INTERNATIONAL PHD PROJECTS IN APPLIED NUCLEAR PHYSICS AND INNOVATIVE TECHNOLOGIESThis project is supported by the Foundation for Polish Science – MPD program, co-financed by the European Union within the European Regional Development Fund
φ meson production in proton-proton collisionsin the NA61/SHINE experiment
at CERN SPS
Antoni Marcinek
H. Niewodniczanski Institute of Nuclear PhysicsPolish Academy of Sciences
Białasówka, 28 April 2017
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 1 / 21
Outline
1 Introduction
2 Analysis methodology
3 Results
4 Summary
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 2 / 21
Introduction
φ = ss meson according to PDG 2014Mass m = (1019.461± 0.019) MeVWidth Γ = (4.266± 0.031) MeVBR(φ→ K+K−) = (48.9± 0.5) %
Goal of the analysisDifferential φ multiplicities in p+p collisions measured in NA61/SHINE→ from invariant mass spectra fits in φ→ K+K− decay channel→ as function of rapidity y and transverse momentum pT
MotivationTo constrain hadron production models→ φ interesting due to its hidden strangeness (ss)
Reference data for Pb+Pb at the same energies
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 3 / 21
NA61/SHINE experiment
General infoFixed target experiment in the North (experimental) Area of CERN SPS
Successor of NA49Beams
hadrons (secondary)ions (secondary and primary)
∼150 physicists→ IFJ PAN group (6 people) since June 2016
Physics active since 2009
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 4 / 21
Physics programme
SHINE = SPS Heavy Ion and Neutrino Experiment
13 20 30 40 80 158
beam momentum (A GeV/c)
p+p
p+Pb
Xe+La 2017
Ar+Sc 2015
2011/12/13
2009/10/11
2012/14/15
Be+Be
Pb+Pb 2016
detailed scan with existing detector
31 120 158 350 400
beam momentum (GeV/c)
p+C
p+C(LT)
p-+C
K-+C 2012
2009/12
2007-10
2007/09/12
2018/19
high stat. with new vertex detector
20 40 158
p+AA=C, Be, Al, etc.
9 - 120 GeV/c
recorded data
pilot (test) data
planned data (approved)
beyond the approved program
Pb+Pb
colli
ding
sys
tem
colli
ding
sys
tem
Heavy ion physicsspectra, correlations, fluctuations
critical point
onset of deconfinement
? EM interactions with spectators
Cosmic rays and neutrinosprecision measurements ofspectra
cosmic rays: Pierre AugerObservatory, KASCADE
neutrinos: T2K, Minerνa,MINOS, NOνA, LBNE
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 5 / 21
Physics programme
SHINE = SPS Heavy Ion and Neutrino Experiment
13 20 30 40 80 158
beam momentum (A GeV/c)
p+p
p+Pb
Xe+La 2017
Ar+Sc 2015
2011/12/13
2009/10/11
2012/14/15
Be+Be
Pb+Pb 2016
detailed scan with existing detector
31 120 158 350 400
beam momentum (GeV/c)
p+C
p+C(LT)
p-+C
K-+C 2012
2009/12
2007-10
2007/09/12
2018/19
high stat. with new vertex detector
20 40 158
p+AA=C, Be, Al, etc.
9 - 120 GeV/c
recorded data
pilot (test) data
planned data (approved)
beyond the approved program
Pb+Pb
colli
ding
sys
tem
colli
ding
sys
tem
Heavy ion physicsspectra, correlations, fluctuations
critical point
onset of deconfinement
? EM interactions with spectators
Cosmic rays and neutrinosprecision measurements ofspectra
cosmic rays: Pierre AugerObservatory, KASCADE
neutrinos: T2K, Minerνa,MINOS, NOνA, LBNE
If not stated otherwise, pictures forp+p @ 158 GeV→√sNN = 17.3 GeV.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 5 / 21
NA61/SHINE detector
liquid H2 target
TPC→ particle tracks in 3Dcurvature→ charge andmomentum
energy loss (dE/dx)→ mass
directly — only charged particles!
