Particle morphology can be described in terms of particle size, shape, internal structure, and surface properties. The terminology used in the literature to describe structured microparticles is inconsistent; this review suggests the following:

A core is the innermost part of the particle. It can be covered by one or more layers that are defined by having distinct composition or properties. A shell is an outer layer that is capable of bearing enough mechanical stress to influence or determine the morphology of the particle. In contrast, a coat is a thin layer incapable of providing structural rigidity. The schematic in Fig. 1b shows an idealized layered sphere with internal structure, comprising a core, a second layer, a shell and a coat. In Fig. 1a the core is replaced with a central void. Fig. 1

Schematic representation of particle morphologies. a Layered with central core. b Layered with central void. c Solid foam, closed cell. d Solid foam, open cell. e Particle with embedded nanoparticles. f Composite shell. g Irregular, with external voids and internal concentration gradients.

With few exceptions, spray dried particles are spherical and their size can be described by their geometric diameter. This is an important parameter which influences the forces particles experience in fluid flows and the packing of particles when they form a powder. The geometric diameter is also often used as a reference, because it is in principle directly accessible by ultramicroscopic techniques.

However, the measurement of geometric diameters by image analysis is not without difficulties. One is rarely interested in the size of just a single particle, but rather the size distribution of all particles in a product. In the context of pharmaceutical applications the mass fraction of a particulate dosage form in a certain particle size interval is often sought. This information can be obtained by image analysis of a sample of the dose, but care has to be taken to ensure that a representative, statistically relevant sample is analyzed. Usually, this can only be accomplished with the help of automated microscopy and image analysis (49). Image analysis techniques generate count-based distributions that have to be converted to a mass-based distribution, e.g. with the help of the Hatch Choate equations (44).

The geometric diameter of particles that are not perfect spheres depends on their orientation. In this case an average geometric diameter can be substituted, e.g. by averaging Feret’s horizontal and vertical diameters (50). If this is carried out on a large number of particles that are randomly oriented, the reported geometrical diameter is the orientation averaged Feret’s diameter.

Other common particle sizing techniques do not measure the geometric diameter directly, but rather derive it from the behavior of the particles in response to the probe used in the analysis. The reported diameters are often confused with geometric diameters, but are in fact equivalent diameters specific to the analytical technique (49). A common example is an optical-equivalent diameter where particles are probed by a light source. Their scattering or absorption behavior is compared to that of a reference geometry, usually a homogeneous sphere. The geometric diameter of the reference sphere is only representative of the diameter of the sample particle if the optical properties of the sample are close to that of the reference. For structured or irregularly shaped particles this is often not the case, as can easily be recognized considering the variety of particle types shown in Fig. 1. The optical particle size for these particles is in reality a function of their geometric diameter and their secondary morphological features (51,52). Equivalent particle diameters based on other properties or behavior such as electrical mobility, diffusion, sedimentation, surface or volume to surface ratio are used frequently and are described in the literature (49,53).

An equivalent diameter of particular importance to respiratory delivery is the aerodynamic diameter, i.e. the diameter of a unit-density sphere that has the same settling velocity as the measured particle. 1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d_{{\text{a}}} = {\sqrt {\frac{{\rho _{{\text{P}}} }}{{\rho ^{ * } }}} }d_{g} $$\end{document}

Equation 1 is a simplified description, valid for spheres in the continuum regime. Treatment and corrections for non-sphericity and transition regime flow can be found in the aerosol science literature (44,49,54). The particle density, ρ_P, in this equation should not be mistaken for the true density of the dried material. The particle density is the mass of the particle divided by the volume of a sphere of diameter d_g and can be significantly lower than the true density, because it includes internal and external voids.

