Intermittently connected mobile networks are wireless networks where most of the time there does not exist a complete path from a source to a destination. Researchers have proposed flooding based schemes for routing in such networks. While flooding based schemes are robust and have a high probability of delivery, they suffer from a huge overhead in terms of bandwidth, buffer space and energy dissipation due to large number of transmissions per packet So flooding based schemes are impractical for resource constrained networks. Controlled replication or spraying methods can reduce this overhead by distributing a small, fixed number of copies to only a few relays, which then independently route each copy towards the destination. These schemes demonstrate a good delay performance without using a lot of resources. There are three important questions in the context of the design of these spraying based schemes: (i) How many copies per packet should be distributed? (ii) How to distribute these copies amongst the potential relays? (iii) How are each of these copies routed towards the destination? The first and the third questions have been studied in detail by different researchers. But, there has been no study which looks at the second question. This paper fills this void. Specifically, we propose a methodology to derive the optimal spraying policy. As a case study, we find the optimal spraying policy for two different spraying based schemes. Finally, we study the optimal policies to infer simple heuristics which achieve expected delays very dose to the optimal.