Data Analysis and Machine Learning

Supervisors: A. Anelli, M. Borsato, M. Calvi, S.Capelli, D.Fazzini, M. Martinelli.
The proposed activities preferably require some knowledge of programming languages (Python, C++)

1. Fun with Particles!
   The LHCb experiment has already discovered more than 60 new particles. Many others may hide within its dataset that is still to be fully exploited. During this project you will have the opportunity to be the first to observe a new particle, determine its spin and parity and add a new item to the Particle Data Book!

2. Study of the reconstruction efficiency of the decay D⁰→K⁰ₛπ⁺π⁻ in the LHCb detector
   The determination of the reconstruction efficiency for the decay D⁰→K⁰ₛπ⁺π⁻ is crucial for the measurement of the D⁰-D̅⁰ mixing variables. Using real data samples collected by the LHCb detector during the 2016–2018 period and simulated data samples, efficiency maps on the Dalitz plot of the decay will be constructed, and their time dependence will be studied. The efficiency maps provided by the simulation will be compared to those determined from the data using Machine Learning algorithms.

3. Study of the decay components of D⁰→K⁰ₛπ⁺π⁻
   Many-body decays proceed to the final state through numerous intermediate states. Studying the relative abundance of these intermediate states is of great importance to understand the physical processes that generated them.
   To measure these quantities, fits to the two-dimensional distribution of the data (Dalitz plot) are performed. Several fitting tools have been developed for this purpose.
   This thesis project proposes to compare the results produced by two different software tools: one, a more mature C++-based framework, and the other, a recently developed Python-based tool that uses TensorFlow — an advanced machine learning platform developed by Google — as its fitting engine.

4. Measurement of the CKM matrix element |Vub| in B⁰ₛ→Kμν decays and applications of Machine Learning techniques.
   The mixing among quarks families is described in the Standard Model by the CKM matrix. The values of its elements can’t be predicted but should be measured. A precise measurement of the Vub element, representing b→u transitions, is possible by studying the B⁰ₛ→Kμν decays recorded by the LHCb experiment. The new data collected in the LHC Run 3, at a much higher Luminosity than in Run 2, will be studied, to prepare a new analysis. Possible topics for a thesis are 1) One important aspect of the measurement is the separation of the signal B⁰ₛ candidates from the background ones, which is done by studying the isolation of the candidates with respect to other charged and neutral particles in the event. 2) Another challenge is the indirect reconstruction of the B⁰ₛ meson momentum, since a fraction of momentum is lost with the neutrino that escapes detection. In both these two cases Machine Learning techniques will be employed and they will be validated by using simulated events and real data. 3) A third project concerns the study of the background due to partially-reconstructed B→J/ψ(μμ)K decays, where one of the muons is not reconstructed because it is outside the detector acceptance. Simulated data will be used to study a specific technique to reduce such background, making use of the muon momentum and the decay kinematic for the identification of the J/ψ resonance.

5. Search for a decay that violates the conservation of the lepton flavour  τ→μμμ with LHCb data.
   One of the Standard Model (SM) assumptions is the conservation of the lepton flavour in particle transitions. Even in SM extensions performed to accommodate neutrino oscillations, the rate for a decay like τ→μμμ would be so extremely low that an observation would provide a clear sign of presence of New Physics. The search of such a decay will be performed using data collected by the LHCb experiment in Run 3. One topic for a thesis concerns the identification of signal candidates using topological, kinematical and particle identification variables. Simulation and real data will be used to train Machine Learning algorithms.  Charmed meson decays Ds→Φ(μμ)π will also be studied both in data and in simulation, as a control channel to calibrate the signal selection.

6. CP-violation in radiative charm decays
   Radiative decays of charm quarks are extremely suppressed in the SM due to its peculiar flavour structure. They are considered a very promising candidate to probe the flavour structure of physics beyond the SM up to very high energy scales.
   The LHCb experiment at the LHC has collected a dataset of several billion charm mesons, the largest in the world. However, disentangling their rare radiative decays from the rest is extremely challenging. Our group has recently demonstrated that the best strategy consists of using photons that convert to electron-positron pairs and is leading the efforts to complete the first LHCb analysis of radiative charm decays.
   Currently, the main task involves developing a fit to the data to statistically subtract residual background contributions. The fit will be optimised and tested extensively on both real and simulated data. The results of the fit will be used to search for a matter-antimatter (CP) asymmetry in these decays.

