was no significant effect on collapse pressures. One possible explanation 

 is that the out-of-roundness is small in each instance relative to radius 

 and shell thickness and, therefore, has a negligible effect on collapse 

 strength. Many of the cylinders had initial flaws in the walls. In 

 addition, the surfaces had varying degrees of imperfections, as is to be 

 expected from shelf stock. Although these flaws and surface imperfections 

 would be expected to have considerable detrimental effect on cylinders 

 under tension or internal pressure, no effect on collapse under external 

 pressure was observed. 



The tests do not demonstrate the meLximiom compressive strength of glass 

 since the cylinders collapsed by elastic instability. The tests do show, 

 however, that compressive stresses of about 100,000 psi may be developed 

 in No. 7740 glass cylinders. Specifically, the maximum calculated circum- 

 ferential stress in Cylinder 20 prior to collapse was 97,000 psi. Tests 

 of thicker long cylinders, short or stiffened cylinders, or spheres would 

 provide the maximum compressive strength of glass as a hull material. 



Glass shows promise of providing higher strength-to-buoyancy ratios 

 for pressure hulls than can be realized by hulls of presently available 

 metals such as high-strength steels, aluminum alloys, and titanium alloys. 

 Although Cylinder 20 collapsed by elastic instability, a hull of steel 

 with a yield strength in excess of 300,000 psi, a hull of aluminum with a 

 yield strength in excess of 100,000 psi, or a hull of titanium with a 

 yield strength of about 200,000 psi would be required to equal its per- 

 formance on a strength-weight basis. 



Glass spheres should provide even higher strength-to-buoyancy ratios 

 than are attainable from cylinders. Figure 2 shows the relative strength- 

 buoyancy ratios for long glass cylinders, stiffened glass cylinders, and 

 glass spheres. The collapse strength of the long glass cylinders are 

 based on the elastic theory of Bresse and Bryan. The predicted collapse 

 strength of the stiffened cylinders is based on the assumption that 

 stiffeners are placed in such a manner that the yield strength of all of 

 the material may be utilized. The solid curves for the elastic buckling 

 of spheres are based on experimental tests of machined aluminum deep 

 spherical shells,^ and the broken curves are based on tests of fabricated 



