (green), K-matrix (magenta) and QMI (blue) amplitude
phases of (top) B + and (bottom) B − decays for states decaying to (left) K + π − and (right) π + π − .
Solid bands show the statistical spread, with progressively lighter shading indicating the combined
statistical and systematic uncertainties, followed by the total uncertainty including that from
the model. The K ∗ (892)0 phase is fixed to zero to serve as a reference, so does not appear here.

28
Candidates / (0.011 GeV/c2)

Candidates / (0.032 GeV/c2)

2500 LHCb K 3π Kη K η' 900 LHCb K ηc DsD
3 fb−1 800 3 fb−1
2000 700
600
1500 500
400
1000
300
500 200
100

1 1.5 2 3 4 5
mK +π − (GeV/c2) mK +π − (GeV/c2)
1 1
Raw asymmetry

Raw asymmetry
0.8 LHCb K 3π Kη K η' 0.8 LHCb K ηc DsD
3 fb−1 3 fb−1
0.6 0.6
0.4 0.4
0.2 0.2
0 0
−0.2 −0.2
−0.4 −0.4 Data Isobar
−0.6 −0.6 Background K-matrix
−0.8 −0.8 QMI
−1 −1
1 1.5 2 3 4 5
mK +π − (GeV/c2) mK +π − (GeV/c2)

Figure 11: Fit projections of each model (top) in the mK + π− region (left) below and (right) above
the open charm threshold, with (bottom) the corresponding asymmetries. Vertical dashed lines
correspond to the opening thresholds of the indicated coupled channels.

7.4 Fit projections

Comparisons of the data and all three fit models, projected onto mK + π− , mπ+ π− and the
cosines of their associated helicity angles, along with the raw asymmetries between B −
and B + decays, can be seen in Figs. 11–14. These models appear to be in good overall
agreement with the data and with each other, both in CP -averaged projections and in the
variation of the asymmetries across the phase space. Additional projections separating the
contributions for various components of the amplitude model are shown in Appendices E, F
and G for the Isobar, K-matrix and QMI approaches, respectively. Further comparisons
of the different S-wave models are given in Appendix H, and comparison of the data and
fit models using weights to enhance components with particular orbital angular momenta
are shown in Appendix I.

29
Candidates / (0.015 GeV/c2)

Candidates / (0.029 GeV/c2)

2000
3000 4π KK ηη ηη' ηηc DsDs/ η' ηc
1800
LHCb
DD LHCb
2500 1600
3 fb−1 1400 3 fb−1
2000
1200
1500 1000
800
1000 600
400
500
200

0.5 1 1.5 2 3 4
mπ +π − (GeV/c2) mπ +π − (GeV/c2)
1 1
Raw asymmetry

Raw asymmetry
0.8 4π KK ηη ηη' 0.8 ηηc DsDs/ η' ηc
0.6 LHCb 0.6 DD LHCb
0.4 3 fb−1 0.4 3 fb−1
0.2 0.2
0 0
−0.2 −0.2
−0.4 −0.4 Data Isobar
−0.6 −0.6 Background K-matrix
−0.8 −0.8 QMI
−1 −1
0.5 1 1.5 2 3 4
mπ +π − (GeV/c2) mπ +π − (GeV/c2)

Figure 12: Fit projections of each model (top) in the mπ+ π− region (left) below and (right) above
the open charm threshold, with (bottom) the corresponding asymmetries. Vertical dashed lines
correspond to the opening thresholds of the indicated coupled channels.
Candidates / (0.02)

Candidates / (0.02)

103 LHCb
3 3 fb−1
LHCb 10
3 fb−1
102
102

10 10

1 1
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ K +π − cosθ K +π −
1 1
Raw asymmetry

Raw asymmetry

0.8 LHCb 0.8 LHCb

3 fb−1 3 fb−1
0.6 0.6
0.4 0.4
0.2 0.2
0 0
−0.2 −0.2
−0.4 −0.4 Data Isobar
−0.6 −0.6 Background K-matrix
−0.8 −0.8 QMI
−1 −1
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ K +π − cosθ K +π −

Figure 13: Fit projections of each model (top) in the cos θK + π− region (left) below and
(right) above the open charm threshold, with (bottom) the corresponding asymmetries.

30
Candidates / (0.02)

Candidates / (0.02)
LHCb 104 LHCb
10 3 3 fb−1 3 fb−1
103

102
102

10
10

1 1
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ π +π − cosθ π +π −
1 1
Raw asymmetry

Raw asymmetry

0.8 LHCb 0.8 LHCb

3 fb−1 3 fb−1
0.6 0.6
0.4 0.4
0.2 0.2
0 0
−0.2 −0.2
−0.4 −0.4 Data Isobar
−0.6 −0.6 Background K-matrix
−0.8 −0.8 QMI
−1 −1
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ π +π − cosθ π +π −

Figure 14: Fit projections of each model (top) in the cos θπ+ π− region (left) below and
(right) above the open charm threshold, with (bottom) the corresponding asymmetries.

31
| AGLASS| 0.8

arg( AGLASS) (°)

LHCb 150 LHCb
0.7 3 fb−1 3 fb−1
0.6 100

0.5 50

0.4 0 *
Isobar K 0(1430)0
0.3 −50 Isobar SVP
0.2 −100 Isobar GLASS
*
K-matrix K 0(1430)0
0.1
−150 K-matrix GLASS K-matrix SVP
0
1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)

Figure 15: (left) Amplitude and (right) phase of the GLASS lineshape, where the band includes
the total uncertainty. Individual contributions from the K0∗ (1430)0 and SVP components are
also shown.

7.5 Evaluation of the B + → K0∗ (1430)0 π + decay properties

The GLASS lineshape as determined from data can be seen in Fig. 15,
for the Isobar and K-matrix approaches. The corresponding over-
all fit fractions are (58.9 ± 2.0 (stat) ± 3.2 (syst) ± 3.0 (model))% and
(47.7 ± 0.5 (stat) ± 1.4 (syst) ± 0.3 (model))% for the Isobar and K-matrix approaches,
respectively. This difference is attributed to interference with each S-wave approach in
π + π − as their total S-wave fit fractions agree to within 1%, as does that obtained with
the QMI approach, reported in Table 6.
The measured resonance parameters of the K0∗ (1430)0 resonance, presented in Ap-
pendices E and F for the Isobar and K-matrix approaches respectively, are practically
identical. The K0∗ (1430)0 fit fraction is likewise expected to be in agreement; however
calculation of the fit fraction from the resonance-sensitive term in Eq. (6) (as was done by
the BaBar collaboration), shows a significant disagreement in Table 5 between the Isobar
and K-matrix approaches. This arises from the ϕNonres term, defined in Eq. (7), in the
GLASS lineshape and impact of the differing values obtained for this parameter in the
Isobar and K-matrix approaches. Considering the K0∗ (1430)0 resonance phase to have
purely a Breit–Wigner form,

mK0∗ Γ(s)
 
ϕK0∗ (s) ≡ arctan , (20)
(m2K0∗ − s)

a double-angle expansion of sin δRes leads to the peaking amplitude best identified with
the K0∗ (1430)0 resonance, which evaluates to
 
1
AK0∗ (s) ≡ cos 2ϕNonres sin 2ϕK0∗ (s) + i cos ϕNonres sin ϕK0∗ (s) e2iδER ,
2 2
(21)
2

with the remainder of AGLASS comprising the slowly varying part (SVP). With this
definition, the B + → K0∗ (1430)0 π + decay properties are given in Table 11. The K0∗ (1430)0
and SVP components of the GLASS lineshape are shown in Fig. 15 as the dashed and
dotted curves, respectively.

32
Table 11: Results obtained in the Isobar and K-matrix approaches for the B + → K0∗ (1430)0 π +
decay properties where the first uncertainty is statistical, the second systematic and the third
from the model.

Observable Isobar K-matrix

F (%) 7.912 ± 0.270 ± 0.267 ± 0.381 6.736 ± 0.208 ± 0.551 ± 0.148
ACP −0.094 ± 0.021 ± 0.038 ± 0.027 −0.121 ± 0.015 ± 0.048 ± 0.013
δ + (◦ ) +119.7 ± 2.6 ± 5.5 ± 3.8 +85.5 ± 1.8 ± 8.7 ± 1.3
δ − (◦ ) +112.4 ± 2.6 ± 2.9 ± 1.8 +77.2 ± 1.9 ± 6.3 ± 1.6

7.6 Alternative ρ0 –ω transition results

In the baseline result, ρ0 –ω mixing is considered to manifest through direct ω → π + π −
decay. A commonly held alternate view is that the ω(782) meson first mixes into the
ρ(770)0 state, which subsequently decays into the π + π − final state. In order to model the
latter, the direct ω contribution is replaced by its product with the ρ0 lineshape as

Tω (s) → Tρω (s) = Tρ (s)Tω (s). (22)

The functional difference between these two paradigms is that the magnitude of Tρω closely
resembles |Tω | while the phase motion shares more similarities with arg(Tρ ). The main
difference is that the phase of Tρω instead rotates by 2π in the same range, [m0 −Γ0 , m0 +Γ0 ],
where Tρ will rotate by π.
The −2 log L value is found to be practically unchanged between these two views of
0
ρ –ω mixing, as shown in Table 12, indicating that there is little prospect of separating
them even in future analyses. Therefore, only the physics parameters that are impacted
by these lineshapes are reported in Table 12. Due to the similarity of fit fractions between
the alternate and baseline models, the systematic and model uncertainties can be directly
imported from the baseline. As expected, the only meaningful differences are in the ρ0 –ω
phases which experience a shift of ∼ −90◦ from the baseline ω phases.

Table 12: Results obtained with the alternative ρ0 –ω transition model. Only parameters
impacted by the change of model are shown, and uncertainties are statistical only. The results
for the ρ0 –ω component should be compared to those for the ω component in the baseline model.

Component Observable Isobar K-matrix QMI

0
ρ(770) F (%) 7.654 ± 0.150 7.922 ± 0.177 7.147 ± 0.248
ACP +0.274 ± 0.019 +0.268 ± 0.021 +0.357 ± 0.036
δ + (◦ ) +2.5 ± 3.0 −4.8 ± 3.2 −1.8 ± 5.4
− ◦
δ () −38.7 ± 2.1 −45.9 ± 2.7 −36.4 ± 5.2
ρ0 –ω F (%) 0.175 ± 0.026 0.190 ± 0.029 0.193 ± 0.029
ACP +0.005 ± 0.129 −0.028 ± 0.136 −0.013 ± 0.168
+ ◦
δ () +2.1 ± 7.5 −1.3 ± 7.4 +2.3 ± 8.2
δ − (◦ ) −20.0 ± 6.1 −28.1 ± 5.9 −15.6 ± 8.0
∆(−2 log L) +0.2 −0.3 −0.2

33
Table 13: Intermediate absolute branching fractions for averaging purposes and upper limits
in cases where the significance of the underlying fit fraction does not exceed three standard
deviations. Branching fractions of the resonance to K + π − or π + π − have been accounted for,
except in cases labelled † where the measured quantity is the product branching fraction. The
first uncertainty is the total obtained from the Isobar approach, while the final uncertainty is
always from the external inclusive B + → K + π + π − branching fractions and any intermediate
branching fractions. A middle uncertainty represents the maximum deviation from the K-matrix
and QMI results. The branching fraction for the χc0 (1P ) component is divided by 100 to have
the same units as the other entries.

Component Branching fraction (×10−6 ) Upper limit

at 90% (95%) CL
B + → K ∗ (892)0 π + 10.54 ± 0.46 ± 0.44 ± 0.26 —
B + → K ∗ (1410)0 π + 9.17 ± 2.39 ± 2.49 ± 1.82 —
B + → K ∗ (1680)0 π + 0.49 ± 0.21 ± 0.23 ± 0.03 < 0.91 (1.02)
B + → ρ(770)0 K + 4.51 ± 0.39 ± 0.29 ± 0.11 —
B + → ω(782)K + 7.59 ± 1.71 ± 0.73 ± 0.67 —
B + → ρ(1450)0 K + † 0.70 ± 0.60 ± 0.40 ± 0.02 —
B + → ρ(1700)0 K + † 0.20 ± 0.22 ± 0.01 ± 0.00 < 0.51 (0.59)
B + → K2∗ (1430)0 π + 3.90 ± 0.57 ± 0.31 ± 0.13 —
B + → f2 (1270)K + 2.21 ± 0.31 ± 0.30 ± 0.09 —
B + → f2′ (1525)K + 3.90 ± 3.84 ± 0.63 ± 0.51 < 9.32 (10.69)
B + → K3∗ (1780)0 π + 1.02 ± 0.27 ± 0.91 ± 0.06 < 2.32 (2.65)
B + → ρ3 (1690)0 K + 1.00 ± 0.26 ± 0.37 ± 0.06 —
B + → χc0 (1P )K + (×10−2 ) 2.03 ± 0.12 ± 0.18 ± 0.09 —
B + → K0∗ (1430)0 π + 7.35 ± 0.50 ± 1.09 ± 0.81 —
B + → (K + π − )SVP π + † 28.69 ± 2.66 ± 6.81 ± 0.70 —
B + → K0∗ (1950)0 π + 1.32 ± 0.27 ± 0.08 ± 0.36 —
B + → f0 (500)K + † 1.06 ± 0.22 ± 0.03 —
B + → f0 (980)K + † 9.55 ± 0.44 ± 0.23 —
B + → f0 (1370)K + † 9.67 ± 1.11 ± 0.23 —
B + → f0 (1500)K + 4.65 ± 0.81 ± 0.32 —
B + → f0 (1710)K + † 0.41 ± 0.34 ± 0.01 —
B + → (ππ–ππ)K + † 3.09 ± 1.14 ± 0.08 —
B + → (ππ–KK)K + † 0.13 ± 0.39 ± 0.00 < 0.73 (0.85)

8 Interpretation
Whilst all S-wave approaches are considered to be equal in this analysis, the Isobar results
are recommended for averaging purposes as every component can be associated to a
physical state. Where applicable, the maximum deviation from the K-matrix and QMI
approaches is assigned as an additional systematic uncertainty. Fit fractions are first
converted to branching fraction measurements as shown in Table 13, which is achieved
through multiplying by the inclusive absolute branching fraction, B(B + → K + π + π − ) =

34
(5.76 ± 0.14) × 10−5 [67]. Intermediate branching fractions, assuming isospin symmetry
where appropriate, are also taken into account wherever possible [63], otherwise the
product of branching fractions is reported. Thus, for an intermediate resonant decay
B + → Rh, the relation between its fit and branching fraction is given by

B(B + → Rh)B(R → h′ h(′) ) = F(B + → Rh)B(B + → K + π + π − ) . (23)

Upper limits based on the Gaussian profiles of the fit fractions are reported for states
not exceeding three standard deviations in significance. Similarly, the quasi-two-body CP
asymmetries are reported in Table 14. Information on the total uncertainty correlations
between all of these measurements can be found in the Supplemental Material [105].
Being based on a much larger data sample, with better separation of signal from
background and improved modelling, these results represent a significant improvement
in precision and in understanding of the B + → K + π + π − decay. Due to the importance
of using the GLASS lineshape as discussed in a companion article [14], these results
should not be averaged with those from previous amplitude analyses of the same decay
process that relied on unitarity-preserving arguments ultimately originating with LASS
data [106, 107]. As a case in point, the result

B(B + → χc0 (1P )K + ) = (2.03 ± 0.12 (Isobar) ± 0.18 (S-wave) ± 0.09 (External)) × 10−4

is in slight tension with the world average B(B + → χc0 (1P )K + ) = 1.51 +0.15 −4


−0.13 × 10 [62],
however the two lowest contributions to this average are from previous B + → K + π + π − am-
plitude analyses; the result of this analysis is in excellent agreement with, and more precise
than, all other previous measurements of this quantity. The branching fraction and CP
asymmetry measurements for B + → ω(782)K + are also in agreement with previous mea-
surements [108, 109], however they are not as precise as those with the ω(782) → π + π − π 0
final state on account of the larger yields available. First upper limits are provided for
decays involving the ρ(1700)0 , K3∗ (1780)0 and ππ–KK intermediate states.
For the penguin-dominated B + → K ∗ (892)0 π + decay, the measured branching fraction
and CP asymmetry in decay are in excellent agreement with results from BaBar and
Belle [106, 107] and with theoretical expectations [16–18, 22, 23, 29–31]. These results
and those for all other excited kaon states comprise particularly crucial input to model-
independent tests for physics beyond the SM via isospin sum rules [32–34] adapted to
excited intermediate states as they provide the purest constraint for the dominant, sensitive
penguin loop amplitude, with the only other contribution coming from weak-annihilation
diagrams. On the other hand, for the B + → ρ(770)0 K + decay, the measured branching
fraction is larger with respect to the previous measurements, and close to the central values
predicted by various theories [16–18, 22, 23, 29–31]. Similarly, the branching fractions of
both the B + → K2∗ (1430)0 π + and B + → f2 (1270)K + decays move towards their theoretical
predictions [19, 24], compared to previous measurements. In the scalar sector, the product
branching fraction and CP asymmetry of B + → f0 (980)K + is in agreement with previous
results and with theory [18, 20, 25].

9 Conclusions
→ K + π + π − decay is performed, based on pp collision
An amplitude analysis of the B +√
data at centre-of-mass energies s = 7 and 8 TeV recorded with the LHCb detector,

35
Table 14: Quasi-two-body CP asymmetries for averaging purposes, where the first uncertainty is
the total obtained from the Isobar approach and a second uncertainty represents the maximum
deviation from the K-matrix and QMI results.