It is straightforward to express the geometric diameter of spray dried, spherical particles as a function of formulation and process variables. The geometric diameter of the dry particle follows from a simple mass balance (55) 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d_{\text{g}} = \sqrt[3]{{\frac{{c_{\text{F}} }}{{\rho _{\text{P}} }}}}d_{\text{D}} $$\end{document}

This equation shows that the geometric diameter can be most effectively controlled by the atomization which determines the diameter of the droplets, d_D. Particle size is also directly affected by changes in the concentration of the feed solution, c_F. If Eqs. 1 and 2 are combined the resulting expression 3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d_{\text{a}} = \sqrt[6]{{\frac{{\rho _{\text{P}} }}{{\rho ^* }}}}\sqrt[3]{{\frac{{c_{\text{F}} }}{{\rho ^* }}}}d_{\text{D}} $$\end{document}shows that the aerodynamic diameter is also primarily determined by the feed solution concentration and the initial droplet diameter. Although somewhat counterintuitive, it is only a weak function of particle density, which only becomes important if it is significantly lower than unit-density. This explains the current trend to particles of very low density on the order of 0.1 g/cm³ in pulmonary delivery, where a small aerodynamic diameter is desirable. Equation 3 also cautions that changes in the droplet size distribution, which may be encountered during process scale-up, or increases in the feed concentration, often initiated to improve the economics of the process, can cause a significant change in the aerodynamic properties of the particles.

Atomizer performance directly influences the particle size of the dried particles. Four different types of atomizers are used for the majority of industrial spray drying applications; rotary atomizers, pressure nozzles, two-fluid nozzles (29,56), and ultrasonic atomizers (57,58). Typical droplet mass median diameters in pharmaceutical spray dryers range from less than 10 μm for pulmonary applications to upwards of 100 μm, which translates to a typical dry particle diameter range of 0.5 to 50 μm. Atomizer performance as a function of solution properties and operating conditions has been described elsewhere (29).

The key to successful particle engineering is understanding and controlling the mechanisms that determine the radial distribution of components during the drying process. Several different driving forces that may be responsible for separation of components have been suggested. Surface activity may lead to preferential adsorption of components on the droplet surface, causing a diffusional flux toward the surface. On the other hand, as the evaporating droplet shrinks its receding droplet surface leads to increasing solute concentrations at the surface. This causes a diffusional flux away from the surface (59). The diffusional motion inside an evaporating droplet must be correctly described to predict the resulting radial distribution of the individual components.

The evaporation of a solution droplet in a spray dryer is a coupled heat and mass transport problem. The process is driven by the difference between the vapor pressure of the solvents and their partial pressure in the gas phase. The rate of evaporation is determined by the balance of energy required to vaporize the solvent and the energy transported to the surface of the droplet. Various numerical models have been presented and reviewed (60–62) that describe the evaporation of multicomponent droplets in detail. Here a simplified analytical treatment is summarized.

In the absence of internal convection the distribution of the chemical components in an evaporating droplet can be described by the non-linear diffusion equation. For constant diffusion coefficients and negligible interactions between the solutes the diffusion equation becomes Fick’s second law of diffusion (42). This equation in its one-dimensional, radially symmetric form has an analytical solution, if the droplet evaporates at a constant evaporation rate (63). The evaporation rate, κ, is defined as (43,64) 4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d^2 \left( t \right) = d_0^2 - \kappa \,t$$\end{document}

The droplet surface area decreases linearly with time. This equation can also be used to approximate the droplet drying time (64) which should be smaller than the residence time of the droplets in the dryer to allow completion of drying. 5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau _{\text{D}} = \frac{{d_0^2 }}{\kappa }$$\end{document}

The solution of the diffusion equation can be rearranged so that it yields the surface enrichment, E_i, which is the surface concentration of component i in relation to its average concentration in the droplet (50) 6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E_i = \frac{{c_{s,i} }}{{c_{m,i} }} = \frac{{\exp \left( {{\text{0}}{\text{.5}}Pe_i } \right)}}{{3\beta _i }},$$\end{document}

Here a dimensionless Peclet number, Pe, has been used to simplify the equation: 7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Pe_i = \frac{\kappa }{{8D_i }}$$\end{document}

β is a function that must be integrated numerically for each Pe number to obtain an exact solution (50). However, an approximate expression for Eq. 6 with an accuracy of ±1% in the range Pe < 20 can be used instead: 8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E_i = \frac{{c_{s,i} }}{{c_{m,i} }} = 1 + \frac{{Pe_i }}{5} + \frac{{Pe_i^2 }}{{100}} - \frac{{Pe_i^3 }}{{4,000}}$$\end{document}

The average concentration follows from a mass balance. 9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$c_{m,i} = c_{0,i} \left( {1 - \frac{t}{{\tau _{\text{D}} }}} \right)^{ - \,\frac{3}{2}} $$\end{document}

