W mass reconstruction and jet calibration in ttbar

W mass reconstruction and jet calibration in ttbar events Wai Ling WU University of Michigan Advisors: Francesco SPANO, Columbia University Tancredi CARLI, CERN 1

Overview l l l Motivation of studying ttbar events Kinematics of W decay Comparison of jets made from towers and topoclusters 8 th August 2007 W. L. WU University of Michigan 2

Why study ttbar events? l l Top mass and W mass are well measured l Top mass ~170 Ge. V l W mass ~80 Ge. V Top abundantly produced at LHC (~1 per second at low luminosity = 1033 cm-2 s-1). Important background for most searches In the final years of commissioning and first year of data taking, serve as important calibration tool 8 th August 2007 W. L. WU University of Michigan 3

Kinematics of W decay W mass ~80 Ge. V d. Rp. T mean: 1. 990 Ge. V W Mean: 8 th August 2007 W. L. WU University of Michigan Excess momentum above W rest mass small Thus not much Lorentz boost for the quarks Therefore, ΔR large and using 0. 4 as cone size in jet algorithm is a reasonable measure 4

Comparison between jets made from towers and topoclusters (1) – the idea Model l Reality I have. . . l l Measurement generation level information and detector level information I can… l l 8 th August 2007 Adjust the model When real data comes, use the model to find reality W. L. WU University of Michigan 5

Comparison between jets made from towers and topoclusters (2) – jets and sample differences l l l Cells in hadronic calorimeter got hit and record the energy and momentum Both samples have the same events, but construct the “energy blobs” differently Towers – geometric algorithm Topoclusters – dynamic algorithm with noise supression Both use H 1 calibration 8 th August 2007 Hadronic Shower W. L. WU University of Michigan 6

Comparison between jets made from towers and topoclusters (3) – matching algorithm - Each quark is matched to all jets - The shortest distance is the match Example: Quark Jet Matching Distance 1 1 0. 0512714 - In case of both quark match to same jet, compare the matching distances and the shorter one is the match. The other quark uses the 2 nd shortest distance jet as match. 1 2 4. 1888058 1 3 2. 0939157 1 4 1. 1173507 2 1 0. 7396227 2 2 4. 5889246 - In this case, q 1 is matched to j 1 and q 2 is matched to j 4 2 3 1. 5489442 2 4 1. 2265931 - Cuts are made in the final stage as to what ΔR is chosen 8 th August 2007 W. L. WU University of Michigan 7

Comparison between jets made from towers and topoclusters (3) – matching algorithm Matching distance distribution for the 1 st quark and the 2 nd quark ΔR from 0. 2 – 1. 0 in steps of 0. 2 is chosen to be the matching radii for performance check 8 th August 2007 W. L. WU University of Michigan 8

Comparison between jets made from towers and topoclusters (4) – W mass is calculated from m = √(E 2 – p 2). Mean: 79. 35 Ge. V RMS: 23 Ge. V Mean: 80. 49 Ge. V Towers – only ΔR < 0. 4 plots are shown 8 th August 2007 W. L. WU University of Michigan 9

Comparison between jets made from towers and topoclusters (5) – W mass is calculated from m = √(E 2 – p 2). Mean: 79. 12 Ge. V RMS: 23 Ge. V Mean: 80. 47 Ge. V Topoclusters – only ΔR < 0. 4 plots are shown 8 th August 2007 W. L. WU University of Michigan 10

Comparison between jets made from towers and topoclusters (6) – difference in W mass l l Difference b/w W Mass from jets and quarks on an event by event basis gives a measure of the expected bias Fitting the distribution with Gaussians with range = (2*RMS, 0. 5*RMS), where the RMS is extracted from the histogram 8 th August 2007 W. L. WU University of Michigan towers mean: -4. 96 Ge. V sigma: 12. 88 Ge. V topoclusters mean: -5. 22 Ge. V sigma: 12. 78 Ge. V 11

Comparison between jets made from towers and topoclusters (7) – bias The bias performances of the two samples are very similar, with the topocluster sample giving a higher bias in reconstructing the W mass. Typical fractional difference in bias (take ΔR = 0. 4) is -5. 27 e-02. 8 th August 2007 W. L. WU University of Michigan 12

Many thanks to… l l l Dr. Francesco Spano, Dr. Tancredi Carli, the Columbia Nevis group CERN University of Michigan l l l Prof. Jean Krisch Prof. Homer Neal Dr. Steven Goldfarb Jeremy Herr Ford Motor Company National Science Foundation 8 th August 2007 W. L. WU University of Michigan 13

Questions? 8 th August 2007 W. L. WU University of Michigan 14