Determining the polarization vector at a specific point in AGS and EIC

• Frank Rathmann (Brookhaven National Laboratory)
• Zhengqiao Zhang

Room: Virtual EC/NPP - Cross Directorate 1. Proposal title: Determining the polarization vector at a specific point in AGS and EIC 2. PI(s): Frank Rathmann and Zhengqiao Zhang 3. Abstract: At RHIC, the calibration of absolute polarization relies on elastic proton scattering from a polarized hydrogen gas jet target in the CNI region. To enhance the polarimetry capabilities for the hadron storage ring (HSR) at EIC, an upgrade to the existing system is suggested that aims to determine the Cartesian components of the polarization vector at the specific location of the polarized target within the HSR. The successful implementation of such an upgrade will provide valuable insights into the spin dynamics within the ring, allowing for the experimental exploration of the specific orientation of the spin vector in that location. In addition, the proposed upgrade will lead to greatly reduced systematic uncertainties in the beam polarization components, since the unwanted spin components in the beam will be determined directly, rather than indirectly by maximizing the wanted ones, as is currently the case. Similarly, such a system, when implemented in the AGS, will allow not only absolute polarization calibration but also the determination of the polarization vector at a singular point in the AGS, thereby reducing the systematic errors of the unwanted polarization components and thus providing a new tool to improve polarization transmission over the entire AGS ramp. The project's objective is to assess the feasibility of such an enhancement and identify necessary modifications to the polarized target setup. The proposed work, to be conducted by a postdoc over a two-year period, will yield scenarios for AGS and HSR to achieve the goal. These scenarios will involve spin tracking of atoms in the HJET and simulation and calculations for an optimized detector system.

Uncovering New Laws of Nature at the EIC

• Hooman Davoudiasl (Brookhaven National Laboratory)

Room: Virtual PO Primary Investigator: Hooman Davoudiasl Other Investigators: Sally Dawson, Robert Szafron Abstract: Open fundamental questions - such as the origin of ordinary matter, or the identity of the invisible substance that dominates galactic structures, i.e. dark matter - strongly motivate new particles and interactions. In this proposal, we point out that the Electron Ion Collider (EIC) can not only provide a powerful tool for studying the structure of atomic nuclei and their constituents, as originally envisioned, but also great opportunities to probe and possibly discover new physics that goes beyond the standard paradigm, potentially leading to a revolution in our understanding of the laws that govern the Universe.

Exploring the Universality of Gluon Saturation at RHIC and the EIC

• Xiaoxuan Chu (BNL)

Room: Virtual PO

Novel radiation-hard fiber optic sensors for environmental and radiation-dose monitoring

• Ketevi Adikle Assamagan (BNL)

Room: Virtual PO PI: Ketevi A. Assamagan (BNL) Co-PI: Elke Aschenauer (BNL) Co-PI: Simon Connell (University of Johannesburg)

High bandwidth, high resolution beam telescope system for advanced particle detectors R&D

• Haider Abidi (CERN)
• Shaochun Tang (BNL)

Room: Virtual PO PI: Haider Abidi, Shaochun Tang OTHER INVESTIGATORS: Eric Buschmann and Hucheng Chen

Real-time learning on heterogeneous devices for detector calibration

• Yihui Ren
• Haider Abidi (CERN)

Room: Virtual PO/NPP - DS/CSI - Cross Directorate PI(s): Haider Abidi (PO/NPP), Yihui Ren (CSI) - Ray Abstract: The goal of this project is to leverage machine-learning inspired auto-differentiation techniques to create a system that can perform real-time (100ns ~ 1s) detector calibration and response correction using heterogeneous devices. Data collected at detectors requires dedicated calibration pre-processes before it can be used. These computationally expensive processes are used to correct for time-dependent effects such as a detector moving in physical space or its response changing with temperature. This project proposes an real-time low-computation alternative approach by implementing an adaptive self-learning, on-the-fly system using heterogeneous system-on-chip devices. The programmable logic will allow the system to connect directly to the detector trigger system, while the optimized engines and AI cores will provide the low latency computation to detect reference physics processes and perform the necessary calibration steps. This information can then be used in real-time to update the detector response. This system will serve as a proof-of-principle that can be used to perform real-time track alignment for beam telescope systems and beam monitor detectors, and can be extended to perform track alignment or calorimeter calibration at trigger level for large scale physics detectors such as the ATLAS, sPHENIX, EIC and future Higgs factory detectors.