With these equations, the surface concentrations of the components can be predicted at any time in the evaporation process as a function of their Peclet number. A very useful parameter, τ_sat,i, the time for a component to reach saturation at the surface, can now be derived. 10a\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau _{{\text{sat}},i} = \tau _{\text{D}} \left( {1 - \left( {S_{0,i} E_i } \right)^{\frac{2}{3}} } \right)$$\end{document}10b\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S_{0,i} = \frac{{c_{0,i} }}{{c_{{\text{sol}},i} }},$$\end{document}where the initial saturation of the components, S_0,i, is the second dimensionless parameter that directly influences the particle formation process.

Components may not crystallize immediately when their solubility limit is reached at the surface. Further characteristic times are associated with crystal nucleation, crystal growth, and polymorph transitions. For components that do not readily crystallize these characteristic times are frequently longer than the “precipitation window” defined by 11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau _{{p,i}} = \tau _{D} - \tau _{{{\text{sat}},i}} = \frac{{d^{2}_{0} }}{\kappa }{\left( {S_{{0,i}} E_{i} } \right)}^{{\frac{2}{3}}} .$$\end{document}

In this case a fully or partially amorphous solid is formed. Here another parameter becomes useful: τ_t,i, the time it would take a component to reach a concentration at the surface that is equal to its true density in the dry state. While this is obviously a hypothetical case in a multi-component droplet, this characteristic time can still be used for a rough estimate of the time at which an amorphous shell at the surface is formed. 12a\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau _{t,i} = \tau _D \,\left( {1 - \left( {P_{0,i} \cdot E_i } \right)^{\frac{2}{3}} } \right),$$\end{document}12b\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_{0,i} = \frac{{c_{0,i} }}{{\rho _{t,i} }}$$\end{document}

The characteristic times given above are useful for particle design, because they allow an estimate of the sequence of precipitation events at the surface and the time that is available for crystallization of components. The equations can only be evaluated if a reasonable estimate for the droplet evaporation rate can be found. Calculation of evaporation rates as a function of the processing parameters is not a trivial problem in a realistic spray dryer. Here a simple approximation is given that uses the following equation (45) for single droplet evaporation 13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\kappa = 8D_{\text{g}} \frac{{\rho _{\text{g}} }}{{\rho _{\text{l}} }}\left( {Y_{\text{s}} \left( {T_{\text{e}} } \right) - Y_\infty } \right).$$\end{document}

Please see Table I for nomenclature. Although it neglects Stephan flow (65), this expression is still a good approximation as long as the partial pressure of the evaporating solvent remains small, i.e. for droplet temperatures that are much lower than the boiling point of the solvent. At high droplet temperatures the effects of Stephan flow should be included. Adequate expressions for the latter case are given in the literature (43,50,66). To find the mass fraction at the surface of the droplet, Y_s, the equilibrium temperature, T_e, of the evaporating droplet must be determined. This can be done by iteratively solving the equation 14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{\Delta \hat Hv}}{{\hat c_p }} = - \frac{{Le \cdot \left( {T_\infty - T_e } \right)}}{{Y_s \left( {T_e } \right) - Y_\infty }}$$\end{document}in combination with a Clausius–Clapeyron relationship. The Lewis number, Le, is the ratio of the thermal diffusivity and the mass diffusivity in the gas phase (42). However, in the most common case of droplets evaporating in dry air or nitrogen the iteration can be avoided and the problem becomes very simple. A useful first approximation for the equilibrium temperature is an empirical correlation (60) for the wet bulb temperature of the droplet as a function of gas temperature, T_G, and boiling temperature of the solution, T_b. 15\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T_{{\text{wb}}} = 137\left( {\frac{{T_{\text{b}} }}{{373.15}}} \right)^{0.68} \log \left( {T_{\text{G}} } \right) - 45$$\end{document}Table I

Nomenclature

Once the droplet temperature is determined, the vapor pressure or mass fraction at the surface of the droplet can be found using an Antoine equation (50) or tabulated data. Finally, the evaporation rate follows according to Eq. 13, and the Peclet numbers and the characteristic times for the components can be calculated. Frequently, no data is available for the diffusion coefficients of the solutes, but an estimate based on the Stokes–Einstein equation is usually sufficient.

