Effect of Milling Temperatures on Surface Area, Surface Energy and Cohesion of Pharmaceutical

Effect of Milling Temperatures on Surface Area, Surface Energy and Cohesion of Pharmaceutical Powders

Umang V. Shaha, Zihua Wanga, Dolapo Olusanmib, Ajit S. Narangb, Munir A. Hussainb, Michael J.Tobync, Jerry Y. Y. Henga*

aSurfaces and Particle Engineering Laboratory (SPEL), Department of Chemical Engineering, Imperial College London, South Kensington Campus, London SW7 2AZ, UK

bBristol-Myers Squibb Co., One Squibb Drive, New Brunswick, NJ 08902, USA

cBristol-Myers Squibb Pharmaceuticals, Reeds Lane, Moreton, Wirral CH46 1QW, UK

*Corresponding Author:

Phone: +44-(0)207-594-0784

Fax: +44-(0)207-594-5700

Web:

Abstract

Particle bulk and surface properties are influenced by the powder processing routes. This study demonstrates the effect of milling temperatures on the particle surface properties, particularly surface energy and surface area, and ultimately on powder cohesion. An Active Pharmaceutical Ingredient (API) of industrial relevance (brivanib alaninate, BA) was used to demonstrate the effect of two different, but most commonly used milling temperatures (cryogenic vs. ambient). The surface energy of powders milled at both cryogenic and room temperatures increased with increasing milling cycles. The increase in surface energy could be related to the generation of surface amorphous regions. Cohesion for both cryogenic and room temperature milled powders was measured and found to increase with increasing milling cycles. For cryogenic milling, BA had a surface area ~5× higher than the one obtained at room temperature. This was due to the brittle nature of this compound at cryogenic temperature. By decoupling average contributions of surface area and surface energy on cohesion by salinization post-milling, theaverage contribution of surface energy on cohesion for powders milled at room temperature was 83% and 55% at cryogenic temperature.

Keywords: Milling Temperature, Milling Time, Cohesion, Silanisation, Surface Energy, Surface Area

1. Introduction

The design and development of pharmaceutical dosage forms rely on the knowledge of particle properties of active pharmaceutical ingredients (APIs) and excipients. Investigating the role of particle processing routes (i.e. drying, crystallisation, roller compaction, spray drying, milling, blending, etc.) on particle bulk and surface properties, has attracted much research interests over the past two decades (Burnettet al., 2012; Dujardinet al., 2012; Iyeret al., 2013; Otteet al., 2011; Feeleyet al., 1998; Yamauchiet al., 2011; Meieret al., 2009). Milling is one the most common particle size reduction step. The effect of milling parameters, such as milling frequency, impact velocity, milling time on particle bulk and surface properties, is well documented in the literature (Otteet al., 2011; Adrjanowiczet al., 2011; Forcinoet al., 2010; Vegtet al., 2009; Shariareet al., 2011; Perkinset al., 2009; Feeleyet al., 1998; Kwanet al., 2004; Kwanet al., 2003; Leleux and Williams, 2014; Newman and Zografi, 2014).

Kwan et al. investigated the effect of milling frequency and propensity of breakage for microcrystalline cellulose and α-lactose monohydrate. A single particle test was used to investigate the role of milling intensityon both microcrystalline cellulose and α-lactose monohydrate particle breakage. With an increase of single particle impact velocity and milling frequency, a higher propensity for breakage for α-lactose monohydrate compared to the microcrystalline cellulose was observed (Kwanet al., 2003). For α-lactose monohydrate, a linear relationship between ball milling time and amount of fine particles generated was proposed (Young et al., 2007). Furthermore, amorphous content was also found to increase with increasing milling time (Young et al., 2007). Milling is reported to result in higher surface energy of salbutamol and the increase in surface energy was suggested to result in poor powder flow characteristics (Feeleyet al., 1998).

