Modelling complex particle interactions in dry powder inhaler

Dry powder inhaler (DPI) delivers active pharmaceutical ingredient (API) particles to the airways and/or lungs [1]. Larger carrier particles are utilized to agglomerate with small API particles so as to improve fluidization, and then deagglomerate to release API particles for drug delivery [2]. Agglomeration and deagglomeration in DPI are affected by complex particle interactions, including van der Waals’ cohesion [3] and electrostatic interaction [4]. Computational fluid dynamics coupled with discrete element method (CFD-DEM) has the advantage of explicitly capturing these particle-particle interactions but is computationally costly [5]. In order to make the simulation affordable, we undertake a two-step approach, particle softening and particle coarsening. First, we soften the particles in the simulation so as to reduce simulation time step. Second, we develop a coarsening approach in which one simulates a subset of representative API particles while still tracking all the carrier particles.

Towards this end, we conduct micro-scale tests using a carrier particle decorated with many API particles. These tests involve collision of a decorated carrier with a wall or another decorated carrier. In these simulations we consider both the van der Waals cohesion and electrostatic interaction between charged carrier and API particles. Through these tests, we examine if, and if so how, the model parameters should be adjusted such that particle softening and coarsening do not lead to falsification of the fraction of API released. For example, it is known that the Hamaker constant must be modified upon softening to ensure that the effective coefficient of restitution is preserved in head-on collision [6]. We find that some modification to rolling friction is also necessary when one coarsens the simulations. In a surprising result, we find that attractive interaction between the API particles translates to an effective increase in the carrier-API attraction, which would be underestimated upon coarsening; a method to correct for this effect is also proposed.

We complement these microscale tests with full DPI simulations to assess the performance of our coarse-grained modeling approach.