Component ACP
B + → K ∗ (892)0 π + −0.038 ± 0.021 ± 0.010
B + → K ∗ (1410)0 π + −0.116 ± 0.192 ± 0.563
B + → K ∗ (1680)0 π + −0.412 ± 0.749 ± 0.960
B + → ρ(770)0 K + +0.280 ± 0.036 ± 0.080
B + → ω(782)K + +0.011 ± 0.171 ± 0.034
B + → ρ(1450)0 K + −0.005 ± 0.486 ± 0.268
B + → ρ(1700)0 K + −0.711 ± 0.999 ± 0.039
B + → K2∗ (1430)0 π + −0.084 ± 0.169 ± 0.092
B + → f2 (1270)K + −0.354 ± 0.089 ± 0.075
B + → f2′ (1525)K + −0.827 ± 0.312 ± 0.069
B + → K3∗ (1780)0 π + −0.936 ± 0.744 ± 0.035
B + → ρ3 (1690)0 K + +0.603 ± 0.151 ± 0.126
B + → χc0 (1P )K + +0.043 ± 0.032 ± 0.031
B + → K0∗ (1430)0 π + −0.094 ± 0.051 ± 0.026
B + → (K + π − )SVP −0.026 ± 0.031 ± 0.002
B + → K0∗ (1950)0 π + +0.050 ± 0.113 ± 0.036
B + → f0 (500)K + −0.547 ± 0.191
B + → f0 (980)K + −0.054 ± 0.023
B + → f0 (1370)K + −0.075 ± 0.093
B + → f0 (1500)K + −0.046 ± 0.108
B + → f0 (1710)K + −0.474 ± 0.674
B + → (ππ–ππ)K + −0.079 ± 0.084
B + → (ππ–KK)K + +0.371 ± 0.464

corresponding to an integrated luminosity of 3 fb−1 . The detailed description of the

analysis procedure presented here complements shorter companion articles focussed on the
CP -conserving implications prompted by an improved understanding of strong interaction
dynamics [14], and the consequent first observations of different sources of CP -violation
effects in this channel [15]. The results showcase the effectiveness of three distinct yet
complementary approaches to describing the large S-wave contribution to this decay:
Isobar, K-matrix, and Quasi-Model-Independent. Overall, good agreement is found
between all three models and the data, indicating significant progress in understanding
the intricate dynamics governing charmless hadronic three-body decays. One important
improvement is the modelling of the Kπ S-wave, in the Isobar and K-matrix approaches,
with the GLASS lineshape rather than the unitarity-preserving LASS function used
in previous amplitude analyses of the same process. As a consequence, the branching
fractions and quasi-two-body CP asymmetries associated with each intermediate state
reported from this analysis should not be averaged with previous measurements.
Both direct and transition-mediated contributions from the ω(782) resonance are

36
considered, with results presented using alternative models. The data do not provide
significant discriminating power between these two cases. Despite the similarities between
the two models, their distinctions are clarified, particularly in the phase shifts seen in
the ρ0 –ω coupling. As for the results themselves, the precision on the B + → χc0 (1P )K +
branching fraction and CP asymmetry is greatly improved, offering the prospect for a
tighter constraint on contaminating penguin amplitudes in measurements of the B 0 –B 0
mixing phase from B 0 → χc0 (1P )K 0 decays. Similar improvements are noted in most
other contributions, each of which comprises fundamental input to model-independent
isospin sum-rule tests of new physics in loop-mediated decays. Finally, the first upper-limit
measurements are made with regards to contributions from the ρ(1700)0 and K3∗ (1780)0
resonances, and the ππ–KK rescattering contribution.
On the technical front, it is interesting to note that in this analysis, the isobar approach
with a sophisticated description of the S-wave contributions, yields a similar, if not better,
fit quality while having fewer parameters than the K-matrix approach. This is also the
first attempt with particle decay at extracting multiple partial waves in a quasi-model-
independent manner. Systematic biases induced by fixing the contents of one bin to the
Isobar prediction, and arising only in the presentation of individual waves, are found to
be under control.
These results establish a solid foundation for future analyses and provide valuable
input to phenomenological work on the strong interaction mechanisms underscoring the
remarkably large CP violation observed in charmless hadronic three-body decays of the
charged B meson, and on B-meson decays in general.

Acknowledgements
We express our gratitude to our colleagues in the CERN accelerator departments for
the excellent performance of the LHC. We thank the technical and administrative staff
at the LHCb institutes. We acknowledge support from CERN and from the national
agencies: ARC (Australia); CAPES, CNPq, FAPERJ and FINEP (Brazil); MOST and
NSFC (China); CNRS/IN2P3 and CEA (France); BMFTR, DFG and MPG (Germany);
NKFIH (Hungary); INFN (Italy); NWO (Netherlands); MNiSW and NCN (Poland);
MEC/IFA (Romania); MICIU and AEI (Spain); SNSF and SER (Switzerland); NASU
(Ukraine); STFC (United Kingdom); DOE NP and NSF (USA). We acknowledge the
computing resources that are provided by ARDC (Australia), CBPF (Brazil), CERN,
IHEP and LZU (China), IN2P3 (France), KIT and DESY (Germany), INFN (Italy),
SURF (Netherlands), Polish WLCG (Poland), IFIN-HH (Romania), PIC (Spain), CSCS
(Switzerland), GridPP (United Kingdom), and NSF (USA). We are indebted to the
communities behind the multiple open-source software packages on which we depend.
Individual groups or members have received support from RTP (Australia), FWO Odysseus
grant G0ASD25N (Belgium), Key Research Program of Frontier Sciences of CAS, CAS
PIFI, CAS CCEPP (China); Minciencias (Colombia); EPLANET, Marie Sklodowska-
Curie Actions, ERC and NextGenerationEU (European Union); A*MIDEX, ANR, IPhU
and Labex P2IO, and Région Auvergne-Rhône-Alpes (France); Alexander-von-Humboldt
Foundation (Germany); ICSC (Italy); Severo Ochoa and Marı́a de Maeztu Units of
Excellence, GVA, XuntaGal, GENCAT, InTalent-Inditex and Prog. Atracción Talento
CM (Spain); the Leverhulme Trust, the Royal Society and UKRI (United Kingdom).

37
Appendices
A B +-candidate mass fit
The mass distributions of the selected B candidates, together with the results of the fits
to these distributions, are shown in Fig. 16. These complement the corresponding plots
with a logarithmic y-axis scale shown in Fig. 1.

Candidates / (9 MeV/c2)
Candidates / (9 MeV/c2)

LHCb LHCb
3 fb−1 3 fb−1
10000 Data 10000
B+, B0 4-body
B0s 4-body B+ B−
B±→ K ± π+ π−
Combinatorial
5000 B± → π ± π + π − 5000
B± → π ± K + K −
B±→ K ±η' (958)

5200 5400 5600 5800 5200 5400 5600 5800

mK +π +π − (MeV/c2) mK −π −π + (MeV/c2)

Figure 16: Mass distributions for (left) B + and (right) B − candidates reconstructed in the
K + π + π − and K − π − π + final state, respectively, for the combined 2011 and 2012 data taking
samples, along with the results of the fit.

38
B Background model
The background model transformed from the square Dalitz plot to the traditional
Dalitz plot are shown. The combinatorial, B + → π + π + π − , B + → π + K + K − and
B + → K + η ′ (π + π − γ) background models corresponding to Figs. 4–7, respectively, are
shown in Figs. 17–20.

Figure 17: Dalitz-plot distributions for the (left) B + and (right) B − combinatorial background
models, scaled to represent their respective yields in the signal region.

Figure 18: Dalitz-plot distributions for the misidentified B + → π + π + π − background models in

(left) B + and (right) B − samples, scaled to represent their respective yields in the signal region.

Figure 19: Dalitz-plot distributions for the misidentified B + → π + K + K − background models in

(left) B + and (right) B − samples, scaled to represent their respective yields in the signal region.

39
Figure 20: Dalitz-plot distributions for the partially reconstructed B + → K + η ′ (π + π − γ) back-
ground models in (left) B + and (right) B − , scaled to represent their respective yields in the
signal region.

40
C Systematic uncertainties
The systematic uncertainties for the CP -averaged fit fractions, broken down by category,
are summarised in Tables 15–17 for the Isobar, K-matrix and QMI approaches, respectively,
while those for the quasi-two-body CP asymmetries are recorded in Tables 18–20. For the
relative phases of B + (B − ) decays, the systematic uncertainties are similarly presented
in Tables 21–23 (24–26). The shorthand labels representing each category of systematic
uncertainty are defined throughout Section 6.

Table 15: Systematic uncertainties on the fit fractions for the Isobar approach given in percent.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (892) 0.003 0.105 0.007 0.455 0.162 0.163 0.156 0.009
K ∗ (1410)0 0.016 0.060 0.005 0.083 0.106 0.040 0.047 0.025
K ∗ (1680)0 0.017 0.022 0.001 0.059 0.022 0.064 0.021 0.020
ρ(770)0 0.172 0.034 0.004 0.528 0.160 0.246 0.106 0.282
ω(782) 0.006 0.003 0.001 0.022 0.023 0.013 0.004 0.008
ρ(1450)0 0.024 0.099 0.037 0.049 0.032 0.268 0.152 1.005
ρ(1700)0 0.009 0.041 0.008 0.018 0.061 0.076 0.028 0.371
K2∗ (1430)0 < 0.001 0.168 0.043 0.192 0.006 0.060 0.153 0.036
f2 (1270) 0.065 0.106 0.014 0.185 0.094 0.041 0.044 0.129
f2′ (1525) 0.004 0.017 0.002 0.046 0.006 0.005 0.004 0.014
K3∗ (1780)0 < 0.001 0.030 0.006 0.014 0.013 0.015 0.023 0.007
ρ3 (1690)0 0.017 0.018 0.011 0.037 0.032 0.022 0.038 0.066
χc0 (1P ) 0.005 0.053 0.005 0.061 0.045 0.005 0.028 0.004
(K + π − )Res 0.107 0.383 0.050 1.030 0.672 0.254 0.272 0.919
(K + π − )ER 0.098 0.403 0.105 0.395 0.318 0.479 0.965 0.364
K0∗ (1950)0 0.019 0.043 0.013 0.057 0.010 0.012 0.107 0.026
f0 (500) 0.003 0.104 0.044 0.162 0.053 0.122 0.103 0.223
f0 (980) 0.032 0.180 0.045 0.404 0.138 0.151 0.069 0.424
f0 (1370) 0.074 0.403 0.045 0.342 0.303 0.480 1.524 0.363
f0 (1500) 0.004 0.023 0.012 0.104 0.023 0.034 0.133 0.228
f0 (1710) 0.009 0.035 < 0.001 0.033 0.017 0.019 0.078 0.580
ππ–ππ 0.193 0.285 0.021 0.908 0.320 0.307 0.847 1.198
ππ–KK 0.028 0.034 0.010 0.112 0.008 0.023 0.656 0.014

41
Table 16: Systematic uncertainties on the fit fractions for the K-matrix approach given in
percent.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (892) 0.062 0.094 0.070 0.354 0.121 0.588 0.046 0.012
∗ 0
K (1410) 0.010 0.069 0.012 0.018 0.153 0.033 0.036 0.010
∗ 0
K (1680) 0.008 0.041 0.009 0.059 0.024 0.092 0.024 0.002
0
ρ(770) 0.036 0.096 0.040 0.496 0.192 0.357 0.058 0.082
ω(782) < 0.001 0.003 < 0.001 0.025 0.028 0.010 0.002 0.002
ρ(1450)0 0.033 0.075 0.032 0.036 0.006 0.254 0.055 0.059
0
ρ(1700) 0.005 0.044 0.003 0.040 0.036 0.103 0.010 0.034
∗ 0
K2 (1430) 0.050 0.235 0.047 0.098 0.031 0.026 0.077 0.030
f2 (1270) < 0.001 0.093 < 0.001 0.184 0.086 0.131 0.021 0.085
′
f2 (1525) 0.002 0.009 0.002 0.028 0.005 0.003 < 0.001 0.011
∗ 0
K3 (1780) 0.010 0.059 0.009 0.037 0.010 0.132 0.029 0.005
0
ρ3 (1690) 0.013 0.020 0.013 0.034 0.026 0.035 0.009 0.028
χc0 (1P ) 0.003 0.052 0.003 0.011 0.058 0.065 0.005 0.005
+ −
(K π )Res 0.051 0.301 0.042 0.304 0.341 0.079 0.115 0.093
+ −
(K π )ER 0.357 0.700 0.422 0.741 0.324 0.382 0.253 0.176
∗ 0
K0 (1950) 0.011 0.052 0.011 0.019 0.013 0.304 0.093 0.008
+ −
π π S-wave 0.160 0.237 0.179 0.909 0.312 1.007 0.120 0.031

Table 17: Systematic uncertainties on the fit fractions for the QMI approach given in percent.

B mass Eff Resol Bkg QMI bias Barrier Lineshape Extra

∗ 0
K (892) 0.007 0.100 0.001 0.045 0.283 0.284 0.021 0.022
K ∗ (1410)0 0.001 0.063 < 0.001 0.009 0.337 0.020 0.034 0.030
K ∗ (1680)0 0.001 0.048 < 0.001 0.022 0.098 0.002 0.013 0.013
ρ(770)0 0.013 0.090 0.004 0.215 0.490 0.151 0.034 0.028
ω(782) < 0.001 0.003 0.002 0.011 0.024 0.004 < 0.001 0.002
ρ(1450)0 0.041 0.136 0.025 0.089 0.427 0.003 0.090 0.049
ρ(1700)0 0.014 0.044 0.006 0.010 0.218 0.009 0.003 0.002
K2∗ (1430)0 0.008 0.115 0.007 0.046 0.040 0.225 0.031 0.030
f2 (1270) 0.020 0.095 0.003 0.312 0.080 0.223 0.037 0.021
f2′ (1525) < 0.001 0.012 < 0.001 0.014 0.015 0.004 0.001 0.018
K3∗ (1780)0 0.003 0.010 < 0.001 0.144 0.006 0.150 0.009 0.051
ρ3 (1690)0 0.001 0.018 < 0.001 0.021 0.027 0.084 0.001 0.013
χc0 (1P ) < 0.001 0.053 0.007 0.136 0.007 0.008 0.001 0.011
S-wave 0.074 0.134 0.029 0.233 0.497 0.232 0.164 0.066

42
 Table 18: Systematic uncertainties on the ACP values for the Isobar approach.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (892) < 0.001 0.015 < 0.001 0.005 0.009 0.001 0.002 0.001
K ∗ (1410)0 0.027 0.057 0.021 0.089 0.020 0.036 0.101 0.049
K ∗ (1680)0 0.024 0.085 0.043 0.673 0.007 0.022 0.214 0.058
ρ(770)0 < 0.001 0.008 < 0.001 0.006 0.013 0.005 0.017 0.016
ω(782) 0.003 0.014 0.001 0.007 0.045 0.002 0.012 0.042
ρ(1450)0 0.053 0.034 0.004 0.114 0.003 0.011 0.097 0.422
ρ(1700)0 0.034 0.044 < 0.001 0.071 0.067 0.061 0.077 0.972
K2∗ (1430)0 0.022 0.059 0.026 0.126 0.011 0.018 0.070 0.019
f2 (1270) 0.005 0.019 0.006 0.029 0.005 0.015 0.003 0.066
f2′ (1525) 0.028 0.059 0.004 0.032 0.193 0.017 0.014 0.156
K3∗ (1780)0 0.082 0.221 0.100 0.583 0.050 0.010 0.371 0.028
ρ3 (1690)0 0.005 0.019 < 0.001 0.100 0.017 0.042 0.008 0.063
χc0 (1P ) 0.001 0.004 0.002 0.001 0.013 0.006 0.003 0.004
(K + π − )Res 0.009 0.021 0.009 0.027 0.009 0.003 0.026 0.001
(K + π − )ER 0.002 0.004 0.001 0.011 0.008 0.002 0.015 0.024
K0∗ (1950)0 0.007 0.028 0.006 0.026 0.013 0.010 0.037 0.019
f0 (500) 0.008 0.031 0.026 0.037 0.018 0.067 0.136 0.079
f0 (980) 0.002 0.009 0.002 0.004 0.005 0.002 0.007 0.008
f0 (1370) 0.017 0.031 0.012 0.025 0.005 0.002 0.037 0.054
f0 (1500) 0.006 0.019 0.002 0.004 0.023 0.004 0.035 0.069
f0 (1710) 0.008 0.036 < 0.001 0.050 0.013 0.028 0.029 0.662
ππ–ππ 0.001 0.018 0.004 0.034 0.005 0.028 0.050 0.026
ππ–KK 0.057 0.170 0.046 0.137 0.206 0.059 0.198 0.078

43
 Table 19: Systematic uncertainties on the ACP values for the K-matrix approach.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (892) < 0.001 0.015 < 0.001 0.004 0.006 0.001 < 0.001 < 0.001
∗ 0
K (1410) 0.010 0.040 0.010 0.022 0.005 0.061 0.025 0.032
∗ 0
K (1680) 0.066 0.202 0.066 0.489 0.030 0.040 0.205 0.070
0
ρ(770) < 0.001 0.006 < 0.001 0.009 0.009 0.003 0.003 0.002
ω(782) < 0.001 0.016 0.002 0.019 0.056 0.003 0.002 0.015
0
ρ(1450) 0.005 0.040 0.001 0.137 0.003 0.030 0.043 0.046
0
ρ(1700) 0.002 0.043 0.001 0.074 0.070 0.007 0.025 0.044
∗ 0
K2 (1430) 0.028 0.083 0.027 0.121 0.029 0.009 0.044 0.007
f2 (1270) 0.007 0.024 0.007 0.025 0.003 0.014 0.002 0.020
′
f2 (1525) 0.008 0.041 0.008 0.084 0.203 < 0.001 0.010 0.169
∗ 0
K3 (1780) 0.104 0.342 0.105 0.490 0.141 0.001 0.193 0.024
0
ρ3 (1690) 0.006 0.021 0.006 0.064 < 0.001 0.027 0.009 0.068
χc0 (1P ) < 0.001 0.004 0.001 0.002 0.010 0.004 < 0.001 0.004
+ −
(K π )Res 0.009 0.025 0.009 0.035 0.018 0.001 0.012 0.003
+ −
(K π )ER < 0.001 0.006 < 0.001 0.007 0.013 0.001 0.002 0.002
∗ 0
K0 (1950) 0.006 0.020 0.008 0.003 0.021 0.004 0.013 0.006
+ −
π π S-wave 0.018 0.025 0.020 0.100 0.031 < 0.001 0.013 0.003

Table 20: Systematic uncertainties on the ACP values for the QMI approach.