New Physics searches with quantum tomography at the LHC

• Robert Szafron (Brookhaven National Laboratory)

Room: Virtual PO PI: Robert Szafron Other Investigators: Viviana Cavaliere, Sally Dawson, Marc-André Pleier, Alessandro Tricoli Abstract: The LHC data analysis is limited by the uncertainties and inefficiencies of the conventional methods of particle detection and reconstruction. Quantum tomography is a novel technique that uses quantum correlations and entanglement to reconstruct the quantum state of a system from a set of measurements. Quantum tomography has been successfully applied to various quantum systems, such as photons, atoms, and ions, but its potential for high-energy physics has not been fully explored. In this project, we propose to use quantum tomography to enhance the sensitivity and precision of the LHC experiments and to search for new physics signatures that may otherwise be missed by the standard techniques. We will develop theoretical models and experimental methods for quantum tomography of the LHC collisions. We expect that quantum tomography will provide a new perspective and a powerful tool for the LHC physics program and will open new avenues for the discovery of new physics at the energy frontier.

Vector meson production at the LHC (on the way to the EIC)

• Peter Steinberg (BNL)
• Peter Steinberg (BNL)
• Zhoudunming Tu (BNL)

Room: Virtual PO PI: Peter Steinberg Co PI: Zhoudunming Tu Description of Project: One of the key scientific topics of the upcoming Electron-Ion Collider (EIC) at BNL is the exclusive production of coherent vector mesons off of nuclei, which are sensitive to both the nuclear gluon momentum and spatial distributions. To advance this physics before the EIC era, it is proposed to study vector meson (particularly J/psi) production at the LHC using ultraperipheral data from the ATLAS experiment taken during the Run 3 heavy ion periods. As of 2023, a first dataset has now been acquired with a newly commissioned low-multiplicity trigger, and data and MC studies are underway. First discussions have already begun on how to optimize the program for a longer Pb+Pb run, which may occur in 2025 (or even late 2024). A postdoc dedicated to this dataset (and also working part time on ePIC) would significantly advance the data analysis, and make a major contribution to first ATLAS results on this topic. They could also contribute to the design of the trigger and analysis strategy for the next long ion run. Results from this analysis would seed further phenomenology pertinent to EIC physics and more generally improve our understanding of how vector mesons can be used to elucidate nPDF and saturation physics. Expected Results: Given the immediate availability of a high quality dataset, the first part of the postdoc’s time would be spent developing the analysis framework, filling in needed simulation studies, and making a major impact on the data analysis effort. A publication should be feasible by late 2024 or early 2025, if not before. At the same time, the postdoc will take responsibility for optimizing a trigger strategy for the longer heavy ion run expected for late 2025, with a view toward a fast publication based on the first paper, with the full Run 3 dataset.

Ion Beam Generation by Picosecond Laser

• Sergey Kondrashev (BNL)

Room: Virtual AD PI: Dr. Sergey Kondrashev (C-AD) Co-PI: Dr. Takeshi Kanesue (C-AD) We propose to study ion beam generation by 5 mJ/5 ps 10 kHz rep-rate laser available in Ion Source Group of C-AD. Such laser was recently delivered to BNL and commissioned on-site. The primary goal of this project is to develop a source of singly charged ions of any solid element of Periodic Table suitable for “slow” and, possibly, “fast” injection into Electron Beam Ion Source (EBIS). We will measure ion yields generated by such ps-laser for different target elements and target irradiation conditions (laser spot size, target irradiation angle, target translation speed). These data will allow us to specify and optimize ion source geometry and parameters. We will couple optimized source of singly charged ions to isotope separator previously developed in our group. If proved to be successful, this method will allow us to enhance output of ion beams of highly charged ions generated by EBIS for different solid-state elements with 2 or more stable isotopes. This enhancement would make it possible to feed any nucleus and provide the high versatility that may be required in the Electron Ion Collider (EIC). Taking advantage of high rep-rate of the available ps-laser, we will also investigate generation of quasi cw beams of highly charged ions which can be of interest for different applications. In particular, we will study the possibility of generation of intense quasi cw beam of lithium 3+ ions. Such beam can be used as a driver for compact neutron generator. If successful, this will be the first realization of quasi cw laser source of highly charged ions.

