Upload
noel-edwards
View
217
Download
1
Embed Size (px)
Citation preview
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 1 ● 03:44:44 PM
Aart HeijboerUniversity of Pennsylvania
previous result: PRL 97, 062003 (2006)current result: PRL 97, 242003 (2006)
PRD in preparation
outline: motivation what do we look for / previous results the measurement and how we improved it the observation
Observation of Bs oscillations at CDF
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 2 ● 03:44:44 PM
Introduction & motivation
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 3 ● 03:44:44 PM
What do we measure?There is a diagram which allows a B
s meson to
transform into its own anti-particle
time
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 4 ● 03:44:44 PM
mass difference
MH = M + M/2
ML = M - M/2
Because of a coupling between the two states, we get: two separate mass eigenstates oscillation between B
s and B
s
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 5 ● 03:44:44 PM
classical version
test
Energy eigenstates (normal modes)
time
By measuring the frequency,we are measuring the strength
of the coupling
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 6 ● 03:44:44 PM
motivation
)(
1)1(
2/1
)(2/14
23
22
32
O
AiA
A
iA
VVV
VVV
VVV
V
tbts
cbcscd
ubusud
CKM
td
most uncertain parametersand
To get at Vtd: measure m
d (B
d oscillations)
hard to calculate (lattice QCD). uncertainty: 10%
1.210.05
q=s or d
If we measure ms, we
can form the ratio whichallows for much more accuratemeasurement of V
ts/V
td(hep-lat/0510113)
pretty wellknown
parameterization assuming unitarity
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 7 ● 03:44:44 PM
motivation
2005 Limit on ms was
already helping to measureCKM matrix...
Plot Combines many different measurements to constrain and .
Check that all measurementsare consistent: i.e. that CKMmatrix is unitary.
Vtd=A3(1--i)
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 8 ● 03:44:44 PM
motivation
Effect of measuring ms to
the level of a few %=> shrinking the orange band
Plot Combines many different measurements to constrain and .
Check that all measurementsare consistent: i.e. that CKMmatrix is unitary.
Vtd=A3(1--i)
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 9 ● 03:44:44 PM
motivation
Effect of measuring ms to
the level of a few %=> shrinking the orange band
Plot Combines many different measurements to constrain and .
Check that all measurementsare consistent: i.e. that CKMmatrix is unitary.
Vtd=A3(1--i)
Turn the argument around: standard model prediction:
Go measure if SM is right. If the answer is no.....
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 10 ● 03:44:44 PM
motivation: new physics
New particles can enhance the Bs B
s transition
harnik et al. hep-ph/0212180
supersymmetry with largemixing between squarks
ms is sensitive to new physics
Phys.Lett.B596:229-239,2004
Z'
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 11 ● 03:44:44 PM
history
UA1 resultPLB 186 (1997) 247
1987 UA1: neutral B-mesons mix Argus measures fraction of B
d mesons
that mix (~18%) => ~half of the Bs mesons must mix => ms is large compared to the decay time => need to resolve rapid oscillations: hard!
=1
1+ (m)-2
1
2
fraction of mixed events
Bs case:=0.5 -> no handleon ms
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 12 ● 03:44:44 PM
time
How do we look for mixing?
Make a Bs meson and measure
its flavor when it decays Identify the flavor at production (tagging) Look at oscillation probability vs proper
decay time.
