QCD dynamics at small quark and gluon momentum fractions or large total energy, which plays a major role for HERA, the Tevatron, RHIC and LHC physics, is still poorly understood. For one of the simplest processes, namely bb production, next-to-leading-order perturbation theory fails. We show that the combination of two recently developed theoretical concepts, the k ⊥ -factorization and the next-to-leadinglogarithmic-approximation BFKL vertex, gives perfect agreement with data. One can therefore hope that these concepts provide a valuable foundation for the description of other high energy processes.Existing QCD calculations describe many high energy observables which involve partonic transverse momentum rather poorly. This is also true for the theoretically especially clean case of bb production, which was investigated experimentally at Fermilab [1]. Since central quark-antiquark production at √ s = 1.8 TeV is sensitive to very small gluon momentum fraction x ≈ 10 −2 − 10 −4 , one probes the gluon content of the nucleon at small x, which is a central issue of current research. We reconsider this process and combine as essential new ingredients the k t -factorization scheme with the next-to-leadinglogarithmic-approximation (NLLA) BFKL production vertex derived in [2]. The k t -factorization approach for the description of high energy processes [3-6] differs strongly from the conventional NLO collinear approximation (e.g. [7]) because it takes the non-vanishing transverse momenta of the scattering partons into account. The usual gluon densities are replaced by unintegrated gluon distributions which depend on the transverse momentum k t . These together with the k t -factorization form a basis for a general calculation scheme for high energy (i.e. small x). The standard collinear approximation has the advantage of being closely related to the operator product expansion. It is, however, only justified for the processes dominated by x = O(1). In application to processes governed by small x the k t -factorization approach has the advantage that its approximations correspond to the dominant kinematics. Essential small x contributions are included in the Born approximation which in the collinear approach are accounted for in higher orders only. This is well known from the case of structure functions where the DGLAP evolution is appropriate for x = O(1) and the BFKL evolution for small x.While the k t -factorization formalism is very attractive theoretically, its phenomenological usefullness has been mostly tested in the case of the structure function F 2 [9,10]. The NLLA BFKL vertices are just the ones needed to treat semi-hard central production at collider energies in this approach.In our calculation we use one particular element of the NLLA BFKL formalism [2,8], namely the effective vertex for quark-antiquark production. Thus our calculation can be seen as a first phenomenological application of this vertex which decides whether the NLLA BFKL formalism can be hoped to converge.One special aspect of the reaction we in...