Deep Neural Network-Based Reconstruction of SeaQuest E906 Data for Boer-Mulders Function Extraction

The Boer-Mulders function determines the distribution of transversely polarized quarks in an unpolarized proton. A non-zero Boer-Mulders function corresponds to a certain handedness of the nucleon and causes an azimuthal asymmetry in Drell-Yan scattering. To extract the unpolarized Drell-Yan Boer-Mulders function from the SeaQuest E906 data, we will utilize reconstruction techniques based on deep neural networks (DNN). This method has the potential to maximize statistics while minimizing systematic uncertainty in measuring the angular-dependence coefficients $lambda$, $mu$, and $ u$. Lower statistical errors enable an improved regression to obtain these coefficients across more bins, facilitating a broader extraction of the Boer-Mulders function with sensitivity to the $p_T$ dependence of $mu$. SeaQuest, located at Fermilab, was a fixed-target experiment devised to detect the Drell-Yan process in $p+p$ and $p+d$ reactions. We will discuss the process of developing the DNN-based reconstruction as well as present preliminary results.