Empirical evaluation of Uncertainty Quantification in Retrieval-Augmented Language Models for Science