Microstructure and lactose crystallization properties in spray dried nanoemulsions

Microstructure and lactose crystallization properties in spray dried nanoemulsions

Accepted Manuscript Title: Microstructure and lactose crystallization properties in spray dried nanoemulsions Author: P.G. Maher M.A.E. Auty Y.H. Roos...

943KB Sizes 0 Downloads 37 Views

Accepted Manuscript Title: Microstructure and lactose crystallization properties in spray dried nanoemulsions Author: P.G. Maher M.A.E. Auty Y.H. Roos L.M. Zychowski M.A. Fenelon PII: DOI: Reference:

S2213-3291(14)00023-9 http://dx.doi.org/doi:10.1016/j.foostr.2014.10.001 FOOSTR 21

To appear in: Received date: Revised date: Accepted date:

17-7-2014 3-10-2014 9-10-2014

Please cite this article as: P.G. MaherM.A.E. AutyY.H. RoosL.M. ZychowskiM.A. Fenelon Microstructure and lactose crystallization properties in spray dried nanoemulsions (2014), http://dx.doi.org/10.1016/j.foostr.2014.10.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Ac ce p

te

d

M

an

us

cr

ip t

*Graphical Abstract

Page 1 of 33

Maher Pg. 1 Highlights

te

d

M

an

us

cr

ip t

x Powders prepared from nanoemulsions crystallized more quickly than those prepared from conventional emulsions x Polarised light microscopy was used to confirm onset of crystallization and was consistent with dynamic vapour absorption data. x CLSM and cryo-SEM showed the distribution of fat globules within the particles in situ. Cryo-SEM was particularly useful for characterising nano-sized fat droplets (<200 nm) which were below the resolution limit of confocal microscopy. x It is suggested that reduced protein in the continuous phase with lactose contributed to the observed increase in crystallization rate in powders prepared from nanoemulsions

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12

Page 2 of 33

Maher Pg. 2 12

Title: Microstructure and lactose crystallization properties in spray dried nanoemulsions

13 14

P.G. Maher1, 2, M.A.E. Auty1*, Y.H. Roos2, L.M. Zychowski1, 2 M.A. Fenelon1,

15

1

16

Fermoy, Co. Cork, Ireland and 2Food Technology, School of Food and Nutritional Sciences,

17

University College Cork, Cork, Ireland.

cr

ip t

Food Chemistry & Technology Department, Teagasc Food Research Centre, Moorepark,

18 Author’s Affiliation and Address:

20

Patrick G. Maher, Research Student, Food Chemistry & Technology Department, Teagasc

21

Food Research Centre, Moorepark, Cork, Ireland.

22

Mark A. Fenelon, Principal Research Officer, Food Chemistry & Technology Department,

23

Teagasc Food Research Centre, Moorepark, Cork, Ireland.

24

Yrjö H. Roos, Professor, School of Food and Nutritional Sciences, University College Cork,

25

Cork, Ireland.

26

Lisa M. Zychowski, Research Student, Food Chemistry & Technology Department, Teagasc

27

Food Research Centre, Moorepark, Cork, Ireland.

28

Mark A.E. Auty, Senior Research Officer, Food Chemistry & Technology Department,

29

Teagasc Food Research Centre, Moorepark, Cork, Ireland.

an

M

d te

Ac ce p

30

us

19

31

*Contact information for Corresponding Author:

32

Dr. Mark Auty, Food Chemistry & Technology Department, Teagasc Food Research Centre,

33

Moorepark, Fermoy, Co. Cork, Ireland. Tel: +3532542442.

34

email: [email protected].

35 36

Page 3 of 33

Maher Pg. 3 Abstract

38

The objective of this study was to characterize lactose crystallization behaviour and

39

microstructure of spray dried nanoemulsions with different fat globule sizes (FGS). Powders

40

of the same composition (57.7% w/w lactose, 12.3% w/w sodium caseinate, 27.7% w/w

41

sunflower oil, and 2.3% w/w water) but different FGS (mean diameters 1100 nm and 160

42

nm) prior to spray drying were manufactured. Differences in lactose crystallization were

43

studied using dynamic vapour sorption (DVS) and polarized light microscopy (PLM).

44

Crystallization kinetics was modelled using the Avrami and Yang equations. Results showed

45

that lactose crystallized in three dimensions and more rapidly in powders with a smaller FGS.

