The phase behavior of multi-component hydrate systems formed in the ocean has been studied using a sea-going in situ Raman spectrometer. Results from a field study using a synthetic hydrate sample and preliminary measurements of naturally occurring thermogenic hydrates at Barkley Canyon are reported and compared. This work follows early field trials using simple and binary synthetic hydrates and an expedition to Hydrate Ridge, studying natural hydrates primarily composed of methane. This refined Raman technique was combined with visual observations utilizing remotely operated vehicles as experimental platforms in an oceanic laboratory. Traditional laboratory Raman analysis was also performed on hydrate samples following recovery. A synthetic hydrate was first studied in Monterey Bay to characterize the formation, evolution, and dissociation of multi-component gas hydrates formed in the deep ocean as a guide to understanding complex behavior of natural hydrate systems. The gas mixture (92.5% CH4, 3.8% C2H6, 1.9% C3H8, 0.4% i-C4H10, 0.4% n-C4H10, 0.1% i-C5H12, 0.1% n-C5H12, and 0.1% neo-C5H12) was chosen to simulate a typical natural gas composition. Regions of sII and mixed sI-sII hydrate were measured in situ after 65 days of formation. The presence of a heterogeneous inter-mixed sI-sII hydrate suggests a formation mechanism where hydrate growth proceeds from localized pockets of isolated gas, such as gas bubbles separated from the bulk vapor phase by hydrate membranes, where sI formation follows the depletion of sII-forming gases within these individual hydrate reactors. Unstable conditions during recovery led to significant hydrate dissociation, leaving only ice and a compositionally homogenous sII hydrate phase, with no evidence of the sI that had been observed at the seafloor. To investigate whether this process may operate in natural systems, a Raman investigation of the natural thermogenic hydrates at Barkley Canyon was carried out. Massive seafloor hydrate outcrops varied in appearance from yellow, due to oil staining, to white. While the yellow hydrate had significant interference due to fluorescence, detailed analysis of the in situ Raman data was possible for the white hydrate. This white hydrate was mainly sII, but the presence of sI was also detected. The heterogeneity in both structure and composition of these natural hydrates was similar to the synthetic hydrate formed in Monterey Bay. These similarities suggest that a controlled seafloor experiment was able to simulate a natural thermogenic hydrate with reasonable accuracy. Further development of theory and testing is needed to determine if the 'individual hydrate reactor' mechanism was indeed the pathway to growth in this natural system. (c) 2007 Elsevier B.V All rights reserved.