The emerging energy harvesting technology, which has been successfully integrated into Wireless Sensor Networks, enables sensor batteries to be charged using renewable energy sources. In the meantime, the problem of Minimum Latency Aggregation Scheduling (MLAS) in battery-powered WSNs has been well studied. However, because sensors have limited energy harvesting capabilities, captured energy is limited and varies greatly between nodes. As a result, all previous MLAS algorithms are incompatible with Battery-Free Wireless Sensor Networks (BF-WSNs). We investigate the MLAS problem in BF-WSNs in this paper. To make the best use of the harvested energy, we build an aggregation tree that leverages the energy harvesting rates of the sensor nodes with an intuitive explanation. The aggregation tree, which determines sender-receiver pairs for data transmission, is one of the two important phases to obtain a low data aggregation latency in the BF-WSNs.