A conventional monopulse radar system uses three beams; sum beam, elevation difference beam and azimuth difference beam, which require different layers of weights to synthesize each beam independently. Since the multi-layer structure increases hardware complexity, many simplified structures based on a single layer of weights have been suggested. In this work, we introduce a new technique for finding disjoint and fully covering sets of weight vectors, each of which constitutes a sparse subarray, forming a single beam. Our algorithm decomposes the original non-convex optimization problem for finding disjoint weight vectors into a sequence of convex problems. We demonstrate the convergence of the algorithm and show that the interleaved array structure is able to meet difficult beam constraints.