Performancetotal acceptance ∼ 80 %momentum resolutionσ(p)/p2 ∼ 10−4 GeV−1
track reconstruction efficiency > 95 %
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 6 / 21
Data selection
Eventsinelastic
in the target
with well measured main vertex
TPC tracksfrom main vertex
well reconstructed
number of points in TPCs→accurate dE/dx and momentum
with dE/dx corresponding tokaons (PID cut)
) [MeV]-,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
1
10
210
310
410
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
no PID cut
with PID cut
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 7 / 21
Kaon candidate selection — PID cut
1
10
210
310
(p/GeV)10
log-1 0 1 2
dE/d
x
1
1.5
2+e +π+
K p
1
10
210
310
(p/GeV)10
log-1 0 1 2
dE/d
x
1
1.5
2
Selection done with dE/dx
Accept tracks in ±5 % band around kaon Bethe-Bloch curve(area between black curves in right picture)
Losses due to efficiency of this selection corrected with tag-and-probe method
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 8 / 21
Signal extractionphase space binning, invariant mass spectrum
0
0.2
0.4
0.6
0.8
1
y-1 0 1 2 3
[GeV
]Tp
0
0.5
1
1.5
2
SignalConvolution of:
relativistic Breit-WignerfrelBW(minv;mφ,Γ) resonance shape
q-Gaussian fqG(minv;σ, q) broadeningdue to detector resolution
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ 0.098 MeV± = 0.991 σ
79± = 9187 bkgN
53± = 2494 pN
0.071 MeV± = 1019.623 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 1.6 /2χ
BackgroundObtained with the event mixing method:
Kaon candidate taken from the current event is combined with candidates fromprevious 500 events to create φ candidates in the mixed events spectrum
Fitting function
f(minv) = Np · (frelBW ∗ fqG)(minv;mφ,Γ, σ, q) +Nbkg ·B(minv)
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 9 / 21
Signal extractionphase space binning, invariant mass spectrum
SignalConvolution of:
relativistic Breit-WignerfrelBW(minv;mφ,Γ) resonance shape
q-Gaussian fqG(minv;σ, q) broadeningdue to detector resolution
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ 0.098 MeV± = 0.991 σ
79± = 9187 bkgN
53± = 2494 pN
0.071 MeV± = 1019.623 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 1.6 /2χ
BackgroundObtained with the event mixing method:
Kaon candidate taken from the current event is combined with candidates fromprevious 500 events to create φ candidates in the mixed events spectrum
Fitting function
f(minv) = Np · (frelBW ∗ fqG)(minv;mφ,Γ, σ, q) +Nbkg ·B(minv)
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 9 / 21
Signal extractiontag-and-probe method → ATLAS, LHCb
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
500
1000
1500
Entries = 128572[0.0,1.6) GeV∈T
[0.0,2.1), p∈Tag: y
ndf = 1.9 /2χ
at least one K pass PIDTag:
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ
0.12 MeV± = 0.70 σ
0.031± = 0.899 ε 183± = 2952 φN
154± = 9746 bkg,pN
504± = 123748 bkg,tN
0.083 MeV± = 1019.566 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 2.0 /2χ
both K pass PIDProbe:
Goal: to remove bias of Nφ due to PID cut efficiency εSimultaneous fit of 2 spectra:
tag — at least one track in the pair passes PID cut
Nt = Nφε(2− ε)probe — both tracks pass PID cut
Np = Nφε2
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 10 / 21
Normalization and corrections
0
0.2
0.4
0.6
0.8
1
y-1 0 1 2 3
[GeV
]Tp
0
0.5
1
1.5
2
d2n
dpT dy = NφNev ∆pT ∆y ×
c∞ · cbkg · cMC
BR(φ→ K+K−)c∞ ∼ 1.06 — extrapolation of the resonance curve
cbkg = 1.05 — unaccounted-for effects in thebackground description by event mixing
0 0.5 1 1.5
corr
ectio
n
1
1.5
2
2.5
3
[0.0,0.3)∈y
0 0.5 1 1.5
1
1.5
2
2.5
3
[0.6,0.9)∈y0 0.5 1 1.5
1
1.5
2
2.5
3
[0.3,0.6)∈y
0 0.5 1 1.5
1
1.5
2
2.5
3
[0.9,1.5)∈y
total
geom
non-geom
[GeV]T
p0 0.5 1 1.5
1
1.5
2
2.5
3
[1.5,2.1)∈y
Monte Carlo correction
cMC =N genφ
N genev
/N selφ
N selev
registration efficiency
trigger bias
losses due to vertex cuts
reconstruction efficiency
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 11 / 21
Uncertainties
StatisticalMINUIT/HESSE (symmetric)
Systematic bin-independent
Source value [%]
BR(φ→ K+K−) 1fitting constraints 2resonance theory 3background 5
Total (quadratic) 6
0 0.5 1 1.5
unce
rtai
nty
[%]
0
20
40
60
[0.0,0.3)∈y
0 0.5 1 1.50
20
40
60
[0.6,0.9)∈y0 0.5 1 1.50
20
40
60
[0.3,0.6)∈y
0 0.5 1 1.50
20
40
60
[0.9,1.5)∈y
statistical
total systematic
tag-and-probe
event cuts
track cuts
bin-independent
[GeV]T
p0 0.5 1 1.50
20
40
60
[1.5,2.1)∈y
p+p @ 158 GeV
Total systematic uncertainty =√∑
σ2i
For p+p @ 40 GeV additional bin-independent 3 % due to cMC averaging
Statistical uncertainty dominates
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 12 / 21
Double differential spectra: p+p @ 158 GeV
0 0.5 1 1.5
]-1
dy)
[GeV
T(d
p /
n2 d
-510
-410
-310
-210[0.0,0.3)∈y
0 0.5 1 1.5-510
-410
-310
-210[0.6,0.9)∈y
0 0.5 1 1.5-510
-410
-310
-210[0.3,0.6)∈y
0 0.5 1 1.5-510
-410
-310
-210[0.9,1.5)∈y
EPOS 1.99
Pythia 6
UrQMD 3.4
[GeV]T
p0 0.5 1 1.5
-510
-410
-310
-210[1.5,2.1)∈y
√sNN = 17.3 GeV
MC normalization:∫model =
∫data
Pythia describes spectra shapes best, UrQMD slightly too long tail,EPOS clearly too short tail
Fit pT e−mT /T → extrapolation to pT =∞→ tail < 1 %Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 13 / 21
Double differential spectra: p+p @ 158 GeV
0 0.5 1 1.5
]-1
dy)
[GeV
T(d
p /