B mass Eff Resol Bkg QMI bias Barrier Lineshape Extra

∗ 0
K (892) 0.001 0.017 < 0.001 0.009 0.012 < 0.001 0.001 0.002
K ∗ (1410)0 < 0.001 0.033 0.002 0.049 0.083 0.003 0.024 0.038
K ∗ (1680)0 0.020 0.082 0.005 0.134 0.047 0.002 0.003 0.077
ρ(770)0 0.003 0.006 0.001 0.021 0.041 0.001 0.008 0.005
ω(782) 0.003 0.010 0.001 0.009 0.002 < 0.001 0.001 0.004
ρ(1450)0 0.002 0.075 0.005 0.314 0.076 0.004 0.035 0.058
ρ(1700)0 0.008 0.063 < 0.001 0.395 0.083 0.001 0.004 0.058
K2∗ (1430)0 0.003 0.037 0.001 0.093 0.019 < 0.001 0.001 0.007
f2 (1270) 0.003 0.020 0.002 0.112 0.017 0.001 0.003 0.005
f2′ (1525) 0.006 0.119 0.003 0.027 0.401 0.004 0.018 0.126
K3∗ (1780)0 0.001 0.027 < 0.001 0.145 0.073 < 0.001 0.002 0.004
ρ3 (1690)0 0.004 0.024 0.004 0.131 0.030 0.005 0.003 0.091
χc0 (1P ) 0.001 0.006 0.001 0.017 0.008 < 0.001 0.002 0.008
S-wave 0.001 0.002 < 0.001 0.014 0.011 < 0.001 0.001 0.002

44
Table 21: Systematic uncertainties on the B + relative phases for the Isobar approach given in
degrees.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (1410) 0.7 2.2 0.1 7.9 1.2 0.3 4.4 1.1
K ∗ (1680)0 0.6 4.5 0.6 18.2 4.1 3.9 15.7 3.0
ρ(770)0 0.1 1.5 0.3 1.5 0.6 0.6 1.5 0.3
ω(782) 0.2 0.3 0.2 1.0 2.5 1.6 0.8 1.1
ρ(1450)0 0.6 4.1 0.6 6.1 2.4 0.7 0.8 2.3
ρ(1700)0 0.1 3.1 0.3 5.2 0.3 4.5 4.1 2.4
K2∗ (1430)0 0.5 2.1 0.4 1.1 0.3 0.5 1.9 0.4
f2 (1270) 0.1 1.4 0.1 2.5 0.2 0.2 1.0 2.3
f2′ (1525) 0.8 6.5 0.7 16.4 1.8 3.0 3.1 5.5
K3∗ (1780)0 0.3 1.4 0.3 10.0 1.3 3.4 1.3 3.5
ρ3 (1690)0 1.3 7.9 0.8 9.2 0.9 3.1 3.6 9.4
χc0 (1P ) 0.4 0.9 0.6 0.6 0.9 0.7 1.6 0.2
(K + π − )Res 0.9 1.6 1.0 7.9 2.9 2.4 0.5 1.3
(K + π − )ER 0.9 1.3 0.6 2.2 < 0.1 0.6 2.9 2.3
K0∗ (1950)0 0.1 3.2 0.4 5.8 0.5 1.2 4.8 2.0
f0 (500) 1.2 1.3 0.3 6.7 0.3 0.9 3.4 1.8
f0 (980) 0.4 0.6 0.2 2.6 2.0 2.1 2.1 0.8
f0 (1370) 0.3 1.2 0.1 5.0 1.7 0.6 1.6 1.5
f0 (1500) 0.4 1.6 0.1 3.8 0.5 1.3 1.3 1.6
f0 (1710) 0.4 1.1 0.2 6.6 0.5 0.4 1.5 4.7
ππ–ππ 1.6 3.0 1.4 6.7 2.1 0.2 3.7 1.9
ππ–KK 1.1 5.4 0.1 3.2 15.7 7.1 24.3 7.2

45
Table 22: Systematic uncertainties on the B + relative phases for the K-matrix approach given
in degrees.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (1410) 0.6 2.1 0.7 8.1 1.6 0.6 0.9 1.1
K ∗ (1680)0 1.7 3.2 2.0 16.0 5.0 2.5 2.2 3.1
ρ(770)0 < 0.1 1.1 0.1 4.9 1.9 1.1 0.6 0.9
ω(782) 0.3 0.6 0.4 3.5 0.8 2.0 0.4 1.4
ρ(1450)0 0.6 3.7 0.6 3.9 2.0 4.0 0.7 2.2
ρ(1700)0 0.2 3.4 0.2 1.6 < 0.1 6.0 2.5 1.2
K2∗ (1430)0 0.5 1.9 0.5 1.6 0.9 0.7 0.4 0.4
f2 (1270) 0.1 1.5 0.1 2.3 0.7 0.6 0.6 2.4
f2′ (1525) 0.9 5.5 0.9 13.0 3.6 3.0 1.4 2.4
K3∗ (1780)0 0.5 1.3 0.5 8.4 0.9 2.8 0.8 2.3
ρ3 (1690)0 1.1 6.9 1.0 8.0 1.3 1.5 1.5 10.1
χc0 (1P ) 0.4 0.9 0.6 1.0 1.2 0.7 0.3 0.2
(K + π − )Res 0.9 1.6 1.0 7.9 2.9 2.4 0.5 1.3
(K + π − )ER 0.4 1.8 0.3 4.6 1.8 0.8 0.3 0.3
K0∗ (1950)0 0.8 3.4 0.9 4.6 1.4 0.6 4.6 1.3

Table 23: Systematic uncertainties on the B + relative phases for the QMI approach given in
degrees.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (1410) 1.1 3.2 0.3 12.2 3.2 0.1 1.1 1.0
K ∗ (1680)0 0.5 8.3 0.7 60.1 3.1 1.3 4.2 17.1
ρ(770)0 0.2 2.9 < 0.1 1.2 8.4 < 0.1 0.5 1.1
ω(782) 0.1 2.4 0.2 2.7 4.6 < 0.1 0.6 2.0
ρ(1450)0 0.1 2.3 < 0.1 4.4 7.4 0.2 0.2 5.5
ρ(1700)0 0.7 2.1 0.6 3.0 0.2 0.1 1.0 2.1
K2∗ (1430)0 0.1 2.8 < 0.1 2.1 0.2 0.1 0.2 1.9
f2 (1270) 0.2 2.6 0.1 0.5 3.0 < 0.1 0.4 3.6
f2′ (1525) 0.2 8.0 0.2 9.2 9.0 0.3 0.5 1.5
K3∗ (1780)0 0.1 1.1 0.1 16.5 0.4 0.1 0.3 3.5
ρ3 (1690)0 0.1 10.3 0.3 1.0 4.5 < 0.1 1.5 14.8
χc0 (1P ) 0.1 1.9 < 0.1 1.7 0.4 < 0.1 0.1 0.7

46
Table 24: Systematic uncertainties on the B − relative phases for the Isobar approach given in
degrees.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (1410) 1.0 1.7 0.7 1.4 2.5 1.6 3.6 0.4
K ∗ (1680)0 9.8 19.6 10.0 25.5 5.4 2.6 37.3 5.5
ρ(770)0 1.5 2.3 1.5 1.3 1.1 0.9 4.7 6.5
ω(782) 1.6 2.5 1.8 1.6 1.9 1.3 5.5 6.3
ρ(1450)0 0.8 1.4 0.8 0.9 2.9 2.9 2.8 14.2
ρ(1700)0 1.1 5.5 0.7 15.0 0.9 18.0 3.4 44.8
K2∗ (1430)0 0.1 1.4 0.2 2.1 0.1 0.4 0.3 0.4
f2 (1270) 0.1 1.3 0.2 2.1 1.3 0.4 0.6 4.4
f2′ (1525) 0.4 17.1 0.3 18.5 16.3 1.1 4.2 54.9
K3∗ (1780)0 2.6 9.6 3.2 8.1 3.3 20.1 11.1 8.4
ρ3 (1690)0 0.2 0.8 0.2 5.2 2.1 0.1 1.3 27.8
χc0 (1P ) < 0.1 1.2 0.8 0.3 1.4 < 0.1 0.7 0.6
(K + π − )Res 0.2 0.8 < 0.1 1.7 2.2 0.3 1.2 1.4
(K + π − )ER 0.5 1.2 0.3 1.1 0.3 0.6 1.7 2.0
K0∗ (1950)0 1.1 4.6 0.9 4.3 0.4 0.1 8.6 0.4
f0 (500) 1.3 1.7 1.0 6.5 < 0.1 0.2 5.9 10.1
f0 (980) 0.5 0.9 0.7 0.8 1.6 2.0 2.9 3.8
f0 (1370) 0.5 0.7 0.3 3.2 1.8 0.7 2.3 1.7
f0 (1500) 0.3 1.0 0.2 1.4 1.4 1.0 1.6 6.3
f0 (1710) 0.8 3.6 0.6 6.0 0.4 1.8 5.9 38.7
ππ–ππ 1.3 2.1 1.4 2.1 1.6 2.9 2.7 8.8
ππ–KK 0.2 3.5 < 0.1 2.8 0.4 3.6 12.9 8.3

47
Table 25: Systematic uncertainties on the B − relative phases for the K-matrix approach given
in degrees.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (1410) < 0.1 1.2 < 0.1 0.5 1.2 1.5 0.8 0.1
K ∗ (1680)0 8.9 25.8 8.8 24.7 16.4 2.2 11.7 1.9
ρ(770)0 1.5 3.3 1.4 1.4 1.5 1.1 1.7 1.0
ω(782) 1.6 3.4 1.8 1.5 1.3 1.5 1.7 1.0
ρ(1450)0 1.0 2.1 0.8 0.8 1.9 2.4 1.8 1.9
ρ(1700)0 0.1 6.4 < 0.1 7.5 1.7 24.7 6.0 7.7
K2∗ (1430)0 < 0.1 1.4 < 0.1 2.4 0.1 0.6 0.9 0.2
f2 (1270) 0.1 2.0 0.1 2.4 1.4 0.6 0.8 1.8
f2′ (1525) 1.1 15.5 1.2 15.4 20.6 2.7 1.8 8.6
K3∗ (1780)0 2.4 14.4 2.4 3.9 6.0 17.6 4.0 0.8
ρ3 (1690)0 0.1 1.4 0.1 4.7 2.5 0.3 0.9 0.8
χc0 (1P ) 0.2 1.0 0.8 0.4 1.0 0.5 0.1 0.1
(K + π − )Res 0.4 2.3 0.4 4.4 3.9 2.4 0.9 1.3
(K + π − )ER < 0.1 1.3 0.1 3.2 0.9 0.3 0.5 0.3
K0∗ (1950)0 1.5 6.2 1.6 3.4 1.8 0.1 6.7 0.5

Table 26: Systematic uncertainties on the B − relative phases for the QMI approach given in
degrees.

B mass Eff Resol Bkg Fit bias Barrier Lineshape Extra

∗ 0
K (1410) 0.3 0.5 0.1 3.1 4.1 0.1 0.6 0.5
K ∗ (1680)0 0.5 1.7 0.1 2.5 13.7 < 0.1 0.7 1.3
ρ(770)0 0.4 2.0 0.3 2.9 4.7 < 0.1 0.2 0.6
ω(782) 0.4 1.8 0.4 2.9 2.7 < 0.1 0.4 0.7
ρ(1450)0 0.2 1.7 0.1 1.2 4.6 < 0.1 1.6 0.9
ρ(1700)0 0.1 9.2 0.9 27.3 7.8 0.4 2.0 7.5
K2∗ (1430)0 < 0.1 2.1 < 0.1 0.9 0.1 < 0.1 0.2 0.4
f2 (1270) < 0.1 1.2 0.1 0.8 3.8 < 0.1 < 0.1 1.6
f2′ (1525) 0.2 9.2 0.6 8.4 19.4 0.2 1.7 30.0
K3∗ (1780)0 1.3 8.8 0.3 6.2 0.9 < 0.1 3.3 16.8
ρ3 (1690)0 0.1 2.9 0.1 0.4 4.4 < 0.1 0.8 1.0
χc0 (1P ) < 0.1 1.1 0.2 1.6 0.7 0.1 0.2 0.1

48
D Goodness-of-fit comparison
As an indication of fit quality, χ2 goodness-of-fit indicators in the square Dalitz plot are
produced with an adaptive binning procedure that requires set minimum numbers of
signal candidates per bin when integrated over charge. The uncertainty is statistical only,
assuming the number of entries per bin is Poisson distributed. A comparison between the
three S-wave approaches can be found in Table 27, in which the number of bins NBin and
free parameters NPar are also recorded. Signed χ2 distributions for each S-wave approach
with at least 30 signal candidates per bin can be seen in Fig. 21.

Table 27: Goodness-of-fit indicators for each S-wave model with various adaptive binning
schemes.

Model Minimum B + : χ2 B − : χ2 NBin − NPar − 1

entries/bin
Isobar 15 5221.4 5760.6 3600 − 98 − 1
30 2879.4 2923.1 1764 − 98 − 1
45 2194.5 2074.0 1225 − 98 − 1
K-matrix 15 5163.7 5772.6 3600 − 107 − 1
30 2836.3 2945.0 1764 − 107 − 1
45 2156.0 2098.3 1225 − 107 − 1
QMI 15 5026.2 5621.4 3600 − 445 − 1
30 2702.4 2853.2 1764 − 445 − 1
45 2025.0 2040.8 1225 − 445 − 1

49
 1 30 1 30
θ'

θ'
Signed χ 2

Signed χ 2
0.9 0.9
0.8 20 0.8 20
0.7 10 0.7 10
0.6 0.6
0.5 0 0.5 0
0.4 0.4
0.3 −10 0.3 −10
LHCb LHCb
0.2 3 fb−1 −20 0.2 3 fb−1 −20
0.1 B+ 0.1 B−
0 −30 0 −30
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
m' m'
1 30 1 30
θ'

θ'
Signed χ 2

Signed χ 2
0.9 0.9
0.8 20 0.8 20
0.7 10 0.7 10
0.6 0.6
0.5 0 0.5 0
0.4 0.4
0.3 −10 0.3 −10
LHCb LHCb
0.2 3 fb−1 −20 0.2 3 fb−1 −20
0.1 B+ 0.1 B−
0 −30 0 −30
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
m' m'
1 30 1 30
θ'

θ'
Signed χ 2

Signed χ 2
0.9 0.9
0.8 20 0.8 20
0.7 10 0.7 10
0.6 0.6
0.5 0 0.5 0
0.4 0.4
0.3 −10 0.3 −10
LHCb LHCb
0.2 3 fb−1 −20 0.2 3 fb−1 −20
0.1 B+ 0.1 B−
0 −30 0 −30
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
m' m'
Figure 21: Signed χ2 distributions indicating the agreement between the (top) Isobar, (middle) K-
matrix and (bottom) QMI models and the data for (left) B + and (right) B − decays.

50
E Additional Isobar results
The lineshape parameters determined in the baseline model are given in Table 28. Various
projections of the data and fit results that serve to highlight the individual contributions
of each partial wave are shown here with the colour legend displayed in Fig. 22. Mass
projections with negative associated values of the cosine of the helicity angle, which
removes the bulk of the reflections, are shown in Figs. 23 and 24. The helicity structure
enhanced in the regions below and above the open-charm threshold are given in Figs. 25
and 26 for the K + π − and π + π − systems, respectively.

Table 28: Lineshape parameters determined in the Isobar S-wave approach, where the uncer-
tainties are statistical, systematic and model-induced, respectively.

Component Parameter Value

GLASS m0 1.461 ± 0.002 ± 0.001 ± 0.001 GeV/c2
Γ0 0.216 ± 0.004 ± 0.004 ± 0.002 GeV
a 6.900 ± 0.244 ± 0.771 ± 0.700 c/GeV
r −2.199 ± 0.051 ± 0.172 ± 0.082 c/GeV
ϕNonres 3.373 ± 0.011 ± 0.015 ± 0.022 rad
ϕER 0.386 ± 0.014 ± 0.040 ± 0.025 rad
f0 (980) m0 0.961 ± 0.002 ± 0.002 ± 0.002 GeV/c2
gππ 0.117 ± 0.003 ± 0.005 ± 0.003 GeV2/c4
gK K̄/ππ 2.571 ±0.123 ± 0.182 ± 0.178

Data
Background
S-wave
K +π − P-wave
π +π − P-wave
K +π − D-wave
π +π − D-wave
K +π − F-wave
π +π − F-wave
Figure 22: Colour legend of each partial wave shown in the subsequent figures of this section.

