Physical characterization of component particles included in dry powder inhalers. I. Strategy review and static characteristics


  • Anthony J. Hickey,

    Corresponding author
    1. Division of Molecular Pharmaceutics, School of Pharmacy, University of North Carolina, Campus Box #7360, 1310 Kerr Hall, Kerr Hall, Chapel Hill, North Carolina 27599-7360
    • Division of Molecular Pharmaceutics, School of Pharmacy, University of North Carolina, Campus Box #7360, 1310 Kerr Hall, Kerr Hall, Chapel Hill, North Carolina 27599-7360. Telephone: (919) 962-0223; Fax: (919) 966-0197.
    Search for more papers by this author
  • Heidi M. Mansour,

    1. Division of Molecular Pharmaceutics, School of Pharmacy, University of North Carolina, Campus Box #7360, 1310 Kerr Hall, Kerr Hall, Chapel Hill, North Carolina 27599-7360
    Search for more papers by this author
  • Martin J. Telko,

    1. Division of Molecular Pharmaceutics, School of Pharmacy, University of North Carolina, Campus Box #7360, 1310 Kerr Hall, Kerr Hall, Chapel Hill, North Carolina 27599-7360
    Search for more papers by this author
  • Zhen Xu,

    1. Division of Molecular Pharmaceutics, School of Pharmacy, University of North Carolina, Campus Box #7360, 1310 Kerr Hall, Kerr Hall, Chapel Hill, North Carolina 27599-7360
    Search for more papers by this author
  • Hugh D.C. Smyth,

    1. Division of Molecular Pharmaceutics, School of Pharmacy, University of North Carolina, Campus Box #7360, 1310 Kerr Hall, Kerr Hall, Chapel Hill, North Carolina 27599-7360
    Current affiliation:
    1. Currently, College of Pharmacy, University of New Mexico, Albuquerque, NM 87131.
    Search for more papers by this author
  • Tako Mulder,

    1. DMV-Fonterra Excipients, Goch, Germany
    Search for more papers by this author
  • Richard McLean,

    1. Pfizer Inc., Sandwich, Kent, UK
    Search for more papers by this author
  • John Langridge,

    1. DMV-Fonterra Excipients, Goch, Germany
    Search for more papers by this author
  • Dimitris Papadopoulos

    1. Pfizer Inc., Sandwich, Kent, UK
    Search for more papers by this author


The performance of dry powder aerosols for the delivery of drugs to the lungs has been studied extensively in the last decade. The focus for different research groups has been on aspects of the powder formulation, which relate to solid state, surface and interfacial chemistry, bulk properties (static and dynamic) and measures of performance. The nature of studies in this field, tend to be complex and correlations between specific properties and performance seem to be rare. Consequently, the adoption of formulation approaches that on a predictive basis lead to desirable performance has been an elusive goal but one that many agree is worth striving towards. The purpose of this paper is to initiate a discussion of the use of a variety of techniques to elucidate dry particle behavior that might guide the data collection process. If the many researchers in this field can agree on this, or an alternative, guide then a database can be constructed that would allow predictive models to be developed. This is the first of two papers that discuss static and dynamic methods of characterizing dry powder inhaler formulations. © 2007 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 96: 1282–1301, 2007


The behavior of particles is the foundation on which dry powder inhaler (DPI) performance is built. Interest in particles may be viewed as one of the oldest scientific activities since the degree of subdivision, or size, is all that separates nanoparticles from boulders.1 Pharmaceutical powders are, in the first instance, of greatest importance for oral dosage forms where particle behavior is crucial to the processes involved in tablet and capsule manufacture and ultimately in their dissolution and drug availability. This interest extends to aerosol products that are highly dependent on the physico-chemical and performance characteristics of the particles that deliver drugs to the lungs.

Aerosol particles are prepared in respirable sizes (<5 µm) and exhibit unique properties based on the forces of interaction which are known to occur.2 Classically dry powder aerosol formulations are prepared as blends with large lactose carrier particles.3 The forces give rise to particle–particle and particle–surface interactions as depicted in the schematic shown in Figure 1.3–5 This is a simplified view of the possible nature of particle interactions but illustrates the likely complexity of a situation in which particle forces are confounded based on surface physical and chemical heterogeneity. The usual particulate interactions of electrostatic, capillary and van der Waals (proximity) forces and mechanical interlocking3, 4 occur but are confounded with the additional complexities associated with the solid state of the presence of amorphous regions, impurities (in lactose mostly proteins and some fats) and specific polar/non-polar regions (dispersive, acid/base energetics).

Figure 1.

Causes of the various types of interactions between fines and carrier particles.

One of the first manifestations of the effect of various forces of interaction on particles is that observed in packing (bulk and tapped density) differences and variability in flow. Flow has been shown to be a key property of the DPI formulation as it aids in metering, fluidization, and dispersion.6, 7 In itself it is not a variable, but a dependent variable based on particle size, shape and forces of interaction. The interplay of these factors is difficult to address directly but it is possible that using complementary techniques a framework can be constructed from which the nature of particles can be used to predict powder performance.

Figure 2 outlines a number of methods that may be used to characterize powders intended for use in inhalers. For ease of reference these techniques have been divided into local and bulk surface analysis, static and dynamic bulk characteristics and performance measures. In the last decade a number of publications have appeared which attempt to correlate certain properties with efficiency of inhaler performance. However, it is clear that the quantity of data required to conduct a thorough assessment may prohibit comprehensive assessment in a single study or by a single research group.