Laboratory Activities

The Electromagnetic Calorimeter towards the Upgrade II of LHCb
Supervisors: M. Calvi, L. Martinazzoli, M. Pizzichemi, M.Salomoni
For the LHCb Upgrade II the electromagnetic calorimeter will be completely redesigned to sustain a much larger radiation dose and particle occupancy. The detector is required to measure electrons and photons with a time resolution of a few tens of picoseconds, a spatial resolution of a few millimeters and a similar energy resolution as the actual Shashlik calorimiter.
The thesis work consists in the evaluation of the performance of different calorimeter prototypes through the analysis of the data collected at test beams at SPS accelerator at CERN and Monte Carlo simulations.
A further thesis project is dedicated to the optimisation of the geometrical parameters of the calorimeter. By using a detailed Geant4 simlation, the energy, space and time resolutions of various configurations will be compared.

1. Study of the performance of different prototypes of Electromagnetic Calorimeter on test beams
   The thesis project is based on the analysis of experimental data collected on a test-beam performed at the SPS facility at CERN and at DESY, and their comparison between Monte Carlo simulations. Different prototypes of "Spaghetti Calorimenters" will be studied to evaluate their performance based on energy, time, spatial and angular resolution.

2. Optimization of Electromagnetic Calorimeter's performance with Monte Carlo simulations
   the student will optimize the geometric parameters of modules of electromagnetic calorimeters based on SPACAL technology. The energy, spatial and time resolutions will be analysed using a detailed MC simulation developed with the Geant4 toolkit.

3. Study of the Particle Identication (PID) capability of an Electromagnetic Calorimeter based on SPACAL technology
   Analysis of the capability of modules of an electromagnetic calorimeter to distinguish electrons, muons, charged pions and charged kaons, according to the different profile of eneergy deposition, both in the particles' flight direction and in the transverse plane. The study will be performed on a detailed simulation developed with the Geant4 toolkit.

Data Analysis

Supervisors: M. Borsato, M. Calvi, Sim. Capelli, P. Carniti, D. Fazzini, C. Gotti, V. Kirsebom, M. Martinelli, G. Pessina, M. Pizzichemi, M.Salomoni, D.Trotta

1. Search for lepton flavor violating decay τ→μμμ

The conservation of lepton flavor in particle transitions is one of the assumptions of the Standard Model. Even in its extensions that include massive neutrinos and neutrino oscillations, the rate of this decay is so low that its eventual observation would be a clear signal of physics beyond the Standard Model. The thesis proposes to initiate the search for this decay using data collected by LHCb during Run 3. The signal selection will be defined using kinematic and topological variables with Machine Learning techniques. Particular attention will be devoted to optimizing the correct identification of muons, based on information from all sub-detectors. The decay of the charmed meson Ds→ϕ(μμ)π will be used as a control channel by comparing its properties in data and simulation.

2. Measurement of the matrix element Vub from the decay Bₛ →Kμν

The CKM matrix representing quark mixing is strongly diagonal, and the coupling of the b quark to the u quark is highly suppressed. Measuring the magnitude of the Vub element is therefore particularly difficult, and improving the precision of currently available results is important. The use of the decay Bₛ →Kμν, accessible only at LHCb, is particularly advantageous due to the possibility of using precise QCD calculations for its estimation. The thesis proposes to study this decay with Run 2 data for the measurement of the differential decay rate. The branching ratio of the decay will be measured in different q² bins, and various models for describing the decay rate will be explored, using different form factor estimates obtained from Lattice QCD calculations and different data fitting strategies. Possible stays at CERN to collaborate with theoretical experts in the field could be funded through the Erasmus+ program.

3. Measurement of mixing and CP violation in multi-body D⁰ decays

Following the recent observation of direct CP violation in D⁰→KK and D⁰→ππ decays, experimental interest has shifted towards the search for CP violation mediated by the oscillation process between D⁰ and D̅⁰ mesons. Observing such a phenomenon, besides clarifying the theoretical landscape of CP violation in charm decays, could also provide hints of possible effects beyond the Standard Model. To carry out this research, the thesis proposes to study the D⁰→Kₛπ⁺π⁻ decay using the “bin-flip” technique, which first enabled the observation of the mass difference between the D⁰ eigenstates. This measurement has revealed systematic effects that could limit future precision in upcoming LHC runs. To correct these effects, the thesis will explore a correction technique for the simulated sample based on machine learning methods and using D⁰→π⁺π⁻ as a control sample.