The role of milling conditions, such as milling intensity, type of mill, milling temperature and compression, on the change in solid-state polymorphic transformation was extensively reviewed by Brittain(Brittain, 2002). Using Indomethacin as a model compound, Brittain demonstrated that γ-indomethacin, which is the most stable form, converted to the metastable α-form upon ball milling at 30 °C, whereas the α-form remained unchanged. Further reduction in milling temperature to 4 °C resulted in the generation of the amorphous phase (Brittain, 2002). More recently, the role of milling temperature on α-glucose was investigated by Dujardin et al. and reported that for milling at -15 °C, α-glucose was found to be amorphous, whereas room temperature milling, α-glucose was structurally invariant.No polymorphic transformations as a result of room temperature milling were observed. It was argued that milling at temperatures below the glass transition temperature of α-glucose resulted in an amorphous phase. However, room temperature milled α-glucose, showed some traces of amorphous material; it was found to convert rapidly into the crystalline form. Structure evaluation of glucose post milling revealed a competition between amorphisation and thermally activated recrystallisation process. Recrystallisation was reported to be governed by molecular mobility in the amorphous state, hence sensitive to milling temperature (Dujardinet al., 2012;Dujardinet al., 2008). Adrjanowicz et al. demonstrated that milling at cryogenic temperature can result in chemical decomposition of furosemide. Chemical stability of furosemide was reported to be effected by the generation of the amorphous phase under cryogenic milling conditions(Adrjanowicz et al., 2011). A higher reactivity of the amorphous phase was linked to the acid catalysed hydrolysis of the amorphous phase of furosemide, resulting in undesirable chemical changes. It was also argued that at room temperature milling, generation of the amorphous phase is not possible due to its low glass transition temperature (Adrjanowicz et al., 2011).

Olusanmi et al. reported the effect of temperature and humidity on the breakage behaviour of sucrose and aspirin particles. The breakage behaviour of both sucrose and aspirin was found to be insensitive to humidity at room temperature conditions. At a constant impact velocity, the extent of impact breakage for particles was found to increase with increasing temperature. At a constant temperature and impact velocity, the extent of breakage of aspirin was found to be higher compared to sucrose.. Furthermore, Olusanmi et al.’s findings revealed the fracture toughness of aspirin decreases with increasing temperature (Olusanmi et al., 2010).

Niwa et al. compared the ultra cryo-milling with the commonly used jet milling technique. In the ultra cryo-milling system, zirconia beads were used as milling media with liquid nitrogen as a dispersion medium. A greater specific surface area (SSA) of the milled powder was observed in the ultra cryo-milling system. Moreover, ultra cryo-mill did not change the crystallinity of the testing materials (phenytoin, ibuprofen and salbutamol sulfate) (Niwa et al., 2010).

However, whilst it has been reported in the literature that milling temperature can influence molecular and bulk solid-state properties as well as breakage behaviour of particles, no study has investigated the role of milling temperature, particularly at the commonly used milling temperatures (cryogenic and ambient), on surface properties and its relation to powder flow for an API. This study reports the role of milling time and temperature (cryogenic and room temperature) on surface properties, i.e. surface energy and surface area, and cohesion of an API Brivanib Alaninate (BA, BMS-582664). BA is an ester prodrug containing an α-amino acid (L-alanine) promoiety (Zhaoet al., 2012; Naranget al., 2015; Badawyet al., 2014). SSA and surface energy are quality attributes of BA to optimise by controlling processing parameters for improving particle performance. Here cohesion is seen as a possible detrimental attribute.An approach to decouple the contribution of surface energy from surface area on cohesion was also undertaken.

1. Materials

Brivanib Alaninate (BA) was received from Bristol-Myers Squibb Pharmaceuticals (New Brunswick, NJ, USA) and used without further purification. n-heptane (≥99.0), n-octane (≥99.0%), n-nonane (≥99.0%), n-decane (≥99.0%) and dichlorodimethylsilane (>99.5%,) were purchased from Sigma Aldrich (Dorset, UK). Methanol (>99.5%), ethyl acetate (>99.5%), dichloromethane (>99.0%), n-hexane (>99.0%), and cyclohexane (>99.0%) were received from VWR BDH (Prolabo, Lutterworth, UK). All chemicals were used as received.

1. Methods

3.1 Milling of BA crystals

A Retsch Cryo Mill (Retsch Technology GmbH, Haan, Germany) was used for cryogenic milling of BA crystals. Recrystallised BA (1 g) was charged in a 50 mL grinding jar containing one grinding ball of 25 mm diameter. For milling at cryogenic temperature, the grinding jar was constantly cooled during the pre-cooling, milling and intermediate cooling steps with liquid nitrogen, flowing around the mill jacket at a constant pressure. In the pre-cooling step, the mill was cooled down to the temperature of -70 °C. Intermediate cooling between two consecutive milling cycles was performed at a vibrational frequency of 5 Hz for 40 s (Shah et al., 2014a). Milling was performed at a vibrational frequency of 25 Hz for 10 s. BA crystals were milled using typical set up for n=1, 3, 6, 9, ….., 21 milling cycle. No cooling was provided to the grinding jar during room temperature (RT) milling.Post milling, samples were collected and stored in the glass vials for further characterisation and processing.