51
Candidates / (0.011 GeV/c2)

Candidates / (0.011 GeV/c2)

104 LHCb K 3π Kη K η' 104 LHCb K 3π Kη K η'
3 fb−1 3 fb−1
103 B+ 103 B−
2
10 102
10 10
1 1
−1 −1
10 10
−2
10 10−2
10−3 10−3
1 1.5 1 1.5
mK +π − (GeV/c2) mK −π + (GeV/c2)
Candidates / (0.032 GeV/c2)

Candidates / (0.032 GeV/c2)

104 LHCb K ηc DsD 104 LHCb K ηc DsD
10 3 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1
−1 −1
10 10
10−2 10−2
10−3 10−3
2 3 4 5 2 3 4 5
mK +π − (GeV/c2) mK −π + (GeV/c2)

Figure 23: Fit projections of mK + π− for (left) B + and (right) B − (top) below and (bottom) above
the open charm threshold, enhanced with cos θK + π− < 0. Vertical dashed lines correspond to
the opening thresholds of the indicated coupled channels.
Candidates / (0.015 GeV/c2)

Candidates / (0.015 GeV/c2)

LHCb 4π KK ηη ηη' LHCb 4π KK ηη ηη'

103 103
3 fb−1 3 fb−1
102 B+ 102 B−

10 10

1 1

10−1 10−1

10−2 10−2

10−3 10−3
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ −π + (GeV/c2)
Candidates / (0.029 GeV/c2)

Candidates / (0.029 GeV/c2)

LHCb ηηc DD DsDs/ η' ηc LHCb ηηc DD DsDs/ η' ηc

103 103
3 fb−1 3 fb−1
102 B+ 102 B−

10 10

1 1
−1 −1
10 10

10−2 10−2

10−3 10−3
2 3 4 2 3 4
mπ +π − (GeV/c2) mπ −π + (GeV/c2)

Figure 24: Fit projections of mπ+ π− for (left) B + and (right) B − (top) below and (bottom) above
the open charm threshold, enhanced with cos θπ+ π− < 0. Vertical dashed lines correspond to the
opening thresholds of the indicated coupled channels.

52
 105 105
Candidates / (0.02)

Candidates / (0.02)
LHCb LHCb
104 3 fb−1 104 3 fb−1
103 B+ 103 B−

102 102
10 10
1 1
10−1 10−1
10−2 10−2
−3
10 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ K +π − cosθ K −π +
Candidates / (0.02)

Candidates / (0.02)
104 LHCb 104 LHCb
103 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1

10−1 10−1
−2
10 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ K +π − cosθ K −π +

Figure 25: Fit projections of cos θK + π− for (left) B + and (right) B − (top) below and (bot-
tom) above the open charm threshold in mK + π− .

105 105
Candidates / (0.02)

Candidates / (0.02)

LHCb LHCb
4
10 3 fb−1 104 3 fb−1
103 B+ 103 B−
2
10 102
10 10
1 1
10−1 10−1
10−2 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ π +π − cosθ π −π +
Candidates / (0.02)

Candidates / (0.02)

104 LHCb 104 LHCb

10 3 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1
−1 −1
10 10
10−2 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ π +π − cosθ π −π +

Figure 26: Fit projections of cos θπ+ π− for (left) B + and (right) B − (top) below and (bot-
tom) above the open charm threshold in mπ+ π− .

53
F Additional K-matrix results
The lineshape parameters determined in the baseline model are given in Table 29. Various
projections of the data and fit results that serve to highlight the individual contributions
of each partial wave are shown here with the colour legend displayed in Fig. 27. Mass
projections with negative associated values of the cosine of the helicity angle, which
removes the bulk of the reflections, are shown in Figs. 28 and 29. The helicity structure
enhanced in the regions below and above the open-charm threshold are given in Figs. 30
and 31 for the K + π − and π + π − systems, respectively.

Table 29: Lineshape parameters determined in the K-matrix S-wave approach, where the
uncertainties are statistical, systematic and model-induced, respectively.

Component Parameter Value

GLASS m0 1.460 ± 0.002 ± 0.002 ± 0.001 GeV/c2
Γ0 0.217 ± 0.004 ± 0.004 ± 0.001 GeV
a 6.967 ± 0.234 ± 0.722 ± 0.135 c/GeV
r −2.180 ± 0.036 ± 0.118 ± 0.077 c/GeV
ϕNonres 3.483 ± 0.008 ± 0.036 ± 0.006 rad
ϕER 0.592 ± 0.009 ± 0.034 ± 0.015 rad

Data
Background
S-wave
K +π − P-wave
π +π − P-wave
K +π − D-wave
π +π − D-wave
K +π − F-wave
π +π − F-wave
Figure 27: Colour legend of each partial wave shown in the subsequent figures of this section.

54
Candidates / (0.011 GeV/c2)

Candidates / (0.011 GeV/c2)

104 LHCb K 3π Kη K η' 104 LHCb K 3π Kη K η'
3 fb−1 3 fb−1
103 B+ 103 B−
2
10 102
10 10
1 1
−1 −1
10 10
−2
10 10−2
10−3 10−3
1 1.5 1 1.5
mK +π − (GeV/c2) mK −π + (GeV/c2)
Candidates / (0.032 GeV/c2)

Candidates / (0.032 GeV/c2)

104 LHCb K ηc DsD 104 LHCb K ηc DsD
10 3 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1
−1 −1
10 10
10−2 10−2
10−3 10−3
2 3 4 5 2 3 4 5
mK +π − (GeV/c2) mK −π + (GeV/c2)

Figure 28: Fit projections of mK + π− for (left) B + and (right) B − (top) below and (bottom) above
the open charm threshold, enhanced with cos θK + π− < 0. Vertical dashed lines correspond to
the opening thresholds of the indicated coupled channels.
Candidates / (0.015 GeV/c2)

Candidates / (0.015 GeV/c2)

LHCb 4π KK ηη ηη' LHCb 4π KK ηη ηη'

103 103
3 fb−1 3 fb−1
102 B+ 102 B−

10 10

1 1

10−1 10−1

10−2 10−2

10−3 10−3
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ −π + (GeV/c2)
Candidates / (0.029 GeV/c2)

Candidates / (0.029 GeV/c2)

LHCb ηηc DD DsDs/ η' ηc LHCb ηηc DD DsDs/ η' ηc

103 103
3 fb−1 3 fb−1
102 B+ 102 B−

10 10

1 1
−1 −1
10 10

10−2 10−2

10−3 10−3
2 3 4 2 3 4
mπ +π − (GeV/c2) mπ −π + (GeV/c2)

Figure 29: Fit projections of mπ+ π− for (left) B + and (right) B − (top) below and (bottom) above
the open charm threshold, enhanced with cos θπ+ π− < 0. Vertical dashed lines correspond to the
opening thresholds of the indicated coupled channels.

55
 105 105
Candidates / (0.02)

Candidates / (0.02)
LHCb LHCb
104 3 fb−1 104 3 fb−1
103 B+ 103 B−

102 102
10 10
1 1
10−1 10−1
10−2 10−2
−3
10 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ K +π − cosθ K −π +
Candidates / (0.02)

Candidates / (0.02)
104 LHCb 104 LHCb
103 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1

10−1 10−1
−2
10 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ K +π − cosθ K −π +

Figure 30: Fit projections of cos θK + π− for (left) B + and (right) B − (top) below and (bot-
tom) above the open charm threshold in mK + π− .

105 105
Candidates / (0.02)

Candidates / (0.02)

LHCb LHCb
4
10 3 fb−1 104 3 fb−1
103 B+ 103 B−
2
10 102
10 10
1 1
10−1 10−1
10−2 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ π +π − cosθ π −π +
Candidates / (0.02)

Candidates / (0.02)

104 LHCb 104 LHCb

10 3 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1
−1 −1
10 10
10−2 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ π +π − cosθ π −π +

Figure 31: Fit projections of cos θπ+ π− for (left) B + and (right) B − (top) below and (bot-
tom) above the open charm threshold in mπ+ π− .

56
G Additional QMI results
Various projections of the data and fit results that serve to highlight the individual
contributions of each partial wave are shown here with the colour legend displayed in
Fig. 32. Mass projections with negative associated values of the cosine of the helicity
angle, which removes the bulk of the reflections, are shown in Figs. 33 and 34. The helicity
structure enhanced in the regions below and above the open-charm threshold are given in
Figs. 35 and 36 for the K + π − and π + π − systems, respectively.

Data
Background
S-wave
K +π − P-wave
π +π − P-wave
K +π − D-wave
π +π − D-wave
K +π − F-wave
π +π − F-wave
Figure 32: Colour legend of each partial wave shown in the subsequent figures of this section.

57
Candidates / (0.011 GeV/c2)

Candidates / (0.011 GeV/c2)

104 LHCb K 3π Kη K η' 104 LHCb K 3π Kη K η'
3 fb−1 3 fb−1
103 B+ 103 B−
2
10 102
10 10
1 1
−1 −1
10 10
−2
10 10−2
10−3 10−3
1 1.5 1 1.5
mK +π − (GeV/c2) mK −π + (GeV/c2)
Candidates / (0.032 GeV/c2)

Candidates / (0.032 GeV/c2)

104 LHCb K ηc DsD 104 LHCb K ηc DsD
10 3 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1
−1 −1
10 10
10−2 10−2
10−3 10−3
2 3 4 5 2 3 4 5
mK +π − (GeV/c2) mK −π + (GeV/c2)

Figure 33: Fit projections of mK + π− for (left) B + and (right) B − (top) below and (bottom) above
the open charm threshold, enhanced with cos θK + π− < 0. Vertical dashed lines correspond to
the opening thresholds of the indicated coupled channels.
Candidates / (0.015 GeV/c2)

Candidates / (0.015 GeV/c2)

LHCb 4π KK ηη ηη' LHCb 4π KK ηη ηη'

103 103
3 fb−1 3 fb−1
102 B+ 102 B−

10 10

1 1

10−1 10−1

10−2 10−2

10−3 10−3
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ −π + (GeV/c2)
Candidates / (0.029 GeV/c2)

Candidates / (0.029 GeV/c2)

LHCb ηηc DD DsDs/ η' ηc LHCb ηηc DD DsDs/ η' ηc

103 103
3 fb−1 3 fb−1
102 B+ 102 B−

10 10

1 1
−1 −1
10 10

10−2 10−2

10−3 10−3
2 3 4 2 3 4
mπ +π − (GeV/c2) mπ −π + (GeV/c2)

Figure 34: Fit projections of mπ+ π− for (left) B + and (right) B − (top) below and (bottom) above
the open charm threshold, enhanced with cos θπ+ π− < 0. Vertical dashed lines correspond to the
opening thresholds of the indicated coupled channels.

58
 105 105
Candidates / (0.02)

Candidates / (0.02)
LHCb LHCb
104 3 fb−1 104 3 fb−1
103 B+ 103 B−

102 102
10 10
1 1
10−1 10−1
10−2 10−2
−3
10 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ K +π − cosθ K −π +
Candidates / (0.02)

Candidates / (0.02)
104 LHCb 104 LHCb
103 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1

10−1 10−1
−2
10 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ K +π − cosθ K −π +

Figure 35: Fit projections of cos θK + π− for (left) B + and (right) B − (top) below and (bot-
tom) above the open charm threshold in mK + π− .

105 105
Candidates / (0.02)

Candidates / (0.02)

LHCb LHCb
4
10 3 fb−1 104 3 fb−1
103 B+ 103 B−
2
10 102
10 10
1 1
10−1 10−1
10−2 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ π +π − cosθ π −π +
Candidates / (0.02)

Candidates / (0.02)

104 LHCb 104 LHCb

10 3 3 fb−1 103 3 fb−1
B+ B−
102 102
10 10
1 1
−1 −1
10 10
10−2 10−2
10−3 10−3
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
cosθ π +π − cosθ π −π +

Figure 36: Fit projections of cos θπ+ π− for (left) B + and (right) B − (top) below and (bot-
tom) above the open charm threshold in mπ+ π− .

59
H Additional S-wave comparisons
Isobar and K-matrix S-wave results projected onto the one-dimensional QMI S-waves
across the full K + π − and π + π − mass ranges are shown in Figs. 37 and 38, respectively.
These can alternatively be expressed in two dimensions on the Argand plane as shown in
Figs. 39 and 40.

K +π − S-wave: arg(( A−S+A+S)/2) (°)

K +π − S-wave: |( A−S+A+S)/2|2

K ηc DsD 150
10−1 100

50
K ηc DsD
10−2 0

QMI −50

10−3 Isobar −100

(a) LHCb (b) LHCb
3 fb−1 K-matrix −150 3 fb−1
1 2 3 4 5 1 2 3 4 5
mK +π − (GeV/c2) mK +π − (GeV/c2)
1
K +π − S-wave: (| A−S|2−| A+S|2)/(| A−S|2+| A+S|2)

K +π − S-wave: arg( A−S)−arg( A+S) (°)

0.8 K ηc DsD 150

0.6 100
0.4
50
0.2
0 0
−0.2
−50
−0.4
−0.6 −100
(c) LHCb (d) LHCb
−0.8 3 fb−1 −150 3 fb−1 K ηc DsD
−1
1 2 3 4 5 1 2 3 4 5
mK +π − (GeV/c2) mK +π − (GeV/c2)

Figure 37: Comparison of the K + π − S-wave results for the (a) CP -averaged magnitude squared,
(b) CP -averaged phase, (c) asymmetry of the magnitude-squared and (d) phase difference
highlighting the charm region, where the vertical grey bands show the total QMI uncertainty
without accounting for an additional systematic effect arising from the fixed-bin values set from
the Isobar result. Vertical dashed lines correspond to the opening thresholds of the indicated
coupled channels.

60
 π +π − S-wave: arg(( A−S+A+S)/2) (°)
π +π − S-wave: |( A−S+A+S)/2|2

1 (a) LHCb ηηc D D DsDs 150 (b) LHCb ηηc D D DsDs

3 fb−1 η' η 3 fb−1 η' η
c 100 c
−1
10
50

0
10−2
−50
−3
10 −100
−150
10−4
1 2 3 4 1 2 3 4
mπ +π − (GeV/c2) mπ +π − (GeV/c2)
1
π +π − S-wave: (| A−S|2−| A+S|2)/(| A−S|2+| A+S|2)

π +π − S-wave: arg( A−S)−arg( A+S) (°)

0.8 (c) LHCb ηηc D D DsDs 150 (d) LHCb ηηc D D DsDs
0.6
3 fb−1 η' η 3 fb−1 η' η
c 100 c
0.4
50
0.2
0 0
−0.2
−50
−0.4 QMI
−0.6 −100 Isobar
−0.8 −150 K-matrix
−1
1 2 3 4 1 2 3 4
mπ +π − (GeV/c2) mπ +π − (GeV/c2)

Figure 38: Comparison of the π + π − S-wave results for the (a) CP -averaged magnitude squared,
(b) CP -averaged phase, (c) asymmetry of the magnitude-squared and (d) phase difference
highlighting the charm region, where the vertical grey bands show the total QMI uncertainty
without accounting for an additional systematic effect arising from the fixed-bin values set from
the Isobar result. Vertical dashed lines correspond to the opening thresholds of the indicated
coupled channels.

61
 1.8 1.8 1.8
0.6 0.6 0.6
1.6 1.6 1.6
S + AS )/2)

S + AS )/2)

S + AS )/2)
1.4 1.4 1.4
+

+
0.4 0.4 0.4
1.2 1.2 1.2

mK +π− (GeV/c2)

mK +π− (GeV/c2)

mK +π− (GeV/c2)
K +π − S-wave: Im((A−

K +π − S-wave: Im((A−

K +π − S-wave: Im((A−
0.2 0.2 0.2
1.0 1.0 1.0

0.8 0.8 0.8

0.0 0.0 0.0
0.6 0.6 0.6
−0.2 −0.2 −0.2
0.4 0.4 0.4

LHCb 0.2 LHCb 0.2 LHCb 0.2

−0.4 3 fb−1 −0.4 3 fb−1 −0.4 3 fb−1
0.0 0.0 0.0
−0.25 0.00 0.25 0.50 −0.25 0.00 0.25 0.50 −0.25 0.00 0.25 0.50
K +π − S-wave: Re((A− +
S + AS )/2) K +π − S-wave: Re((A− +
S + AS )/2) K +π − S-wave: Re((A− +
S + AS )/2)
1.8 1.8 1.8
+ + +
0.6 B 0.6 B 0.6 B
1.6 1.6 1.6

1.4 1.4 1.4

0.4 0.4 0.4
S)

S)

S)
K +π − S-wave: Im(A+

K +π − S-wave: Im(A+

K +π − S-wave: Im(A+
1.2 1.2 1.2
mK +π− (GeV/c2)

mK +π− (GeV/c2)

mK +π− (GeV/c2)
0.2 0.2 0.2
1.0 1.0 1.0

0.8 0.8 0.8

0.0 0.0 0.0
0.6 0.6 0.6
−0.2 −0.2 −0.2
0.4 0.4 0.4

LHCb 0.2 LHCb 0.2 LHCb 0.2

−0.4 3 fb−1 −0.4 3 fb−1 −0.4 3 fb−1
0.0 0.0 0.0
−0.25 0.00 0.25 0.50 −0.25 0.00 0.25 0.50 −0.25 0.00 0.25 0.50
K +π − S-wave: Re(A+
S) K +π − S-wave: Re(A+
S) K +π − S-wave: Re(A+
S)
1.8 1.8 1.8
− − −
0.6 B 0.6 B 0.6 B
1.6 1.6 1.6

1.4 1.4 1.4

0.4 0.4 0.4
S)

S)

S)
K +π − S-wave: Im(A−

K +π − S-wave: Im(A−

K +π − S-wave: Im(A−

1.2 1.2 1.2

mK +π− (GeV/c2)

mK +π− (GeV/c2)

mK +π− (GeV/c2)
0.2 0.2 0.2
1.0 1.0 1.0

0.8 0.8 0.8

0.0 0.0 0.0
0.6 0.6 0.6
−0.2 −0.2 −0.2
0.4 0.4 0.4

LHCb 0.2 LHCb 0.2 LHCb 0.2

−0.4 3 fb−1 −0.4 3 fb−1 −0.4 3 fb−1
0.0 0.0 0.0
−0.25 0.00 0.25 0.50 −0.25 0.00 0.25 0.50 −0.25 0.00 0.25 0.50
K +π − S-wave: Re(A−
S) K +π − S-wave: Re(A−
S) K +π − S-wave: Re(A−
S)

Figure 39: Argand diagrams of the (top) CP -averaged, (middle) B + and (bottom) B − S-wave
results in K + π − below the open-charm threshold for the (left) Isobar, (middle) K-matrix and
(right) QMI approaches. White curves indicate central values while smoothed bands cover
the total uncertainty at 1σ. In the QMI plots, cubic splines are provided purely for visual
guidance, and the shorter error bars indicate the total QMI uncertainty without accounting for
an additional systematic effect arising from the fixed-bin values set from the Isobar result.