4. Measurement of CP violation in D⁰→K⁺π⁻π⁺π⁻ decays

The doubly Cabibbo-suppressed decay D⁰→K⁺π⁻π⁺π⁻ offers an unexplored territory for the search for effects beyond the Standard Model. In fact, this is a channel where CP violation effects are expected to be null in the Standard Model, and their observation would provide clear evidence of exotic processes. The thesis proposes to perform the first measurement of CP violation in this channel using the triple product technique. The work will involve defining the optimal event selection (also using Machine Learning techniques), fitting the model to the data, and evaluating the sensitivity of the measurement using real data collected by the LHCb experiment.

5. CP violation in radiative charm decays

Radiative decays of charm quarks are extremely suppressed in the SM due to its peculiar flavour structure. They are considered a very promising candidate to probe the flavour structure of physics beyond the SM up to very high energy scales. The LHCb experiment at the LHC has collected a dataset of several billion charm mesons, the largest in the world. However, disentangling their rare radiative decays from the rest is extremely challenging. Our group has recently demonstrated that the best strategy consists of using photons that convert to electron-positron pairs and is leading the efforts to complete the first LHCb analysis of radiative charm decays. Currently, the main task involves developing a fit to the data to statistically subtract residual background contributions. The fit will be optimised and tested extensively on both real and simulated data. The results of the fit will be used to search for a matter-antimatter (CP) asymmetry in these decays.

Innovative Software Techniques for Particle Reconstruction

1. New strategies for signal identification using Graph Neural Networks

The upcoming LHC runs will bring significant increases in luminosity, pushing the limits of the trigger, selection, and analysis of events collected by the LHCb detector. A new AI-based strategy has been devised using Graph Neural Networks (GNNs) to perform a Deep Full Event Interpretation (DFEI) of events. The algorithm aims to reconstruct the decay chains of all beauty hadrons present in the event, providing their full hierarchical structure. It can be used at the trigger level for the inclusive selection of interesting events and for determining the sub-sample of particles to be saved. A second use of DFEI is in the offline selection of signal candidates and the rejection of background candidates. Background candidates in standard selection algorithms often come from multi-track decay chains that are only partially reconstructed or from random combinations of tracks originating from different hadrons.
This project has two goals:
A) To explore the use of DFEI to recognize such background types by reconstructing the full decay chains of heavy hadrons. The algorithm will be used for the selection of semileptonic decays of beauty hadrons and compared with other isolation strategies.
B) To add new functionalities to the algorithm, extending it to the selection of neutral particles and charm particles, retraining and re-optimizing the GNN for use at the trigger level.

2. Implementation and development of a tracking reconstruction algorithm based on RETINA in a state-of-the-art FPGA array

Field Programmable Gate Arrays (FPGAs) are programmable boards capable of executing specific algorithms with high speed and efficiency. One algorithm particularly suited for this technology is RETINA, inspired by how neurons process retinal information to reconstruct images.
It has been demonstrated that similar algorithms are highly effective in reconstructing particle tracks at very fast rates, making them valuable for trigger applications in High Energy Physics experiments. LHCb is already using RETINA for clustering in its vertex detector and aims to extend its use to the downstream tracking stations, enhancing the experiment’s physics capabilities for Run 4 and beyond.
A test-bed is currently operational at CERN to analyze LHCb Run 3 data in parallel. Furthermore, a mini-array of FPGAs will soon be installed at the University of Milano-Bicocca, enabling further studies using technology optimized for Run 4 data-taking.
This project offers a unique opportunity to contribute from the very beginning and gain hands-on experience in a rapidly evolving field at the intersection of artificial intelligence and high-performance computing.

3. Development of Innovative Techniques for Cherenkov Photon Reconstruction in the LHCb RICH

Charged particles produced in High-Energy Physics experiments emit photons via the Cherenkov effect when passing through a radiator material. The emission angle of these photons relative to the particle’s trajectory depends on the particle’s mass. In RICH (Ring Imaging CHerenkov) detectors, these photons are focused by a system of mirrors onto a plane of photodetectors, forming rings whose radii are determined by the emission angle. Reconstructing these rings enables the identification of the originating particle.
The goal of this thesis is to develop a Machine Learning-based algorithm for Cherenkov ring reconstruction. Data and simulations from the LHCb experiment will be used to evaluate the algorithm’s performance, comparing it to those currently used within the collaboration. Cutting-edge Nvidia A100 and A6000 GPUs will be used for development and testing.