Figure 2.

Methods and analyses to establish general physicochemical properties (PCA), surface analyses, bulk characteristics and performance properties.

The following sections describe the principles behind specific techniques and include analysis of a range of lactose particles intended as carriers for drug (e.g., albuterol sulfate25) to illustrate the application of these techniques and the conclusions that can be drawn. Figure 3 shows the basis for inclusion of lactose in albuterol DPI formulations. Lactose is intended to act as a diluent, allowing metering of small quantities of drug in larger, manageable quantities of lactose.26 In addition, since lactose carrier particles are large (>50 µm) they act as fluidizing agents in the dispersion of the drug and when acted on by airflow shear forces facilitate the generation of primary particles of drug that can be carried on the inspiratory flow into the lungs of patients. The dispersion process is recognized to be a highly complex process that is far from being completely understood. It is generally recognized that the complex nature of the dispersion process is not attributed to any single specific factor. A body of knowledge suggests that fine lactose (i.e., small particles that are in the respirable size range) may occupy “active sites” on the surfaces of larger lactose particles. This is one factor that may influence dispersion. This paper is the first in a two-part sequence that discuss the range of methods, outlined in Figure 2, that might be employed to evaluate blend formulations intended for use in DPIs. In the following sections of this paper, a summary of specific methods employed for the static characterization of albuterol-lactose blends27–31 is given. Dynamic characterization of dry powder aerosols is described in the second paper25 of this two-paper series.

Figure 3.

Schematic diagram illustrating the detachment of fines from carrier particles as the static powder is dilated and aerosolized (left to right).


Solid State Characterization

In any formulation, including dry powder inhalation pharmaceutical aerosols, the solid-state32 of the drug and excipient are important aspects of the physical and chemical stability, pharmaceutical and therapeutic performance of the drug product. Briefly, it is important to consider crystalline solids,33–36 crystalline polymorphs,37, 38 crystal hydrates,39 amorphous,40–43 liquid crystals,44 nanocrystals, and solid-state phase transitions.1–20 Indeed, the subject of a local polymorphism with single crystals is extremely topical and the subject of some very recent debate.45–48 Solid-state phase transitions include polymorphic interconversions, polyhydrate interconversions, hydration-dehydration conversion, and order-to-disorder transformations. Additionally, these solid-state phase transitions can be thermodynamic transitions and kinetic transitions.

It is important to recognize that a number of other partially ordered/disordered phases49 are known to exist in the solid-state that can be of pharmaceutical significance given the chemistry of many pharmaceuticals and effects of pharmaceutical processing, and they include quasi-crystals, plastic crystals, disordered nanocrystals (i.e., non-amorphous disordered solids lacking long-range crystalline order),50 liquid crystals (thermotropic and lyotropic thermodynamically stable phases observed in surfactants, biopolymers, and biomaterials), and polyamorphism.51 It has been demonstrated that the amorphous form of an aerosolized drug has a tremendous effect on absorption in the deep lung,52 and, hence, it quite likely that the degree of order/disorder present in the solid aerosol particle, particularly at the surface of the aerosolized particle, has significant therapeutic and biological effects.

Water53–56 associated with the sugar carrier may interact with the drug through the interface between the drug and carrier; presence of a little water can have significant effects on drug surface molecular mobility/transformation/instability.57 A disordered phase, such as an amorphous phase having similar mechanical and physical properties of a supercooled liquid existing at temperatures below its thermodynamic crystallization temperature but has not been given sufficient time to anneal and crystallize to its thermodynamically stable ordered phase, inherently has a higher degree of molecular mobility.58, 59 Additionally, disordered phases, such as an amorphous40 phase or non-amorphous disordered nanocrystals, often originate at surfaces and interfaces, especially in pharmaceutical particles due to processing effects. Gas and vapor absorption occurs into these disordered regions located at the surface, and absorption phenomenon is known to accelerate where disordered surface regions are present. Water vapor absorption provides water molecules for participation in hydrolysis reactions, and oxygen gas absorption provides oxygen molecules for participation in oxidation reactions in the solid-state. Molecular mobility is furthered increased as water acts as a plasticizer, and further physical60 and chemical instability61 occurs over pharmaceutically relevant time-scales.

The classical methods of evaluating the solid state are X-ray powder diffraction14, 18, 37, 50, 62–65 and thermal analyses.39, 49, 66–70 X-ray diffraction of solid-state materials gives important insight, based on the degree of long-range order present, into the extent and nature of the crystallinity, microstructure, and nanocrystallinity with a limit of detection of about 10% to give a signal.63 Thermal analysis indicates polymorphs, hydrates, binding interactions, amorphous and thermotropic and lyotropic phase transitions, in general, based on the gain and loss of enthalpy (heat) that is, order-to-disorder (e.g., melting) and disorder-to-order (e.g., crystallization) phase transitions also with a limit of detection of around 10%.63 Thermal analysis has been employed in investigating lactose crystallinity,71 amorphous character,72, 73 physical studies on DPIs,74 water vapor-solid interactions,75, 76 and assessment of powder surface energetics.77, 78 Crystallization of lactose from the amorphous state79 and albuterol-lactose particulate interactions80 have been observed using atomic force microscopy (AFM), a powerful surface and nanoimaging analytical technique that will be presented in-depth. The presence of water molecules in the bulk powder can be assessed by Karl Fischer analysis to about 0.1% water content but more sophisticated approaches employ water vapor sorption,81–86 particularly to assess the presence of amorphous material63, 87 and to probe the nature of water-pharmaceutical solid interactions53–56, 60, 86, 88 at the molecular level, including phase transitions that are induced by the level of hydration present in the solid-state structure.86 This is a particularly important approach to the assessment of hygroscopic solids,89, 90 and aerosols.74, 91–98