3.2 Silanisation of milled BA

Milled BA powders was methylated using a silanisation protocol reported in the literature (Al-Chalabi et al., 1990). In a typical process, 500 mg of powder was added to a 50 mL 5% (v/v) solution of dichlorodimethyl silane in cyclohexane. The mixture was refluxed at 80 °C for 24 hours. Then, the reaction mixture is allowed to cool to room temperature and filtered using a general-purpose laboratory filter paper (Whatman, UK) followed by drying in a vacuum oven at 80 °C for 4 hours. Post silanisation, the silanised BA powders was stored in a glass vial at ambient conditions.

3.3 Surface energy analysis

Surface energy heterogeneity of powders was measured using a Surface Energy Analyser (SEA, Surface Measurement Systems Ltd., London, UK). Approximately 300 mg BA powders was packed into silanised glass columns (4 mm diameter, 30 mm length). Helium at a flow rate of 10 sccm was used as a carrier gas for all injections for the columns packed with un-silanised BA. For silanised BA, it was observed that the retention time was lower for all dispersive and mono-polar probes. To improve the resolution for determining time difference between retention time and dead probe time (tR – t0), which ultimately results in improved resolution for calculation of net retention volume, a lower carrier gas flow rate was used (3 sccm) for columns packed with silanised BA.Methane was used as a dead time probe.For determining dispersive surface energy, a series of dispersive n-alkane probes were injected at a range of concentrations to achieve fractional surface coverages (n/nm) ranging from 0.007 to 0.1.For surface energy measurements to be representative of the entire material surface properties, typical fractional coverages used for analysis ranges from n/nm=0.02 to 0.05(Gamble et al., 2013; Gamble et al., 2012; Shah et al., 2014a; Shah et al., 2014b). Considering the range of fractional surface coverages typically used to provide material representative surface energy, n/nm=0.04 was selected for analysis of both silanised and unsilanised materials.The dispersive surface energy was calculated using the Schultz method (Schultz et al., 1987).

For acid-base surface energy determination, mono-polar probes, dichloromethane (mono-polar acidic probe) and ethyl acetate (mono-polar basic probe were injected at different concentrations with a target to achieve the same fractional coverages, and calculated using the van Oss-Chaudhary-Good method (Van Oss et al., 1998; Das at al., 2012). Different empirical models have been proposed in the last three decades for determination of acid-base properties of polar probes. A detailed review of the various approaches is presented elsewhere (Etzler, 2003). For polar probes, van Oss et al. used water with +: - equal to unity (Van Oss et al., 1988), resulting in a “basicity catastrophe” phenomena. Della Volpe and Siboni proposed a different ratio suggesting water as acidic rather than amphoteric (Volpe and Siboni, 1997). Despite differing in absolute values, the acidity/basicity order is similar for both approaches (Clint and Dunstan, 2001). For IGC, the + or - values for the monopolar probes requires further validation, and the acid-base surface energy values for the solid surface onlyoffers a description of the relative acid-base nature (Etzler, 2003; Ho and Heng, 2013).Details for the SEA approach and potential applications are detailed elsewhere (Ho and Heng, 2013).

3.4 Surface area analysis

Approximately 300 mg of milled BA and silanised milled BA were conditioned under helium purge at 40 °C for at least 12 hours. Post-conditioning, the sample mass was measured and used for surface area determination. A fully automated Micromeritics Tristar 3000 (Micromeritics Instruments Corporation, Norcross, USA) system was used for the measurement of nitrogen isotherms at -195.8 °C. The surface area was calculated using the BET model based on the linear region of the nitrogen adsorption isotherm (from p/p0 = 0.05 to 0.3) using the Micromeritics Analysis Software (Micromeritics Instruments Corporation, Norcross, USA).

3.5 Uniaxial compression test

A uniaxial compression test, established for determining the unconfined yield stress for soil mechanics application (Head, 1994) and recently adapted for pharmaceutical powders (Shah et al., 2014b), was used for determination of powder cohesion. BA powder was poured into a 5 mm diameter evacuable IR die (Specac Ltd., Slough, UK) and consolidated to prepare a loosely packed powder compact. At least three different compacts were prepared, consolidated at three different consolidation loads. Post consolidation, the confinement was removed and the compact was carefully ejected. An L/D ratio of at least 2:1 was maintained for the compacts. Ejected compacts were subject to uniaxial compression load with a 36 mm diameter cylindrical aluminium probe and a 50 N load cell operated in a displacement compression mode. Yield load was measured for all compacts prepared at different consolidation loads. The yield stress was plotted as a function of consolidation stress. A linear regression can be fitted to the yield stress as a function of consolidation stress. This linear regression line was extrapolated and cohesion was determined from its intercept with the y-axis (i.e. at zero consolidation stress). The angle of internal friction can be calculated from the gradient of regression line. Theoretical principles of this method can be found in the literature (Head, 1994; Shah et al., 2014b).

