We give examples of smooth Calabi-Yau 3-folds in P 6 of low degree, up to the first difficult case, which occurs in degree 17. In this case we show the existence of three unirational components of their Hilbert scheme, all having the same dimension 23 + 48 = 71.The constructions are based on the Pfaffian complex, choosing an appropriate vector bundle starting from their cohomology table. This translates into studying the possible structures of their Hartshorne-Rao modules.We also give a criterium to check the smoothness of 3-folds in P 6 . Licensed to Univ of British Columbia. Prepared on Thu Jul 2 05:43:28 EDT 2015 for download from IP 142.103.160.110. License or copyright restrictions may apply to redistribution; see http://www.ams.org/license/jour-dist-license.pdf 210 FABIO TONOLIi.e. subschemes defined locally by the 2r × 2r Pfaffians of an alternating map ϕ from a vector bundle of odd rank 2r + 1 to a twist of its dual. In particular, a Pfaffian subscheme in P n has the following resolution:where the map ψ is locally given by the 2r × 2r Pfaffians of ϕ and ψ t is the transpose of ψ. Being Pfaffian, this subscheme is automatically subcanonical, in the sense that its canonical bundle is the restriction of a multiple of O P n (1). A recent result of Walter [Wal96] shows that under a mild additional hypothesis every subcanonical Gorenstein codimension 3 subscheme X in P n is Pfaffian (see [EPW01] for a description of the non-Pfaffian case), and therefore one can attempt to get its equations, starting by constructing its Pfaffian resolution.In this paper we apply this method to build examples of smooth Calabi-Yau 3-folds in P 6 . In order to build a Pfaffian resolution of a subcanonical Gorenstein codimension 3 subscheme X, Walter shows an explicit way to choose an appropriate vector bundle, starting from its Hartshorne-Rao modules H i * (I X ): this is a precise hint for constructing a resolution. But to find out what are the possible structures for such modules is the hard part in the construction: indeed, from the invariants of X one can deduce only the "minimal" possible Hilbert functions of its Hartshorne-Rao modules, and their module structures remain obscure. In this sense the problems met in the constructions are the same as in the codimension 2 cases, except that here the range of examples where the construction is straightforward (and their Hilbert scheme component unirational) is rather short.We construct examples of smooth Calabi-Yau 3-folds in P 6 having degree d in the range 12 ≤ d ≤ 17. Such a bound can be better understood by looking at hyperplane sections of the desired 3-folds. Since a hyperplane section of a Calabi-Yau 3-fold is a canonical surface, a lower bound on the degree d of the desired 3-fold can be obtained easily by the Castelnuovo inequality: if the canonical map of a surface S is birational, then K 2 S ≥ 3p g − 7, cf. [Cat97, p.24]. This gives d ≥ 11. Furthermore, the case d = 11 is interesting, but no smooth examples were found and we believe that they don't exist: every Calabi-Yau...