Evslin, JarahMinasian, Ruben2008-04-142011-09-072008-04-142011-09-072008-04-14https://openscience.sissa.it/handle/1963/2625We add to the mounting evidence that the topological B model's normalized holomorphic three-form has integral periods by demonstrating that otherwise the B2-brane partition function is ill-defined. The resulting Calabi-Yau manifolds are roughly fixed points of attractor flows. We propose here that any admissible background for topological strings requires a quantized (twisted) integrable pure spinor, yielding a quantized (twisted) generalized Calabi-Yau structure. This proposal would imply in particular that the A model is consistent only on those Calabi-Yau manifolds that correspond to melting crystals. When a pure spinor is not quantized, type change occurs on positive codimension submanifolds. We find that quantized pure spinors in topological A-model instead change type only when crossing a coisotropic 5-brane. Quantized generalized Calabi-Yau structures do correspond to twisted K-theory classes, but some twisted K-theory classes correspond to either zero or to multiple structures.215269 bytesapplication/pdfen-USPreprint