Several corrections, e.g. for droplet–droplet interactions or for convective flow around the drying droplets would need to be introduced to describe the actual situation in a spray dryer. Adequate models describing droplet evaporation in spray dryers have been developed in combustion research (67) and the food industry (68,69). Modeling of flow fields in spray dryers has been accomplished (34). The usefulness of the simple approach presented here lies in its potential to provide a quick overview of general trends for a particular particle design problem. It allows an assessment of the relative importance of various process parameters and material properties. This treatment cannot describe the processes that occur during the solidification of the particles, such as precipitation or buckling of shells (70), and it does not account for surface activity. However, it can approximate the radial distribution of all components prior to solidification. This may be predictive of the structure of the dry particle, as has been pointed out in an excellent study on the drying of milk powder (59). Hence, this model allows an estimate of which component is likely to act as a shell former.

A comprehensive theoretical description of particle formation during spray drying has not been published. The drying process of micro-droplets proceeds through non-equilibrium states where material properties are frequently unknown and experimentally difficult to access. Therefore, much of what is known about microparticle drying has been derived experimentally. It is worthwhile to review the experimental techniques that have been used to study droplet evaporation and particle formation. When interpreting the results of these studies it is important to consider the inherent limitations of the particular methods.

Various experimental techniques for the investigation of droplet drying and particle formation have been introduced. Studies using actual processing equipment are difficult to execute and interpret, because of the complexity of the two-phase flow in spray dryers, the difficulty of installing adequate analytical instrumentation into a development dryer, and the large number of processing and formulation variables. Consequently, most of the techniques study more or less simplified systems that are models of the actual situation in a dryer.

Arguably the most simplified system is a single droplet in quiescent gas. Various levitation techniques, using optical, acoustic, or electrodynamic forces can be used to freely levitate single microdroplets (71,72). Non-contact probing of the evaporating droplets allows measurement of evaporation rates with minimal interference. The shortcoming of the technique is in the time-scales that can be probed. It is difficult to introduce and levitate microdroplets in the experimental apparatus quickly enough to allow observation of a phenomenon with a characteristic time on the order of 10 ms. Most studies on freely levitated droplets have used either larger droplets or liquids with very low vapor pressure that evaporate so slowly that the initial droplet capture and stabilization becomes short compared to the time of the measurement. A second disadvantage of this method is that shrinking and solidifying droplets experience a significant change in properties such as refractive index, charge state, or sphericity. This can lead to difficulties in maintaining a stable levitation until the end of the evaporation process.

An interesting variant of single droplet levitation uses the Leidenfrost phenomenon to trap droplets on a concave hot plate (70,73). This technique was successfully used to study shell buckling during particle formation. A drawback of this approach is that the flow and temperature fields near the droplet are not similar to those of a free flowing droplet in a spray dryer.

The latter complication has been addressed by studying droplets suspended from thin filaments. This technique was used to observe the particle formation process of millimeter-sized droplets; particle density and morphology as a function of latent heat of crystallization, solubility, and drying rate were investigated (74). A recent study used a similar experimental set-up with an added capability to measure the surface adhesion of single, partially dry particles as a function of particle composition and moisture content (69). A refined filament suspension technique was applied to monitor the drying process of milk droplets (75–78). The morphology of dried particles was studied using suspended droplets for more than forty years (62,79–84), and much useful information has been derived from these studies. However, the use of a filament to suspend droplets has limitations. Heat and mass transfer can be affected by heat conduction between droplet and filament. The technique requires relatively large droplets, typically in the millimeter diameter range. This is problematic, because the processes involved in droplet drying scale differently with diameter. Droplet behavior observed on millimeter sized droplets may not be representative of microdroplets that are of interest in pharmaceutical applications.

Realistically sized droplets have often been studied in small scale spray dryers (55,85). These studies have shed light on many individual aspects of the drying process. The importance of the ratio between droplet evaporation rate and diffusional motion of the solutes was recognized and the concept of the Peclet number has been used to explain low density particle morphology (86). It was also found that feed solution concentration (87) and solubility of the excipients affect particle morphology (88–90). Several studies show that precipitation kinetics and crystallization play an important role and may be affected by evaporation rate (91,92). It has also been pointed out that drying temperature has a strong effect on particle morphology (93,94).

