Probing the Creativity of Large Language Models: Can models produce divergent semantic association?