Surface Analyses

Scanning Electron Microscopy (SEM)

SEM is recognized as unique tool in the visual examination of particles and their surfaces. The resolution is of the order of nanometers (magnifications in the range 20–100,000×). A fine beam of electrons of medium energy (5–50 keV) scans a gold-palladium coated sample producing secondary electrons, backscattered electrons, light or cathodoluminescence and X rays. The latter allow for X-ray microanalysis for specific elements. SEM is routinely used for imaging particles in the micron and smaller size range and for examining the surfaces of larger particles. The resolution allows identification of specific surface geometric features that are indicative of structural phenomena.

Atomic Force Microscopy (AFM)

AFM offers a unique opportunity to examine surface structure of a variety of materials with mesoscopic scale resolution (10−6–10−9 m), and quantify the individual particle and excipient interaction by direct force measurement in a variety of environmental conditions.99–101

The relevant adhesive/cohesive force components considered in DPI system are intermolecular van der Waals force, capillary force, and electrostatic force.102 The adhesion force measurement of powder includes vibration,103 centrifugation,104–106 and impact separation,107–109 before the advent of direct measurement of colloid probe technique.110, 111 Bulk adhesion leads to global adhesion characteristics,112 but DPI requires the knowledge of adhesion forces from microscopic analysis for the purpose of elucidating the fundamental mechanism such as work of adhesion and surface energetics,113, 114 and the prediction115 of formulation stability, and redispersion.

There are several factors that link the seemingly straightforward AFM colloid probe technique to the force measurement. They include the following: (1) the cantilever tip consistency; (2) the physical and chemical properties of colloid probe and substrate surfaces; (3) environmental issues, such as temperature and relative humidity; and (4) the contact area determination or normalization.

Perhaps the most important influence of adhesion/cohesion forces is related to the colloid probe and substrate surface roughness, which is mostly characterized by the root mean square deviation (Rrms) in a given area. There was evidence that adhesion of drug particles to carrier surfaces increased with surface roughness116 if particle deformation was negligible, or decreased,117 or an optimum surface roughness existed;118 but the adhesion still relies on the true contact area of interaction. The adhesion force distributed more widely when a rough surface was used because the asperity radius or the effective contact area is more scattered.119–121 The influence of surface roughness on the force distribution between single particles and both smooth and rough substrates has been reported.117, 122–124

Recently, the surface chemistry influence including morphology (amorphous/crystalline surface domains,125 polymorphs126, 127), surface free energy and work of adhesion,113, 128, 129 and hydrophobicity130 were studied extensively. By scanning simultaneously the AFM topographic and phase image, Price et al.131 studied the physical transformation of lactose crystal by milling process, and observed morphology and physico-mechanical changes (amorphous recrystallization) on the surface of crystalline material in accordance to the elevated humidity. Hooton et al.132 compared different polymorphs of sulfathiazole (I, III, IV) and their corresponding surface energy and work of adhesion (based on Johnson-Kendall-Roberts (JKR) theory133) determined against highly orientated pyrolytic graphite (HOPG) and polymorph particles. Ward et al.134 utilized both AFM and Raman microscopy to identify and map surface amorphous domains of sorbitol. Besides the AFM phase imaging, the analysis of the adhesion data over a given surface area (force volume scans135, 136) will give a distribution of adhesion 3-D profile.137 The introduction of the ternary system138 such as different surface fines139 or force control agents140 represents a surface property modification to produce a higher fine particle fraction (FPF) by enhancing aerosolization. Forces were examined using colloid probe technique and the cohesive-adhesive balance approach (CAB).141 Measuring interparticulate forces in liquid, together with the surface energetics measurement such as contact angle and inverse gas chromatography can be used for the characterization or prediction of suspension stability of pressurized metered dose inhalers (pMDIs).114, 142, 143

Particulate adhesion is a dynamic process with the increase/decrease of relative humidity (RH%). The capillary forces arise from moisture condensation in the gap between two contiguous surfaces, and become dominant as the humidity increases. Both increase and decrease of particulate forces were reported at elevated RH depending on three different contact asperity scenario (nano-contact128), possible long range electrostatic interaction,144 or morphology change (local recrystallization and particle fusing of micronized sulbutamol sulphate125), (amorphorization of zanamivir145) induced by moisture. The thickness of the adsorbed water146 affects the adhesion force and depends on the hydrophobicity of surface material.117

Both heterogeneous asperity (geometric variations in the contact zone) and heterogeneous surface energy will cause the logarithmic normal distribution of the forces.124, 147 Once the adhesion/cohesion forces are determined, drug particle surface energetics and interparticulate forces can be correlated.114, 131 The key issue for quantifying surface energies and work of adhesion by AFM is the characterization of the contact area between the probe and the substrate surface. One approach148 describes a probe tip self-imaging technique in which the geometry of the probe is recorded reversely. Association with the Young's modulus, of both probe and substrate, can then be used to estimate the contact area. Work of adhesion can be calculated according to the JKR theory.

