Free resolutions of certain codimension three perfect radical ideals

Abstract

Introduction. The original conception of this work included a complete study of certain codimension three perfect ideals that arise naturally in a set-up, with the purpose of enriching the present collection of such ideals. The ones we consider are always radical and their syzygy-theoretic properties are relatively easy to describe. From the viewpoint of the theory of graded resolutions, they share common numerical features with (the projecting cones of) certain codimension two and three varieties in projective space. Their resolutions being seldom pure, they seemed largely convenient for testing the sharpness of some calculations made in the pure case (cf. [19], [25], [26]). Moreover, they yield a large class suited for testing recently developed concepts, such as the notion of strong obstruction, that have its natural place in the theory of linkage ([loc. cit.]) or the more general theory of residual intersection developed in [16]. The authors did not pursue this line of thought, but it seemed worthwhile pointing it out. A brief description of the present contents is as follows. In the first section one obtains the explicit free resolution of a large class of codimension three ideals that appear as presentation ideals of normal cones associated to determinantal ideals fixing a submatrix. As it turns out, the resolutions of these presentation ideals are given by a mapping cylinder of a suitable map to the resolution of the original determinantal ideals. The second section is concerned with deriving some ideal-theoretic results from the knowledge of the presentation given in section one. This procedure, by and large, follows the techniques developed by various authors ([17], [14], [23] and [24]). In the third section one gives the resolution of the determinantal ideal Jr associated to a map m ~ R (m > n), fixing r c m, in the case of the data r = n - 2 and m = n + 2. This case not only extends the previous known ones [1], [2], but mainly gives some genuine clue for the general situation. The construction follows closely the spirit of the scandinavian complex [9] in that one is led to tensor suitable complexes and then to cut down the resulting size by an use of trace maps. This method seems to generalize well in various contexts [21]. Last one establishes exact values for the codimension of the complete intersection locus of the ideals considered in the previous sections. The reason for doing so is that the known estimates to present [11], [27] would give no real grasp, in this case, of the locus.