Particle Detectors for LHCb

1. Study of radiation resistance and timing resolution of Silicon Photomultipliers (SiPMs) irradiated with neutrons and cooled with liquid nitrogen
   Supervisors: M. Calvi, P. Carniti, C. Gotti, G. Pessina, D. Trotta

In the coming years, the LHCb RICH detectors will be upgraded to operate under higher luminosity conditions, enabling the collection of more data. The current photodetectors will be replaced by Silicon Photomultipliers (SiPMs), solid-state devices sensitive to single photons. In LHCb, these photodetectors will have to survive high doses of radiation, particularly high neutron fluences (over 10¹³ n/cm²). To mitigate radiation effects, SiPMs will be cooled to very low temperatures, down to liquid nitrogen levels (77 K).
This thesis will focus on studying the single-photon response of SiPMs irradiated with neutrons and cooled to cryogenic temperatures. Key quantities to be characterized include their timing resolution, that is, the ability to determine the arrival time of Cherenkov photons with precision at the level of tens or hundreds of picoseconds.

2. Study of the performance of various electromagnetic calorimeter prototypes from Monte Carlo simulations and beam tests
   Supervisors: M. Calvi, A. Minotti, M. Salomoni, M. Pizzichemi

For the LHCb Upgrade II, the electromagnetic calorimeter will be completely redesigned to cope with a significant increase in radiation dose and occupancy. The detector must provide measurements of electrons and photons with a timing resolution of a few tens of picoseconds, a spatial resolution of a few millimeters, and an energy resolution compatible with the current Shashlik technology. For this reason, several prototypes of electromagnetic calorimeters using SPACAL (SPAghetti CALorimeter) technology are under development.
The thesis work will involve analyzing experimental data collected from beam tests at the SPS at CERN and at DESY, and comparing the results with those obtained from Monte Carlo simulations developed with the Geant4 toolkit. Various prototypes based on SPACAL technology will be studied, evaluating their performance in terms of energy, timing, spatial, and angular resolution, and simulating different configurations.
In preparation for the next phase of the calorimeter upgrade, the performance of the first SpaCal modules produced for Run 4 of LHCb will also be evaluated, with particular attention to energy resolution, by collecting and analyzing data through beam tests.
Stays at CERN are possible to participate in the experimental activities and beam measurements of the prototypes.

CERN offers every year some fellowships to host university students that are interested in High Energy Physics research. The fellowships are granted for a period of a few weeks and the activities take place during the summer. The requirements for the participation are reference letters, a curriculum vitae and the results of the exams. The deadline of the call is on the 31th January each year.

The University of Milano Bicocca offers a doctoral course in Physics and Astronomy. Access to the course is limited to the students who successfully pass a public selection whose call is usually made in April/May at this address. Every year PhD fellowships are offered by both the University of Milano Bicocca and INFN. Typically, one fellowship is reserved for candidates with a foreign degree equivalent to the Italian master's degree "Laurea Magistrale". A number of fellowships are assigned on projects that are freely chosen by the students, depending on the ranking. Other fellowships are assigned to specific research projects.

⚠️ Call (Session I - deadline 15/05/2024)

INFN Fellowship for non-Italian Researchers

INFN offers every year about 20 post-doctoral fellowships for NON-Italian researchers. Each fellowship is granted for 12 months and is renewable for a further 12 months. The call usually opens in September and its deadline is in November. The participation requirements are a plan of the research activity to be done in one of the laboratories and sections of INFN and three reference letters. The section of Milano Bicocca is among those where the research activity can be undertaken. Please get in touch with us for help devising the research plan.

⚠️ 2024 Call - deadline 15/11/2024

University of Milano Bicocca Postdoctoral Fellowships

The Physics department of the University of Milano Bicocca offers post-doctoral fellowship with a duration of 24 months renewable for a further 24 months to carry out research activities. The call generally has a deadline in November. To participate a research plan is required. The successful candidates are selected after an interview with a commission appointed by the Physics department.

⚠️ 2023 Call - deadline 23/11/2023

Individual Marie Curie Fellowship

The European Commission, within the scope of the activities to support the research and the movement of talents, offers some individual fellowships to perform your own research activity in foreign institutes with respect to the one of origin. The University of Milano Bicocca is recognised as a hosting institute. The Department of Physics has already hosted many researchers in the past and in particular in our group. The deadline for this call is usually in the month of September.

⚠️ 2024 Call - deadline 11/11/2024

Cariplo foundation Fellowship

The Cariplo Foundation offers some individual fellowships for early-career researchers "Giovani Ricercatori" (no more than 2 years after PhD). The grant has a budget of up to 200k euros for a project of up to 36 months.

⚠️ 2024 Call - deadline 22/03/2024