Inverse Gas Chromatography (IGC)

IGC is a technique for studying solids using gas chromatography principles. A solid analyte is packed into or coated onto a chromatography column and a series of nonpolar and polar probe gases are eluted. Interactions between the gaseous probe molecules and the stationary phase result in a characteristic net retention volume, which is used in the determination of the free energy of adsorption and other thermodynamic surface parameters.

The technique has been used to study the adhesional properties of polymers,149 fibers,150 and composite materials.151 More recently IGC has been applied to pharmaceuticals, such as in the study of DPI152–154 and pMDI155 formulations, for which adhesional properties are thought to play a crucial role. IGC can be used for determination of surface energy and surface acid/base properties which directly influence adhesional properties. The technique for determining surface energy and acid/base interactions of surfaces by IGC are based on the work of Schultz et al.151 though some subsequent experimenters have deviated from the method.

Surface free energy is due to Lifshitz-van der Waals (LW) forces and specific acid-base interactions,156 which contribute to intermolecular forces independently.157, 158 Thus, total surface free energy, γS, can be represented as the sum of dispersive and specific (nondispersive) interactions as

equation image(1)

where γmath image designates the dispersive surface free energy, and γmath image the specific surface energy.

The dispersive component, γmath image, can be determined from the retention volumes of n-alkanes,159, 160 based on the following equation151

equation image(2)

where NA is Avogadro's number, A is the effective surface area of the probe molecule, γmath image and γmath image are dispersive free energies of interacting solid and probe, and C a constant that depends on the chosen reference state. Given that surface area and γmath image increase linearly for the homologous series of alkanes, a plot of RTlnVN versus equation image yields a line with slope equation image.

Specific free energy is determined from the retention volumes of polar probes. Using the same RT lnVN versus equation image plot, the specific free energy of adsorption is the difference between RT lnVN of the probe and the n-alkane line. Specific free energy data for different probes can be combined into two parameters related to the character of the interacting surface by way of the acid/base approach to molecular interactions.161 Based on this approach specific interactions are classified as either electron donor or electron acceptor type interactions. Donor and (adjusted) acceptor numbers, DN and AN*, represent the ability of a probe to donate or accept electrons from reference acceptors and donors.161, 162 According to this approach, the surface can be characterized by acid and base parameters via the equation

equation image(3)

where ΔHsp is the specific enthalpy of adsorption and KA and KB are the acid (acceptor) and base (donor) parameters of the studied surface, respectively. Many publications in the pharmaceutical literature152–154, 163–171 have used an alternative expression based on surface free energy rather than enthalpy given by the equation

equation image(4)

Use of Eq. (4) instead of Eq. (3) greatly simplifies the experiment, allowing the experimenter to determine KA and KB from data at a single temperature by plotting ΔGmath image/AN* versus DN/AN* for a number of probes. Determination of KA and KB from Eq. (3) requires experiments to be performed at different temperatures so that specific enthalpy of adsorption, ΔHmath image, can first be determined from the temperature dependence of ΔGmath image.

While the distinction between Eqs (3) and (4) appears trivial, data in our lab suggests it is an important one.172



Lactose monohydrate (Respitose™) batches of two milled batches (designated as ML A and ML B) and six sieved batches (designated as SV A, SV B, SV C, SV D, SV E, and SV F) were obtained from DMV-Fonterra Excipients. Table 1 indicates the physical properties of these excipients as supplied by the manufacturer.

Table 1. Physicochemical Properties of Eight Batches of Lactose as Supplied by the Manufacturer*
  • *

    Batches selected from a Principal Components Statistical Analysis of this data, which is supplied for the convenience of the reader.

Protein (Kjeldahl N* 6.24)2721889313619922512479
E 10% 1 cm, 400 nm0.0160.0080.0080.0130.0070.0080.0080.010
Sulphated ash0.
UV-absorption 210–220 nm0.0480.0370.0470.0490.0490.0630.0380.038
UV-absorption 270–300 nm0.0120.0140.0160.0150.0150.0140.0090.012
Acid value0.
Specific rotation55.355.355.354.955.
Particle size (Malvern; µm)
Specific surface area (m2/g)0.340.430.440.460.410.300.890.87

Two milled batches (ML A and ML B) and two sieved batches (SV A and SV D) lactose were used for the IGC experiments. Alkane probes used were hexane (99 + %, Aldrich, Milwaukee, WI), heptane (99 + %, Aldrich), octane (99.5 + %, Fluka, Bochs, Germany), nonane (99 + %, Aldrich), and decane (99 + %, Aldrich). Polar probes were chosen to cover a wide DN/AN* range; the probes were tetrahydrofuran (THF) (EM Science, affiliate of Merck KGaA, Darmstadt, Germany, 99.99%), chloroform (100%, Mallinkrodt), acetone (99.7%, Mallinkrodt, Phillipsburg, NJ), ethyl acetate (99.9%, Mallinkrodt), diethyl ether (99%+, Acros, Morris Plains, NJ), and ethanol (100%, Aaper, Shelbyville, KY).