B s/B
s
time
Fourier component of asymmetry:Mixing amplitude
asymmetry=N
unminx-N
mix/N
mixing frequency
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 13 ● 03:44:44 PM
history-II: looking for oscillations
LEP, SLD, CDF RUN-Ihave tried, but have not seen a convincingsignal: lower limits on m
s
2005 world average:compatible with mixingbut not significant enough
We know Bs mixes, now try tosearch for the right frequency m
s
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 14 ● 03:44:44 PM
history: recent results
D0 resultlarge amplitude around m
s = 19 ps-1
CDF's previous resultamplitude compatible with 1 around 17.3
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 15 ● 03:44:44 PM
history: recent results
D0 result A double sided 90% confidence
level interval. (17-20 ps-1) p-value = 5% (now 8%)
CDF's previous result p-value = 0.2% (3) linguistics “under the assumption that
this is signal, we measure:”
When are we sure we are not seeing a fluctuation?convention: 5: p-value = 5.7 x 10-7 => ~no doubt => publish “Observation”
PRL 97, 062003 (2006)
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 16 ● 03:44:44 PM
Upgrading the analysis
signal selectionuse neural networkuse particle-ID on kaon-candidatesadd partially reconstructed decaysadd some trigger paths
taggingkinematic information same side taggercombine opposite side taggers optimally
Previous analysis had: precise measurement of ms, but the significance
was not enough (0.2%).Goal for new analysis: remove all doubt and go for the conclusive, final, 'Observation'reblind the analysis & improve until we have a chance to see 5
most likely
contain 90% ofexperiments
previousresult
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 17 ● 03:44:44 PM
1) Getting the signals
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 18 ● 03:44:44 PM
The Tevatron
1.2 fb-1 collected,1 fb-1 in 'good' runs
~2/3 of the datacollected in 2005
no new data added since previous result
both are 1 fb-1
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 19 ● 03:44:44 PM
The CDF detector
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 20 ● 03:44:44 PM
The CDF detector
momentum &dE/dx
decaytime
PID
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 21 ● 03:44:44 PM
The signals we are looking for
lss lDB 0
0sB
sD
W
b cl
s
l
s
Semileptonic
0sB
sD
W
b cd
s
u
s
Hadronic
ss DB0
Bs Momentum is measured
Bs mass used for good S/N
Small branching ratio: low yield
Missing momentum ()Need to rely on D
s mass
Large branching ratio: high yield
Bs /Bs
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 22 ● 03:44:44 PM
A typical B event at a hadron collider
=> look for displaced tracks
Trigger on events with twodisplaced (d>120 m) tracks
dvery fast reconstruction of
silicon data at L2 (20s latency) by dedicated hardware: SVT
cdf
d ()
Displaced track trigger
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 23 ● 03:44:44 PM
Hadronic signals
Added partially reconstructed modesin the Ds-> mode.
Bs/B
s
undetected particles have smallmomentum => decay is
almost fully reconstructed
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 24 ● 03:44:44 PM
Hadronic signals
reflection of less severe because we checkthat the PID of the kaon. Allows looser veto based onM
KK - MD+
Added one new decay modeB
s -> six pions
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 25 ● 03:44:44 PM
hadronic signal selection
signal selection previously doneusing cuts on kinematic variables,vertex displacement, fit quality....
select signal using NNinputs: O(30) variables includingall of the above & P
T of tracks,
Ds's 's, K*, PID
particle-ID likelihoodsuse measurement of time-of-flightand dE/dx in the COT to distinguishkaons from pions.
careful to make sure NN is not trained to select mass. (e.g. R notused as input variable)
gain signal lose onlybackground
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 26 ● 03:44:44 PM
hadronic yields
8700 signal events (was 3700)improved S/B too
equivalent to an increase in sample size of a factor 2.5
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 27 ● 03:44:44 PM
Semileptonic signals
improvements: using particle ID to identify kaons
reduces combinatorial backgroundeliminates need for explicit D+ rejection
62k Bs signal events(was 37k)
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 28 ● 03:44:44 PM
Proper decay time and resolution
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 29 ● 03:44:44 PM
Proper time reconstruction: hadronic
Bs/B
s
L xy
piece of cake!
Bs mass from the PDG, PT from adding up the moment of all the decay tracks
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 30 ● 03:44:44 PM
Proper time: semileptonic
Bs/B
s
L xy
● k-factor PDF obtained from Monte Carlo● Take advantage of variation with l D
s mass.
● high mlDs
means small missing pT)
● can not measure pT for semileptonic
● large uncertainty in ct for high Lxy● correct for missing p
T on average:
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 31 ● 03:44:44 PM
Proper time resolution
Effect of non-zero ct error asymmetry: attenuation of the oscillation
Smearing of decay time causes attenuation of asymmetry signal: Have to know
ct to measure the mixing amplitude
How to measure ct?