n2 d
-510
-410
-310
-210[0.0,0.3)∈y
0 0.5 1 1.5-510
-410
-310
-210[0.6,0.9)∈y
0 0.5 1 1.5-510
-410
-310
-210[0.3,0.6)∈y
0 0.5 1 1.5-510
-410
-310
-210[0.9,1.5)∈y
EPOS 1.99
Pythia 6
UrQMD 3.4
[GeV]T
p0 0.5 1 1.5
-510
-410
-310
-210[1.5,2.1)∈y
√sNN = 17.3 GeV
MC normalization:∫model =
∫data
Pythia describes spectra shapes best, UrQMD slightly too long tail,EPOS clearly too short tail
Fit pT e−mT /T → extrapolation to pT =∞→ tail < 1 %
First 2D (y vs pT ) φ production measurements for p+p @ 158 GeV
+1 bin in y, +1 bin in pT compared to 2×1D NA49
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 13 / 21
Double differential spectra: p+p @ 80 GeV
0 0.5 1
]-1
dy)
[GeV
T(d
p /
n2 d -310
[0.0,0.3)∈y
0 0.5 1
-310
[0.6,0.9)∈y0 0.5 1
-310
[0.3,0.6)∈y
[GeV]T
p0 0.5 1
-310
[0.9,1.5)∈y
EPOS 1.99
Pythia 6
UrQMD 3.4
√sNN = 12.3 GeV
MC normalization:∫model =
∫data
Pythia describes spectra shapes best, UrQMD slightly too long tail,EPOS clearly too short tail
Fit pT e−mT /T → extrapolation to pT =∞→ tail < 4 %Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 14 / 21
Double differential spectra: p+p @ 80 GeV
0 0.5 1
]-1
dy)
[GeV
T(d
p /
n2 d -310
[0.0,0.3)∈y
0 0.5 1
-310
[0.6,0.9)∈y0 0.5 1
-310
[0.3,0.6)∈y
[GeV]T
p0 0.5 1
-310
[0.9,1.5)∈y
EPOS 1.99
Pythia 6
UrQMD 3.4
√sNN = 12.3 GeV
MC normalization:∫model =
∫data
Pythia describes spectra shapes best, UrQMD slightly too long tail,EPOS clearly too short tail
Fit pT e−mT /T → extrapolation to pT =∞→ tail < 4 %
First φ production measurements for p+p @ 80 GeV
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 14 / 21
Rapidity
y0 0.5 1 1.5 2
dy /dn
0
2
4
6-3
10×
NA49
NA61/SHINE
EPOS 1.99
Pythia 6
UrQMD 3.4
y0 0.5 1 1.5
dy /dn
0
2
4
6-3
10×
EPOS 1.99
Pythia 6
UrQMD 3.4
p+p @ 158 GeV p+p @ 80 GeV
√sNN = 17.3 GeV
√sNN = 12.3 GeV
MC normalization:∫model =
∫data
EPOS and UrQMD shape comparable to data, Pythia slightly narrower
Fit Gaussian e−y2/2σ2y → extrapolation to y =∞→
tails: 3 % for 158 GeV, 7 % for 80 GeV
NA61/SHINE consistent with NA49
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 15 / 21
Transverse mass spectra at midrapidity
[GeV]0 - mTm0 0.2 0.4 0.6
]-2
dy)
[GeV
Tdm
T(m /
n2 d -310
-210
[0.0,0.3)∈y
[GeV]0 - mTm0 0.2 0.4
]-2
dy)
[GeV
Tdm
T(m /
n2 d-3
10
-210
[0.0,0.3)∈y
p+p @ 158 GeV p+p @ 80 GeV
√sNN = 17.3 GeV
√sNN = 12.3 GeV
Thermal fit resultspbeam [GeV] Tφ [MeV] Tπ− [MeV]
158 150± 14± 8 159.3± 1.3± 2.680 148± 30± 17 159.9± 1.5± 4.1
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 16 / 21
Single differential spectra: p+p @ 40 GeV
[GeV]T
p0 0.5 1
]-1
dy)
[GeV
T(d
p /
n2 d
-310
[0.0,1.5)∈y
y0 0.5 1 1.5
dy /dn
0
1
2
3
4-3
10×
EPOS 1.99
Pythia 6
UrQMD 3.4
√sNN = 8.8 GeV
MC normalization:∫model =
∫data
yUrQMD agrees with data, EPOS bit too narrow, Pythia even narrower
extrapolation tail 5 %
pT
Pythia agrees best, UrQMD similar, EPOS spectrum too short tail
extrapolation tail < 1 %
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 17 / 21
Single differential spectra: p+p @ 40 GeV
[GeV]T
p0 0.5 1
]-1
dy)
[GeV
T(d
p /
n2 d
-310
[0.0,1.5)∈y
y0 0.5 1 1.5
dy /dn
0
1
2
3
4-3
10×
EPOS 1.99
Pythia 6
UrQMD 3.4
√sNN = 8.8 GeV
MC normalization:∫model =
∫data
yUrQMD agrees with data, EPOS bit too narrow, Pythia even narrower
extrapolation tail 5 %
pT
Pythia agrees best, UrQMD similar, EPOS spectrum too short tail
extrapolation tail < 1 %
First φ production measurements for p+p @ 40 GeV
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 17 / 21
Reference data for Pb+Pb: σy = width of dn/dy
beamy
2 2.5 3
yσ
0.6
0.8
1
1.2
1.4-π+K-
K
Λφ
beamy
2 2.5 3
yσ
0.6
0.8
1
1.2
1.4
coalescence-
K+
Kp+pPb+Pb
EPOS 1.99Pythia 6UrQMD 3.4
Comparison of particles / reactionsAll but φ in Pb+Pb:σy proportional to ybeam with the same rate of increase
two new φ points in p+p emphasize peculiarity of φ in Pb+Pb
CoalescenceFor p+p only 40 GeV compatible with production through K+ K− coalescence
collision energy
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 18 / 21
Reference data for Pb+Pb: total yield
[GeV]NNs5 10 15 20
⟩π⟨ / ⟩φ⟨
1000
1
2
3
4
5
p+p
Pb+Pb
[GeV]NNs5 10 15 20
doub
le r
atio
(P
b+P
b) /
(p+
p)
1
2
3
4⟩π⟨ / ⟩φ⟨
⟩+π⟨ / ⟩+K⟨
⟩-π⟨ / ⟩-K⟨
φ/π ratio increases with collision energy
Production enhancement in Pb+Pb about 3×, independent of energyEnhancement systematically larger than for kaons, comparable to K+
→ for K− consistent with strangeness enhancement in parton phase(square of K− enhancement)
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 19 / 21
Comparison with world data and models
[GeV]NNs10 20 30 40 50
⟩φ⟨
0
0.01
0.02
0.03
0.04
world data
NA61/SHINE
EPOS 1.99Pythia 6UrQMD 3.4HRG
[GeV]NNs10
2103
10 410
dy (
y =
0)
/dn
0
0.01
0.02
0.03 world data
NA61/SHINE
EPOS 1.99Pythia 6UrQMD 3.4
p+p world dataResults consistent with world data, much more accurate
ModelsEPOS close to data, Pythia underestimates experimental data,UrQMD underestimates ∼ 2×, HRG (thermal) overestimates ∼ 2×EPOS rises too fast with
√sNN
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 20 / 21
Summary
ResultsDifferential multiplicities of φ mesons in p+p:158 GeV first 2D (y and pT ), more accurate than 2×1D (y or pT ) NA4980 GeV 2D, first at this energy40 GeV 2×1D, first at this energy
Comparison with experimental dataResults consistent with p+p world data, but much more accurate!