Pharmaceutical particles are first characterized in terms of their physico-chemical properties, some of which are general, and are shown in the materials section for the powders described. However, additional properties may be studied more closely, as they have some significance for the performance of powders in inhalers. In particular, these characteristics relate to polymorphic behavior, physical and chemical transitions, including both thermodynamic and kinetic, and, hence, potential physical and chemical instabilities.

Solid State Characterization

X-ray powder diffraction: (Philips 1720 CuK X-ray) patterns were obtained to evaluate the crystallinity of the lactose (1–2 g) batches as well as assess the presence of the polymorph, b-anhydrous lactose.

Differential scanning calorimetry (DSC): employed approximately 18 mg samples of each of the eight lactose batches sealed in tared aluminum pans and scanned at 5°C/min from 50 to 270°C using a Perkin Elmer DSC 6 (Norwalk, CT). Thermograms were processed and analyzed using the accompanying software, Pyris Thermal Analysis Instrument Control and Data Analysis Software (v.3.01).

Surface Analysis

Scanning Electron Microscopy (SEM)

Scanning electron microscopy (SEM, model 6300, JEOL, Peabody, NY) was employed for imaging of morphology, size, and surface characteristics of sieved and milled lactose particles. An electron beam with an accelerating voltage of 15 kV was used at 600× magnification. The powders were placed on double-sided adhesive carbon tabs (Ted Pella, Inc., Redding, CA) which was adhered to aluminum stubs (Ernest F. Fullam, Inc., Latham, NY) and were coated with a gold-palladium alloy thin film (150–250 Å) using a sputter-coater (Polaron 5200, Structure Probe Supplies, West Chester, PA) operating at 2.2 kV, 20 mV, 0.1 torr (under argon) for 90 s. Analysis of the SEM micrographs was performed using NIH Image J software (National Institutes of Health, NIH, Bethesada, MD).

Micrographs were subjected to image analyses to facilitate estimation of the relative abundance of surface fine particles on a series of powder systems. The images used were at various magnification levels allowing good resolution of the fine particles while maximizing the field of view for adequate inclusion of the large carrier particles. Image processing included modifying the threshold value of the 8-bit digital image such that the surface fine particles could be distinguished from the textured background of the large carrier particles. Digitized image corrected for background threshold. Note: surface particles become dominant feature of image.

Because of the shadowing effects from SEM imaging and the presence of large particle edge effects, the particle size and counting analysis was only performed on particles that were comprised of 1–50 pixels. This would correspond to particle projected areas of approximately 0.2–10 µm2 (∼5 pixels/µm).

Atomic Force Microscopy (AFM)

Images were acquired using the Topometrix Explorer AFM (ThermoMicroscopes, Sunnyvale, CA) under ambient conditions (23–25°C; 35–40% RH). Steel stubs (Ted Pella, Inc.) were used for mounting samples. Cantilever mounting glue was superglue. Sample mounting glue was MikroStik™ (Ted Pella, Inc.). Silicon nitride cantilever tips (non-contact tips without coating) were PPP-NCL (Nanosensors™, Neuchatel, Switzerland) with the following specifications: Thickness: 7 µm (range: 6–8 µm); Mean width: 38 µm (range: 30–45 µm); Length: 225 µm (range: 215–235 µm); Force constant: 48 N/m (range: 21–98 N/m); Average Resonance Frequency: 190 kHz. Nanotopographic images were obtained for all batches (six sieved and two milled). Scan rates were done at 5 and 10 µm/s in non-contact acquisition mode. Scan ranges were 10 µm × 10 µm, 5 µm × 5 µm, and 1 µm × 1 µm. Images with resolution set at 100, 300, 400, and 500 were obtained. Best images were obtained by using a combination of a slower scanning rate and higher resolution. Sensitivity to electromagnetic waves and vibration increased significantly at these optimal settings. ThermoMicroscopes SPM lab analysis software (ThermoMicroscopes) and Gwyddion software were used in analyzing the AFM micrographic images.

Sample preparation was carried out by mounting to steel discs at or near their plane of maximum stability by using the following procedure:

  • (1)A small amount of powder was dropped from a height (0.5 m) onto clean overhead projector transparencies.
  • (2)Powder sample discs were painted with MikroStik™ adhesive until excess solvent had visibly evaporated.
  • (3)The disc was inverted (adhesive side down) on a position on the transparency that contained a dilute region of powder.
  • (4)Un-adhered particles were removed by gentle tapping of disc on bench.
  • (5)This method achieves good particle dilution (easy particle optical identification during AFM imaging), and particles are typically adhered in the plane of imaging (surface facing up) to facilitate non-contact topographical imaging.

