HL-LHC Ian Lim September 15, 2017 About Me Stanford University, - PowerPoint PPT Presentation

Vertex Reconstruction at the HL-LHC Ian Lim September 15, 2017

About Me • Stanford University, B.S. in Physics with Honors (2017) • Interests in cats, baking, and high-energy physics • SULI intern, autumn 2017 (August 21-December 8) • Working with Maurice, Ben, and Simone on vertex reconstruction for the HL-LHC upgrade 2

How well do existing vertexing algorithms perform at large mu? And how can they be improved? 3

Plan of Attack 1. Intro to vertexing 2. The problem, in more detail 3. My studies so far 4. Future work 4

What is Vertexing? • Primary vertices are locations of proton-proton collisions in the detector • Two main parts – position reconstruction and track association • How well can we determine where a collision happened in space? • Given the tracks left in our detector by collision products, how well can we associate them to the correct vertex? 5

Definitions: Events, Hard Scatter, and Pile-Up • Today, protons are collided in bunches with a period of 25 ns. • Each bunch crossing (event) results in about 20-40 actual collisions ( μ ). • One special “hard scatter” vertex and many others (pile -up) • We can also categorize events by how well we reconstruct vertices. Image credit to Ariel Schwartzman, SLAC. Hard scatter tracks in red and pile-up tracks in blue. 6

Based on https://arxiv.org/pdf/1611.10235.pdf 7

Central problems of vertexing • At the HL-LHC, we expect μ ~ 200 – a tenfold increase! • With increased vertex density, performing a clean reconstruction becomes significantly harder. • Hard scatter is obscured by 10x more pile-up • More tracks to assign • Greater likelihood of merging • But the detector will be upgraded as well (e.g. pseudorapidity |η|< 4, vs. | η |<2.5 now) 8

Current approach (my work) • First, benchmark existing algorithms on simulated μ =200 data • Vertex reconstruction efficiency • Event classification and distribution • Next, determine new/alternate metrics for measuring performance • Bias introduced by misidentified tracks • Finally, implement and test new vertexing procedures. 9

Truth and reco vertices/event 10

Reconstruction efficiency + distribution 11

Event classification vs. μ 12

Vertex density and event classification 13

Hard scatter z-resolution 14

Spread of hard scatter contributions 15

Future work • Continuing to develop interesting metrics for vertexing • Working on improved vertexing methods • Use of high η tracks from the forward region? • Track clustering methods (cf. Meloni 2017, arXiv:1705.00022v1) • Vertex candidate substructure (cf. Schwartzman) 16

Thank you for your time! 17