Some researchers have used a chain of monodisperse free falling droplets to study heat and mass transfer, drying, and particle formation processes (80,95–98). With this technique the effect of drying rates on particle formation and the formation of internal voids in particles have been studied. The droplet generator commonly used in this technique produces closely spaced droplets, which may lead to droplet–droplet interactions (99–101), because the evaporation process is determined by gas phase transport processes. Furthermore, the studies were still carried out on relatively large droplets with a diameter >170 μm (102).

Recently, an improved droplet chain technique has been introduced (50) that is based on a droplet-on-demand generator. In this technique the droplet distance is a free variable and can be chosen so that droplets do not measurably interact. It is also capable of generating droplets in the diameter range useful for respiratory delivery. Basic particle formation mechanisms have been derived with this technique, which are categorized in the next section.

Low density or hollow particles are advantageous for several applications, but specifically for pulmonary drug delivery where they improve dispersibility and delivery efficiency by lowering the aerodynamic diameter of the particles. Several particle design strategies have been developed to create voids in particles to lower the density. A hollow sphere can be created by using a shell forming excipient in combination with suitable processing conditions as described above. Irregular spherical particles may have external void space in folds or dimples. A corrugated surface further assists in dispersibility by minimizing contact areas. The design of such wrinkled particles is reviewed in the next section and creation of internal void space by alternative methods in the following paragraphs.

The term microencapsulation has been used in various ways by researchers and developers in food and pharmaceutical sciences. It often simply means that a substance is incorporated into a microparticle and no further description of the particle structure and its effect on the performance of the product is attempted. While in the food industry microencapsulation is often associated with retention of flavor or aroma substances during drying and storage (205–207), pharmaceutical applications usually apply microencapsulation to control the release, improve bioavailability, e.g. through mucoadhesion (208–211), or to mask the taste of an active pharmaceutical ingredient (212,213). The term microencapsulation implies that particles have an outer layer that contains or protects the interior, but this is only one among several particle morphologies that have been reported. Encapsulation in a wider sense can also be accomplished with a homogeneous particle where the matrix itself delays dissolution and release, or with composite particles, where smaller embedded particles or capsules may be covered by additional layers with a barrier function (114,214–216).

Modulating the release of volatile or soluble components from a microparticle in the presence of an external driving force is a difficult task. The release rate from a microparticle is influenced by complex interactions between body fluids, the encapsulant, and the encapsulated material (217). Dissolution, erosion and diffusion of material in the particle not only depend on material properties such as solubility and diffusion coefficients, but also on the morphology of the particle. The rate of diffusion through a layer of encapsulant, for example, depends on its thickness and surface area. In the case of microparticles the layer thickness tends to be variable and on the order of 1 μm or less, and the specific surface area is very large, in particular if pores are present. The small length scales associated with microparticles favor rapid release or an unacceptable burst effect unless the associated diffusion coefficients and solubilities are very low. Hence, materials investigated for microencapsulation of spray dried particles are typically high molecular weight, slowly biodegradable polymers such as polylactide (89,114,118,214,218–224), polylactide-co-glycolide, polycaprolactone (219,225–228), carbomers (229), and methacrylate copolymers (213,230–237), or polysaccharides (238) such as chitosan (215,216,229,239–257), cellulose derivatives (132,213,215,216,229,237,243,245,258), and hyaluronan (238,259), or cross-linked proteins (260). Polylactide-co-glycolide, PLGA, has been studied most intensively and has been used to encapsulate small molecule (94,117,220,224,261–273) and protein (58,219,274–280) therapeutics and various antigens (218,223,227,281–287) for vaccination purposes.

Detailed reviews of different microencapsulation strategies for the purpose of controlled release and adjuvanticity in therapy and vaccination are available (33,249,288–295). This review focuses on a different application of microencapsulation; the use of shells or coats to improve the dispersibility of powders for pulmonary delivery. One approach in particular has attracted significant attention, the use of leucine and trileucine as dispersibility enhancers.