Inverse Gas Chromatography: experiments were conducted with a Hewlett-Packard 5890 Series II gas chromatograph with flame ionization detector. The chromatograph was modified to allow installation of straight 205 mm, 4 mm ID glass columns. Dry N2 was employed as carrier gas at a flow rate of 30 mL/min. Lactose monohydrate was packed into deactivated glass columns and plugged with silanated glass wool. Packed columns were allowed to equilibrate for 6 h after a temperature change before injections were made. Injections were made with a 1 µL-Hamilton syringe; injection volumes were <0.01 µL (based on detector response). Each injection was made at least three times and averaged; the relative standard deviations in the retention times of single probes on a given column were <1% in each case. In addition, each batch was examined with several packed columns. Dead-time was calculated using the retention times of heptane, octane, and nonane.173 Data were analyzed on MS Excel. Probe surface areas are taken from Schultz et al., 1987. Other probe properties were obtained from chemistry handbooks.174


Solid State Characterization

XRPD data are shown in Figure 4 for all lactose batches which are in good agreement with previous reports.31 Since there are no distinctions between them and the data overlay they are not identified individually.

Figure 4.

X-ray powder diffractograms of six sieved (SV) and two milled (ML) lactose batches.

DSC data are shown in Figure 5 and are in good agreement with previous reports.30, 175 A dehydration peak occurs at approximately around 140°C and a melt peak at about 200°C followed by the decomposition peak. In the case of a sieved sample (SV-94) shown at the top of the plot there is a small, almost indistinguishable peak at about 240°C which would indicate the presence of β-lactose.31

Figure 5.

Differential scanning calorimetry (DSC) figures for six sieved (SV) and two milled (ML) lactose batches.

Lactose batches had similar XRPD and DSC profiles. This appears to indicate that the presence of polymorph and/or amorphous content was not detectable within the limits of these analytical techniques, that is, under 10% (w/w) content.

Surface Analysis

SEM imaging and image analysis of sieved lactose indicate the following: (1) relatively uniform particle shape distribution; (2) few particle aggregates; (3) relatively narrow size distribution: (4) few single particles under 10 µm; and (5) smooth irregularly shaped asymmetric cubic morphology with some nanocrevices. Contrastingly, SEM micrograph analysis of milled lactose indicate the following: (1) relatively non-uniform particle shape distribution; (2) significantly more particle aggregates (stronger surface and interfacial interactions); (3) relatively wide particle size distribution; (4) many particles under 10 microns; and (5) irregularily shaped morphology, increased surface roughness, nanocrevices, and surface fines. These results are in excellent agreement with SEM imaging analysis on sieved and milled lactose reported previously.27, 31

Additionally, SEM was employed to study the presence of surface fines that may have significance for the performance of the powder systems. Various groups have investigated the use of ternary blends to modify the interactions between the carrier particles and micronized drug particles. This approach relies on the hypothesis that the “inert” fine particles added to the carrier system will occupy highly charged or energetic sites and thereby improve the deaggregation kinetics of drug particles during dispersion from a DPI device. Our observations that there may be significant differences between the powder systems in the quantity of surface fines and that surface fines may be associated with certain crystal faces requires investigation so that these variables can be included in the statistical analysis and interpretation of relative powder performance.

There are few documented methods for the quantification of surface fines. However, an alternative approach used to complement the image analysis technique involved the washing of the lactose particles and measurement of size using a laser diffraction instrument (Malvern Mastersizer).

Figure 6A,B shows SEM and digitized images of a lactose surface illustrating the manner in which surface features are highlighted. There appeared to be a trend to variation but of no statistically significant difference. Consequently the data are not presented. Figure 7A shows fines estimation by Malvern Laser Diffraction in saturated isopropanol (data supplied by the manufacturer) and washing with oil and decanting the small particles for assessment by laser diffraction. Clear differences between milled and sieved samples but not between batches prepared in the same way are observed. Also a difference in the number of particles observed in IPA and oil for sieved particles was observed despite the same general trend to similarities between batches. This difference may be explained by the different physical properties of the suspending media. Figure 7B shows that the particle size defining the tenth percentile of the distribution was larger for sieved particles (∼20–35 µm) than for milled (∼5 µm) consistent with the smaller number of fines in the sieved sample than milled.

Figure 6.

A: Scanning electron micrograph of lactose particle: B: digital image of particle; (C) sieved lactose; and (D) milled lactose.

Figure 7.

A: Fines analysis (particles <5 µm) and B: Fines analysis particle size of the tenth percentile of the distribution (d10) obtained by laser diffraction.

AFM imaging, currently, is labor intensive and operator dependent. Significant sample preparation time, microscope setup, scanning time, and image analysis hinders the capturing of large data sets. Combined with relatively small surface area mapping the issue of obtaining statistically relevant samples is apparent.

That is, there is a need to obtain information on topography and roughness from a number of particles for each batch (inter-particle variability), different areas on the same particle face (intra-particle variability), and different crystal faces (intra-particle-face variability). AFM images collected in the present studies (e.g., ML B) show evidence of tip contamination from surface fines despite scanning in non-contact mode. Particularly “dusty” samples are likely to result in tip-fine particle contamination. Scanning at lower amplitudes with higher set-points (degree of interaction of cantilever with surface) can minimize this phenomenon but also gives rise to decreased image resolution. Many studies in the literature have used modified lactose (decanted, compressed, air-jet treated) to avoid this.

Figure 8 is a representative image obtained by AFM showing clear geometric features (roughness) on the surface. AFM image analysis of sieved lactose indicates a relatively smooth surface, at the nanometer level, with some nanocrevices. AFM image analysis of milled lactose indicates detailed visibility of highly irregular surface morphology with many nanocrevices (surface defects) that serve as high surface energy sites.

Figure 8.