62
 1.8 1.8 1.8
0.8 0.8 0.8
1.6 1.6 1.6
S + AS )/2)

S + AS )/2)

S + AS )/2)
0.6 1.4 0.6 1.4 0.6 1.4
+

+
0.4 1.2 0.4 1.2 0.4 1.2

mπ+π− (GeV/c2)

mπ+π− (GeV/c2)

mπ+π− (GeV/c2)
π +π − S-wave: Im((A−

π +π − S-wave: Im((A−

π +π − S-wave: Im((A−
0.2 1.0 0.2 1.0 0.2 1.0

0.8 0.8 0.8

0.0 0.0 0.0
0.6 0.6 0.6
−0.2 −0.2 −0.2
0.4 0.4 0.4
−0.4 LHCb 0.2 −0.4 LHCb 0.2 −0.4 LHCb 0.2
3 fb−1 3 fb−1 3 fb−1
−0.6 0.0 −0.6 0.0 −0.6 0.0
−1.0 −0.5 0.0 −1.0 −0.5 0.0 −1.0 −0.5 0.0
π +π − S-wave: Re((A− +
S + AS )/2) π +π − S-wave: Re((A− +
S + AS )/2) π +π − S-wave: Re((A− +
S + AS )/2)
1.8 1.8 1.8
+ + +
0.8 B 0.8 B 0.8 B
1.6 1.6 1.6
0.6 1.4 0.6 1.4 0.6 1.4
S)

S)

S)
π +π − S-wave: Im(A+

π +π − S-wave: Im(A+

π +π − S-wave: Im(A+
0.4 1.2 0.4 1.2 0.4 1.2
mπ+π− (GeV/c2)

mπ+π− (GeV/c2)

mπ+π− (GeV/c2)
0.2 1.0 0.2 1.0 0.2 1.0

0.8 0.8 0.8

0.0 0.0 0.0
0.6 0.6 0.6
−0.2 −0.2 −0.2
0.4 0.4 0.4
−0.4 LHCb 0.2 −0.4 LHCb 0.2 −0.4 LHCb 0.2
3 fb−1 3 fb−1 3 fb−1
−0.6 0.0 −0.6 0.0 −0.6 0.0
−1.0 −0.5 0.0 −1.0 −0.5 0.0 −1.0 −0.5 0.0
π +π − S-wave: Re(A+S) π +π − S-wave: Re(A+S) π +π − S-wave: Re(A+S)
1.8 1.8 1.8
− − −
0.8 B 0.8 B 0.8 B
1.6 1.6 1.6
0.6 1.4 0.6 1.4 0.6 1.4
S)

S)

S)
π +π − S-wave: Im(A−

π +π − S-wave: Im(A−

π +π − S-wave: Im(A−

0.4 1.2 0.4 1.2 0.4 1.2

mπ+π− (GeV/c2)

mπ+π− (GeV/c2)

mπ+π− (GeV/c2)
0.2 1.0 0.2 1.0 0.2 1.0

0.8 0.8 0.8

0.0 0.0 0.0
0.6 0.6 0.6
−0.2 −0.2 −0.2
0.4 0.4 0.4
−0.4 LHCb 0.2 −0.4 LHCb 0.2 −0.4 LHCb 0.2
3 fb−1 3 fb−1 3 fb−1
−0.6 0.0 −0.6 0.0 −0.6 0.0
−1.0 −0.5 0.0 −1.0 −0.5 0.0 −1.0 −0.5 0.0
π +π − S-wave: Re(A−S) π +π − S-wave: Re(A−S) π +π − S-wave: Re(A−S)

Figure 40: Argand diagrams of the (top) CP -averaged, (middle) B + and (bottom) B − S-wave
results in π + π − below the open-charm threshold for the (left) Isobar, (middle) K-matrix and
(right) QMI approaches. White curves indicate central values while smoothed bands cover
the total uncertainty at 1σ. In the QMI plots, cubic splines are provided purely for visual
guidance, and the shorter error bars indicate the total QMI uncertainty without accounting for
an additional systematic effect arising from the fixed-bin values set from the Isobar result.

63
I Legendre-weighted mass projections
Contributions from different partial waves and their interference can be visualised through
weighting data in the mass by Legendre polynomials ⟨Pn ⟩, of the cosine of the associated
helicity angle. In this way, CP violation mediated by the interference between partial
waves can also be seen by weighting the B − and B + samples separately, and then taking
their difference, ∆⟨Pn ⟩. These are shown in Figs. 41 and 42 for mK + π− and Figs. 43
and 44 for mπ+ π− , both below the open charm threshold.
As all known resonances decaying to either the K + π − or π + π − final state in the
charmless region are already included, areas in which all models disagree with the data
hint at where non-S-wave amplitude models could be improved in future analyses. This
could occur either at theoretical level to incorporate rescattering effects in higher partial
waves or through further experimental innovation again with quasi-model-independent
solutions. The inclusion of cc loop contributions may also provide an explanation [110,111].

64
 400
〈P1〉/(0.022 GeV/c2)

∆〈P1〉/(0.022 GeV/c2)
200
LHCb 120 LHCb
3 fb−1 100 3 fb−1
0
80
−200
60
−400
40
−600 20
−800 0
−1000 −20
−1200 −40
−1400 −60
−1600 −80
−1800
1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)
〈P2〉/(0.022 GeV/c2)

∆〈P2〉/(0.022 GeV/c2)
LHCb 100 LHCb
1600
3 fb−1 80 3 fb−1
1400
60
1200
1000 40
800 20
600 0
400 −20
200 −40
0 −60
1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)
〈P3〉/(0.022 GeV/c2)

∆〈P3〉/(0.022 GeV/c2)

LHCb 80 LHCb
100
3 fb−1 60 3 fb−1
50
40
0
20
−50
0
−100
−20
−150
−40
−200
−60
−250
1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)
〈P4〉/(0.022 GeV/c2)

∆〈P4〉/(0.022 GeV/c2)

150 LHCb 120 LHCb

3 fb−1 100 3 fb−1
100
80
50 60
0 40
20
−50
Isobar 0
−100 K-matrix −20
−150 QMI −40
1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)

Figure 41: (left) Background-subtracted data weighted by the first four Legendre polynomials in
cos θK + π− below the open charm threshold and integrated over the B charge, along with each
signal model. (right) Differentials of these distributions between B − and B + .

65
〈P5〉/(0.022 GeV/c2)

∆〈P5〉/(0.022 GeV/c2)
100 LHCb 60 LHCb
3 fb−1 3 fb−1
80 40
60
20
40
20 0

0 −20
−20
1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)

80
〈P6〉/(0.022 GeV/c2)

∆〈P6〉/(0.022 GeV/c2)
60 LHCb LHCb
3 fb−1 60 3 fb−1
40
40
20 20

0 0

−20 −20

−40
1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)
〈P7〉/(0.022 GeV/c2)

∆〈P7〉/(0.022 GeV/c2)

60
LHCb 40 LHCb
40 3 fb−1 3 fb−1
30
20 20
0 10
−20 0
−40 −10
−60 −20

1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)
60 60
〈P8〉/(0.022 GeV/c2)

∆〈P8〉/(0.022 GeV/c2)

40 LHCb LHCb
3 fb−1 50 3 fb−1
20 40
0
30
−20
20
−40
−60 10
−80 Isobar 0
−100 K-matrix −10
−120 QMI −20
1 1.5 1 1.5
mK +π − (GeV/c2) mK +π − (GeV/c2)

Figure 42: (left) Background-subtracted data weighted by the next four Legendre polynomials
in cos θK + π− below the open charm threshold and integrated over the B charge, along with each
signal model. (right) Differentials of these distributions between B − and B + .

66
〈P1〉/(0.029 GeV/c2)

∆〈P1〉/(0.029 GeV/c2)
400 LHCb LHCb
3 fb−1 200 3 fb−1
200
0 100
−200
0
−400
−600 −100
−800
−1000 −200
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ +π − (GeV/c2)
〈P2〉/(0.029 GeV/c2)

∆〈P2〉/(0.029 GeV/c2)
400
LHCb LHCb
300 3 fb−1 100 3 fb−1
200
50
100
0
0
−100 −50
−200 −100
−300
−150
−400
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ +π − (GeV/c2)
〈P3〉/(0.029 GeV/c2)

∆〈P3〉/(0.029 GeV/c2)

250 LHCb LHCb

3 fb−1 60 3 fb−1
200 40
20
150
0
100 −20

50 −40
−60
0
−80
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ +π − (GeV/c2)
〈P4〉/(0.029 GeV/c2)

∆〈P4〉/(0.029 GeV/c2)

200 30
LHCb LHCb
150 3 fb−1 20 3 fb−1
10
100
0
50
−10
0 −20
−50 −30
Isobar
−100 −40
K-matrix
−50
−150 QMI
−60
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ +π − (GeV/c2)

Figure 43: (left) Background-subtracted data weighted by the first four Legendre polynomials in
cos θπ+ π− below the open charm threshold and integrated over the B charge, along with each
signal model. (right) Differentials of these distributions between B − and B + .

67
〈P5〉/(0.029 GeV/c2)

∆〈P5〉/(0.029 GeV/c2)
180 LHCb 30 LHCb
160 3 fb−1 3 fb−1
20
140
10
120
0
100
80 −10
60 −20
40 −30
20 −40
0 −50
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ +π − (GeV/c2)
〈P6〉/(0.029 GeV/c2)

∆〈P6〉/(0.029 GeV/c2)
180 40
LHCb LHCb
160 3 fb−1 3 fb−1
140 20
120
100 0
80
60 −20
40
20 −40
0
−20 −60
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ +π − (GeV/c2)
〈P7〉/(0.029 GeV/c2)

∆〈P7〉/(0.029 GeV/c2)

LHCb 60 LHCb
100 3 fb−1 3 fb−1
40
80
20
60
0
40

20 −20

0 −40

−60
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ +π − (GeV/c2)
〈P8〉/(0.029 GeV/c2)

∆〈P8〉/(0.029 GeV/c2)

LHCb Isobar LHCb

40
100 3 fb−1 K-matrix 3 fb−1
80 30
QMI
20
60
10
40
0
20
−10
0
−20
−20 −30
−40 −40
0.5 1 1.5 0.5 1 1.5
mπ +π − (GeV/c2) mπ +π − (GeV/c2)

Figure 44: (left) Background-subtracted data weighted by the next four Legendre polynomials
in cos θπ+ π− below the open charm threshold and integrated over the B charge, along with each
signal model. (right) Differentials of these distributions between B − and B + .

68
Supplemental material
In addition to the results presented in the main body, a supplementary collection of files
recording the complete expressions of the amplitude models in each S-wave approach is
provided as *.[Link], in both their raw and convention-independent forms as they are
impractical to publish either in the main text or appendices. These files are given in the
JavaScript Object Notation (JSON) format, which is both machine and human readable,
containing parameter names, central values, total uncertainties and their corresponding
two-dimensional arrays representing correlations between parameters. Additional steering
files that drive the Laura++ Dalitz-plot software package to generate pseudoexperiments
based on the Isobar and K-matrix approaches are also included as [Link] and
[Link], respectively.

References
[1] N. Cabibbo, Unitary symmetry and leptonic decays, Phys. Rev. Lett. 10 (1963) 531.

[2] M. Kobayashi and T. Maskawa, CP -violation in the renormalizable theory of weak

interaction, Prog. Theor. Phys. 49 (1973) 652.

[3] M. E. Shaposhnikov, Standard model solution of the baryogenesis problem, Phys.

Lett. B277 (1992) 324, Erratum ibid. B282 (1992) 483.

[4] LHCb collaboration, R. Aaij et al., Measurement of CP violation in the phase

space of B ± → K ± π + π − and B ± → K ± K + K − decays, Phys. Rev. Lett. 111 (2013)
101801, arXiv:1306.1246.

[5] LHCb collaboration, R. Aaij et al., Measurement of CP violation in the phase space
of B ± → K + K − π ± and B ± → π + π − π ± decays, Phys. Rev. Lett. 112 (2014) 011801,
arXiv:1310.4740.

[6] LHCb collaboration, R. Aaij et al., Measurement of CP violation in the three-

body phase space of charmless B ± decays, Phys. Rev. D90 (2014) 112004,
arXiv:1408.5373.

[7] LHCb collaboration, R. Aaij et al., Direct CP violation in charmless three-body

decays of B ± mesons, Phys. Rev. D108 (2023) 012008, arXiv:2206.07622.

[8] LHCb collaboration, R. Aaij et al., Search for direct CP violation in charged
charmless B → P V decays, Phys. Rev. D108 (2023) 012013, arXiv:2206.02038.

[9] LHCb collaboration, R. Aaij et al., Amplitude analysis of B ± → π ± K + K − decays,

Phys. Rev. Lett. 123 (2019) 231802, arXiv:1905.09244.

[10] LHCb collaboration, R. Aaij et al., Observation of several sources of CP violation in

B + → π + π + π − decays, Phys. Rev. Lett. 124 (2020) 031801, arXiv:1909.05211.

[11] LHCb collaboration, R. Aaij et al., Amplitude analysis of the B + → π + π + π − decay,

Phys. Rev. D101 (2020) 012006, arXiv:1909.05212.

69
[12] R. H. Dalitz, On the analysis of τ -meson data and the nature of the τ -meson, Phil.
Mag. Ser. 7 44 (1953) 1068.
[13] E. Fabri, A study of τ -meson decay, Nuovo Cim. 11 (1954) 479.
[14] LHCb collaboration, R. Aaij et al., Resolution of outstanding puzzles and observation
of several intermediate states in B + → K + π + π − decays, LHCb-PAPER-2025-068,
in preparation.
[15] LHCb collaboration, R. Aaij et al., Observation of several sources of CP violation
in B + → K + π + π − decays, LHCb-PAPER-2025-069, in preparation.
[16] M. Beneke and M. Neubert, QCD factorization for B → P P and B → P V decays,
Nucl. Phys. B675 (2003) 333, arXiv:hep-ph/0308039.
[17] H.-Y. Cheng and C.-K. Chua, Revisiting charmless hadronic Bu,d decays in QCD
factorization, Phys. Rev. D80 (2009) 114008, arXiv:0909.5229.
[18] H.-Y. Cheng, C.-K. Chua, and Z.-Q. Zhang, Direct CP violation in charmless
three-body decays of B mesons, Phys. Rev. D94 (2016) 094015, arXiv:1607.08313.
[19] H.-Y. Cheng and K.-C. Yang, Charmless hadronic B decays into a tensor meson,
Phys. Rev. D83 (2011) 034001, arXiv:1010.3309.
[20] H.-Y. Cheng, C.-K. Chua, K.-C. Yang, and Z.-Q. Zhang, Revisiting charm-
less hadronic B decays to scalar mesons, Phys. Rev. D87 (2013) 114001,
arXiv:1303.4403.
[21] S. Kränkl, T. Mannel, and J. Virto, Three-body non-leptonic B decays and QCD
factorization, Nucl. Phys. B899 (2015) 247, arXiv:1505.04111.
[22] H.-n. Li and S. Mishima, Penguin-dominated B → P V decays in NLO perturbative
QCD, Phys. Rev. D74 (2006) 094020, arXiv:hep-ph/0608277.
[23] J. Chai et al., Charmless two-body B meson decays in the perturbative QCD factor-
ization approach, Chin. Phys. C46 (2022) 123103, arXiv:2207.04190.
[24] Y. Li et al., Quasi-two-body decays B(s) → P f2 (1270) → P ππ in the perturbative
QCD approach, Phys. Rev. D98 (2018) 056019, arXiv:1807.02641.
[25] W. Wang, Y.-L. Shen, Y. Li, and C.-D. Lu, Study of scalar mesons f0 (980) and
f0 (1500) from B → f0 (980)K and B → f0 (1500)K decays, Phys. Rev. D74 (2006)
114010, arXiv:hep-ph/0609082.
[26] Y.-L. Shen, W. Wang, J. Zhu, and C.-D. Lu, Study of K0∗ (1430) and a0 (980)
from B → K0∗ (1430)π and B → a0 (980)K decays, Eur. Phys. J. C50 (2007) 877,
arXiv:hep-ph/0610380.
[27] C.-H. Chen and H.-n. Li, Three-body nonleptonic B decays in perturbative QCD,
Phys. Lett. B561 (2003) 258, arXiv:hep-ph/0209043.
[28] W.-F. Wang, H.-C. Hu, H.-n. Li, and C.-D. Lü, Direct CP asymmetries of three-body
B decays in perturbative QCD, Phys. Rev. D89 (2014) 074031, arXiv:1402.5280.

70
[29] W. Wang, Y.-M. Wang, D.-S. Yang, and C.-D. Lu, Charmless two-body B(s) →
V P decays in soft collinear effective theory, Phys. Rev. D78 (2008) 034011,
arXiv:0801.3123.

[30] G. Bell, M. Beneke, T. Huber, and X.-Q. Li, Two-loop current–current operator
contribution to the non-leptonic QCD penguin amplitude, Phys. Lett. B750 (2015)
348, arXiv:1507.03700.

