Salinity is a significant barrier to the distribution of diatoms, and though it is generally understood that diatoms are ancestrally marine, the number of times diatoms independently colonized fresh waters and the adaptations that facilitated these colonizations remain outstanding questions in diatom evolution. Resolving the exact number of freshwater colonizations will require large-scale phylogenetic reconstruction with dense sampling of marine and freshwater taxa. A more tractable approach to understanding the marine--freshwater barrier is to study a group of diatoms with high diversity in each habitat. The "centric" diatom order Thalassiosirales affords an excellent opportunity to study the origin and evolution of diatoms in fresh waters. Thalassiosirales is a well-supported monophyletic group common in marine, brackish, and freshwater habitats. Thalassiosirales species historically are classified into the marine Thalassiosiraceae or freshwater Stephanodiscaceae, reflecting the more generally held hypothesis that diatoms are naturally split along marine--freshwater lines. The fossil record suggests that Stephanodiscaceae traces to a single colonization of freshwater in the mid-Miocene, and in addition, Stephanodiscaceae species share a suite of complex cell wall characters, which has been interpreted as corroborating evidence for their monophyly. I reconstructed the phylogeny of Thalassiosirales and used the phylogeny to test these and other hypotheses and to address a number of other problems related to the marine--freshwater boundary in diatoms. Phylogenetic analyses showed strong evidence for multiple colonizations of freshwater and reject all previous colonization hypotheses. Results further show that part of Stephanodiscaceae is an early diverging lineage within Thalassiosirales, indicating that these two distantly related and separately derived Stephanodiscaceae lineages independently evolved a similar set of complex morphological features upon or shortly after the colonization of fresh waters. Finally, marine and freshwater diatoms, including Thalassiosirales, show several important differences in silicon physiology. In addition to containing an order of magnitude more silica in their cell walls, freshwater diatoms have a drastically lower enzymatic affinity for silicic acid, the dissolved form of silica used by diatoms. I sequenced the silicon transporter genes from marine and freshwater Thalassiosirales and show that physiological differences are not due to differences in the coding sequence. / text