3.6 Differential Scanning Calorimetry

Specific heat capacity of milled BA powders was measured using a differential scanning calorimetry (DSC). When a sample is subjected to a linear temperature ramp, the rate of heat flow in the material is proportional to its instantaneous specific heat. The change of specific heat capacity, Cp at the glass transition region is related to presence of amorphous content, reported to be linearly proportional to the amorphous content (Hurtta and Pitkänen, 2004). Detailed method, including mathematical derivations can be referred elsewhere (O'Neill, 1966). This approach is reported to be able of estimate the degree of amorphous content with overall sensitivity of ±5% (Saleki-Gerhardt et al., 1994). Thermograms of the samples were carried using TA Q2000 differential scanning calorimeter (TA Instruments, New Castle, DE, USA) connected to the Platinum software. 5-10mg of samples was loaded in the aluminum pans that are hermitically sealed using TA supplied crimping tool. Thermal behaviours of the samples were studied under nitrogen purge at the heating rate of 10 oC/ min, covering temperature range from 20 oC to 250 oC. Thermograms were analysed using Universal Analysis software supplied as a part of Advantage Software v5.5.3 (TA Instruments, New Castle, DE, USA).

Error bars plotted in all figures are standard deviation of at least three repeat measurements.

1. Results and Discussion

4.1 Effect of processing on the surface properties and cohesion of BA

4.1.1 Effect of milling time and temperature on surface energy and surface area of milled powders

For BA milled at cryogenic and RT temperatures, Fig. 1 and Fig. 2 show the dispersive (γd) and the acid base (γAB) surface energy as a function of milling cycles, respectively, at an isostere at fractional surface coverage (n/nm) 0.04.For BA milled at both cryogenic and RT milling temperatures, γd and γAB were found to increase with increasing milling cycles. Fig. 3 shows the change in specific heat capacity (ΔCp) for phase transition, which can be directly correlated to the amorphous content present, as a function of number of milling cycles for powders milled at RT and cryogenic temperature.

The glass transition temperature (Tg) of BA was characterised using DSC and found to be 40oC.Two different mechanisms may prevail for surface amorphisation of BA milled at room and cryogenic temperature.For milling at room temperature, Tg of BA was found to be below milling temperatures, resulting in the process of local thermal melting followed by a rapid quench. In the case of milling at cryogenic temperature Tg of BA is above milling temperatures. In such cases surface amorphisation can result from an accumulation of defects in milled crystals resulting in highly defective crystalline phase which can become physically less stable than the metastable liquid and amorphises spontaneously at the temperature of milling. Detailed mechanisms for both the possible mechanisms are detailed elsewhere (Descamps et al., 2007; Fecht, 1992; Willart et al., 2004).

It is evident from Fig. 3 that the amount of amorphous content increases with increasing milling cycles. The amorphous state is reported to have higher surface energy compared to the crystalline state, suggesting that the increase in surface energy with increasing milling cycles can be attributed to the presence of surface amorphous content(Brum and Burnett, 2011; Burnett et al., 2012).

For milling at cryogenic temperatures, the γd for milling cycle-1 to milling cycle-6 was found to be very similar (within experimental error margins), whereas γAB for milling cycle-1 to milling cycle-3 was found to be within experimental error margins. γd and γAB were observed to increase with increasing milling cycles from cycle-6 to cycle-21. Milling at cryogenic temperatures provides a temperature blanket and is known to delay the amorphous phase formation upon milling (Shah et al., 2014a; Ward and Schultz, 1995).

For BA milled at cryogenic temperature, no amorphous content was detected for cycle-1 milled powders, whereas a small amount of amorphous content was detected for powders milled with cycle-3 (see Fig. 3). The increase in γd and γAB for BA after cycle-1 milling at cryogenic temperature can be attributed to the particle breakage and other disorder caused at the surface.It has previously been reported that milling can also cause surface defects and crystals fractures, depending on milling conditions, effecting surface energetics (Balard et al., 2008; Chamarthy and Pinal, 2008; Feeley et al., 1998; Heng et al., 2006; Ho et al., 2012).

For powders milled at cycle-1 and 3, surface energy was found to be very similar. However, small amount of amorphous was observed at the powders milled at cycle-3. Amorphous content was found to increase from powders milled from cycle-3 to cycle-21. A significant increase in γAB for powders milled with cycle-6 compared to powders milled with cycle-3 can be observed.