Emphasize peculiarity of longitudinal expansion (σy) in Pb+Pb
Confirm enhancement in Pb+Pb, independent of energy in considered range,similar to kaons
Comparison with modelsEach describes well either pT or y shape, but not both
None is able to describe total yields
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 21 / 21
BACKUP
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 1 / 38
Vertex z cut choice
z [cm]-620 -600 -580 -560 -540
entr
ies
0
1000
2000
z [cm]-620 -600 -580 -560 -540
entr
ies
0
5
10
15
20
310×
loose vertex z cutAccepts windows of LHT.
Small cMC → no in-target events removed due to vertex z resolution.
Requires EMPTY target subtraction to remove background from windows.
tight vertex z cutRemoves interactions in windows of LHT.
Large cMC → in-target events removed due to vertex z resolution.
Negligible EMPTY target contribution (no windows)→ no EMPTY subtraction.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 2 / 38
Empty target statistics — 158 GeV
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
50
100
Entries = 5334[0.0,1.6) GeV∈T
[0.0,2.1), p∈Tag: y
ndf = 0.8 /2χ
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
5
10
15
20
Entries = 478
= 4.27 MeVΓ
0.39 MeV± = 0.85 σ
0.14± = 0.95 ε 36± = 119 φN
30± = 380 bkg,pN
103± = 5110 bkg,tN
0.44 MeV± = 1019.43 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 0.7 /2χ
EMPTY target subtraction requires division of these stats in the same bins asfor FULL target analysis→ clearly not feasible.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 3 / 38
Example 1D y binning fit to constrain ε
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 30559[0.0,1.6) GeV∈T
[0.9,1.5), p∈Tag: y
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
50
100
Entries = 3362
= 4.27 MeVΓ
= 0.99 MeVσ
0.056± = 0.973 ε 82± = 768 φN
80± = 2804 bkg,pN
247± = 29559 bkg,tN
= 1019.62 MeVφm
q = 1.50
[0.0,1.6) GeV∈T
[0.9,1.5), p∈Probe: y
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 4 / 38
Example 2D binning fits
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
50
100
Entries = 7880[0.2,0.4) GeV∈T
[0.9,1.5), p∈Tag: y
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
10
20
30
Entries = 786
= 4.27 MeVΓ
= 0.99 MeVσ
0.048± = 0.980 ε 23± = 185 φN
37± = 610 bkg,pN
116± = 7570 bkg,tN
= 1019.62 MeVφm
q = 1.50
[0.2,0.4) GeV∈T
[0.9,1.5), p∈Probe: y
) = 7.9φN(σ /φN
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
10
20
Entries = 931[1.2,1.6) GeV∈T
[0.9,1.5), p∈Tag: y
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
5
10
Entries = 146
= 4.27 MeVΓ
= 0.99 MeVσ
0.055± = 0.974 ε 7.1± = 23.2 φN
17± = 133 bkg,pN
40± = 908 bkg,tN
= 1019.62 MeVφm
q = 1.50
[1.2,1.6) GeV∈T
[0.9,1.5), p∈Probe: y
) = 3.2φN(σ /φN
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 5 / 38
Integral value vs integration cut-off
lower integration limit [MeV]600 800 1000
]re
f [
% N
ref
N -
N
0
0.005
0.01
0.015
RelBW⊗q-Gaus RelBW⊗q-Gaus
upper integration limit [MeV]1000 2000 3000
]re
f [
% N
ref
N -
N
0
2
4
RelBW⊗q-Gaus RelBW⊗q-Gaus
lower integration limit [MeV]600 800 1000
bias
[%
]
0
0.005
0.01
0.015
RelBW⊗q-Gaus RelBW⊗q-Gaus
upper integration limit [MeV]1000 2000 3000
bias
[%
]
2
4
6
RelBW⊗q-Gaus RelBW⊗q-Gaus
N — integral using given limit, Nref— integral using edges of minv histogram aslimits.