Detector design for a far forward liquid argon TPC at the high luminosity LHC

• Milind Diwan (BNL)

Room: Virtual PO/ATRO - Cross Directorate

DEMP as a probe to the 𝚲 hyperon polarization puzzle

• Zhoudunming Tu (BNL)

Room: Virtual PO PI: Zhoudunming Tu (Kong) co-PIs: Francesco Bossu (IRFU, Saclay), Abhay Deshpande (SBU/BNL), Wenliang Li (CFNS/SBU), Silvia Niccolai (IJCLab, France) Abstract: In the 1970s, an unexpected transverse Λ polarization in unpolarized proton-Beryllium collisions was discovered, which initiated extensive studies on spin phenomena in high-energy physics over the past five decades. Despite the emergence of numerous promising theoretical models, the foundational mechanism driving this polarization phenomenon remains elusive to this day. Notably, in both longitudinally and transversely target-polarized lepton-hadron and hadron-hadron collisions, the Λ hyperon exhibits no discernible polarization relative to the initial parton spin direction. Unraveling the mystery of how the Λ hyperon obtains its spin has evolved into one of the most important questions in addressing this puzzle. A recent experimental proposal aims to answer this question with a distinct and unequivocal approach, employing an exclusive channel known as Deep Exclusive Meson Production (DEMP). In this LDRD proposal, we plan to make the first Λ polarization measurement in DEMP based on the newly-recorded data with longitudinally polarized target from Run Group C at CLAS12 Collaboration. We will also perform a quantitative feasibility study based on the EIC second detector and its far-forward detector subsystem, where the energy range is 100 times higher. The anticipated result at CLAS12 will provide instrumental constraint, if not a definitive answer, to the puzzle of Λ hyperon polarization. This will open a new set of physics measurements at the EIC to further understand the nonperturbative QCD.

Understanding Single Event Burnout to empower future silicon detectors

• Gabriele D'Amen (Brookhaven National Laboratory (US))

Room: Virtual PO - Not Cross Directorate (per Gabriele D'Amen) PI: Gabriele D'Amen (Physics Department) Additional investigators:       Gabriele Giacomini (Instrumentation Division)       Alessandro Tricoli (Physics Department) Abstract: "Silicon sensors with gain are employed in multiple applications such as High Energy Physics, Nuclear science, and many other fields. Often these sensors are required to be able to withstand enormous amount of radiation while maintaining acceptable performances. Over the years, much attention has been focused on the improvement of the radiation hardness of these devices, and on the mitigation of permanent damages. Specifically, particles interacting with highly biased sensors can produce irreversible damages known as Single Event Burnouts (SEBs). This effect represents one of the main limitations of silicon technology in high fluence scenarios. The current consensus is that SEB events are more likely when the particle deposits high amount of energy in the interaction with silicon. We plan to expose irradiated silicon sensors to a beam of 29 MeV protons produced at the BNL Tandem Van de Graaff accelerator. Protons of this energy deposit a high amount of energy in silicon in their interaction, potentially increasing the probability of SEBs. This work will shed a light into the mechanism of SEB and permanent radiation damages in general and drive the evolution of future silicon detectors in high-radiation environments."

