Guessing models and generalized Laver diamond

We analyze the notion of guessing model, a way to assign combinatorial properties to arbitrary regular cardinals. Guessing models can be used, in combination with inaccessibility, to characterize various large cardinals axioms, ranging from supercompactness to rank-to-rank embeddings. The majority of these large cardinals properties can be defined in terms of suitable elementary embeddings j : V_γ → V_λ. One key observation is that such embeddings are uniquely determined by the image structures j[V_γ] ≺ V_λ. These structures will be the prototypes guessing models. We shall show, using guessing models M, how to prove for the ordinal κ_M = j_M (crit(j_M )) (where π_M is the transitive collapse of M and j_M is its inverse) many of the combinatorial properties that we can prove for the cardinal j (crit(j )) using the structure j [V_γ ] ≺ V_j (γ ) . κ_M will al- ways be a regular cardinal, but consistently can be a successor and guessing models M with κ_M = w_2 exist assuming the proper forcing axiom. By means of these models we shall introduce a new structural property of models of PFA: the existence of a “Laver function” f : w_2 → H_w_2 sharing the same features of the usual Laver functions f : κ → H_κ provided by a supercompact cardinal κ. Further applications of our analysis will be proofs of the singular cardinal hypothesis and of the failure of the square principle assuming the existence of guessing models. In particular the failure of square shows that the existence of guessing models is a very strong assumption in terms of large cardinal strength. Key words: Large cardinals, Forcing axioms, Laver diamond