Fits with y∞ − a/|x−mφ|b to obtain y∞— value of relative difference whenlimit is infinite. This allows to calculate correction / bias of the integral for eachvalue of limit.
cR = Nref +NR
Nref= y∞
100 % + 1 cL = NL +Nref
Nref
Reference lower limit for rel. Breit-Wigner already gives at least ‰ accuracy.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 6 / 38
Bias / correction due to integration cut-offlower integration limit [MeV]
600 800 1000
]re
f [
% N
ref
N -
N
0
0.005
0.01
0.015
RelBW⊗q-Gaus RelBW⊗q-Gaus
upper integration limit [MeV]1000 2000 3000
]re
f [
% N
ref
N -
N
0
2
4
RelBW⊗q-Gaus RelBW⊗q-Gaus
lower integration limit [MeV]600 800 1000
bias
[%
]
0
0.005
0.01
0.015
RelBW⊗q-Gaus RelBW⊗q-Gaus
upper integration limit [MeV]1000 2000 3000
bias
[%
]
2
4
6
RelBW⊗q-Gaus RelBW⊗q-Gaus
c∞ = N∞Nref
= NL +Nref +NR
Nref
= NL +Nref +NR +Nref −Nref
Nref
= cL + cR − 1 .
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 7 / 38
Introduction to systematic studies
Biases in this analysis may arise as consequence of1 wrong choice of analytical parametrizations for resonance shape and detector
resolution effect2 choice of integration range of signal parametrization curve to obtain the yield3 unaccounted effects in background description4 constraints used in fitting5 wrong assumptions associated with kaon selection efficiency6 improper MC corrections of detector effects
First 2 points — methods used up to now are changed (changes centralvalues): Voigtian + integration in broad range→ q-Gaussian ⊗ relativisticBreit-Wigner + correction to integral
Other points — systematic uncertainties are estimated using improved methodsfor signal extraction.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 8 / 38
Signal parametrization choiceGaus⊗BW q-Gaus⊗BW
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ 0.12 MeV± = 1.38 σ
77± = 9237 bkgN
50± = 2444 pN
0.071 MeV± = 1019.717 φm
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 1.4 /2χ
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ 0.097 MeV± = 0.978 σ
78± = 9198 bkgN
53± = 2483 pN
0.071 MeV± = 1019.712 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 1.4 /2χ
Gaus⊗RelBW q-Gaus⊗RelBW
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ 0.12 MeV± = 1.38 σ
77± = 9230 bkgN
51± = 2451 pN
0.071 MeV± = 1019.626 φm
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 1.6 /2χ
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ 0.098 MeV± = 0.991 σ
79± = 9187 bkgN
53± = 2494 pN
0.071 MeV± = 1019.623 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 1.6 /2χ
Initially used Voigtian = Gaus⊗BW due to technical convenienceUsing MC decided to change Gaussian→ q-Gaussian (explained later)For φ relativistic Breit-Wigner (used in NA49) better than non-relativistic, whichyields couple % sub-threshold production.Change in χ2 / ndf due to effects in background (explained later)
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 9 / 38
Quantitative comparison of signal parametrizations
[GeV]T
p0 0.5 1 1.5
Rel
BW
]⊗
) [%
q-G
aus
∞c φ(N∆
-10
-5
0 [0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
Rel
BW
]⊗
) [%
q-G
aus
∞c φ(N∆
-10
-5
0 [0.3,0.6)∈y
BW⊗q-Gaus
RelBW⊗Gaus
BW⊗Gaus
[GeV]T
p0 0.5 1 1.5
Rel
BW
]⊗
) [%
q-G
aus
∞c φ(N∆
-10
-5
0 [0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
Rel
BW
]⊗
) [%
q-G
aus
∞c φ(N∆
-10
-5
0 [0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
Rel
BW
]⊗
) [%
q-G
aus
∞c φ(N∆
-10
-5
0 [1.5,2.1)∈y
Yields are corrected integrals in (−∞,+∞) (explained later)Old parametrization yielded up to 10% underestimated results.About 2% due to detector resolution modelAbout 5% due to resonance model
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 10 / 38
Background distortions
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
500
1000
1500
Entries = 128572[0.0,1.6) GeV∈T
[0.0,2.1), p∈Tag: y
ndf = 1.9 /2χ
at least one K pass PIDTag:
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ
0.12 MeV± = 0.70 σ
0.031± = 0.899 ε 183± = 2952 φN
154± = 9746 bkg,pN
504± = 123748 bkg,tN
0.083 MeV± = 1019.566 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 2.0 /2χ
both K pass PIDProbe:
Underestimation of background for high minv in Tag→ K∗0
Underestimation for low minv, overestimation of background for high minv inProbe→ electrons
Using MC (next slides) — up to 10% systematic effect
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 11 / 38
MC vs data
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
10
20
30
310×Entries = 2735448[0.0,1.6) GeV∈
T[0.0,2.1), p∈Tag: y
ndf = 15.7 /2χ
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
5
10
310×Entries = 352111
= 4.27 MeVΓ
0.024 MeV± = 0.593 σ
0.0058± = 0.9122 ε 865± = 76501 φN
813± = 291436 bkg,pN
2299± = 2632288 bkg,tN
0.016 MeV± = 1019.433 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 12.1 /2χ
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
500
1000
1500
Entries = 128572[0.0,1.6) GeV∈T
[0.0,2.1), p∈Tag: y
ndf = 1.9 /2χ
at least one K pass PIDTag:
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
100
200
300
400
Entries = 11681
= 4.27 MeVΓ
0.12 MeV± = 0.70 σ
0.031± = 0.899 ε 183± = 2952 φN
154± = 9746 bkg,pN
504± = 123748 bkg,tN
0.083 MeV± = 1019.566 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 2.0 /2χ
both K pass PIDProbe:
Mock PID cuts tuned in MC (top) to have similar shapes as in data (bottom).
