Pulmonary drug delivery is the focus of much research and development because of its great potential to produce maximum therapeutic benefit. Among the available options the dry powder inhaler (DPI) is the preferred device for the treatment of an increasingly diverse number of diseases. However, as drug delivery from a DPI involves a complicated set of physical processes and the integration of drug formulations, device design and patient usage, the engineering development of this medical technology is proving to be a great challenge. Currently there is large range of devices that are either available on the market or under development, however, none exhibit superior clinical efficacy. A major concern is the inter- and intra-patient variability of the drug dosage delivered to the deep lungs. The extent of variability depends on the drug formulation, the device design and the patient's inhalation profile. This article reviews recent advances in DPI technology and presents the key factors which motivate and constrain the successful engineering of a universal, patient-independent DPI that is capable of efficient, reliable and repeatable drug delivery. A strong emphasis is placed on the physical processes of drug powder aerosolisation, deagglomeration, and dispersion and on the engineering of formulations and inhalers that can optimise these processes.

Sparse-Lagrangian filtered density function (FDF) simulations using a generalized multiple mapping conditioning mixing model and density coupling via a conditional form of the equivalent enthalpy method are performed for both constant density and variable density turbulent jet diffusion flames. The consistency between the sparse-Lagrangian FDF for the reactive species and the Eulerian large eddy simulation (LES) for velocity along with the accuracy of the reactive species predictions relative to the exact equilibrium solution are presented in detail. The sensitivity to the number of particles used in the simulations, the mixing localization structure, chemistry and numerical time step are all investigated. The analysis shows that consistency between the FDF and LES fields is relatively insensitive to the sparseness of the particle distributions and other model parameters but that the reactive species are strongly dependent on the degree of mixing localization in the LES mixture fraction space. An algorithm is developed to control the localization for any sparse distribution of particles with inter-particle distances within the inertial range, and it is shown that reactive species predictions are sensitive to the mixing distance in a reference mixture fraction space while there is very low sensitivity to the number of particles used in the simulations.