Every pharmaceutical product must remain efficacious and safe during its shelf life. For microparticle products this requirement tends to be complex, because not only the active and the excipients need to remain chemically stable, but also the physical stability of the dosage form must be accomplished. Microparticles produced by spray drying are often in an energetically unfavorable state, because of their large surface area. Also, they frequently have not reached an equilibrium state, because of the short characteristic times of the drying process. Given sufficient mobility, microparticles tend to convert to an energetically more favorable state, either by crystallization (316–320), polymorph transition (321), crystal growth, or fusion of particles (322). All of these processes usually lead to product failure. Chemical stability can be lost through many routes depending on the nature of the active. Some adverse influences like light exposure or moisture and oxygen induced reactions can be mitigated through external protection, e.g. desiccants or oxygen scavengers. However, it is usually preferable to plan for chemical stability during particle design. For degradation that is driven by air–liquid interface exposure, aggregation, or unfolding of proteins there is no other viable option. For biotechnology application particle design also may need to consider biological stability of live organisms. The reader is referred to recent reviews for an introduction into the various stabilization approaches that have been developed (323) and the underlying concepts of physical forms of dry dosage forms (41,314), specifically the nature of amorphous solids (324–326). This review focuses on the particular issues arising in stabilization of spray dried microparticles.

Physical stability of microparticles can be accomplished by designing either a fully crystalline or an amorphous particle with high viscosity (327). Viscosity in an amorphous glass is characterized by the glass transition temperature. Amorphous glasses possess high viscosity if kept well below the glass transition temperature, which is typically designed to be 40 to 50°C above the storage temperature. This strategy, known as glass stabilization, has also been used to provide chemical stability for the active pharmaceutical ingredient, in particular for biologicals. Excipients used for glass stabilization are usually saccharides, polyols, or organic salts that form glasses of high viscosity. Embedding molecules in a rigid amorphous matrix slows degradation reactions by hindering translational motion, but it may not be sufficient to slow fast local rotational and vibrational motions of the molecules. A water replacement strategy may be necessary to achieve stability (106,328). This strategy uses excipients capable of replacing the hydrogen bonds that were broken due to desiccation, or capable of slowing down fast dynamics in glasses. In case of cell or virus formulations, the situation is further complicated by the fact that the inside of the organism can only be protected if stabilizers are transported across its membrane and the difference between interior and exterior environment does not cause damage to the integrity of the cell or virus. Amino acid (35) and saccharide excipients such as glucose (329), sorbitol (106), lactose (330), sucrose (331–337), or trehalose (106,125,182,282,283,332,334,337,338) have been found to be particularly effective in desiccoprotection and have been used in spray drying applications.

When designing stable microparticles manufactured by spray drying, one must consider several challenges that are specific to the microparticle formation process: As discussed above, components in a microparticle tend to separate due to their different diffusional velocities if their Peclet numbers are sufficiently different. It is clearly disadvantageous if stabilizers separate from the molecule they are supposed to stabilize. This mechanism has mostly been neglected in discussions about glass stabilization of microparticles. Rapid droplet evaporation tends to produce dry particles where the composition is a function of the radius. Hence, the glass transition temperature is also expected to be different for the core and the shell of a layered particle (84).

A second mechanism that can lead to unfavorable separation of components is surface activity. A surfactant that is successful in keeping air–liquid interface sensitive molecules away from the surface may unintentionally produce a coat with a low glass transition temperature on the surface that causes unacceptably high cohesive forces between the particles. Alternatively, a surface active protein may at least partially be enriched on the surface, leading to different stabilizer-protein ratios on the surface versus the core of the particle (131).

A third separation mechanism is the sequential precipitation of components due to their different characteristic times to saturation. The situation is further complicated by the fact that the precipitation window for typical spray drying applications is on the order of the characteristic times for crystal nucleation, crystal growth, or polymorph transitions.