[31] H.-Y. Cheng, C.-W. Chiang, and A.-L. Kuo, Updating B → P P , V P decays in the
framework of flavor symmetry, Phys. Rev. D91 (2015) 014011, arXiv:1409.5026.

[32] H. J. Lipkin, Y. Nir, H. R. Quinn, and A. Snyder, Penguin trapping with isospin
analysis and CP asymmetries in B decays, Phys. Rev. D44 (1991) 1454.

[33] M. Gronau, A precise sum rule among four B → Kπ CP asymmetries, Phys. Lett.
B627 (2005) 82, arXiv:hep-ph/0508047.

[34] M. Gronau, D. Pirjol, and J. Zupan, CP asymmetries in B → Kπ, K ∗ π, ρK decays,

Phys. Rev. D81 (2010) 094011, arXiv:1001.0702.

[35] B. Bhattacharya, M. Imbeault, and D. London, Extraction of the CP-violating

phase γ using B → Kππ and B → KKK decays, Phys. Lett. B728 (2014) 206,
arXiv:1303.0846.

[36] B. Bhattacharya et al., Charmless B → P P P decays: The fully-symmetric final

state, Phys. Rev. D89 (2014) 074043, arXiv:1402.2909.

[37] E. Bertholet et al., Extraction of the CKM phase γ using charmless 3-body decays
of B mesons, Phys. Rev. D99 (2019) 114011, arXiv:1812.06194.

[38] B. Bhattacharya et al., Charmless B → P P P decays: The fully antisymmetric final

state, Phys. Rev. D109 (2024) 013001, arXiv:2308.16240.

[39] M. Jung, Determining weak phases from B → J/ψP decays, Phys. Rev. D86 (2012)
053008, arXiv:1206.2050.

[40] K. De Bruyn and R. Fleischer, A roadmap to control penguin effects in Bd0 → J/ψKS0
and Bs0 → J/ψϕ, JHEP 03 (2015) 145, arXiv:1412.6834.

[41] R. Garcia-Martin et al., Pion-pion scattering amplitude. IV. Improved analysis with
once subtracted Roy-like equations up to 1100 MeV, Phys. Rev. D83 (2011) 074004,
arXiv:1102.2183.

[42] J. R. Peláez and A. Rodas, Dispersive πK → πK and ππ → KK amplitudes

from scattering data, threshold parameters, and the lightest strange resonance κ or
K0∗ (700), Phys. Rept. 969 (2022) 1, arXiv:2010.11222.

[43] LHCb collaboration, A. A. Alves Jr. et al., The LHCb detector at the LHC, JINST
3 (2008) S08005.

[44] LHCb collaboration, R. Aaij et al., LHCb detector performance, Int. J. Mod. Phys.
A30 (2015) 1530022, arXiv:1412.6352.

71
[45] R. Aaij et al., Performance of the LHCb Vertex Locator, JINST 9 (2014) P09007,
arXiv:1405.7808.

[46] R. Arink et al., Performance of the LHCb Outer Tracker, JINST 9 (2014) P01002,
arXiv:1311.3893.

[47] M. Adinolfi et al., Performance of the LHCb RICH detector at the LHC, Eur. Phys.
J. C73 (2013) 2431, arXiv:1211.6759.

[48] A. A. Alves Jr. et al., Performance of the LHCb muon system, JINST 8 (2013)
P02022, arXiv:1211.1346.

[49] R. Aaij et al., The LHCb trigger and its performance in 2011, JINST 8 (2013)
P04022, arXiv:1211.3055.

[50] V. V. Gligorov and M. Williams, Efficient, reliable and fast high-level triggering
using a bonsai boosted decision tree, JINST 8 (2013) P02013, arXiv:1210.6861.

[51] N. Grieser et al., The LHCb stripping project: Sustainable legacy data processing for
high-energy physics, Comput. Softw. Big. Sci. 9 (2025) 21, arXiv:2509.05294.

[52] T. Sjöstrand, S. Mrenna, and P. Skands, PYTHIA 6.4 physics and manual, JHEP
05 (2006) 026, arXiv:hep-ph/0603175.

[53] T. Sjöstrand, S. Mrenna, and P. Skands, A brief introduction to PYTHIA 8.1,

Comput. Phys. Commun. 178 (2008) 852, arXiv:0710.3820.

[54] I. Belyaev et al., Handling of the generation of primary events in Gauss, the LHCb
simulation framework, J. Phys. Conf. Ser. 331 (2011) 032047.

[55] D. J. Lange, The EvtGen particle decay simulation package, Nucl. Instrum. Meth.
A462 (2001) 152.

[56] P. Golonka and Z. Was, PHOTOS Monte Carlo: A precision tool for QED corrections
in Z and W decays, Eur. Phys. J. C45 (2006) 97, arXiv:hep-ph/0506026.

[57] Geant4 collaboration, J. Allison et al., Geant4 developments and applications, IEEE
Trans. Nucl. Sci. 53 (2006) 270.

[58] Geant4 collaboration, S. Agostinelli et al., Geant4: A simulation toolkit, Nucl.

Instrum. Meth. A506 (2003) 250.

[59] M. Clemencic et al., The LHCb simulation application, Gauss: Design, evolution
and experience, J. Phys. Conf. Ser. 331 (2011) 032023.

[60] LHCb collaboration, R. Aaij et al., Measurement of the relative branching fraction
′ ′
of B + → h+ h + h − decays, Phys. Rev. D102 (2020) 112010, arXiv:2010.11802.

[61] L. Breiman, J. H. Friedman, R. A. Olshen, and C. J. Stone, Classification and

regression trees, Wadsworth international group, Belmont, California, USA, 1984.

[62] Particle Data Group, S. Navas et al., Review of particle physics, Phys. Rev. D110
(2024) 030001.

72
[63] Particle Data Group, P. A. Zyla et al., Review of particle physics, Prog. Theor. Exp.
Phys. 2020 (2020) 083C01.

[64] T. Skwarnicki, A study of the radiative cascade transitions between the Upsilon-prime
and Upsilon resonances, PhD thesis, Institute of Nuclear Physics, Krakow, 1986,
DESY-F31-86-02.

[65] ARGUS collaboration, H. Albrecht et al., Exclusive hadronic decays of B mesons, Z.

Phys. C48 (1990) 543.

[66] BESIII collaboration, M. Ablikim et al., Precision study of η ′ → γπ + π − decay

dynamics, Phys. Rev. Lett. 120 (2018) 242003, arXiv:1712.01525.

[67] Heavy Flavor Averaging Group, Y. Amhis et al., Averages of b-hadron, c-hadron, and
τ -lepton properties as of 2021, Phys. Rev. D107 (2023) 052008, arXiv:2206.07501,
updated results and plots available at [Link]

[68] S. Mandelstam, J. E. Paton, R. F. Peierls and A. Q. Sarker, Isobar approximation

of production processes, Ann. Phys. 18 (1962) 198.

[69] D. J. Herndon, P. Söding, and R. J. Cashmore, Generalized isobar model formalism,

Phys. Rev. D11 (1975) 3165.

[70] J. J. Brehm, Unitarity and the isobar model: Two-body discontinuities, Ann. Phys.
108 (1977) 454.

[71] J. M. Blatt and V. F. Weisskopf, Theoretical nuclear physics, Springer, New York,
1952.

[72] J. D. Jackson, Remarks on the phenomenological analysis of resonances, Nuovo Cim.

34 (1964) 1644.

[73] C. Zemach, Three pion decays of unstable particles, Phys. Rev. 133 (1964) B1201.

[74] C. Zemach, Use of angular momentum tensors, Phys. Rev. 140 (1965) B97.

[75] G. J. Gounaris and J. J. Sakurai, Finite width corrections to the vector meson
dominance prediction for ρ → e+ e− , Phys. Rev. Lett. 21 (1968) 244.

[76] BaBar collaboration, B. Aubert et al., Improved measurement of the CKM angle γ
in B ∓ → D(∗) K (∗∓) decays with a Dalitz plot analysis of D decays to KS0 π + π − and
KS0 K + K − , Phys. Rev. D78 (2008) 034023, arXiv:0804.2089.

[77] P. Estabrooks, Where and what are the scalar mesons?, Phys. Rev. D19 (1979)
2678.

[78] D. Aston et al., A study of K − π + scattering in the reaction K − p → K − π + n at

11 GeV/c, Nucl. Phys. B296 (1988) 493.

[79] E791 collaboration, E. M. Aitala et al., Model independent measurement of S-wave

K − π + systems using D+ → Kππ decays from Fermilab E791, Phys. Rev. D73
(2006) 032004, Erratum ibid. D74 (2006) 059901, arXiv:hep-ex/0507099.

73
[80] D. V. Bugg, The mass of the σ pole, J. Phys. G34 (2007) 151,
arXiv:hep-ph/0608081.

[81] S. M. Flatté, Coupled-channel analysis of the πη and KK systems near KK threshold,

Phys. Lett. B63 (1976) 224.

[82] J. R. Pelaez, A. Rodas, and J. R. de Elvira, The f0 (1370) controversy from dispersive
meson-meson scattering data analyses, Phys. Rev. Lett. 130 (2023) 051902, Erratum
ibid. 132 (2024) 239901, arXiv:2206.14822.

[83] K. M. Watson, The effect of final state interactions on reaction cross sections, Phys.
Rev. 88 (1952) 1163.

[84] J. H. Alvarenga Nogueira et al., CP violation: Dalitz interference, CP T , and final

state interactions, Phys. Rev. D92 (2015) 054010, arXiv:1506.08332.

[85] I. Bediaga, T. Frederico, and O. Lourenço, CP violation and CPT invari-

ance in B ± decays with final state interactions, Phys. Rev. D89 (2014) 094013,
arXiv:1307.8164.

[86] R. A. Garrote, J. Cuervo, P. C. Magalhães, and J. R. Peláez, Dispersive ππ → KK

amplitude and giant CP violation in B to three light-meson decays at LHCb, Phys.
Rev. Lett. 130 (2023) 201901, arXiv:2210.08354.

[87] S. U. Chung et al., Partial wave analysis in K-matrix formalism, Annalen Phys. 4
(1995) 404.

[88] R. H. Dalitz and S. F. Tuan, The phenomenological representation of K-nucleon

scattering and reaction amplitudes, Annals Phys. 10 (1960) 307.

[89] E. P. Wigner, Resonance reactions and anomalous scattering, Phys. Rev. 70 (1946)
15.

[90] E. P. Wigner and L. Eisenbud, Higher angular momenta and long range interaction
in resonance reactions, Phys. Rev. 72 (1947) 29.

[91] I. J. R. Aitchison, The K-matrix formalism for overlapping resonances, Nucl. Phys.
A189 (1972) 417.

[92] V. V. Anisovich and A. V. Sarantsev, K-matrix analysis of the (IJ P C =

00++ )-wave in the mass region below 1900 MeV, Eur. Phys. J. A16 (2003) 229,
arXiv:hep-ph/0204328.

[93] FOCUS collaboration, J. M. Link et al., Dalitz plot analysis of Ds+ and D+
decay to π + π − π + using the K-matrix formalism, Phys. Lett. B585 (2004) 200,
arXiv:hep-ex/0312040.

[94] F. Krinner et al., Ambiguities in model-independent partial-wave analysis, Phys.

Rev. D97 (2018) 114008, arXiv:1710.09849.

[95] BaBar collaboration, B. Aubert et al., An amplitude analysis of the decay B ± →

π ± π ± π ∓ , Phys. Rev. D72 (2005) 052002, arXiv:hep-ex/0507025.

74
 [96] LHCb collaboration, R. Aaij et al., Measurement of B 0 , Bs0 , B + and Λ0b production
asymmetries in 7 and 8 TeV proton-proton collisions, Phys. Lett. B774 (2017) 139,
arXiv:1703.08464.

[97] LHCb collaboration, R. Aaij et al., Measurement of the track reconstruction efficiency
at LHCb, JINST 10 (2015) P02007, arXiv:1408.1251.

[98] L. Anderlini et al., The PIDCalib package, LHCb-PUB-2016-021, CERN, 2016.

[99] J. Back et al., Laura++ : A Dalitz plot fitter, Comput. Phys. Commun. 231 (2018)
198, arXiv:1711.09854, the software version used in this analysis is available from
[Link]

[100] F. James and M. Roos, Minuit: A system for function minimization and analysis
of the parameter errors and correlations, Comput. Phys. Commun. 10 (1975) 343.

[101] R. Brun and F. Rademakers, ROOT: An object oriented data analysis framework,
Nucl. Instrum. Meth. A389 (1997) 81.

[102] J. Rademacker, P. d’Argent, and J. Dalseno, Mint2, doi: 10.5281/zenodo.2585535.

[103] LHCb collaboration, R. Aaij et al., Updated branching fraction measurements of

0
B(s) → KS0 h+ h′− decays, JHEP 11 (2017) 027, arXiv:1707.01665.

[104] LHCb collaboration, R. Aaij et al., Amplitude analysis of Bs0 → KS0 K ± π ∓ decays,
JHEP 06 (2019) 114, arXiv:1902.07955.

[105] See supplemental material at [link] for complete results files including correlation
matrices of the total uncertainties, as well as Laura++ steering files.

[106] Belle collaboration, A. Garmash et al., Evidence for large direct CP violation in
B ± → ρ(770)0 K ± from analysis of the three-body charmless B ± → K ± π ± π ∓ decay,
Phys. Rev. Lett. 96 (2006) 251803, arXiv:hep-ex/0512066.

[107] BaBar collaboration, B. Aubert et al., Evidence for direct CP violation from Dalitz-
plot analysis of B ± → K ± π ∓ π ± , Phys. Rev. D78 (2008) 012004, arXiv:0803.4451.

[108] BaBar collaboration, B. Aubert et al., Branching fraction and CP -violation charge
asymmetry measurements for B-meson decays to ηK ± , ηπ ± , η ′ K, η ′ π ± , ωK, and
ωπ ± , Phys. Rev. D76 (2007) 031103, arXiv:0706.3893.

[109] Belle collaboration, V. Chobanova et al., Measurement of branching fractions and

CP violation parameters in B → ωK decays with first evidence of CP violation in
B 0 → ωKS0 , Phys. Rev. D90 (2014) 012002, arXiv:1311.6666.

[110] B. El-Bennich et al., CP violation and kaon-pion interactions in B → Kπ + π −

decays, Phys. Rev. D79 (2009) 094005, Erratum ibid. D83 (2011) 039903,
arXiv:0902.3645.

[111] L. A. Heuser et al., Understanding large localized CP violation in B ± → K ± π + π −

using dispersive methods, Phys. Rev. Lett. 136 (2026) 111901, arXiv:2508.10989.

75
 LHCb collaboration
R. Aaij38 , M. Abdelfatah69 , A.S.W. Abdelmotteleb57 , C. Abellan Beteta51 ,
F. Abudinén59 , T. Ackernley61 , A.A. Adefisoye69 , B. Adeva47 , M. Adinolfi55 ,
P. Adlarson87 , C. Agapopoulou14 , C.A. Aidala89 , Z. Ajaltouni11 , S. Akar11 ,
K. Akiba38 , P. Albicocco28 , J. Albrecht19,g , R. Aleksiejunas81 , F. Alessio49 ,
P. Alvarez Cartelle56,47 , R. Amalric16 , S. Amato3 , J.L. Amey55 , Y. Amhis14 ,
L. An6 , L. Anderlini27 , M. Andersson51 , P. Andreola51 , M. Andreotti26 ,
S. Andres Estrada44 , A. Anelli31,p , D. Ao7 , C. Arata12 , F. Archilli37 , Z. Areg69 ,
M. Argenton26 , S. Arguedas Cuendis9,49 , L. Arnone31,p , M. Artuso69 ,
E. Aslanides13 , R. Ataı́de Da Silva50 , M. Atzeni65 , B. Audurier12 , J.A. Authier15 ,
D. Bacher64 , I. Bachiller Perea50 , S. Bachmann22 , M. Bachmayer50 , J.J. Back57 ,
Z.B. Bai8 , P. Baladron Rodriguez47 , V. Balagura15 , A. Balboni26 , W. Baldini26 ,
Z. Baldwin79 , L. Balzani19 , H. Bao7 , J. Baptista de Souza Leite2 ,
C. Barbero Pretel47,12 , M. Barbetti27 , I.R. Barbosa70 , R.J. Barlow63,† ,
M. Barnyakov25 , S. Barsuk14 , W. Barter59 , J. Bartz69 , S. Bashir40 , B. Batsukh82 ,
P.B. Battista14 , A. Bavarchee80 , A. Bay50 , A. Beck65 , M. Becker19 ,
F. Bedeschi35 , I.B. Bediaga2 , N.A. Behling19 , S. Belin47 , A. Bellavista25 ,
I. Belov29 , I. Belyaev36 , G. Benane13 , G. Bencivenni28 , E. Ben-Haim16 ,
R. Bernet51 , A. Bertolin33 , F. Betti59 , J. Bex56 , O. Bezshyyko88 ,
S. Bhattacharya80 , M.S. Bieker18 , N.V. Biesuz26 , A. Biolchini38 , M. Birch62 ,
F.C.R. Bishop10 , A. Bitadze63 , A. Bizzeti27,q , T. Blake57,c , F. Blanc50 ,
J.E. Blank19 , S. Blusk69 , J.A. Boelhauve19 , O. Boente Garcia49 , T. Boettcher90 ,
A. Bohare59 , C. Bolognani19 , R. Bolzonella26,m , R.B. Bonacci1 , A. Bordelius49 ,
F. Borgato33,49 , S. Borghi63 , M. Borsato31,p , J.T. Borsuk86 , E. Bottalico61 ,
S.A. Bouchiba50 , M. Bovill64 , T.J.V. Bowcock61 , A. Boyer49 , C. Bozzi26 ,
J.D. Brandenburg91 , A. Brea Rodriguez50 , N. Breer19 , C. Breitfeld19 ,
J. Brodzicka41 , J. Brown61 , D. Brundu32 , E. Buchanan59 , M. Burgos Marcos84 ,
C. Burr49 , C. Buti27 , J.S. Butter56 , J. Buytaert49 , W. Byczynski49 ,
S. Cadeddu32 , H. Cai75 , Y. Cai5 , A. Caillet16 , R. Calabrese26,m , L. Calefice45 ,
M. Calvi31,p , M. Calvo Gomez46 , P. Camargo Magalhaes2,a , J.I. Cambon Bouzas47 ,
P. Campana28 , A.C. Campos3 , A.F. Campoverde Quezada7 , Y. Cao6 , S. Capelli31,p ,
M. Caporale25 , L. Capriotti26 , R. Caravaca-Mora9 , A. Carbone25,k ,
L. Carcedo Salgado47 , R. Cardinale29,n , A. Cardini32 , P. Carniti31 , L. Carus22 ,
A. Casais Vidal65 , R. Caspary22 , G. Casse61 , M. Cattaneo49 , G. Cavallero26 ,
V. Cavallini26,m , S. Celani49 , I. Celestino35,t , S. Cesare49,o , A.J. Chadwick61 ,
I. Chahrour89 , H. Chang4,d , M. Charles16 , Ph. Charpentier49 ,
E. Chatzianagnostou38 , R. Cheaib80 , M. Chefdeville10 , C. Chen57 , J. Chen50 ,
S. Chen5 , Z. Chen7 , A. Chen Hu62 , M. Cherif12 , A. Chernov41 ,
S. Chernyshenko53 , X. Chiotopoulos84 , G. Chizhik1 , V. Chobanova44 ,
M. Chrzaszcz41 , V. Chulikov28,49,36 , P. Ciambrone28 , X. Cid Vidal47 , G. Ciezarek49 ,
P. Cifra49 , P.E.L. Clarke59 , M. Clemencic49 , H.V. Cliff56 , J. Closier49 ,
C. Cocha Toapaxi22 , V. Coco49 , J. Cogan13 , E. Cogneras11 , L. Cojocariu43 ,
S. Collaviti50 , P. Collins49 , T. Colombo49 , M. Colonna19 , A. Comerma-Montells45 ,
L. Congedo24 , J. Connaughton57 , A. Contu32 , N. Cooke60 , G. Cordova35,t ,
C. Coronel66 , I. Corredoira 12 , A. Correia16 , G. Corti49 , G.C. Costantino61 ,
J. Cottee Meldrum55 , B. Couturier49 , D.C. Craik51 , N. Crepet14 , M. Cruz Torres2,h ,
M. Cubero Campos9 , E. Curras Rivera50 , R. Currie59 , C.L. Da Silva68 , X. Dai4 ,
E. Dall’Occo49 , J. Dalseno44,47 , C. D’Ambrosio62 , J. Daniel11 , G. Darze3 ,
A. Davidson57 , J.E. Davies63 , O. De Aguiar Francisco63 , C. De Angelis32,l ,