2
2
)( t
t
m
eD
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 32 ● 03:44:44 PM
Measuring the proper time resolution
Cannot measure the ct resolutiondirectly on data (no prompt peak in theB
s signal due to trigger)
Solution: construct events that look like a B but are known to come from the PV..
real D
some pionfrom the underlyingevent (i.e. from PV)
True <ct> must be zero; compare errorwith predicted error from the vertex fit.
And study dependence on kinematicvariables, isolation, 2 of fit etc..
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 33 ● 03:44:44 PM
proper time resolution
period for ms = 18 ps-1
● partially reconstructed hadronic almost as good as the fully rec.● time resolution = ¼ oscillation period
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 34 ● 03:44:44 PM
3) Flavour tagging
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 35 ● 03:44:44 PM
Flavor tagging
Bs/B
s
L xy
We need to know the flavor of the Bs at production.
Opposite side tag:look at the decay products of the other b quark in the event:
LeptonKaonCharge of the b-'jet'
the two b quarks fragment independently: can calibrate opposite side taggers with B
d & B
uother B often outside acceptance
Same side kaon tag: look at particles produced in B meson formation (K in case of Bs)
Very powerful (high acceptance)but cannot calibrate on B
d & B
uNeed to rely on MC
K
e/K
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 36 ● 03:44:44 PM
A tagger is characterized by: efficiencyD: dilution = 1-2 x mistag rate (large dilution is good)
Dampens asymmetry: Effective statistical power / figure of merit = D2
x D
We must know what D is to measure A
D<1 dampens the oscillation signal
Flavor tagging II
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 37 ● 03:44:44 PM
OST: improvementprevious analysis:
In case multiple tagger gave a decision, used the best onenew analysis
combine all tagger decision, per-event predicted dilutions and kinematicvariables into a single Neural Network, producing tag decision & dilution.
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 38 ● 03:44:44 PM
●Fragmentation into Bs tends to
produce an additional kaon.●Charge of K identifies B
s flavor at production
●B0 and B+ mostly accompanied by pions●Use combined likelihood from time-of-flight detector and dE/dx in COT identify Kaons
Same side kaon tagging
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 39 ● 03:44:44 PM
Same side kaon tagging
tagger not optimized for these modes, but valuable checkthat MC predicts the right dilution
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 40 ● 03:44:44 PM
Same side kaon: improvementsSame side kaon: improvements
Previously based only on particle IDKinematic variables can also be used for
for selecting the tagging trackCombine in Neural network
PID likelihood, R, pT p
Trel, p
Lrel,...
improvement in D2 based on predicted dilutions for events: semileptonic data: +8% hadronic data : +0% monte carlo : +6%
40
1.8%OST
3.7(4.8)%
D2
SST
Tag
final tagger performance
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 41 ● 03:44:44 PM
Blind Amplitude Scan
(July 25)
With all the improvements:
A < 0.2 → good chance of seeing 5.
unblind!
(old value: 0.28)
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 42 ● 03:44:44 PM
Results
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 43 ● 03:44:44 PM
Results
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 44 ● 03:44:44 PM
Results
depth of likelihood dip=-17.26
what is the significance of such a dip?
old value
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 45 ● 03:44:44 PM
significance
“Observation”
Randomize the tagger decisions of the data events and repeat the likelihood scanmany many, many, many, many, many, many, many, many, many, many, many, many, many, many, many, many, many, many, many, many, many, many times
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 46 ● 03:44:44 PM
comparing taggers
Opposite side tagger Same side kaon tagger
Both taggers see the oscillations
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 47 ● 03:44:44 PM
Comparing datasetspartially reconstructed hadronic modes
semileptonic modes
fully reconstructed hadronic modes
all see the signal!