A good example for this effect is the variety of physical forms of mannitol that have been observed in spray dried microparticles. Whether mannitol crystallizes during spray drying or subsequently on storage depends on several factors. In a salmon calcitonin–mannitol system the mannitol remained amorphous when the formulation contained less than 50% mannitol, but crystallized when more mannitol was added (321). Similar observations were made during the development of a spray dried monoclonal antibody. Mannitol crystallinity depended on formulation composition (182) and crystallization on stability could be retarded by adding an inhibitor (339). An insulin–mannitol formulation produced crystalline mannitol, whereas an otherwise similar RNAse-mannitol formulation produced amorphous mannitol (340). The sensitivity of the mannitol crystallinity to composition and processing conditions can be explained by similar characteristic times for crystallization and droplet evaporation in typical pharmaceutical applications. This hypothesis is supported by the observation that different polymorphs and an amorphous mannitol fraction were simultaneously present in a salmon calcitonin–mannitol formulation at high mannitol fractions (321,341). According to the Ostwald step rule, crystallization is initiated by forming the least stable polymorph with subsequent transitions to the most stable form. However, if the process cannot proceed to equilibrium, because the loss of mobility occurs faster than the transition to the most stable form, the resulting solid state of the powder is a mixture of non-equilibrium states, which were arrested depending on the length of the precipitation window.

Partially crystalline systems or presence of more than one polymorph has also been observed for other spray dried substances like glycine (342), sucrose (34), polyethylene glycol and lactose (343,344), cromolyn sodium (345), or acetazolamide (346). Particles may also have separated amorphous phases (347), a possible explanation for the destabilizing effect of sucrose at high sucrose to protein ratios. Partially crystalline systems in particular are undesirable for two reasons. Firstly, since the more stable solid form has already been nucleated, the tendency to convert the remaining less stable fraction during storage is increased. This may lead to inferior stability. Once the amorphous fraction starts to crystallize, the water contained in it is mostly expelled and increases the mobility in the remaining amorphous fraction through plasticization. This accelerates the crystallization (348), which usually causes rapid failure. Secondly, a solid dosage form whose physical state is sensitive to small changes in the timing of the drying process is hard to commercialize. Scale-up becomes very difficult, because the longer residence time typical for larger dryers is likely to alter the physical state and, thus the stability of the product (34).

The same mechanisms that lead to the challenges outlined above present opportunities, which can be exploited to design structured microparticles with unique properties. Because of the typically short precipitation window, crystallization of substances during spray drying often leads to very small crystals. While this can pose a stability risk due to the tendency of nanocrystals to grow on stability by an Ostwald ripening mechanism, the unique properties of nanocrystals can also be beneficial, e.g. leading to an increase in bioavailability (349) for low solubility molecules. Spatial separation of material with high and low viscosity can be used to design structured microparticles that possess much better physical stability than a homogeneous particle with the same composition. Typical examples are gel particles, known as dry emulsions, where liquid nanodroplets are stably encapsulated in an amorphous matrix with high glass transition temperature (48,129). Another example of advanced glass stabilization is shown in Fig. 10. An active pharmaceutical ingredient with low glass transition temperature was physically and chemically stabilized in an amorphous, structured microparticle. The glass transition temperature of the homogeneously mixed formulation was estimated to be suppressed to approximately 5°C at the 60% RH stress condition due to plasticization. The storage temperature was at least 20°C above the glass transition temperature, which in theory should lead to physical instability of the powder in a matter of minutes. On the contrary, the aerosol performance of this powder was still largely unchanged after 7 days of exposure. The most likely explanation for this observation is that the particles have a core–shell structure like the one shown in Fig. 1g. The calculation of the glass transition temperature assumed a well-mixed glass and a homogeneous particle, while in reality the particle had a core with low viscosity encapsulated by a stable shell of high viscosity. Exceptional physical stability of the powder was the result. Fig. 10

Particles designed for high physical stability, containing 56% of an encapsulation agent forming the shell, 20% saccharide, 20% active pharmaceutical ingredient with low glass transition temperature, and 4% inorganic salt.

The examples given in this review show the level of sophistication that can be achieved with a seemingly simple process such as spray drying. Even spray drying from a homogeneous solution, if properly understood, can be used to design particles that have a multifunctional internal structure. Potential further advances in particle engineering may be achieved by combining spray drying with additional processing steps before, during, or after the drying step.

The most common modification to a simple solution spray drying process is the preparation of a suspension or colloid prior to spray drying. Two-phase systems are often necessary because of solubility limitations of active and excipients, but they are also employed to transfer some of the structural properties of the two-phase system onto the final dry particles. A typical example is spray drying from an o/w emulsion to manufacture PulmoSpheres™, where the droplet size of the dispersed phase influences the void size in the dried particles. Ordinary emulsions, mostly o/w emulsions (46–48), and immiscible aqueous (359) and non-aqueous (360) two-phase systems have been studied. Multiple emulsions have also been successfully spray dried, e.g. an O/W/O/W triple emulsion for the encapsulation of orange oil (361). Spray drying of sols (362–365) or suspended solids is a variation of this theme.

