The poor heat and mass transfer occurring in the H2O-LiBr absorber is one of the main limitations towards reducing the size of absorption chillers. Previous research shows that, as droplets form at the base of the absorber pipes, considerably large mass transfer coefficients can be obtained. However, modeling this phase of droplet formation has not been fully explored, as most of the research focuses on the film flow process preceding the droplet formation. The present study focuses on two aspects. Firstly, the dynamics of the droplet formation is investigated, with a focus on the effect of the solid surface shape on the droplet formation. A model to describe the droplet profile geometry was developed using the Euler-Lagrange equation and validated against experimental tests. Several pin geometries were tested and the results have shown that a 120 degrees rhomboidal geometry is more suitable to increase the liquid-vapor interface area, while lowering the risk of droplet coalescence. Secondly, an analytical heat and mass transfer model based on the Fourier Series method has been developed to study the influence of pin size on the absorption process in an adiabatic absorber. The results show that the optimum width of the 120 degrees rhomboidal pin is found at 6 mm, which maximizes the water absorbed during the droplet formation phase, without excessive use of material. The common assumption that treats the forming droplet as a half sphere fails to capture changes in the pin-droplet interaction which adversely affects the model accuracy. The proposed model shows that for pin widths smaller than 6 mm, the absorption process is impaired by the lower surface area exposed to the water vapor, resulting in up to 67% less mass absorbed obtainable and a decrease in the cooling power obtainable. In more practical terms, a rated 3 kW of cooling power is used as a case study with a 6 mm pin which reduces the absorber volume by 30% with respect to an absorber with 3 mm wide pins. This shows that by treating the droplet formation phase with a more physically-sound geometrical domain can improve the accuracy of the absorption models and can be easily implemented with any programming language to help in the design optimization of the absorber.