The International Space Station Alpha's (ISSA) thermal radiators are designed to tolerate ammonia freezing conditions. The cold case thermal design environment for ISSA is -92.8°C (-135°F). This environment is below the freezing point of ammonia, the Active Thermal Control System's (ATCS) working fluid, Tfreeze = -78°C (-108°F). Ammonia contracts 10% by volume when it freezes. Liquid ammonia can fill in this 10% volume and hard pack the individual flow tubes in the radiator. A hard packed flow tube filled with frozen ammonia would have to be able to tolerate this 10% volume increase when the ammonia thaws.The ISSA radiator flow tube design accommodates the volume change of thawing ammonia by three mechanisms. The most severe condition will arise if the center of a flow tube thaws while the ends remain frozen; thus, any increase in pressure has no axial relief. The first mechanism to accommodate the volumetric expansion is the straining of the flow tube when exposed to high pressures. At the design ammonia thaw pressure the flow tube volume will increase by 5.5%. Second, the frozen/liquid ammonia will be compressed by 2.4% at the high ammonia thaw pressures. Thirdly, the melting temperature increases at high pressures, -71°C (-96°F), and the volumetric expansion of the solid->liquid phase change is reduced to 7.5% at the thaw pressure. An additional expansion of 0.4% occurs due to liquid density variation for temperatures above the melting point -67.8°C (-90.1°F). These conditions are reached for a maximum radiator flow tube pressure of 87 MPa (12.6 ksi). The flow tube design has adequate structural margin to accommodate this high pressure.The radiator panel is designed with variable flow tube spacing to ensure all of the flow tubes within a single panel do not freeze simultaneously. The flow tubes on the edge of the panel with the longest fin length freeze first. Their flow, and corresponding heat load, are then transferred to the remaining non-frozen flow tubes.The radiator flow tube design-to-freeze approach and panel freezing pattern have been tested. Testing of the design-to-freeze approach verified that the analysis contained in this paper is accurate and slightly conservative. Testing of the flow tube freezing pattern has also been tested and been shown to correlate with computer models of the panel.