Risk Mitigation for DUNE AI/ML

• Haiwang Yu (Brookhaven National Laboratory)

Room: Virtual PO/CSI - Cross Directorate PI: Haiwang Yu (PI, PO), Yihui Ren (co-PI, CSI) Other Investigators: Brett Viren (PO), Chao Zhang (PO), Yi Huang (CSI) Abstract: We propose an efficient, automated Human-Instructed-Machine-Learning method (InstructLArML) to reduce the systematic uncertainties introduced by AI/ML algorithms in DUNE analysis. To increase physics sensitivities, DUNE plans to use AI/ML algorithms extensively. AI/ML algorithms exhibit superior performance in a range of tasks when trained and tested on simulation. However, given the persistence of data-simulation-discrepancies (DSD), the powerful feature extraction of AI/ML raises substantial concerns about systematic uncertainties in rigorous physics analyses. These DSD related AI/ML systematic uncertainties pose significant risk to DUNE’s physics goals. To fully address these concerns and to maximize DUNE’s physics potential, we plan to build a system on three foundational blocks, (1) a method to estimate DSD related systematic uncertainties (Risk Estimation); (2) a method to build AI/ML algorithms with reduced DSD related systematic uncertainties (Risk Mitigation); (3) a method to understand and reduce DSD (Simulation Augmentation). A pilot study on (1) was conducted in another LDRD project and can be adopted for DUNE. We plan to study (2) in this proposal. So that a potential follow-up ECA can study (3) and then build a Responsible AI System for DUNE based on them. The proposed InstructLArML method in this LDRD serves as a Risk Mitigation approach by embedding human knowledge into AI/ML models. The underlying hypothesis is that a small dataset labeled by human experts can effectively fine-tune a pre-trained large AI/ML model to significantly increase its robustness against DSD. It is designed based on two key AI/ML techniques: Latent Representation Learning (LRL) and Active Learning (AL). Where we expect LRL can prevent catastrophic forgetting, and AL can significantly reduce the statistics needed for expensive human labeling. This design is inspired by (1) the human feedback method used in training InstructGPT, and (2) the human expert hand-scan of real data events, which leads to the advanced Wire-Cell reconstruction paradigm. The proposed R&D includes: (1) Build the InstructLArML with three components, an encoder, a value estimator, and an uncertainty estimator, for a neutrino tagging task; (2) Simulate multiple datasets with variations in flux, generator, detector response, noise models, etc. One with large statistics serves as “simulation”, others with smaller statistics as “data”; (3) Test the effectiveness of the uncertainty estimator (AL aspect of this proposal); (4) Test whether the fine-tuning works with a much smaller human-labeled dataset (LRL aspect); (5) Summarize key results and conclusions. While developing the Wire-Cell reconstruction for MicroBooNE, we accumulated expertise in the hand scan of real data events and method validations. During the development of the LS4GAN LDRD project, we gained experience in detecting data-simulation differences and understanding the systematic uncertainties caused by those differences. Our CSI team members have a good understanding of AI/ML Uncertainty Quantification (UQ) with previous projects.

A Novel Prompt Signal Processing System for DUNE

• Michael Kirby (Brookhaven National Lab)

Room: Virtual PO PI: Michael Kirby The goal of this proposal is to develop prompt processing algorithms and infrastructure for the DUNE raw data stream to enhance the physics capabilities of the experiment. Because of the location of the Far Detector deep underground, raw data recorded by¬ DUNE consists of noisy current time-series data induced by drifting electrons. For most use cases, the first steps of the data analysis are noise removal and the reconstruction of electrons. This reduces the data volume by roughly two to three orders of magnitude (x100 or x1000). These algorithms need to be able to dynamically remove noise, adapt to detector configuration, and operate at multiple energy scales for region of interest selection of beam neutrino interactions, atmospheric neutrino interactions, and lower energy searches. This processing would offer two primary advantages: Firstly, the short turn-around time between recording and analyzing the data allows to spot problems early and potentially take countermeasures. Secondly, the resulting data will have a drastically smaller footprint on storage, minimizing the number of costly read-backs from tape and reducing the cost of active storage for analysis. This smaller data volume will enable rapid development of novel techniques of event reconstruction and identification. The research will study the deployment of the BNL-lead WireCell Toolkit algorithms on distributed computing (both HTC and HPC) to determine the optimal resource configuration, and model the interplay between data volume reduction, the downstream event data model, and the improvement in physics capabilities.