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 12 / 38
MC dirty vs clean
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
10
20
30
310×Entries = 2735448[0.0,1.6) GeV∈
T[0.0,2.1), p∈Tag: y
ndf = 15.7 /2χ
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
5
10
310×Entries = 352111
= 4.27 MeVΓ
0.024 MeV± = 0.593 σ
0.0058± = 0.9122 ε 865± = 76501 φN
813± = 291436 bkg,pN
2299± = 2632288 bkg,tN
0.016 MeV± = 1019.433 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 12.1 /2χ
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
10
20
30
310×Entries = 2216843[0.0,1.6) GeV∈
T[0.0,2.1), p∈Tag: y
ndf = 1.3 /2χ
) [MeV]-
,K+(Kinvm1000 1020 1040 1060 1080
entr
ies
0
5
10
310×Entries = 250372
= 4.27 MeVΓ
0.022 MeV± = 0.696 σ
0.0051± = 0.8893 ε 830± = 81894 φN
668± = 186206 bkg,pN
2085± = 2132134 bkg,tN
0.015 MeV± = 1019.472 φm
q = 1.50
[0.0,1.6) GeV∈T
[0.0,2.1), p∈Probe: y
ndf = 1.3 /2χ
In cleaned sample (no electrons, nothing from K∗0) no background problemsobserved.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 13 / 38
Source of low minv effect in MC — electrons
) [MeV]-
,K+(Kinvm980 1000 1020 1040 1060 1080
entr
ies
0
2000
4000
6000
[0.0,3.0) GeV∈T
[-20.0,20.0), p∈Probe: y
same events
mixed events
[0.0,3.0) GeV∈T
[-20.0,20.0), p∈Probe: y
) [MeV]-,e+(einvm0 20 40 60 80 100
entr
ies
0
2000
4000
6000
[0.0,3.0) GeV∈T
[-20.0,20.0), p∈Probe: y
same events
mixed events
[0.0,3.0) GeV∈T
[-20.0,20.0), p∈Probe: y
Picture compatible with correlation due to Coulomb interaction (studied e.g. asbackground effect in HBT correlations for kaons and pions)
Effect stronger for electrons as compared to hadrons due to lower mass ofelectrons?
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 14 / 38
Systematics
[GeV]T
p0 0.5 1 1.5
) [%
cle
an]
∞c φ(N∆
-10
-5
0
[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
) [%
cle
an]
∞c φ(N∆
-10
-5
0
[0.3,0.6)∈y
data-like±data-like w/o e*0data-like w/o K
[GeV]T
p0 0.5 1 1.5
) [%
cle
an]
∞c φ(N∆
-10
-5
0
[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
) [%
cle
an]
∞c φ(N∆
-10
-5
0
[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
) [%
cle
an]
∞c φ(N∆
-10
-5
0
[1.5,2.1)∈y
Up to 10% systematic effect coming mostly from K∗0→ assign 10% systematicuncertainty bin independent
Although in most bins effect about 5%, assigning bigger uncertainty possiblytakes into account mismatches between MC and data
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 15 / 38
Resolution model
MC with Γ = 0 (20M pp@158 events) provides insight into the effect of detectorresolution on minv.It turns out that the default choice of Gaussian model is not optimal→ testedalso Lorentz and q-Gaussian:
) [MeV]-,K+(Kinvm1016 1018 1020 1022
entr
ies
0
100
200
300
400
Entries = 2204
0.0089 MeV± = 0.8318 σ 0.013 MeV± = 1019.475 φm
[0.6,0.8) GeV∈T
[0.0,0.3), p∈y
ndf = 5.2 /2χGaussian
) [MeV]-,K+(Kinvm1016 1018 1020 1022
entr
ies
0
200
400
Entries = 2204
0.022 MeV± = 1.049 σ 0.011 MeV± = 1019.484 φm
[0.6,0.8) GeV∈T
[0.0,0.3), p∈y
ndf = 8.2 /2χBreit-Wigner
) [MeV]-,K+(Kinvm1016 1018 1020 1022
entr
ies
0
100
200
300
400
Entries = 2204
0.016 MeV± = 0.584 σ 0.012 MeV± = 1019.475 φm
0.030±q = 1.390
[0.6,0.8) GeV∈T
[0.0,0.3), p∈y
ndf = 0.7 /2χ
q-Gaussian
Black dots are the same in all 3 pictures.Each model has a location parameter mφ and width parameter σ.q-Gaussian has additional shape parameter q:
q = 2⇔ shape = Lorentzq → 1 shape→ Gaussian
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 16 / 38
Choice of model
[GeV]T
p0 0.5 1 1.5
ndf
/2 χ
1
10
[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
ndf
/2 χ
1
10
[0.3,0.6)∈y
Gaussian
Breit-Wigner
q-Gaussian
[GeV]T
p0 0.5 1 1.5
ndf
/2 χ
1
10
[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
ndf
/2 χ
1
10
[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
ndf
/2 χ
1
10
[1.5,2.1)∈y
q-Gaussian clearly favoured.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 17 / 38
Parameters stability: σ
[GeV]T
p0 0.5 1 1.5
[M
eV]
σ
0.5
1
1.5
2 [0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
[M
eV]
σ
0.5
1
1.5
2 [0.3,0.6)∈y
Gaussian
Breit-Wigner
q-Gaussian
[GeV]T
p0 0.5 1 1.5
[M
eV]
σ
0.5
1
1.5
2 [0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
[M
eV]
σ
0.5
1
1.5
2 [0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
[M
eV]
σ
0.5
1
1.5
2 [1.5,2.1)∈y
Weighted sample standard deviation 7–9% σ fitted in full phase space,depending on model, with smallest for Gaussian.
