Nucleation Mechanism Most common model: ¯Methane solvation shell is dodecahedral hydrate cage ¯Nucleation occurs by aggregation of dodecahedral water cages Other models based on long-range ordering of methane molecules 0.6 ps

Mechanism: Long Lived Fragments

Cluster Growth Large, rapid variations in cluster size Not dependent on clustering distance

Hydrate Inhibitors Thermodynamic ¯ e.g. methanol, glycols (salt) ¯very large volumes (50% w/w water) Kinetic ¯Since very late 1980’s ¯Affect nucleation / growth – not thermodynamics ¯small volumes (parts per thousand) ¯lactam polymers (PVP, PVCap & copolymers) ¯small is better (1–2 kDa) ¯others include quaternary ammonium salts

Fish Anti-freeze Proteins From Arctic Winter Flounder Main ice growth face (Lal et al., Faraday Discussions, 1993)

Relax ca. 1 ns Adding PVP

Water in Hydrate Environment

Methane-Methane RDFs For methane within the water film With PVP ; without PVP

Inhibitors & Nucleation? Without PVP With PVP 0.6 ns10.5 ns40.2 ns 10.8 ns5.4 ns0.9 ns

Hydrate Inhibitors in Motion

Inhibitors: Selected Conformations Typically no contact between inhibitor and hydrate cluster ca. 0.5 ns after adding inhibitor

Bridging between inhibitor and hydrate cluster Inhibitors: Selected Conformations

Conclusions Molecular Simulations can provide important information to inform coarse-graining ¯Parameterise models ¯Identify mechanisms to inform model-formulation ¯Identify regimes for different mechanisms – e.g. effect of different sub-coolings on crystal growth/additives ¯Oswald’s step rule – which unstable polymorphs never restructure? ¯Nature of the match between additive and crystal

Acknowledgements Modelling ¯Changman Moon ¯Mark Storr ¯Rob Hawtin ¯Dorothy Duffy Funding ¯EPSRC ¯British Council ¯ICI ¯RF Rogaland Computer Time ¯Materials Consortium ¯Warwick Centre for Scientific Computing