Currents Across Time: Quantum Computing for EIC and muon g − 2

• Raza Sufian (Brookhaven National Laboratory)
• Swagato Mukherjee (Brookhaven National Laboratory)

Room: Virtual PO/CSI - Cross Directorate PI: Swagato Mukherjee (NT, BNL). Co-PIs: Raza Sufian (RBRC, BNL), Taku Izubuchi (HET/RBRC, BNL), Kwangmin Yu (CSI, BNL), Kazuki Ikeda (SBU). Examining correlations among electromagnetic currents at different times has the potential to tackle crucial nuclear and high-energy physics questions beyond the capabilities of traditional lattice field theory simulations. A prime example is understanding scattering cross sections in the Electron Ion Collider (EIC), where the hadronic structure of emitted photons hinges on correlations in electromagnetic currents at distinct times. These unequal-time correlations also determine the hadronic contributions to the vacuum polarization and light-by-light scattering amplitude in the time-like region. These are relevant to the search for beyond Standard Model physics through measurements of the muon’s anomalous magnetic moment (g−2). The same quantum chromodynamics (QCD) corrections also are integral parts of the di-lepton production and light-by-light scatterings measured in the ultra-peripheral heavy-ion collisions, such as those at the Relativistic Heavy-Ion Collider (RHIC). Quantum computing offers a unique avenue to understand the dynamics of electromagnetic currents at a large scale. This proposal aims to assess and demonstrate the feasibility of simulating the time evolution of electromagnetic currents on quantum computers. To enhance practicality, the study focuses on lower-dimensional quantum field theory simulations. We also will explore machine learning techniques to encode essential physical degrees of freedom within a reduced number of qubits and shorter-depth quantum circuits, aligning with current quantum device capabilities. This interdisciplinary initiative involves participants from BNL’s Nuclear and High Energy Theory groups, Computational Science Initiative, RIKEN-BNL Research Center, and Stony Brook University. Success in this endeavor could pioneer a transformative approach, providing answers to pivotal scientific questions aligned with BNL’s multi-disciplinary core missions.

Improving Optical Modeling and Reconstruction for DUNE Phase-II

• Jay Hyun Jo (Brookhaven National Laboratory)

Room: Virtual PO Early Career Award (ECA) eligible Abstract: The Deep Underground Neutrino Experiment (DUNE) is the flagship particle physics experiment in the U.S. With Liquid Argon Time Projection Chamber (LArTPC)'s excellent particle tracking and energy calorimetry capabilities, DUNE aims to answer many fundamental questions about the Universe such as why matter dominates antimatter. The Phase-II far detector modules of DUNE, labeled FD3/4, are essential for achieving DUNE’s primary physics goal: attaining a sensitivity to CP violation exceeding 3σ across over 75% of the range of possible values of the unknown 𝛿CP by 2035. With anticipated endorsement from the Particle Physics Project Prioritization Panel (P5) this December, the research and development of FD3/4 emerge as pivotal undertakings for the neutrino physics community. The design for the Phase-2 FD3/4 detectors remains under development, offering opportunities for the integration of next-generation LArTPC detectors that promise enhanced physics capabilities and more cost-effective construction methods. The proposal aims to enhance optical modeling and reconstruction within DUNE's FD3/4 detectors. A significant element of a proposed FD3/4 design is expanding light coverage to around 75%, improving DUNE physics sensitivity as well as expanding its physics reach. However, in order to fully leverage the increased light information from the light-enhanced FD3/4, numerous novel studies in lesser-known optical transportation models as the optical properties of the liquid argon have significant uncertainty. The abundance of light signals in the FD3/4 facilitates the enhancement of optical modeling using calibration techniques that utilize charge signals, and implementing these optical simulations will further improve event reconstruction by effectively leveraging the light signal. The results of this project will inform the configuration optimization for the FD3/4 photon detector system, making the proposed light enhanced FD3/4 design more competitive and compelling, by improving physics sensitivity and reducing costs.

