The logarithmic conformal field theory describing critical percolation is further explored using Watts' determination of the probability that there exists a cluster connecting both horizontal and vertical edges. The boundary condition changing operator which governs Watts' computation is identified with a primary field which does not fit naturally within the extended Kac table. Instead a "shifted" extended Kac table is shown to be relevant. Augmenting the previously known logarithmic theory based on Cardy's crossing probability by this field, a larger theory is obtained, in which new classes of indecomposable rank-2 modules are present. No rank-3 Jordan cells are yet observed. A highly non-trivial check of the identification of Watts' field is that no Gurarie-Ludwig-type inconsistencies are observed in this augmentation. The article concludes with an extended discussion of various topics related to extending these results including projectivity, boundary sectors and inconsistency loopholes.
D RIDOUTHere, we wish to restrict ourselves to the conformal field theory description of critical percolation. Much of the interest in this conformal field theory stems from the simple realisation that it must be logarithmic (see [13] for an early statement to this effect). Indeed, it is almost universally agreed that critical percolation corresponds to a theory with vanishing central charge c (though [14] offers a dissenting opinion). A standard argument [15,16] then proves that any c = 0 conformal field theory built from irreducible Virasoro modules is trivial 1 . The alternative -that the theory is built from reducible but indecomposable Virasoro modules -leads to so-called logarithmic conformal field theory [17,18].Intertwined with this story is that of the c → 0 "catastrophe" [13,[19][20][21][22]. Here, one asks what happens to the operator product expansion of a primary field and its conjugate when c → 0. The standard form of this expansion shows that the coefficient of the energy-momentum tensor T (z) diverges unless the dimension of the primary field also tends to 0. Resolving this issue involves modifying the operator product expansion by adding "partner fields", and one quickly finds that this modification also leads to logarithmic conformal field theory. We mention that the derivation of the standard operator product expansion between a primary field and its conjugate breaks down at c = 0 [23] (because the energy-momentum tensor is null), so the "catastrophe" alluded to above is merely an expression of the subtlety involved in making sense of limits such as c → 0.Exactly the same problem occurs with the ∂ 2 T (z) and : T (z) T (z) : terms as c → −22 5 , unless the primary field dimension tends to 0 or −1 5 [23, 24]. The stage is now set for studying critical percolation via investigating c = 0 logarithmic conformal field theories. However, the number of such theories is probably infinite: Aside from the theories describing percolation and the other c = 0 statistical model, self-avoiding walks, there are theories con...