Atomic force microscopic image of the surface of lactose particle: (A) sieved lactose; and (B) milled lactose.

IGC—Two sieved (SV A and SV D) and two milled (ML A and ML B) Respitose batches were evaluated with respect to dispersive surface free energy at 60°C, 45°C, and 30°C (in that order) with at least three replicates. Variations in dispersive surface energy were slightly, with an average dispersive free energy of 41.7 ± 1.0 mJ/m2. DSC ascertained that no bulk physical changes had taken place as a consequence of the IGC experiment. While there do not appear to be any significant differences in the dispersive surface free energies of the four batches at any of the temperatures, there do appear to be differences between the milled and the sieved batches when studied across the temperature range. This is evident in Figure 9. The slopes, which represent surface entropy, are larger for the sieved than the milled batches. As a result, higher surface enthalpies are obtained for the SV than the ML batches (106–110 mJ/m2 vs. 98–101 mJ/m2).

Figure 9.

Dispersive free energies of two sieved (SV) and two milled (ML) lactose monohydrate versus temperature. Relative standard deviation was <2.5% in each case (n = 4).

Examination of the specific interactions reveals differences between the milled and the sieved batches, as well as among the two sieved and two milled batches.172 These differences are presented in Table 2. No standard deviations are available, since only one best-fit line was obtained from the average enthalpy data obtained from four columns per batch. However, the differences in the enthalpy data from column to column were small, with RSD <3% for most probes.

Table 2. Surface Acid/Base Constants with corresponding Correlation Parameters for Sieved (SV) and Milled (ML) Lactose Monohydrate Batches, in Accordance With Schultz et al.151
Lactose BatchKAKBR2
  1. Table modified from data reported in Ref.172

SV A0.1460.4630.991
SV D0.1460.3540.996
ML A0.1670.3790.998
ML B0.1580.3310.998

The batches appear to be quite similar, which is expected given that they are the same material from a single manufacturer (DMV-Fonterra Excipients). Since KA and KB values are unitless, differences between KA and KB for a material cannot be interpreted directly as signifying a more acidic or more basic surface. However, comparing KA and KB for different batches allows these comparisons to be made. Lactose is known to be an acidic material and the differences in acidic parameter are small. Nonetheless, the differences observed are real, as the sieved batches are similar to one another (0.146) but differ from the milled batches (0.158 and 0.167). The larger KA of the milled batches may in part be explained by its larger surface area (0.87–0.89 m2 for ML vs. 0.34–0.46 m2 for SV batches). The differences in KB are more marked and might be more indicative of actual material variations. Since lactose is an acidic material, the differences in KB are likely tied to other surface properties, perhaps to the presence of impurities at the surface. However, since KA is obtained from the slope of the line and KB from the intercept, determination of KB is less precise than determination of KB.


Differences were noted in the particles size distribution, particularly with respect to fines (d10 = 30 and 4 µm, respectively) and surface area (∼0.4 and 0.9 m2/g, respectively) of sieved and milled particles. Conventional XRPD and DSC are routinely employed to characterize solids and our data demonstrated that there were few if any differences between the batches of lactose studied. However, this data is important to establish the starting characteristics of any particles employed in comparative studies.

We have illustrated the morphological differences between lactose particles evident on inspection of images obtained by both SEM and AFM. There are a number of additional methods of relevance that are based on the use of AFM. One such method employs plots of cohesive-adhesive balance (CAB) obtained by AFM.115 These provide a direct correlation of the force ratio with the ratio of the thermodynamic work of cohesion/adhesion. This approach requires the measurement across atomic smooth surfaces and bases on the assumption that the surface contact areas are the same, thus being normalized.

AFM provides a unique opportunity to comparatively examine and predict potential pharmaceutical formulation. Recently, Using cohesive pull-off forces between three drugs as a function of RH, Young et al.176 correlated these data with in vitro aerosolization performance to evaluate AFM prediction. Hooton et al.177 also applied the CAB approach in screening the behavior of novel sugar candidates as carriers for DPI formulation.

Most of the adhesive/cohesive data generated for the DPI formulation were pull-off forces in the AFM force curves. They are a mixture of different fundamental forces. However, the force curves are capable of generating more abundant information besides adhesion such as (1) the long-range attractive/repulsive force (electrostatic response) before jump-on-contact with the surface; (2) the elasticity of the sample surface during the contact; (3) Surface deformation (hysteresis).

Due to the limitation of AFM on relatively flat surface, a majority of substrates used are either high-pressure compaction tablets or recrystallized material. These surfaces may not be a true representation of the surfaces in real pharmaceutical formulation.

Because of its prevalent use in DPI and other pharmaceutical formulations, lactose monohydrate has been studied extensively by IGC. Several investigators have used IGC to probe batch-to-batch variation and aid in the choice of lactose for use in the formulation. Yet, most studies to date suffer from a number of short-comings. First, use of the relationship based on Eq (4) rather than (3) yields unreliable data which may be contradictory to the more rigorous approach using Eq (3).172 Moreover, most studies to date have merely been descriptive. Surface energy was linked with dispersive properties and different conclusions were obtained. However, these studies were limited to select parameters and neglected to account for confounding factors. Surface energy in itself is not a property that varies from one batch of material to the next but an indicator of other variability, such as impurity/chemical heterogeniety profile or surface disorder content at the nanometer level (such as amorphicity, liquid crystallinity, nanocrystallinity, polymorphism, etc).