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 48 ● 03:44:44 PM
measurement of ms and |V
td|/|V
ts|
uncertainty of length scale ofdetector
limited by theoryerror
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 49 ● 03:44:44 PM
asymmetry
mtPPPP
Amixumix
mixumix cos
asymmetry
time
stack all oscillation periods on top of each otherand fit for A in bins of proper decay time...
data are well described by theexpected cosine curve.
hadronic data only
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 50 ● 03:44:44 PM
asymmetry
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 51 ● 03:44:44 PM
conclusions
We have 'Observed' Bs mixing p-value = 8x10-8, significance is 5.4
We measure
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 52 ● 03:44:44 PM
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 53 ● 03:44:44 PM
World average back in 1999justforfun
Abbaneo & Boix: JHEP 08(1999)004“p-value = 3%”
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 54 ● 03:44:44 PM
all data before 2004LEP + SLD
World average before Tevatron run II
Amplitude scan: fix ms and measure the amplitude
of the corresponding frequency component (Fourier)
Recent D0 result● 17 <ms <21 ps-1 at 90% CL● p-value = 5%
previous results
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 55 ● 03:44:44 PM
likelihood and amplitude scan
Amplitude: quantity constructed so thatA=1 if the data (i.e. the likelihood) consistent with mixingA=0 everywhere else
Amplitude is 'easy' to understand, but we use thelikelihood to make the measurement, namely:
signatureof a signal
signatureof a signal
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 56 ● 03:44:44 PM
●resolution ~ 100 ps●separates kaons from pions up to 1.5 GeV●crucial for SSKT
Time of flight detector
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 57 ● 03:44:44 PM
Expected CDF Sensitivity
With L00
Without L00
95% CL sensitivity
3 sensitivity
ms
A
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 58 ● 03:44:44 PM
cdf-II error at 15.0/ps : 0.2
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 59 ● 03:44:44 PM
likelihood definition
the probability to measure an event as (un)mixed as a function of the measured decay time:
the probability that an event mixes (does not mix)as a function of the true decay time:
goodness of flavortagger (dilution)
time resolution
(sketch of)
-log likelihood of all events:
If the data are compatible with oscillations, -log(L) has a dip at the right value of ms
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 60 ● 03:44:44 PM
The big questions
How accurately can we measure ms?
determined by shape of the likelihood at minimum How sure are we that what we are seeing is
not just a random fluctuation?formally: What is the probability that an apparent signal at least as significant as the one we see in data would occur at any value of m
s in case there is no signal in reality?
probability that a dipso deep would come
from random fluctuation= 0.2 %
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 61 ● 03:44:44 PM
Opposite side taggers
Measure opposite side tagger dilutions Simultaneous fit to lD+, lD0
and lD* modes
DSeB
tmDSeBt
dt
1:
)cos(1:/
/0
predicteddilution
dilutionscale factor
Dilution predicted onevent-by-event basis(based on P
Trel, lepton-id
etc).
how to check/calibratethe prediction is correct?
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 62 ● 03:44:44 PM
Amplitude scan in sideband
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 63 ● 03:44:44 PM
A nice event
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 64 ● 03:44:44 PM
systematics on amplitude
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 65 ● 03:44:44 PM
systematics on ms
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 66 ● 03:44:44 PM
likelihood definition
the probability to measure an event as (un)mixed as a function of the measured decay time:
the probability that an event mixes (does not mix)as a function of the true decay time:
goodness of flavortagger (dilution)
time resolution
(sketch of)
-log likelihood of all events:
If the data are compatible with oscillations, -log(L) has a dip at the right value of ms
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 67 ● 03:44:44 PM
and cpv
0.000
0.020
0.040
0.060
0.080
0.100
0.120
0.140
0.160
0 1 2 3 4 5 6
Int Luminosity
DG/G stat. Err. a
DG/G (J/psi phi)
DG/G (BsKK)
DG/G (BsKK) +SVT imp
0
0 .1
0 .2
0 .3
0 .4
0 .5
0 .6
0 1 2 3 4 5 6
Int. Lum i (fb-1 )
sigma S(Psi phi)
J/ps i ph i
J/ps i+Ps i'
PSI + TTT
ed2=0.07
sigm
a(de
ltaga
mm
a)
sigm
a(ph
i)
Aart Heijboer ● Observation of Bs oscilations ● Nikhef special seminar ● 01/08/07 ● slide 68 ● 03:44:44 PM
likelihood definition
the probability to measure an event as (un)mixed as a function of the measured decay time:
the probability that an event mixes (does not mix)as a function of the true decay time:
goodness of flavortagger (dilution)
time resolution
(sketch of)
-log likelihood of all events:
If the data are compatible with oscillations, -log(L) has a dip at the right value of ms