Use of high performance computing for finite element analysis of interaction region magnets

• Racquel Lovelace

Room: Virtual EC PI: Racquel Lovelace Abstract: In this proposal we will mitigate risks of a novel superconducting cable winding pattern for interaction region (IR) magnets using High Performance Computing (HPC). Modern colliders, such as EIC, push the limits of conventional magnet technology. This drives the need for modern magnet technology approaches, which are more challenging to simulate and push limits of conventional computers. Using HPC allows us to perform simulations in sufficient detail for a key magnet technology called Canted Cosine Theta (CCT), which has planned uses in modern colliders such as EIC and FCC-ee. We will employ a standard commercial finite element program, COMSOL, which is commonly used for analyzing magnets. 3D simulations with coupled Multiphysics models are limited to RAM size in typical workstations, which will be overcome by HPC. This project aims to demonstrate the feasibility and practicality of this approach by successfully simulating one EIC IR magnet in a coupled Multiphysics 3D simulation.

Semantic Data Compression to Enable Advanced Deep Learning Methods for LArTPC Detectors

• Dmitrii Torbunov (BNL)

Room: Virtual PO/CSI - Cross Directorate PI: Dmitrii Torbunov <dtorbunov@bnl.gov> co-PI: Dossay Oryspayev <doryspaye@bnl.gov> Abstract: Deep Learning (DL)/Artificial Intelligence (AI) methods are finding widespread applications in multiple areas of science and technology. They allow to significantly reduce costs of data analysis, improve its accuracy, and facilitate scientific discovery. The application of DL/AI methods to Liquid Argon Time Projection Chamber (LArTPC) Detectors, crucial in the upcoming DUNE neutrino experiment, holds immense potential. However, the direct application of the state-of-the-art (SotA) DL/AI methods to LArTPC data analysis is met with multiple challenges. One of the largest problems is related to the sheer size of LArTPC images, typically reaching dimensions of 10,000x10,000 pixels, far surpassing the typical image sizes (up to 1000x1000 pixels) used in DL/AI model development. This renders the direct application of SotA AI methods unfeasible due to hardware limitations. To address this challenge, we propose the development of a staged neural compression algorithm tailored for large LArTPC images. This algorithm aims to distill essential features from the data, compressing them into a compact "latent" space. Once compressed, SotA DL/AI methods can be seamlessly applied within this reduced space, overcoming hardware limitations and enabling scientists to leverage cutting-edge AI algorithms for LArTPC data analysis. As an additional benefit, the neural compression techniques introduce semantic meaning to the "latent" space, providing novel avenues for understanding and interpreting LArTPC data.

Advancing PIONEER test of lepton universality through enhanced detector simulation/design and seeding improved BNL capability for fast frontend amplifiers

• Vladimir Tishchenko (BNL)

Room: Virtual PO/ATRO - Cross Directorate PI: Vladimir Tishchenko Other Investigators: Xin Qian (NPP/PO), Chao Zhang (NPP/PO), Prashansa Mukim (ATRO/IO), Grzegorz Deptuch (ATRO/IO), Gabriele Giacomini (ATRO/IO)

Separations of radionuclides by centrifugal partition chromatography methods

• Michael Chimes
• Cathy Cutler (BNL)