Since many DPI formulations are interactive blends of micronized drug and larger lactose carrier particles, adhesional properties are important design considerations. If particles adhere strongly, the inspiratory airflow of the patient during DPI actuation may be insufficient to separate micronized drug from the carrier particles, which may result in poor or variable delivery to the lung. Understanding the adhesional forces of lactose and drug may allow manufacture of drug and choice of excipients to be used to optimize the interactions.

While direct measurement of adhesion, for example, by using centrifugal detachment or atomic force microscopy, is possible, the techniques suffer from poor reproducibility because only select or specific interactions between particles are considered during each measurement. By contrast, IGC probes the surface properties of the entire sample of material.

Solid surfaces, such as those of lactose and drugs, are heterogeneous with varying degrees and distribution of crystallinity, different exposed functional groups and a distribution of surface contaminants. IGC has the additional benefit of probing the most energetic surface sites. When the extremely small concentrations of probe vapor at infinite dilution are injected into the IGC column, the most active surface sites preferentially interact with the probe. This has been cited as the reason why IGC results often do not agree with contact angle measurements.178

If active sites on a large lactose carrier particle are indeed more energetic (high Gibbs surface free energy) then it may preferentially bind a micronized drug particle (s). This process can be regarded much in the same way as a high energy surface, such as a clean pure solid surface or a clean pure aqueous surface, spontaneously adsorbs hydrocarbon impurities existing in the vapor state from air, since there is a thermodynamic driving force to decrease Gibbs surface free energy. Specifically, spontaneous hydrocarbon adsorption onto clean pure solid surface or aqueous surface favorably decreases polar interactions, increases hydrophobicity, and hence, lowers the surface energy.

Thus, many investigators have attempted to link increased surface energy of a solid respirable particle with poor aerosol dispersion from a DPI formulation. While the connection between surface energy of a solid material and aerosol dispersion is conceivable based on first principles, it is a challenge to directly confirm experimentally. This may be attributed to the fact that the surface energy of a solid-state material, such as that used in respirable solid particles, is not an independent parameter but rather the Gibbs surface free energy thermodynamic manifestation of a collective ensemble of other solid-state surface characteristics, such as surface rugosity (a macroscopic property), surface amorphous content (a material property), degree of surface hydrophobicity and surface polarity (material and chemical properties), and the presence of surface impurities (a chemical composition property) that may have been adsorbed onto the solid surface (rendering the solid surface more hydrophobic and lower in Gibb's surface free energy which is a thermodynamically-favorable process) from the ambient vapor environment or absorbed into the solid powder (i.e., present in the bulk and at the surface) during the medicinal chemistry synthesis process. Hence, large scale, experimental design type studies may be necessary to tease out the relationship between surface energy of solid-state respirable particles and dry powder aerosol dispersion, while simultaneously and carefully controlling (over both experimental and pharmaceutically-relevant time-scales) these other important surface characteristics that directly influence surface energy.

The demonstrable differences between sieved and milled lactose as established by a variety of physico-chemical, morphological and surface analytical methods establishes the baseline characteristics of powders that will be subsequently employed in bulk and dynamic analyses of flow, electrostatics and aerodynamic performance with respect to the dispersion of the model drug albuterol.


A number of static and dynamic methods based on surface, bulk and performance methods may be employed to characterize the performance of DPI formulations. Six sieved and two milled batches of α lactose monohydrate were evaluated for their physicochemical properties, surface and bulk morphology.

There were differences in particle size and surface area between the milled and sieved batches of lactose. SEM images showed that there were more fines associated with the milled than sieved lactose batches. Further assessment by image analysis and particle elutriation allowed quantification of these differences. Atomic force microscopy demonstrated that the milled particles exhibited greater surface roughness (nanosurface crevices and adsorbed surface fines) than the sieved particles. Inverse gas chromatography indicated similar dispersive forces at the surface of all lactoses but differences in the polar forces. There are clear indications that the surfaces of milled and sieved particles are different and these differences may be attributed to the different physical, chemical, and material properties of these surfaces resulting in different Gibbs surface free energies. Recognizing also that the surface of given lactose particle is neither chemically nor physically homogenous but rather a composite of various heterogenous regions (both physical and chemical) attributed to the existence of acceptable but significant amounts of various types of residual lipids and protein on the surface from the extraction process from milk (which is both a complex fluid and a biocolloidal dispersion) and effects of pharmaceutical processing to create respirable particles. These differences were investigated with respect to their effects on the dynamic performance properties relative to drug dispersion, which is described in the second paper25 of this two-part series.


Dr. Wallace Ambrose at the UNC School of Dentistry, Dental Research, is acknowledged for access and expert assistance with SEM. Dr. Richard Superfine and Dr. Michael Falvo of the UNC Nanoscience Research Group at the NIH NIBIB Center for Computer Integrated Systems for Microscopy and Manipulation at UNC are acknowledged for expert discussions on AFM. Dr. Michael Chua and Dr. Wendy Soloman are acknowledged for providing access to the UNC School of Medicine, Michael Hooker Medical Microscopy Facility. Martin J. Telko thanks the USP for a graduate research fellowship.