observation of the w b -
DESCRIPTION
Observation of the W b -. Brad Abbott University of Oklahoma On behalf of the D Ø Collaboration. SLAC January 6, 2009. B physics at the Tevatron. Upsilon discovered in 1977 at Fermilab by Lederman et. al. Since then other observations of other B mesons - PowerPoint PPT PresentationTRANSCRIPT
Brad Abbott 1
Observation of the b-
Brad AbbottUniversity of Oklahoma
On behalf of the DØ Collaboration
SLAC January 6, 2009
Brad Abbott 2
B physics at the Tevatron
• Upsilon discovered in 1977 at Fermilab by Lederman et. al.
• Since then other observations of other B mesons – B+, B0, Bs, Bc
+ (pre Tevatron RunII)
– B* (pre Tevatron RunII), – Bd**(Tevatron RunII)– Bs** (Tevatron RunII)
• Little experimentally known about B baryons
• Only one B baryon known for many years
Summer 1977
Brad Abbott 3
B baryonsCounting only quark content 15 b baryons are expected
Charmless b baryons multiplet
Brad Abbott 4
B baryons more recently
• Prior to Run II only 1 b baryon, b(udb), was considered observed (b J/ )
• However in last 2 years 4 new b baryons have been discovered b
+(uub) b-(ddb) (CDF, 2007)
b-(dsb) (D0/CDF 2007)
b-(ssb) (D0 2008)
Brad Abbott 5
(*)-b in October 2006
CDF announces the observation of the b’s with 1.1 fb-1
PRL 99, 202001 (2007)
Brad Abbott 6
Last year: --bb observation observation
(syst) (stat) 0.09 0.28 0.09 0.08 -
R
5.5ln2ln2
BS
B
L
LL
Signal Significance:
)/()(
) /()(
JBR
JBRR
bb
bb
PRL 99, 052001 (2007)
M(b-)
5792.9 2.5 (stat) 1.7 (syst) MeV/c2
PRL 99, 052002 (2007)
Signal significance = 7.8
CDF
M(b)=5.774 ± 0.019 GeV
Brad Abbott 7
B baryonsVarious predictions for b baryon masses
“Know” approximately where to look and what decay modes are expected
B quark section large at Tevatron(pp bb) = 150 ub @ √s=2 TeV
(e+e- bb) = 7 nb @ √s=Mz
(e+e- bb) = 1 nb @ √s=M(4S)
However large backgrounds make detection challenging
Brad Abbott 8
Tough environment
BaBar
~ 5 tracks/event
Brad Abbott 9
DataTevatron is running well: ~ 5 fb delivered to D0/CDF
D0 Efficiency high: ~ 4.5 fb recorded by each experiment
This analysis
1.3 fb-1
Brad Abbott 10
DØ detector
Large coverage
No particle ID(limited /K separation)
Unprescaled dimuontriggers at all luminosities
Brad Abbott 11
Triggers
• Entire b physics program based on muon triggers (single and dimuon triggers)
• Do not have bandwidth available to trigger on displaced vertices
• Many b physics analyses use dimuon triggers (unprescaled) and search for J/ decaying to .
• Dimuon triggers have several Hz final rate
Brad Abbott 12
B physics program at DØ
Over 34 papers published/accepted for publicationCP violation in Bs decaysFlavor oscillations in Bs decaysBs Branching ratio of Bs Ds*Ds*Bc meson mass/lifetimeDirect CP violation….For full list seehttp://www-d0.fnal.gov/Run2Physics/WWW/results/b.htm
Brad Abbott 13
B baryons at DØb J/
Aids in understanding momentum scale
Possible biases in reconstruction
Lambda long lived so gives practicein reconstructing long lived particles
Discovered that official processing of data inefficient for tracks with large impactparameterTakes a significant time to reconstruct each event andDØ cannot afford to increase reconstruction time forall events
Brad Abbott 14
Data reprocessing
p
When tracks are reconstructed, a maximum impact parameter is required to increase the reconstruction speed and lower the rate of fake tracks.
