1. Progress on H/Abb -> 4b’s channel for the FTK physics case ~ 4jets trigger w/ and w/o FTK ~
Kohei Yorita
Young-Kee Kim
University of Chicago
@ the FTK Meeting on May 2nd, 2006

2. Trigger Rate Comparisons
> Requiring 4jets with different Pt cuts.
> Trigger Rate :
Sherpa 2to3 > Sherpa 2to2 > Pythia
> No bb production in Sherpa but gluon
splitting produces b quarks.
bb production : 0.48 mb with no pt cut
~ mb with pt>20 GeV,
< 1% w.r.t. dijet - Need more check.
> Use Sherpa 2to3, yt>25 as a default
- Better kinematics and conservative.
Sample
Xsec(mb)
NEV
Sherpa2to3, yt>20
0.827
~11M
Sherpa2to3, yt>25
0.361
Sherpa2to2, yt>20
0.776
~9.4M
Sherpa2to2, yt>25
0.340
~1M
PYTHIA dijet, pt>20
0.817
300K
1 btag
2 btag
0 btag
3 btag
4 btag

3. Atlfast Pt vs LVL1 8x8 Calo Et
Old : |eta| < 2.8, JETB
New : |eta| < 3.2,
against both JET and JETB
(option)
> Slope became more closer
to 1, and turn-on curve gets
more steeper.
In TDR, LVL1 Pt is set to GeV,
That corresponds to GeV 8x8
Calo Cluster Et, giving ~100Hz of
LVL1 output rate.
Due to yt>25 cut, rates in lower pt region
might not be true. (unlikely to have 2KHz
by 10 GeV of LVL1 8x8 Calo cluster Et.)
Not exactly sure what’s happening.
- need more investigation.
For this talk, everything is based on
Atlfast Pt.
TDR
4jets w/o btag

4. How Much Can We Lower the Pt Threshold at LVL1 ?
Assume that a realistic limit of
LVL1 output rate for our jet
trigger is a few KHz in the default
table. (total 40 -> 100 KHz)
This limits lowering threshold
rather than LVL2 reduction rate
with FTK.
LVL1 :
To give ~1KHz of LVL1 output,
Pt threshold for 4jets is ~40 GeV.
TDR : Pt>55-65 GeV, ~200Hz.
LVL2 :
In any case LVL2 output has to be
an order of ~10Hz.
TDR : Pt> GeV, ~10-20Hz.
So trying to keep 40GeV even
@ LVL2 and reduce LVL2 output
rate to ~10Hz by only requiring
btagging.
1 btag
2 btag
3 btag
4 btag

5. LVL2 Reduction by btagging
Due to contents of b quarks,
The reduction rate is not simple linear dependency
w.r.t. Ru (light jet rejection factor).
- Compare dotted lines with solid lines
Why is real reduction rate higher than slope of 1 ?
(1) 4jets with |eta| < 3.2
(2) 4jets with |eta| < 2.5 & N-btag
So eff. of N(eta<2.5)/N(eta<3.2) is included.
Denominator : N(eta<2.5)
* Need to check bb production.

6. Setting up working points for the cases w/ and w/o FTK
Conditions to be considered :
(1) Pt LVL1 can not be lowered below 40 GeV (limited by LVL1 output rate)
(2) Pt LVL2 can be changed by requiring btag.
But it has to give the same LVL2 rate as Pt>110 GeV w/o btag (TDR) A
~10Hz
LVL1
LVL2 A
4J : rate ~200Hz
4J110 no btag : output rate 9.6 Hz B
4J40 : rate ~1KHz
4J40 w/ 1btag, Ru=100 : output rate 13.4 Hz C
4J40 w/ 2btag, Ru=10 : output rate 6.5 Hz
w/o FTK B C
w/ FTK
LVL1 limit

7. Trigger Rate vs Signal Efficiency
4J110 no btag B
4J40 w/ 1btag, Ru=100 C
4J40 w/ 2btag, Ru=10
In each line from right to left, each point comes
from 20,30,40, …….. GeV Pt cut. 20
1 btag
2 btag 30 40 50
A(110) A
B(40)
C(40)
Signal acceptance is increased by a factor of by btagging,
while keeping LVL2 output rate around 10Hz !

8. Improvement at Trigger Level
4J110 no btag B
4J40 w/ 1btag, Ru=100 C
4J40 w/ 2btag, Ru=10
Btagging efficiency : 50%
@ L = 30 fb^-1 MA
Signal Eff. (%) S
B (LVL2 out rate)
S/B (%)
S/sqrt(S+B)
300
(39.9pb) A
0.24
2884
1.44e+8 (9.6Hz)
2.0e-3 B
4.51
53936
2.01e+8 (13.4Hz)
2.7e-2
3.80 C
2.02
24223
9.82e+7 (6.5Hz)
2.5e-2
2.44
500
(4.18pb)
1.12
1404
9.7e-4
0.12
9.67
12127
6.0e-3
0.86
4.24
5317
5.4e-3
0.54
700
(0.82pb)
2.86
704
4.9e-4
0.06
13.6
3349
1.7e-3
5.95
1464
1.5e-3
0.15
900
(0.21pb)
5.61
353
2.5e-4
0.03
16.6
1044
5.2e-3
0.07
7.28
458
4.7e-3
0.05

9. Optimization to see (Final) Improvements w/ and w/o FTK
Since optimization can be done in offline, (i.e. using offline quality btag),
Only difference between w/o and w/ FTK is 4th jet pt threshold (110 and 40 GeV)
- Looked at up to (300,200,120,120), 10 GeV step and up to 4 btags.
A (w/o FTK)
Eff_b = 50%, Eff_c = 10%, mistag 1%, Start from 4J110
B (w/ FTK)
Eff_b = 50%, Eff_c = 10%, mistag 1%, Start from 4J40
@ L = 30 fb^-1 MA
Optimized cuts S B
S/B (%)
S/sqrt(S+B)
300
A(140,140,110,110), +4b
62.9
13.9
451
7.17
B(130,70,70,40), +4b
1175
1352
86.9
23.4
500
A(160,140,110,110), +4b
29.9
12.9
231
4.6
B(160,120,70,40), +4b
251
489 51
9.2
700
A(220,130,110,110), +4b
11.4
9.9
115
2.5
B(290,200,40,40), +3b
200
1122
17.8
5.5
900
5.1
1.3
B(300,200,40,40), +3b 94
1043
9.0
2.8
Optimized by S/sqrt(S+B), not very stable due to low stat and tight cuts.

10. Very Quick Look w/o FTK w/ FTK *** Note *** This is very preliminary
& optimistic, because
there is no bb production
taken into account.
And also signal cross section
used is slightly larger than
TDR.
Need more study to say
something conclusive.
But relative improvement
looks very nice and promising.
700

11. Summary
> By requiring high quality btagging by FTK at LVL2, we can lower the pt
threshold to 40 from 110 GeV while keeping ~10Hz of LVL2 output rate
and ~1KHz of LVL1 output.
> At Trigger level, this gives significance improvement by a factor of 2-10.
> After optimization, we still do see good improvement with FTK. Statistical
significance gets better by a factor of 2-3 depending on MA.
Things to do :
> Should look at dijet mass distributions.
> Need more careful study on LVL1 parameterization.
> Need to look at bb production in Sherpa.
> Need Ru vs btag efficiency plot from realistic FTK simulation.
> Plan to look at full simulation for signal.
> Make conclusions :
- How much can tan be improved (lowered) in this channel ?
- Can we measure the mass of Higgs by having FTK ?
> Documentation.

12. With Other Final States
Obviously 4b channel
is not discovery channel
But need it for property !