76
F. De Benedetti49 , J. de Boer38 , K. De Bruyn83 , S. De Capua63 , M. De Cian63 ,
U. De Freitas Carneiro Da Graca2,b , E. De Lucia28 , J.M. De Miranda2 , L. De Paula3 ,
M. De Serio24,i , P. De Simone28 , F. De Vellis19 , J.A. de Vries84 , F. Debernardis24 ,
D. Decamp10 , S. Dekkers1 , L. Del Buono16 , B. Delaney65 , J. Deng8 ,
V. Denysenko51 , O. Deschamps11 , F. Dettori32,l , B. Dey80 , P. Di Nezza28 ,
S. Ding69 , Y. Ding50 , L. Dittmann22 , A.D. Docheva60 , A. Doheny57 , C. Dongd,4 ,
F. Dordei32 , A.C. dos Reis2 , A.D. Dowling69 , L. Dreyfus13 , W. Duan73 ,
P. Duda86 , L. Dufour50 , V. Duk34 , P. Durante49 , M.M. Duras86 , J.M. Durham68 ,
O.D. Durmus80 , A. Dziurda41 , S. Easo58 , E. Eckstein18 , U. Egede1 ,
S. Eisenhardt59 , E. Ejopu61 , L. Eklund87 , M. Elashri66 , D. Elizondo Blanco9 ,
J. Ellbracht19 , S. Ely62 , A. Ene43 , J. Eschle69 , T. Evans38 , F. Fabiano14 ,
S. Faghih66 , L.N. Falcao31,p , B. Fang7 , R. Fantechi35 , L. Fantini34,s , M. Faria50 ,
K. Farmer59 , F. Fassin83,38 , D. Fazzini31,p , L. Felkowski86 , C. Feng6 , M. Feng5,7 ,
A. Fernandez Casani48 , M. Fernandez Gomez47 , A.D. Fernez67 , F. Ferrari25,k ,
F. Ferreira Rodrigues3 , M. Ferrillo51 , M. Ferro-Luzzi49 , R.A. Fini24 , M. Fiorini26,m ,
M. Firlej40 , K.L. Fischer64 , D.S. Fitzgerald89 , C. Fitzpatrick63 , T. Fiutowski40 ,
F. Fleuret15 , A. Fomin52 , M. Fontana25,49 , L.A. Foreman63 , R. Forty49 ,
D. Foulds-Holt59 , V. Franco Lima3 , M. Franco Sevilla67 , M. Frank49 ,
E. Franzoso26,m , G. Frau63 , C. Frei49 , D.A. Friday63,49 , J. Fu7 , Q. Führing19,56,g ,
T. Fulghesu13 , G. Galati24,i , M.D. Galati38 , A. Gallas Torreira47 , D. Galli25,k ,
S. Gambetta59 , M. Gandelman3 , P. Gandini30 , B. Ganie63 , H. Gao7 , R. Gao64 ,
T.Q. Gao56 , Y. Gao8 , Y. Gao6 , Y. Gao8 , L.M. Garcia Martin50 ,
P. Garcia Moreno45 , J. Garcı́a Pardiñas65 , P. Gardner67 , L. Garrido45 , C. Gaspar49 ,
A. Gavrikov33 , L.L. Gerken19 , E. Gersabeck20 , M. Gersabeck20 , T. Gershon57 ,
S. Ghizzo29,n , Z. Ghorbanimoghaddam55 , F.I. Giasemis16,f , V. Gibson56 ,
H.K. Giemza42 , A.L. Gilman66 , M. Giovannetti28 , A. Gioventù47 , L. Girardey63,58 ,
M.A. Giza41 , F.C. Glaser22,14 , V.V. Gligorov16 , C. Göbel70 ,
L. Golinka-Bezshyyko88 , E. Golobardes46 , A. Golutvin62,49 , S. Gomez Fernandez45 ,
W. Gomulka40 , F. Goncalves Abrantes64 , I. Gonçales Vaz49 , M. Goncerz41 ,
G. Gong4,d , J.A. Gooding19 , C. Gotti31 , E. Govorkova65 , J.P. Grabowski30 ,
L.A. Granado Cardoso49 , E. Graugés45 , E. Graverini35,u,50 , L. Grazette57 ,
G. Graziani27 , A.T. Grecu43 , N.A. Grieser66 , L. Grillo60 , C. Gu15 , M. Guarise26 ,
L. Guerry11 , A.-K. Guseinov50 , Y. Guz6 , T. Gys49 , K. Habermann18 ,
T. Hadavizadeh1 , C. Hadjivasiliou67 , G. Haefeli50 , C. Haen49 , S. Haken56 ,
G. Hallett57 , P.M. Hamilton67 , Q. Han33 , X. Han22,49 , S. Hansmann-Menzemer22 ,
N. Harnew64 , T.J. Harris1 , M. Hartmann14 , S. Hashmi40 , J. He7,e , N. Heatley14 ,
A. Hedes63 , F. Hemmer49 , C. Henderson66 , R. Henderson14 , R.D.L. Henderson1 ,
A.M. Hennequin49 , K. Hennessy61 , J. Herd62 , P. Herrero Gascon22 , J. Heuel17 ,
A. Heyn13 , A. Hicheur3 , G. Hijano Mendizabal51 , J. Horswill63 , R. Hou8 ,
Y. Hou11 , D.C. Houston60 , N. Howarth61 , W. Hu7,e , X. Hu4 , W. Hulsbergen38 ,
R.J. Hunter57 , D. Hutchcroft61 , M. Idzik40 , P. Ilten66 , A. Iohner10 , H. Jage17 ,
S.J. Jaimes Elles77,48,49 , S. Jakobsen49 , T. Jakoubek78 , E. Jans38 , A. Jawahery67 ,
C. Jayaweera54 , A. Jelavic1 , V. Jevtic19 , Z. Jia16 , E. Jiang67 , X. Jiang5,7 ,
Y. Jiang7 , Y.J. Jiang6 , E. Jimenez Moya9 , N. Jindal91 , M. John64 ,
A. John Rubesh Rajan23 , D. Johnson54 , C.R. Jones56 , S. Joshi42 , B. Jost49 ,
J. Juan Castella56 , N. Jurik49 , I. Juszczak41 , K. Kalecinska40 , D. Kaminaris50 ,
S. Kandybei52 , M. Kane59 , Y. Kang4,d , C. Kar11 , M. Karacson49 ,
A. Kauniskangas50 , J.W. Kautz66 , M.K. Kazanecki41 , F. Keizer49 , M. Kenzie56 ,
T. Ketel38 , B. Khanji69 , S. Kholodenko62,49 , G. Khreich14 , F. Kiraz14 , T. Kirn17 ,
V.S. Kirsebom31,p , S. Klaver39 , N. Kleijne35,t , A. Kleimenova50 , D. Klekots88 ,

77
K. Klimaszewski42 , M.R. Kmiec42 , T. Knospe19 , R. Kolb22 , S. Koliiev53 ,
L. Kolk19 , A. Konoplyannikov6 , P. Kopciewicz49 , P. Koppenburg38 , A. Korchin52 ,
I. Kostiuk38 , O. Kot53 , S. Kotriakhova , E. Kowalczyk67 , O. Kravcov81 ,
M. Kreps57 , W. Krupa49 , W. Krzemien42 , O. Kshyvanskyi53 , S. Kubis86 ,
M. Kucharczyk41 , A. Kupsc87 , V. Kushnir52 , B. Kutsenko13 , J. Kvapil68 ,
I. Kyryllin52 , D. Lacarrere49 , P. Laguarta Gonzalez45 , A. Lai32 , A. Lampis32 ,
D. Lancierini62 , C. Landesa Gomez47 , J.J. Lane1 , G. Lanfranchi28 ,
C. Langenbruch22 , J. Langer19 , T. Latham57 , F. Lazzari35,u , C. Lazzeroni54 ,
R. Le Gac13 , H. Lee61 , R. Lefèvre11 , M. Lehuraux57 , E. Lemos Cid49 , O. Leroy13 ,
T. Lesiak41 , E.D. Lesser49 , B. Leverington22 , A. Li4,d , C. Li4 , C. Li13 , H. Li73 ,
J. Li8 , K. Li76 , L. Li63 , P. Li7 , P.-R. Li74 , Q. Li5,7 , T. Li72 , T. Li73 ,
Y. Li8 , Y. Li5 , Y. Li4 , Z. Lian4,d , Q. Liang8 , X. Liang69 , Z. Liang32 ,
S. Libralon48 , A. Lightbody12 , C. Lin7 , T. Lin58 , R. Lindner49 , H. Linton62 ,
R. Litvinov32 , D. Liu8 , F.L. Liu1 , G. Liu73 , K. Liu74 , S. Liu5 , W. Liu8 ,
Y. Liu59 , Y. Liu74 , Y.L. Liu62 , G. Loachamin Ordonez70 , I. Lobo1 ,
A. Lobo Salvia10 , A. Loi32 , T. Long56 , F.C.L. Lopes2,a , J.H. Lopes3 ,
A. Lopez Huertas45 , C. Lopez Iribarnegaray47 , Q. Lu15 , C. Lucarelli49 ,
D. Lucchesi33,r , M. Lucio Martinez48 , Y. Luo6 , A. Lupato33,j , M. Lupberger20 ,
E. Luppi26,m , K. Lynch23 , S. Lyu6 , X.-R. Lyu7 , G.M. Ma4,d , H. Ma72 ,
S. Maccolini49 , F. Machefert14 , F. Maciuc43 , B. Mack69 , I. Mackay64 ,
L.M. Mackey69 , L.R. Madhan Mohan56 , M.J. Madurai54 , D. Magdalinski38 ,
J.J. Malczewski41 , S. Malde64 , L. Malentacca49 , G. Manca32,l , G. Mancinelli13 ,
C. Mancuso14 , R. Manera Escalero45 , A. Mangalasseri80 , F.M. Manganella37 ,
D. Manuzzi25 , D. Marangotto30,o , J.F. Marchand10 , R. Marchevski50 , U. Marconi25 ,
E. Mariani16 , S. Mariani49 , C. Marin Benito45 , J. Marks22 , A.M. Marshall55 ,
L. Martel64 , G. Martelli34 , G. Martellotti36 , L. Martinazzoli49 , M. Martinelli31,p ,
D. Martinez Gomez83 , D. Martinez Santos44 , F. Martinez Vidal48 ,
A. Martorell i Granollers46 , A. Massafferri2 , R. Matev49 , A. Mathad49 ,
C. Matteuzzi69 , K.R. Mattioli15 , A. Mauri62 , E. Maurice15 , J. Mauricio45 ,
P. Mayencourt50 , J. Mazorra de Cos48 , M. Mazurek42 , D. Mazzanti Tarancon45 ,
M. McCann62 , N.T. McHugh60 , A. McNab63 , R. McNulty23 , B. Meadows66 ,
S.E.R. Medaer49 , D. Melnychuk42 , D. Mendoza Granada16 ,
P. Menendez Valdes Perez47 , F.M. Meng4,d , M. Merk38,84 , A. Merli50,30 ,
L. Meyer Garcia67 , D. Miao5,7 , H. Miao7 , M. Mikhasenko79 , D.A. Milanes85 ,
A. Minotti31,p , E. Minucci28 , B. Mitreska63 , D.S. Mitzel19 , R. Mocanu43 ,
A. Modak58 , L. Moeser19 , R.D. Moise17 , E.F. Molina Cardenas89 , T. Mombächer47 ,
M. Monk56 , T. Monnard50 , S. Monteil11 , A. Morcillo Gomez47 , G. Morello28 ,
M.J. Morello35,t , M.P. Morgenthaler22 , A. Moro31,p , J. Moron40 , W. Morren38 ,
A.B. Morris81,49 , A.G. Morris13 , R. Mountain69 , Z. Mu6 , E. Muhammad57 ,
F. Muheim59 , M. Mulder19 , K. Müller51 , F. Muñoz-Rojas9 , V. Mytrochenko52 ,
P. Naik61 , T. Nakada50 , R. Nandakumar58 , G. Napoletano50 , I. Nasteva3 ,
M. Needham59 , N. Neri30,o , S. Neubert18 , N. Neufeld49 , J. Nicolini49 ,
D. Nicotra84 , E.M. Niel15 , L. Nisi19 , Q. Niu74 , B.K. Njoki49 , P. Nogarolli3 ,
P. Nogga18 , C. Normand47 , J. Novoa Fernandez47 , G. Nowak66 , C. Nunez89 ,
H.N. Nur60 , A. Oblakowska-Mucha40 , T. Oeser17 , O. Okhrimenko53 ,
R. Oldeman32,l , F. Oliva59,49 , E. Olivart Pino45 , M. Olocco19 , R.H. O’Neil49 ,
J.S. Ordonez Soto11 , D. Osthues19 , J.M. Otalora Goicochea3 , P. Owen51 ,
A. Oyanguren48 , O. Ozcelik49 , F. Paciolla35,v , A. Padee42 , K.O. Padeken18 ,
B. Pagare47 , T. Pajero49 , A. Palano24 , L. Palini30 , M. Palutan28 , C. Pan75 ,
X. Pan4,d , S. Panebianco12 , S. Paniskaki49,33 , L. Paolucci63 , A. Papanestis58 ,