Room: Virtual IP Early Career Award (ECA) eligible PIs: Michael Chimes, Cathy S. Cutler Abstract: Separation of neighboring lanthanide elements has historically proven difficult owing to the similarity in their chemical behavior. Two such separations of interest in the context of radiopharmaceutical research and development are terbium from dysprosium and neodymium from praseodymium, for reasons discussed further below. One promising method for the separation of these elements is Centrifugal Partition Chromatography (CPC), which is a liquid-liquid chromatography method that does hundreds of automatic successive extractions within small, engraved cells or chambers on a rotating disk. By pumping the mobile phase from cell to cell through the stationary phase, which is held in place by centrifugal force, no solid support is needed and separations of nuclides/molecules with similar chemistries can be achieved with minimal solvent volumes, thereby reducing waste generated, and good throughput times, increasing productivity and lowering cost. Additionally, the use of CPC could also potentially be used to improve current separation processes performed at BNL. One example of this would be to aid the future scale up of the actinium process to thorium target masses of 100g which is of interest to the DOE Isotope Program in order to meet the potential demand of Ac-225. Terbium-161 (t1/2 = 6.89 d, β- 100%, Eβmax 594 keV) is a highly promising radioisotope for targeted cancer treatment due to its emitted radiation, which consists of both Auger electrons and short-range beta particles. Similarly, neodymium-140 (t1/2 = 3.37 d, e− capture 100%) has also gained increasing attention for clinical applications as its daughter praseodymium-140 (t1/2 = 3.39 min, β+, Eβavg = 1067 keV) is a rapidly decaying positron emitter useful for PET imaging. These radioisotopes can be produced through proton irradiation of dysprosium and praseodymium respectively. Purification of useful Tb/Nd isotopes from the Dy/Pr target material would require efficient chemical separation. Currently, effective research into and use of these isotopes is hindered due to the difficulties present in separation of adjacent lanthanides owing to the similarity in chemical behavior of the lanthanide elements, with separation methods utilizing slight differences in ionic radii between adjacent lanthanides. Work proposed herein will therefore focus on the investigation and optimization of the separation of Tb and Nd from Dy and Pr respectively by CPC, with the potential of extending research to its use in the Ac/Th process in future.

Accelerating Isotope Production with Artificial Intelligence Driven Robotics

• Kim Dohyun (BNL)

Room: Virtual IP PI: Dohyun Kim, Jasmine Hatcher Lamarre Utilizing automation and machine learning (ML) for isotope production minimizes human contact with radioactive elements. This allows for continuous processing of targets, reduced human error and increased efficiency. Methods for automating separations are currently being investigated. However, these methods are limited by the need for the researcher’s to be physically present in the laboratory. We propose to use robotics driven by artificial intelligence (AI) to develop and produce radioisotopes continuously. In many cases, robotic arms are used in automation systems. They have been used in factory automation systems, car assembly lines and even for preparing food. Using a robotic arm in an automation system requires that the object being manipulated is always in the exact same position. However, with a robotic arm using AI/ML can overcome this limitation, using AI/ML we can automate tasks even when the object is not always in the same position, and expand the range of usability. In this proposed study we will develop an ML model which will be trained to identify laboratory glassware, chemicals, etc using image recognition, the AI model will then control a robot arm to carry out the separations. If this system can be used for isotope processing, it can be applied to automate various types of isotope production by employing the existing manual methods without developing new methods for each isotope’s automation system. Even if we develop an automation system for developing each isotope, there are limits to fully automating complex processes, such as producing Ac-225 from start to finish. We can further utilize the results from the current LDRD project “Utilizing AI/ML and automation systems to inform and optimize isotope separations” to develop and validate a new method for isotope separation proposed in the research task.

A study comparing the accuracy of FEA sub-modeling methods against physical testing for the mechanical analysis of a CCT magnet

• Sara Notaro (BNL)

Room: Virtual C-AD/NPP and EIC - Cross Directorate PI: Sara Notaro Abstract: Finite element analysis (FEA) software packages such as COMSOL and ANSYS are used to study the mechanical behavior of a magnet due to magnetic forces. While FEA allows engineers to analyze structures that would otherwise be too complex for classical methods, large scale models require assumptions and simplifications. The Canted Cosine Theta (CCT) magnet for EIC is a challenge for detailed structural analysis due to the size and geometric complexity of the design, and therefore it is necessary to simplify the model by using techniques such as sub-modeling, geometry de-featuring, and material modeling. To study the accuracy of these methods, a simple physical structure is built, tested, and then modeled using FEA software. The simulation results for sub-modeling methods are compared to data obtained through testing of the physical structure. Expected Results: Data collected from both the physical and simulation results will be applied to structural mechanical modeling of the full CCT magnet. These results can then be applied to related magnets, such as collared superconducting magnets.