Zipf's law predicts a power-law relationship between word rank and frequency in language communication systems, and is widely reported in texts yet remains enigmatic as to its origins. Computer simulations have shown that language communication systems emerge at an abrupt phase transition in the fidelity of mappings between symbols and objects. Since the phase transition approximates the Heaviside or step function, we show that Zipfian scaling emerges asymptotically at high rank based on the Laplace transform. We thereby demonstrate that Zipf's law gradually emerges from the moment of phase transition in communicative systems. We show that this power-law scaling behavior explains the emergence of natural languages at phase transitions. We find that the emergence of Zipf's law during language communication suggests that the use of rare words in a lexicon is critical for the construction of an effective communicative system at the phase transition.