But for particles like the b-,
this requirement could result in missing the and proton
tracks from the and - decays
Brad Abbott 15
Reprocessing J/ sample
• Since interesting decays all have a J/ only reconstruct subset of data
• Much smaller sample and can be reconstructed in a reasonable time (few months)
Brad Abbott 16
Increase of reconstruction efficiency
0
SK p
Opening up the IP cut: (Before) ( After )GeV GeV GeV
D0 D0 D0
Brad Abbott 17
Similarities of decays
b J/ b J/
Techniques learned in measuring b led to observation of b in 2007
Apply same idea to searching for b
Use techniques learned from b to help in searching for b
“extra pion”
Brad Abbott 18
Search for the -b(bss)
bss quarks combination Mass is predicted to be
5.94 - 6.12 GeV
0.83<(-b)<1.67 ps
(prediction)
M(-b) > M(b)
Brad Abbott 19
--
b
+
-
-
p
K-
~3 cm
~5 cm
Similar decay topologiesb J/ b J/
Brad Abbott 20
Slight differences
• c() = 7.89 cm• c() = 2.46 cm• Lifetime(b) = 1.42 ps• Lifetime(b)= .84-1.69 ps (predicted)• Mass(b) = 5.792 GeV• Mass(b) = 5.94-6.12 GeV (predicted)
• Much more difficult to reconstruct than
Brad Abbott 21
vs reconstruction →p decays:
– pT(p)>0.7 GeV
– pT()>0.3 GeV
- → decays:– pT()>0.2 GeV
– Transverse decay length>0.5 cm
– Collinearity>0.99
)( p
D0
D0
reconstruction
Black symbols: right sign combinationsGreen/red symbols: wrong sign combinations
Brad Abbott 22
reconstruction
• Need to clean up mass peak
• First clean up peak
• Apply Boosted Decision Trees to further clean up sample
Brad Abbott 23
optimization
Apply a cut on proper decay length significance
( decay length significance > 10)
D0 Before cut
Brad Abbott 24
reconstruction
• Minimum selection cuts: +K vertex
reconstructed– Transverse decay
length significance>4– Proper decay length
uncertainty<0.5 cm
Wrong-sign
events+K+
Right-sign events(+K-)
D0
PDG mass value
Brad Abbott 25
Boosted Decision Trees (BDT)
• 20 input variables and - vertex
quality, decay lengths and decay kinematics
• For training we use MC signal and background from wrong-sign events (J/(K+)).
• Most important variables: pT(K)
pT(p)
pT()
decay length
Brad Abbott 26
- after BDT selection
- signal much
cleaner
D0
Signal yield approximately the sameBackgrounds greatly reduced
Brad Abbott 27
Contamination due to --
• There is a reflection due to assigning a pion as a kaon.
• Remove by requiring M()>1.34.
DØ
DØ
Brad Abbott 28
- after BDT selection and removing decays
Wrong-sign combination events
DØ
Brad Abbott 29
Final b optimization
• Optimized for the efficiency
• Further reduce background (based on level we observe in the wrong-sign combinations.)
• Uncertainty on b proper decay length < .03 cm
b transverse momentum greater than 6 GeV
J/ and in same hemispere
Brad Abbott 30
Mass ResolutionPDGPDG
JJJ MMMMMM ///
= 80 MeV = 34 MeV
“poor man’s” way to reduce mass resolution
Brad Abbott 31
Look where we don’t expect any signal
• After “optimize” J/+ decays by using wrong-sign combination events: <0.03 cm
J/ and in the same hemisphere
pT(J/+)>6 GeV
30 events remain
Brad Abbott 32
Side band control samples
DØ DØ
Brad Abbott 33
Summary of all control samples
No excess is observed in any MC sample after selection criteria applied.