These values used to estimate systematic uncertainties (≈ 1%) associated withassumption of invariant σ in phase space bins.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 18 / 38
Parameters stability: q (only for q-Gaussian)
[GeV]T
p0 0.5 1 1.5
q
1
1.5
2
2.5
3
[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
q
1
1.5
2
2.5
3
[0.3,0.6)∈y
[GeV]T
p0 0.5 1 1.5
q
1
1.5
2
2.5
3
[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
q
1
1.5
2
2.5
3
[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
q
1
1.5
2
2.5
3
[1.5,2.1)∈y
Weighted sample standard deviation 6% q fitted in full phase space.These value used to estimate systematic uncertainties (< 2%) associated withassumption of invariant q in phase space bins.q needs to be fixed to MC average value of 1.5 in fits to data due to backgrounddistortions (q-Gaussian can adapt its shape via q to fit background as signal)
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 19 / 38
Parameters stability: mφ
[GeV]T
p0 0.5 1 1.5
[M
eV]
EP
OS
- m
φm
-0.4
-0.2
0
0.2
0.4 [0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
[M
eV]
EP
OS
- m
φm
-0.4
-0.2
0
0.2
0.4 [0.3,0.6)∈y
Gaussian
Breit-Wigner
q-Gaussian
[GeV]T
p0 0.5 1 1.5
[M
eV]
EP
OS
- m
φm
-0.4
-0.2
0
0.2
0.4 [0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
[M
eV]
EP
OS
- m
φm
-0.4
-0.2
0
0.2
0.4 [0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
[M
eV]
EP
OS
- m
φm
-0.4
-0.2
0
0.2
0.4 [1.5,2.1)∈y
Weighted sample standard deviation ≈ 0.5% Γ, 0.002% mφ fitted in full phasespace, for all models.
Translates into < 0.5% systematic uncertainty.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 20 / 38
Systematics due to constraints on parameters
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-2
0
2
[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-2
0
2
[0.3,0.6)∈y-σ+σ
)σ(statσ - σ)σ(statσ + σ
-φm+φm
)φ
(mstatσ - φm
)φ
(mstatσ + φm-q+q
, others refitted-q
, others refitted+q
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-2
0
2
[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-2
0
2
[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-2
0
2
[1.5,2.1)∈y
„+“, „-“ superscripts — fits redone with listed fixed parametersincreased/decreased by factors obtained from MC study from previous slidesAlso shown refits with fixed parameters shifted by their statistical errors from fitsin full phase spaceSystematic uncertainty: 2%, bin independent or should sum up, or bin by bin?
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 21 / 38
Tag-and-probe systematics
Known sources of systematic error in tag-and-probe:non-constant value of PID efficiency (ε) within phase space binsconstraints on ε in (y, pT ) bins fits if ε non-constant between bins
Known and unknown effects studied by variation of window size aroundBethe-Bloch (range +/- 30% of default/reference window size = ±5%Bethe-Bloch value)Done for 2 cases of fitting strategy:
default — value of ε fitted in y bin in full pT range is used to soft-constrain fits in(y, pT ) binsfree ε in (y, pT ) bins fits
to validate these strategies
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 22 / 38
Fit results: ε in „constrained” strategy
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[0.3,0.6)∈y reference
3.5%±cut
3.8%±cut
4.1%±cut
4.4%±cut
4.7%±cut
5.3%±cut
5.6%±cut
5.9%±cut
6.2%±cut
6.5%±cut
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[1.5,2.1)∈y
Apart from one case in the last y bin, ε changes monotonically with window size
agrees with expectation
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 23 / 38
Fit results: ε in „free” strategy
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[0.3,0.6)∈y reference
3.5%±cut
3.8%±cut
4.1%±cut
4.4%±cut
4.7%±cut
5.3%±cut
5.6%±cut
5.9%±cut
6.2%±cut
6.5%±cut
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
ε
0
0.2
0.4
0.6
0.8
1[1.5,2.1)∈y
Visible problems with monotonic dependence of ε on window size
contrary to expectation — fit instabilities?
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 24 / 38
Fit results: Nφ in „constrained” strategy
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
0
50 [0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
0
50 [0.3,0.6)∈y3.5%±cut
3.8%±cut
4.1%±cut
4.4%±cut
4.7%±cut
5.3%±cut
5.6%±cut
5.9%±cut
6.2%±cut
6.5%±cut
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
0
50 [0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
0
50 [0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
0
50 [1.5,2.1)∈y
Differences between Nφ values for the given and the reference cut aspercentage of results for reference cut
If no systematic error→ all points should cluster at zero; standard deviation =measure of systematic uncertainty
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 25 / 38
Fit results: Nφ in „free” strategy
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-50
0
50
[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-50
0
50
[0.3,0.6)∈y3.5%±cut
3.8%±cut
4.1%±cut
4.4%±cut
4.7%±cut
5.3%±cut
5.6%±cut
5.9%±cut
6.2%±cut
6.5%±cut
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-50
0
50
[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-50
0
50
[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
) [%
ref
eren
ce]
∞c φ(N∆
-50
0
50
[1.5,2.1)∈y
Differences between Nφ values for the given and the reference cut aspercentage of results for reference cutIf no systematic error→ all points should cluster at zero; standard deviation =measure of systematic uncertaintyClearly more spread than in „constrained” case
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 26 / 38
Tag-and-probe systematic uncertainties
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
10
20
30 [0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
10
20
30 [0.3,0.6)∈y
εconstrained
εfree
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
10
20
30 [0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
10
20
30 [0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
10
20
30 [1.5,2.1)∈y
„constrained” strategy yields smaller systematic uncertainties than „free”
Above, together with better behaviour of ε and smaller statistical uncertaintiesclearly favours „constrained” strategy over the „free” one.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 27 / 38
Main vertex Z position cut variations
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-40
-20
0
20
40[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-40
-20
0
20
40[0.3,0.6)∈y
cut width = 13.0 cm
cut width = 13.5 cm
cut width = 14.0 cm
cut width = 14.5 cm
