Statistical analysis of LDMX data
(2023) FYSK03 20231Department of Physics
Particle and nuclear physics
- Abstract
- This thesis explores a statistical framework for analysing LDMX data. It uses pre-built functions in the ROOT framework to calculate likelihoods and confidence levels (CL). Data, background and signal were generated in Monte Carlo simulations with a loose veto on signal in the Hcal with simple systematic uncertainties (for the background and signal) created by varying this selection. The data sets consisted of a 4000 event background and 25 event signal (a strong signal). Four hypotheses were tested with dark photon masses of 1, 10, 100, and 1000 MeV. The pre-bult functions did not work as intended hence another method to find CL$_s$ values needs to be found. Signs of signals were found in three of four hypotheses with the most promising... (More)
- This thesis explores a statistical framework for analysing LDMX data. It uses pre-built functions in the ROOT framework to calculate likelihoods and confidence levels (CL). Data, background and signal were generated in Monte Carlo simulations with a loose veto on signal in the Hcal with simple systematic uncertainties (for the background and signal) created by varying this selection. The data sets consisted of a 4000 event background and 25 event signal (a strong signal). Four hypotheses were tested with dark photon masses of 1, 10, 100, and 1000 MeV. The pre-bult functions did not work as intended hence another method to find CL$_s$ values needs to be found. Signs of signals were found in three of four hypotheses with the most promising being the 1000 MeV followed by the 100 MeV which where observed to a 95\% CL, the 10 MeV was observed to a 68\% CL. Lastly, the 1 MeV signal was not observed at a 68\% CL. (Less)
- Popular Abstract
- You may be familiar with the standard model, the most successful theory in physics ever created and is the closest to a theory of everything that humanity has ever gotten. It describes the world in terms of 17 unique particles. These particles make up everything that we interact with every day, indeed you and me too. It may thus seem odd to claim that most matter is something different called dark matter.
From the best estimates of astrophysics, we know that dark matter makes up 85% of all matter. This is all good and well, but, you may ask, what is this dark matter? That is a brilliant question and a question that particle physicists have spent several decades on. Historically, the search for dark matter has focused on particles... (More) - You may be familiar with the standard model, the most successful theory in physics ever created and is the closest to a theory of everything that humanity has ever gotten. It describes the world in terms of 17 unique particles. These particles make up everything that we interact with every day, indeed you and me too. It may thus seem odd to claim that most matter is something different called dark matter.
From the best estimates of astrophysics, we know that dark matter makes up 85% of all matter. This is all good and well, but, you may ask, what is this dark matter? That is a brilliant question and a question that particle physicists have spent several decades on. Historically, the search for dark matter has focused on particles heavier than the mass of a proton. This may seem like nothing at all for something that is supposed to constitute 85% of all matter, but is quite heavy for a particle. This has led to lighter possible dark matter particles being neglected – experiments designed to find heavy particles are not suited to finding light particles. The Light Dark Matter eXperiment (LDMX) aims to rectify this gap in our search.
The LDMX experiment hopes to produce dark matter by shooting electrons at a block of tungsten where a process known as bremsstrahlung (breaking radiation in German) takes place creating an enormous amount of particles (radiation). The idea of bremsstrahlung is simple – think of how when you are driving a car and break, sound and heat are generated. Bremsstrahlung is the particle equivalent of the noise generated when breaking in your car. The only thing that then remains to do then is to observe the dark matter, which is easier said than done.
It is, in fact, it is reckoned to be so difficult to directly observe dark matter particles that it would be easier to observe all standard model particles instead and then infer the existence of the dark matter. Thinking back to our car breaking, imagine that you have to find how much heat is generated by only looking at the sound spectrum created. Now imagine that you have thousands of cars breaking at roughly the same time, all creating sound. This is what LDMX aims to do for an extremely difficult task, made harder by uncertainties in the experiment. Imagine that the sound spectrum from the cars breaking can be off by 10Hz or that the amplitudes in the spectrum are off. To combat this a statistical analysis is necessary. Essentially being that by looking at the data carefully enough we can determine if a result is significant – unlikely enough to happen by random chance that it can be said to be a discovery.
The hope is that with proper construction and implementation, LDMX will produce dark matter and detect it, consequently, widening humanity’s horizon by shedding light on what has been invisible to our understanding for the better part of a century. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9127478
- author
- Ritter, Noah LU
- supervisor
- organization
- course
- FYSK03 20231
- year
- 2023
- type
- M2 - Bachelor Degree
- subject
- keywords
- LDMX, dark matter, light dark matter, statistics, hypothesis testing, ROOT
- language
- English
- id
- 9127478
- date added to LUP
- 2023-06-20 09:07:22
- date last changed
- 2023-06-20 09:07:22
@misc{9127478,
abstract = {{This thesis explores a statistical framework for analysing LDMX data. It uses pre-built functions in the ROOT framework to calculate likelihoods and confidence levels (CL). Data, background and signal were generated in Monte Carlo simulations with a loose veto on signal in the Hcal with simple systematic uncertainties (for the background and signal) created by varying this selection. The data sets consisted of a 4000 event background and 25 event signal (a strong signal). Four hypotheses were tested with dark photon masses of 1, 10, 100, and 1000 MeV. The pre-bult functions did not work as intended hence another method to find CL$_s$ values needs to be found. Signs of signals were found in three of four hypotheses with the most promising being the 1000 MeV followed by the 100 MeV which where observed to a 95\% CL, the 10 MeV was observed to a 68\% CL. Lastly, the 1 MeV signal was not observed at a 68\% CL.}},
author = {{Ritter, Noah}},
language = {{eng}},
note = {{Student Paper}},
title = {{Statistical analysis of LDMX data}},
year = {{2023}},
}