Design and development of advanced Al-Ti-V alloys for beampipe applications in particle accelerators

Abstract

The present investigation reports the design and development of an advanced material with a high figure of merit (FoM) for beampipe applications in particle accelerators by bringing synergy between computational and experimental approaches. Machine learning algorithms have been used to predict the phase(s), low density, and high radiation length of the designed Al-Ti-V alloys. Al-Ti-V alloys with various compositions for single-phase and dual-phase mixtures, liquidus temperature, and density values are obtained using the Latin hypercube sampling method in TC Python Thermo-Calc software. The obtained dataset is utilized to train the machine-learning algorithms. Classification algorithms such as XGBoost and regression models such as Linear Regression and Random Forest regressor have been used to compute the number of phases, radiation length, and density respectively. The XGBoost algorithms show an accuracy of $98\%$, the Linear regression model shows an accuracy of $94\%$, and the Random Forest regressor model is accurate up to $99\%$. The developed Al-Ti-V alloys exhibit high radiation length as well as a good combination of high elastic modulus and toughness due to the synergistic effect of the presence of hard $Al_3Ti$ phase along with a minor volume fraction of FCC $(Al)_{ss}$ solid solution phase mixture. The comparison of our alloys, alloy-1 ($Al_{75.2}Ti_{22.8}V_{2}$) and alloy-2 ($Al_{89}Ti_{10}V_{1}$) shows an increase in the radiation length by seven-times and a decrease in the density by two to three times as compared to stainless steel 304, the preferred material for constructing beampipes in low-energy particle accelerators. Further, we experimentally verify the elastic modulus of the alloy-1 and compute the FoM equal to 0.416, which is better than other existing materials for beampipes in low-energy experiments.