cut width = 15.0 cm
cut width = 15.5 cm
cut width = 16.0 cm
cut width = 16.5 cm
cut width = 17.0 cm
cut width = 17.5 cm
cut width = 18.5 cm
cut width = 19.0 cm
cut width = 19.5 cm
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-40
-20
0
20
40[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-40
-20
0
20
40[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-40
-20
0
20
40[1.5,2.1)∈y
Differences between normalized and corrected yield values for the given andthe reference cut as percentage of results for reference cut = 18 cm
If no systematic error→ all points should cluster at zero; standard deviation =measure of systematic uncertainty
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 28 / 38
Systematic uncertainty due to vertex Z position cut
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
5
10
15
[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
5
10
15
[0.3,0.6)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
5
10
15
[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
5
10
15
[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
5
10
15
[1.5,2.1)∈y
Magnitude similar to tag-and-probe systematics
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 29 / 38
Number of TPCs points cuts variations
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-20
-10
0
10[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-20
-10
0
10[0.3,0.6)∈y
no_bxby
>10VTPC
>25, nalln
>11VTPC
>26, nalln
>12VTPC
>27, nalln
>13VTPC
>28, nalln
>14VTPC
>29, nalln
>16VTPC
>31, nalln
>17VTPC
>32, nalln
>18VTPC
>33, nalln
>19VTPC
>34, nalln
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-20
-10
0
10[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-20
-10
0
10[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
dy))
[%
ref
eren
ce]
T(d
p /
n2(d∆
-20
-10
0
10[1.5,2.1)∈y
Differences between normalized and corrected yield values for the given andthe reference cut as percentage of results for reference cut =nall > 30, nV TPC > 15nGAP−TPC > 4 not varied; also shown result after removing Bx,By cutIf no systematic error→ all points should cluster at zero; standard deviation =measure of systematic uncertainty
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 30 / 38
Systematic uncertainty due to track quality cuts
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
5
10[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
5
10[0.3,0.6)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
5
10[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
5
10[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
syst
. unc
erta
inty
[%]
0
5
10[1.5,2.1)∈y
Magnitude smaller than for to tag-and-probe and vertex cut systematics
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 31 / 38
Model dependence of MC correction
MC correction may depend onφ model — shape of generated φ spectrum
event model — distributions of other particles and correlations between particles
detector model — geometry, materials, models of interactions with material
Removing φ model dependencecalculate correction is small bins; on application level use
weighting of entries with the correctionaveraging of correction with fit of data spectrum (NA49)
reweight existing MC (Antoni’s pp h- paper)
Reducing systematic uncertainty of correctiondetector model dependence unavoidable; can only improve the model
event model→ find better one, or
factorize correction into accurate large part that doesn’t depend on event modeland smaller that depends
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 32 / 38
Breakdown of the MC correction
Single correction is calculated and applied to data, but one can look howdifferent effects contribute to this correction.
Breakdown realised by sequentially applying selection cuts. For φ it means thatboth K need to pass the given track cut.
Conditions probably are not statistically independent, so change of cutssequence may change the breakdown.
Overall systematic uncertainty might be reduced if correction factorized intodominant, accurate part and subdominant, less accurate part.
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 33 / 38
Breakdown of MC correctionregistration efficiency
Correction
cgeom =(nreg
ngen
)−1
where nreg — spectrum of generated (SimEvent) tracks that pass the cuts:
Number of GEANT points in all TPCs > 30
Number of GEANT points in VTPCs > 15 or GTPC > 4
Supposed to correct for particle registration efficiency (geometry, interactionswith detector, K decays)
Probably does not take into account correctly the K decay effect
No dependence on the model of event production→ candidate to factorize outand calculate from large statistics, well binned, flat phase space MC
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 34 / 38
Breakdown of MC correctiontrigger bias
Correction
cT2 =(nT2
nreg
)−1
where nT2 — spectrum of generated (SimEvent) tracks that pass the cut reg andevents with T2 trigger (no GEANT hits in S4).
Corrects for trigger bias due to S4 killing inelastic events
Expected to be bigger at high energies (many high momentum tracks) andsmaller at low energies
Depends on the model of event production
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 35 / 38
Breakdown of MC correctionvertex cuts bias
Correction
cVertex =(nver
nT2
)−1
where nver — spectrum of generated (SimEvent) tracks that pass the cut reg, T2and events pass all vertex cuts
Corrects for bias due to vertex cuts — removal of low multiplicity events
Expected to be smaller at high energies (large track multiplicities) and bigger atlow energies
Depends on the model of event production
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 36 / 38
Breakdown of MC correctiontrack cuts bias
Correction
cTrack =(nsel
nver
)−1
Corrects for reconstruction efficiency and bin migration, since nsel binnedaccording to the reconstructed momentumReconstruction efficiency
expected to be small for proton-proton due to low track multiplicitiesdepends on the model of event production
Bin migrationdepends on momentum resolution
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 37 / 38
Breakdown of MC correction
[GeV]T
p0 0.5 1 1.5
corr
ectio
n
0.8
1
1.2
1.4[0.0,0.3)∈y
[GeV]T
p0 0.5 1 1.5
corr
ectio
n
0.8
1
1.2
1.4[0.3,0.6)∈y
total
geom
T2
Vertex
Track
[GeV]T
p0 0.5 1 1.5
corr
ectio
n
0.8
1
1.2
1.4[0.6,0.9)∈y
[GeV]T
p0 0.5 1 1.5
corr
ectio
n
0.8
1
1.2
1.4[0.9,1.5)∈y
[GeV]T
p0 0.5 1 1.5
corr
ectio
n
0.8
1
1.2
1.4[1.5,2.1)∈y
Antoni Marcinek (IFJ PAN) φ production in p+p in NA61/SHINE Białasówka, 28 April 2017 38 / 38