Abstract　 　　 

　 Massive stars drive the evolution of galaxies, but how they come into existence is still a fundamental open question. Particularly critical and poorly understood is the very early stage, which determines how initial conditions in molecular clouds lead to the onset of star formation. In this talk, I will present our series studies on infrared dark clouds (IRDCs) aiming to characterize the early stage. Seen as shadows against the Galactic IR background, IRDCs are extremely cold and dense molecular clouds that represent the early stage in massive star formation. Our dedicated observing campaign is carefully designed to resolve the initial fragmentation and at the time to measure the gas temperature and turbulence. This coordinated effort allows us to infer the nature of the fragmentation and to constrain theoretical models. We find that the early stage is characterized by hierarchical fragmentation of IRDCs. The fragmentation at multiple levels leads to structures at multiple spatial scales, ranging from ~1 pc down to ~0.01 pc, where turbulence dominates the fragmentation at all the probed scales. The results are yet to be fully implemented by theoretical models. If time permits, I will also introduce my functional work at ALMA (Atacama Large Millimeter/submillimeter Array), the next generation radio interferometry.