)))((())((/
)))((())((/
))((/0*
pJ
KKJB
pJ
b
S
b
Study known decays using MC samples with statistics larger than data set
Brad Abbott 34
“Open Box” and look at right-sign combinations
Excess of events
79 events selected
Brad Abbott 35
Mass measurement
• Fit:Unbinned log-
likelihood fitGaussian signal, flat
backgroundNumber of
background/signal events are floating parametersNumber of signal events: 17.8 ±
4.9
Mean of the Gaussian: 6.165 ± 0.010(stat) GeV
Width of the Gaussian fixed (MC): 0.034 GeV
Brad Abbott 36
Signal Significance
• Two likelihood fits are performed to estimate the significance:1.Signal + background hypothesis (LS+B)
2.Only background hypothesis (LB)
• We evaluate the significance:
• Significance of the observed signal: 5.4
BS
B
L
LL ln2ln2
Brad Abbott 37
Consistency check: Increase pT(B)
Significance > 6
Brad Abbott 38
Look back plots
D0
D0
See consistent yields
Brad Abbott 39
Another consistency check
We can compare proper decay length of sample with MC sample with a lifetime of 1.54 ps
We do not know the lifetime of the b
We do not have enough events to measure the lifetime of the b
Brad Abbott 40
Alternative Cuts Based Analysis (CBA)
Variable BDT CBApT() (GeV) >0.2 and input to BDT >0.2
pT(p) (GeV) >0.2 and input to BDT >0.7
pT(K) (GeV) input to BDT >0.3
collinearity input to BDT >0.99
Transverse decay length (cm)
input to BDT >0.5
Proper decay length uncertainty (cm)
<0.3 <0.3
Variables selected based on relative importance in BDT
performance
Brad Abbott 41
Cut Based Analysis
Number of signal events: 15.7 ± 5.3
Mean of the Gaussian: 6.177 ± 0.015(stat) GeV
Width of the Gaussian fixed (MC): 0.034 GeV
Signal significance reduced to 3.9 due to increased background
Brad Abbott 42
BDT or Cut Base Analysis
After removing duplicate events and combining analyses we observe 25.5 ± 6.5 events. Significance: 5.4
Cut based and BDT can select different events. Overlap is ~ 50%
Brad Abbott 43
Signal confirmed without BDT.
• BDT vs CBA– Consistent number of
observed signal candidates
– Consistent mass– Consistent
reconstruction efficiencies
– BDT has better background rejection power.
Brad Abbott 44
Systematic uncertainties on the mass
• Fitting models– Linear background instead of flat background gives
negligible change.– Varying the signal Gaussian width between 28 – 40
MeV resulted in a 3 MeV uncertainty• Momentum scale correction:
– Fit to the b mass peak in data gives a 4 MeV uncertainty.
• Event selection:– Varying selection criteria and from the mass shift observed
between the cut-based and BDT analysis gives a 12 MeV uncertainty.
Brad Abbott 45
b mass
M(b)=6.165 ± 0.010(stat) ± .013(stat) GeV
A little on the high side of the predictions1-2
Brad Abbott 46
Production rate
)()(32.080.0)/()(
)/()( 14.022.0 syststat
JBrbf
JBrbf
bb
bb
To determine production rate we normalize to b
)(
)(
)(
)(
)/()(
)/()(
b
b
b
b
bb
bb
N
N
JBrbf
JBrbf
2.06.1)(
)(
b
b
The systematic uncertainty includes contributions from the signal yields as well as selection efficiencies
Determined from MC
Brad Abbott 47
Production rate
8.9)/(
)/(
: (1997) 2799 56, D Rev. Phys. From
J
J
b
b
ps 42.1)( 28.024.0
b
theory
ps 67.1)(83.0 b
ps) 67.1)(( 062.0
ps) 83.0)(( 126.0
)(
)(
b
b
b
b
bf
bf
14.007.0)(
)(
b
b
bf
bf
Spin of b = ½ Spin of b = 3/2
Brad Abbott 48
Summary of b
Number of signal events: 17.8 ± 4.9 (stat) ± 0.8(syst)
Mass: 6.165 ± 0.010(stat) ± 0.013(syst) GeV
Significance= 5.4
arXiv:0808.4142(2008) PRL 101, 232002 (2008)
Brad Abbott 49
Summary of b
14.007.0)(
)(
b
b
bf
bf
Consistent with expectations
)()(32.080.0)/()(
)/()( 14.022.0 syststat
JBrbf
JBrbf
bb
bb
Brad Abbott 50
Future plans
• Reprocess Run IIb data– Lots more data (1.3 pb-1Run IIa vs 3.2 pb-1
Run IIb) – Much higher instantaneous luminosities so
much larger combinatorial background– Layer 0 silicon
• Begin searches for other B baryons
Brad Abbott 51
Other possible B baryons(need a J/ to trigger)
0cb(dcb) 0
c J/ +
cb(ucb) +c J/
0bb(ubb) 0
b J/low rate
-bb(dbb) -
b J/low rate
0cb(scb) 0
c J/
Brad Abbott 52
Backup
Brad Abbott 53
Event display of b candidate
Brad Abbott 54
Boosted Decision Trees (BDT)
• All variables are on - or its decay products.
• For training we use MC signal and background from wrong-sign events (J/(K+)).