Sharing the huge amount of data existed and being produced in specific domain is of great significance for researches. The semantic heterogeneity has been identified as the most important and toughest problem when integrating various sources. This paper proposes a semantic data integration method to manipulate information scattered over many data sources in specific area of science. The method uses OWL ontology to represent the global semantics of domain model and the local semantics of heterogeneous sources respectively. Semantic mappings between global and local ontology are built in a mapping ontology. Semantic heterogeneity among the data sources is resolved by TBox reasoning, and there is no need to migrate data from data sources to ontology instances. Domain-specific Markup Language is used as the uniform format for query results to facilitate data exchange in the domain. The architecture and key components for implementing the method are presented. To further illustrate the method, the paper gives a case study to show how to integrate data semantically in material sciences.