78
M. Pappagallo24,i , L.L. Pappalardo26 , C. Pappenheimer66 , C. Parkes63 ,
D. Parmar79 , G. Passaleva27 , D. Passaro35,t , A. Pastore24 , M. Patel62 ,
J. Patoc64 , C. Patrignani25,k , A. Paul69 , C.J. Pawley84 , A. Pellegrino38 ,
J. Peng5,7 , X. Peng74 , M. Pepe Altarelli28 , S. Perazzini25 , H. Pereira Da Costa68 ,
M. Pereira Martinez47 , A. Pereiro Castro47 , C. Perez46 , P. Perret11 ,
A. Perrevoort83 , A. Perro49 , M.J. Peters66 , K. Petridis55 , A. Petrolini29,n ,
S. Pezzulo29,n , J.P. Pfaller66 , H. Pham69 , L. Pica35,t , M. Piccini34 , L. Piccolo32 ,
B. Pietrzyk10 , R.N. Pilato61 , D. Pinci36 , F. Pisani49 , M. Pizzichemi31,p,49 ,
V.M. Placinta43 , M. Plo Casasus47 , T. Poeschl49 , F. Polci16 , M. Poli Lener28 ,
A. Poluektov13 , I. Polyakov63 , E. Polycarpo3 , S. Ponce49 , D. Popov7,49 ,
K. Popp19 , K. Prasanth59 , C. Prouve44 , D. Provenzano32,l,49 , V. Pugatch53 ,
A. Puicercus Gomez49 , G. Punzi35,u , J.R. Pybus68 , Q. Qian6 , W. Qian7 ,
N. Qin4,d , R. Quagliani49 , R.I. Rabadan Trejo57 , R. Racz81 , J.H. Rademacker55 ,
M. Rama35 , M. Ramı́rez Garcı́a89 , V. Ramos De Oliveira70 , M. Ramos Pernas49 ,
M.S. Rangel3 , G. Raven39 , M. Rebollo De Miguel48 , F. Redi30,j , J. Reich55 ,
F. Reiss20 , Z. Ren7 , P.K. Resmi64 , M. Ribalda Galvez45 , R. Ribatti50 ,
G. Ricart12 , D. Riccardi35,t , S. Ricciardi58 , K. Richardson65 ,
M. Richardson-Slipper56 , F. Riehn19 , K. Rinnert61 , P. Robbe14,49 , G. Robertson60 ,
E. Rodrigues61 , A. Rodriguez Alvarez45 , E. Rodriguez Fernandez47 ,
J.A. Rodriguez Lopez77 , E. Rodriguez Rodriguez49 , J. Roensch19 , A. Rogovskiy58 ,
D.L. Rolf19 , P. Roloff49 , V. Romanovskiy66 , A. Romero Vidal47 , G. Romolini26,49 ,
F. Ronchetti50 , T. Rong6 , M. Rotondo28 , M.S. Rudolph69 , M. Ruiz Diaz22 ,
R.A. Ruiz Fernandez47 , J. Ruiz Vidal84 , J.J. Saavedra-Arias9 , J.J. Saborido Silva47 ,
D. Sahoo80 , N. Sahoo54 , B. Saitta32 , M. Salomoni31,49,p , I. Sanderswood48 ,
R. Santacesaria36 , C. Santamarina Rios47 , M. Santimaria28 , L. Santoro 2 ,
E. Santovetti37 , A. Saputi26,49 , A. Sarnatskiy83 , G. Sarpis49 , M. Sarpis81 ,
C. Satriano36 , A. Satta37 , M. Saur74 , H. Sazak17 , F. Sborzacchi49,28 ,
A. Scarabotto19 , S. Schael17 , S. Scherl61 , M. Schiller22 , H. Schindler49 ,
M. Schmelling21 , B. Schmidt49 , N. Schmidt68 , S. Schmitt65 , H. Schmitz18 ,
O. Schneider50 , A. Schopper62 , N. Schulte19 , M.H. Schune14 , G. Schwering17 ,
B. Sciascia28 , A. Sciuccati49 , G. Scriven84 , I. Segal79 , S. Sellam47 , T. Senger51 ,
M. Senghi Soares39 , A. Sergi29,n , N. Serra51 , L. Sestini27 , B. Sevilla Sanjuan46 ,
Y. Shang6 , D.M. Shangase89 , R.S. Sharma69 , L. Shchutska50 , T. Shears61 , J. Shen6 ,
Z. Shen38 , S. Sheng50 , B. Shi7 , J. Shi56 , Q. Shi7 , W.S. Shi73 , E. Shmanin25 ,
R. Silva Coutinho2 , G. Simi33,r , S. Simone24,i , M. Singha80 , I. Siral50 ,
N. Skidmore57 , T. Skwarnicki69 , M.W. Slater54 , E. Smith65 , M. Smith62 ,
L. Soares Lavra59 , M.D. Sokoloff66 , F.J.P. Soler60 , A. Solomin55 , K. Solovieva20 ,
N.S. Sommerfeld18 , R. Song1 , Y. Song50 , Y. Song4,d , Y.S. Song6 ,
F.L. Souza De Almeida45 , B. Souza De Paula3 , K.M. Sowa40 , E. Spadaro Norella29,n ,
E. Spedicato25 , J.G. Speer19 , P. Spradlin60 , F. Stagni49 , M. Stahl79 , S. Stahl49 ,
S. Stanislaus64 , M. Stefaniak91 , O. Steinkamp51 , Y. Su7 , F. Suljik64 , J. Sun32 ,
J. Sun63 , L. Sun75 , D. Sundfeld2 , W. Sutcliffe51 , P. Svihra78 , V. Svintozelskyi48 ,
K. Swientek40 , F. Swystun56 , A. Szabelski42 , T. Szumlak40 , Y. Tan4 , Y. Tang75 ,
Y.T. Tang7 , M.D. Tat22 , J.A. Teijeiro Jimenez47 , F. Terzuoli35,v , F. Teubert49 ,
E. Thomas49 , D.J.D. Thompson54 , A.R. Thomson-Strong59 , H. Tilquin62 ,
V. Tisserand11 , S. T’Jampens10 , M. Tobin5,49 , T.T. Todorov20 , L. Tomassetti26,m ,
G. Tonani30 , X. Tong6 , T. Tork30 , L. Toscano19 , D.Y. Tou4,d , C. Trippl46 ,
G. Tuci22 , N. Tuning38 , L.H. Uecker22 , A. Ukleja40 , D.J. Unverzagt22 ,
A. Upadhyay49 , B. Urbach59 , A. Usachov38 , U. Uwer22 , V. Vagnoni25,49 ,
A. Vaitkevicius81 , V. Valcarce Cadenas47 , G. Valenti25 , N. Valls Canudas49 ,

79
J. van Eldik49 , H. Van Hecke68 , E. van Herwijnen62 , C.B. Van Hulse47,x ,
R. Van Laak50 , M. van Veghel84 , G. Vasquez51 , R. Vazquez Gomez45 ,
P. Vazquez Regueiro47 , C. Vázquez Sierra44 , S. Vecchi26 , J. Velilla Serna48 ,
J.J. Velthuis55 , M. Veltri27,w , A. Venkateswaran50 , M. Verdoglia32 , M. Vesterinen57 ,
W. Vetens69 , D. Vico Benet64 , P. Vidrier Villalba45 , M. Vieites Diaz47 ,
X. Vilasis-Cardona46 , E. Vilella Figueras61 , A. Villa50 , P. Vincent16 , B. Vivacqua3 ,
F.C. Volle54 , D. vom Bruch13 , K. Vos84 , C. Vrahas59 , J. Wagner19 , J. Walsh35 ,
N. Walter49 , E.J. Walton1 , G. Wan6 , A. Wang7 , B. Wang5 , C. Wang22 ,
G. Wang8 , H. Wang74 , J. Wang7 , J. Wang5 , J. Wang4,d , J. Wang75 ,
M. Wang49 , N.W. Wang7 , R. Wang55 , X. Wang8 , X. Wang73 , X.W. Wang62 ,
Y. Wang76 , Y. Wang6 , Y.H. Wang74 , Z. Wang14 , Z. Wang30 , J.A. Ward57,1 ,
M. Waterlaat49 , N.K. Watson54 , D. Websdale62 , Y. Wei6 , Z. Weida7 ,
J. Wendel44 , B.D.C. Westhenry55 , C. White56 , M. Whitehead60 , E. Whiter54 ,
A.R. Wiederhold63 , D. Wiedner19 , M.A. Wiegertjes38 , C. Wild64 , G. Wilkinson64,49 ,
M.K. Wilkinson66 , M. Williams65 , M.J. Williams49 , M.R.J. Williams59 ,
R. Williams56 , S. Williams55 , Z. Williams55 , F.F. Wilson58 , M. Winn12 ,
W. Wislicki42 , M. Witek41 , L. Witola19 , T. Wolf22 , E. Wood56 , G. Wormser14 ,
S.A. Wotton56 , H. Wu69 , J. Wu8 , X. Wu75 , Y. Wu6,56 , Z. Wu7 , K. Wyllie49 ,
S. Xian73 , Z. Xiang5 , Y. Xie8 , T.X. Xing30 , A. Xu35,t , L. Xu4,d , M. Xu49 ,
R. Xu89 , Z. Xu49 , Z. Xu7 , Z. Xu5 , S. Yadav26 , K. Yang62 , X. Yang6 ,
Y. Yang80 , Y. Yang7 , Z. Yang6 , Z. Yang4 , H. Yeung63 , H. Yin8 , X. Yin7 ,
C.Y. Yu6 , J. Yu72 , X. Yuan5 , Y Yuan5,7 , J.A. Zamora Saa71 , M. Zavertyaev21 ,
M. Zdybal41 , F. Zenesini25 , C. Zeng5,7 , M. Zeng4,d , S.H Zeng55 , C. Zhang6 ,
D. Zhang8 , J. Zhang7 , L. Zhang4,d , R. Zhang8 , S. Zhang64 , S.L. Zhang72 ,
Y. Zhang6 , Y.Z. Zhang4,d , Z. Zhang4,d , Y. Zhao22 , A. Zhelezov22 , S.Z. Zheng6 ,
X.Z. Zheng4,d , Y. Zheng7 , T. Zhou6 , X. Zhou8 , V. Zhovkovska57 , L.Z. Zhu59 ,
X. Zhu4,d , X. Zhu8 , Y. Zhu17 , V. Zhukov17 , J. Zhuo48 , D. Zuliani33,r ,
G. Zunica28 .
1
School of Physics and Astronomy, Monash University, Melbourne, Australia
2
Centro Brasileiro de Pesquisas Fı́sicas (CBPF), Rio de Janeiro, Brazil
3
Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil
4
Department of Engineering Physics, Tsinghua University, Beijing, China
5
Institute Of High Energy Physics (IHEP), Beijing, China
6
School of Physics State Key Laboratory of Nuclear Physics and Technology, Peking University, Beijing,
China
7
University of Chinese Academy of Sciences, Beijing, China
8
Institute of Particle Physics, Central China Normal University, Wuhan, Hubei, China
9
Consejo Nacional de Rectores (CONARE), San Jose, Costa Rica
10
Université Savoie Mont Blanc, CNRS, IN2P3-LAPP, Annecy, France
11
Université Clermont Auvergne, CNRS/IN2P3, LPC, Clermont-Ferrand, France
12
Université Paris-Saclay, Centre d’Etudes de Saclay (CEA), IRFU, Gif-Sur-Yvette, France
13
Aix Marseille Univ, CNRS/IN2P3, CPPM, Marseille, France
14
Université Paris-Saclay, CNRS/IN2P3, IJCLab, Orsay, France
15
Laboratoire Leprince-Ringuet, CNRS/IN2P3, Ecole Polytechnique, Institut Polytechnique de Paris,
Palaiseau, France
16
Laboratoire de Physique Nucléaire et de Hautes Énergies (LPNHE), Sorbonne Université,
CNRS/IN2P3, Paris, France
17
I. Physikalisches Institut, RWTH Aachen University, Aachen, Germany
18
Universität Bonn - Helmholtz-Institut für Strahlen und Kernphysik, Bonn, Germany
19
Fakultät Physik, Technische Universität Dortmund, Dortmund, Germany
20
Physikalisches Institut, Albert-Ludwigs-Universität Freiburg, Freiburg, Germany
21
Max-Planck-Institut für Kernphysik (MPIK), Heidelberg, Germany

80
22
Physikalisches Institut, Ruprecht-Karls-Universität Heidelberg, Heidelberg, Germany
23
School of Physics, University College Dublin, Dublin, Ireland
24
INFN Sezione di Bari, Bari, Italy
25
INFN Sezione di Bologna, Bologna, Italy
26
INFN Sezione di Ferrara, Ferrara, Italy
27
INFN Sezione di Firenze, Firenze, Italy
28
INFN Laboratori Nazionali di Frascati, Frascati, Italy
29
INFN Sezione di Genova, Genova, Italy
30
INFN Sezione di Milano, Milano, Italy
31
INFN Sezione di Milano-Bicocca, Milano, Italy
32
INFN Sezione di Cagliari, Monserrato, Italy
33
INFN Sezione di Padova, Padova, Italy
34
INFN Sezione di Perugia, Perugia, Italy
35
INFN Sezione di Pisa, Pisa, Italy
36
INFN Sezione di Roma La Sapienza, Roma, Italy
37
INFN Sezione di Roma Tor Vergata, Roma, Italy
38
Nikhef National Institute for Subatomic Physics, Amsterdam, Netherlands
39
Nikhef National Institute for Subatomic Physics and VU University Amsterdam, Amsterdam,
Netherlands
40
AGH - University of Krakow, Faculty of Physics and Applied Computer Science, Kraków, Poland
41
Henryk Niewodniczanski Institute of Nuclear Physics Polish Academy of Sciences, Kraków, Poland
42
National Center for Nuclear Research (NCBJ), Warsaw, Poland
43
Horia Hulubei National Institute of Physics and Nuclear Engineering, Bucharest-Magurele, Romania
44
Universidade da Coruña, A Coruña, Spain
45
ICCUB, Universitat de Barcelona, Barcelona, Spain
46
La Salle, Universitat Ramon Llull, Barcelona, Spain
47
Instituto Galego de Fı́sica de Altas Enerxı́as (IGFAE), Universidade de Santiago de Compostela,
Santiago de Compostela, Spain
48
Instituto de Fisica Corpuscular, Centro Mixto Universidad de Valencia - CSIC, Valencia, Spain
49
European Organization for Nuclear Research (CERN), Geneva, Switzerland
50
Institute of Physics, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland
51
Physik-Institut, Universität Zürich, Zürich, Switzerland
52
NSC Kharkiv Institute of Physics and Technology (NSC KIPT), Kharkiv, Ukraine
53
Institute for Nuclear Research of the National Academy of Sciences (KINR), Kyiv, Ukraine
54
School of Physics and Astronomy, University of Birmingham, Birmingham, United Kingdom
55
H.H. Wills Physics Laboratory, University of Bristol, Bristol, United Kingdom
56
Cavendish Laboratory, University of Cambridge, Cambridge, United Kingdom
57
Department of Physics, University of Warwick, Coventry, United Kingdom
58
STFC Rutherford Appleton Laboratory, Didcot, United Kingdom
59
School of Physics and Astronomy, University of Edinburgh, Edinburgh, United Kingdom
60
School of Physics and Astronomy, University of Glasgow, Glasgow, United Kingdom
61
Oliver Lodge Laboratory, University of Liverpool, Liverpool, United Kingdom
62
Imperial College London, London, United Kingdom
63
Department of Physics and Astronomy, University of Manchester, Manchester, United Kingdom
64
Department of Physics, University of Oxford, Oxford, United Kingdom
65
Massachusetts Institute of Technology, Cambridge, MA, United States
66
University of Cincinnati, Cincinnati, OH, United States
67
University of Maryland, College Park, MD, United States
68
Los Alamos National Laboratory (LANL), Los Alamos, NM, United States
69
Syracuse University, Syracuse, NY, United States
70
Pontifı́cia Universidade Católica do Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil, associated to 3
71
Universidad Andres Bello, Santiago, Chile, associated to 51
72
School of Physics and Electronics, Hunan University, Changsha City, China, associated to 8
73
State Key Laboratory of Nuclear Physics and Technology, South China Normal University, Guangzhou,
China, associated to 4
74
Lanzhou University, Lanzhou, China, associated to 5

81
75
School of Physics and Technology, Wuhan University, Wuhan, China, associated to 4
76
Henan Normal University, Xinxiang, China, associated to 8
77
Departamento de Fisica , Universidad Nacional de Colombia, Bogota, Colombia, associated to 16
78
Institute of Physics of the Czech Academy of Sciences, Prague, Czech Republic, associated to 63
79
Ruhr Universitaet Bochum, Fakultaet f. Physik und Astronomie, Bochum, Germany, associated to 19
80
Eotvos Lorand University, Budapest, Hungary, associated to 49
81
Faculty of Physics, Vilnius University, Vilnius, Lithuania, associated to 20
82
Institute of Physics and Technology, Mongolian Academy of Sciences, Ulan Bator, Mongolia, associated
to 5
83
Van Swinderen Institute, University of Groningen, Groningen, Netherlands, associated to 38
84
Universiteit Maastricht, Maastricht, Netherlands, associated to 38
85
Universidad de Ingenierı́a y Tecnologı́a (UTEC), Lima, Peru, associated to 65
86
Tadeusz Kosciuszko Cracow University of Technology, Cracow, Poland, associated to 41
87
Department of Physics and Astronomy, Uppsala University, Uppsala, Sweden, associated to 60
88
Taras Schevchenko University of Kyiv, Faculty of Physics, Kyiv, Ukraine, associated to 14
89
University of Michigan, Ann Arbor, MI, United States, associated to 69
90
Indiana University, Bloomington, United States, associated to 68
91
Ohio State University, Columbus, United States, associated to 68

a
Universidade Estadual de Campinas (UNICAMP), Campinas, Brazil
b
Centro Federal de Educacão Tecnológica Celso Suckow da Fonseca, Rio De Janeiro, Brazil
c
Department of Physics and Astronomy, University of Victoria, Victoria, Canada
d
Center for High Energy Physics, Tsinghua University, Beijing, China
e
Hangzhou Institute for Advanced Study, UCAS, Hangzhou, China
f
LIP6, Sorbonne Université, Paris, France
g
Lamarr Institute for Machine Learning and Artificial Intelligence, Dortmund, Germany
h
Universidad Nacional Autónoma de Honduras, Tegucigalpa, Honduras
i
Università di Bari, Bari, Italy
j
Università di Bergamo, Bergamo, Italy
k
Università di Bologna, Bologna, Italy
l
Università di Cagliari, Cagliari, Italy
m
Università di Ferrara, Ferrara, Italy
n
Università di Genova, Genova, Italy
o
Università degli Studi di Milano, Milano, Italy
p
Università degli Studi di Milano-Bicocca, Milano, Italy
q
Università di Modena e Reggio Emilia, Modena, Italy
r
Università di Padova, Padova, Italy
s
Università di Perugia, Perugia, Italy
t
Scuola Normale Superiore, Pisa, Italy
u
Università di Pisa, Pisa, Italy
v
Università di Siena, Siena, Italy
w
Università di Urbino, Urbino, Italy
x
Universidad de Alcalá, Alcalá de Henares, Spain
†
Deceased

82