We study the minimum latency broadcast scheduling problem, in which a single source has a quantity of data that must be transmitted to all other nodes in a multi-hop network in minimum time. Aside from the obvious application to classical communications, this problem also relates to some more general problems in the field of network science. Previous approaches to scheduling have assumed a simplistic collision model of interference, while others have studied the more realistic physical model of total received interference power. Existing suboptimal approaches for transmitting the data typically assume a collision-free, fixed-rate, single packet transmission. In this work, we devise an optimal approach for broadcast under a physical interference model with fixed-rate, single packet transmission by converting it to a shortest path problem for an unweighted, undirected graph. Since this optimal approach does not scale well, we also consider a suboptimal layered approach which separates the routing and scheduling functions, but relaxes the fixed-rate, single packet assumption. This goes beyond the signal-to-interference-plus-noise (SINR) threshold model to allow for rate adaptation as a function of SINR. We include improvements on previous routing approaches, and we formulate a linear programming approach to the variable-rate scheduling for broadcast. Simulations show that in some special cases, this variable-rate layered approach can even outperform the optimal fixed-rate, single packet approach.