Let f: X → S be a smooth projective curve over a base S, let Gm be the fppf sheaf represented by the multiplicative group. I will discuss the result of Grothendieck and Deligne, dating from the end of the 1960s, that the truncation up to degree 1 of Rf*Gm is Cartier dual to itself. Their proof essentially boils down to the correspondence between sections of f and relative divisors on X, but in order to make this run smoothly, they use the fact that the truncation up to degree 1 of Rf*Gm can be identified in a natural way with the 'Picard stack' (symmetric group-like stack) on S of invertible sheaves on X.