Again, the use of a solid dispersion is often dictated by poor solubility and the dispersed solid may simply be micronized or unprocessed material (366–371). Increasingly, though, the dispersed phase has been produced in yet another processing step and possesses specific properties that translate to the final dried particle. Dispersions of small unilamellar liposomes have been spray dried to produce highly dispersible powders for possible pulmonary delivery of superoxide dismutase (372). Spray dried suspensions of dextran microspheres (373) were investigated in the context of nasal drug delivery, pre-formed microcapsules were spray dried to encapsulate phase change materials (374), and dispersions of porous and non-porous silica micro and nanoparticles were spray dried in combination with indomethacin (375), tolbutamide (376), and simvastatin (377). DNA in combination with lyoprotectants was lyophilized, micronized, and subsequently spray dried as a solid-in-oil dispersion to achieve encapsulation in PLGA (378).

A very exciting development is the design of composite particles spray dried from dispersed nanoparticles (51,52,70,73,86,119,365,379–384) or even mixtures of different nanoparticles (385). The nanoparticles are typically manufactured separately prior to preparation of the dispersion and retain their special properties such as enhanced solubility or targeting capacity for the final product. Some of their disadvantages, e.g. their tendency to agglomerate or their unfavorable aerosol transport properties, are mitigated by the fact that they are now part of a larger microparticle (382). It is interesting that the properties of the product are now determined by the primary morphology of the composite particle, the morphology of its nanoscale substructure, and their interactions. This highlights the importance of proper particle design in these systems.

Rather complex dispersion systems with more than one solid dispersed phase have been studied. Calcium salts were dispersed in an aqueous polymeric suspension and then spray dried to produce microencapsulated particles (386). A second example with multiple pre-processing steps is based on a complex dispersion of polymer nanoparticles and micronized fumed silica that was spray dried to produce nanoparticle-coated microparticles (387–393). In this system the drug may be associated either with the organic nanoparticle phase or the inorganic silicon dioxide phase. Several complex dispersions based on o/w emulsions with additional suspended solid material have also been tried. The solid dispersed phase consisted either of the micronized active (181,184), precipitated ethylcellulose (216), or highly porous magnesium aluminometasilicate particles (394).

In the second class of modified spray drying additional steps are added during atomization or drying. Typical examples are atomizer nozzles that allow the mixing of fluids immediately prior to or simultaneous with the atomization process (395,396), or multi-zonal drying chambers, where the droplet evaporation is followed by a secondary drying or firing step that can change the solid state of the particles (397) or remove placeholder material to create porosity (382,398,399). Sophisticated processes, in which spray drying from complex dispersions is coupled with a multi-zonal dryer with three different temperature zones have been presented (400).

Lastly, post-processing steps can be added after spray drying. While secondary drying is common to adjust the moisture content of powder, more intricate processes such as microencapsulation (401) have also been proposed. Developments in the area of bioactive surface layers seem particularly promising. For example, the surface of spray dried microparticles has been loaded with covalently bonded antibodies (402) or adsorbed DNA (403,404) for immunomodulation purposes.

A thorough review of the large number of combination processes where spray dried microparticles play a role is beyond the scope of this review. It will be interesting to follow whether the added functionality of the particles fabricated with such processes will justify their increased complexity in a commercial setting. It is already obvious that the new class of engineered spray dried particles requires a shift in the thinking and structure of pharmaceutical organizations developing them. Because both process and formulation now equally determine key product parameters, it should be reconsidered whether separate groups working on formulation and process development, optimization and scale-up are adequate. This new class of products demands a unified design approach reflected in an organizational structure that allows constant collaboration of formulators and process engineers in a single group. It also becomes evident that purely empirical approaches to product design are doomed to fail in a situation where the parameter space is defined by a very large number of formulation and process variables. Successful design of engineered particles requires thorough understanding and predictive modeling, so that the early development process can be completed in an acceptable time with a high likelihood of success. Much of the fundamental work, including the measurement of urgently needed material properties remains to be done.