46

PLM images showed a higher rate of crystal formation for smaller FGS powders when stored

47

for up to 4 days at 55% relative humidity. Confocal laser scanning microscopy (CLSM) and

48

cryo-scanning electron microscopy (Cryo-SEM) images indicated the more evenly distributed

49

small fat globules inside powder particles prepared from spray-dried nanoemulsions. The

50

surface of powder particles was uneven and ruptured post lactose crystallization. Crystals of

51

presumptive anhydrous D- and E-lactose (5:3 molar ratio) appeared after humidification.

52

Results showed powder particles of the same composition were altered in lactose

53

crystallization characteristics by changing the FGS of emulsions pre spray drying.

cr

us

an

M

d

te

Ac ce p

54

ip t

37

55

Keywords: Lactose crystallization, dynamic vapour sorption, polarized light microscopy,

56

confocal laser scanning microscopy, cryo-scanning electron microscopy.

57 58 59 60 61

Page 4 of 33

Maher Pg. 4 62

Highlights

63 x Reducing FGS of emulsions pre drying increased rate of lactose crystallization.

65

x Crystallization occurred in three dimensions based on the Avrami exponent.

66

x CLSM and cryo-SEM showed in situ smaller FGS in powders from nanoemulsions.

67

x Needle-like anhydrous lactose crystals were observed by cryo-SEM and CLSM.

68

x Particles were aggregated and lost their shape post lactose crystallization.

us

cr

ip t

64

69

an

70 71

M

72 73

77 78 79 80 81 82

te

76

Ac ce p

75

d

74

83 84 85 86

Page 5 of 33

Maher Pg. 5 87

1. Introduction

88 In dairy products, the degree of lactose crystallization has a significant effect on ingredient

90

and food properties, such as texture and flavour. Crystallization of lactose in milk powders

91

results in lumping and caking which has a negative impact on powder reconstitution (Lai and

92

Schmidt, 1990). Component crystallization also causes the release of entrapped lipids in

93

powders (Fäldt and Bergenståhl, 1996; Shimada et al., 1991) making them more susceptible

94

to oxidation.

us

cr

ip t

89

an

95

Various methods can be used to quantify crystallization of amorphous sugars. Isothermal

97

differential scanning calorimetry (DSC) was used by Roos and Karel (1990) and Kedward et

98

al. (1998) to determine the crystallization of freeze dried lactose and sucrose at low water

99

contents (< 5%) and high temperatures (365 – 400 K). X-ray diffraction (XRD) was used by

100

Jouppila et al. (1997) and Miao and Roos (2005) to monitor crystallization kinetics of lactose,

101

trehalose and lactose/trehalose mixtures. Mahlin et al. (2004) used atomic force microscopy

102

(AFM) to quantify crystallization.

d

te

Ac ce p

103

M

96

104

Lactose crystallization can be measured by gravimetric means, by monitoring changes in

105

mass of powder samples upon humidification (Iglesias and Chirife, 1978; Jouppila and Roos,

106

1994; Lai and Schmidt, 1990). Crystallization results in sorbed water release, which is

107

measured from mass decrease in a humidified powder (Burnett et al., 2004; Burnett et al.,

108

2006). This method is most successfully used at ambient temperatures where the

109

crystallization rate is quite slow and can be easily monitored (Kedward et al., 2000).

110

Page 6 of 33

Maher Pg. 6 Crystallization kinetics can be modelled using various relationships. The Avrami equation

112

(Avrami, 1939, 1940) has been used by several researchers (Arvanitoyannis and Blanshard,

113

1994; Haque and Roos, 2005; Kedward et al., 2000; Mazzobre et al., 2003). Derivative

114

models from the Avrami equation include the Yang equation (Yang et al., 2010) and

115

Urbanovici-Segal equation (Urbanovivi and Segal, 1990). The Urbanovici-Segal equation

116

includes a parameter r (r > 0) that determines how far the model deviates from the Avrami

117

equation (Soutari et al., 2012). Other kinetic models used to determine lactose crystallization

118

are the Williams-Landel and Ferry (WLF) equation (Roos and Karel, 1991) and the Hoffman

119

equation (Arvanitoyannis and Blanshard, 1994). The Arrhenius equation can be used to

120

determine the activation energy (EA) required for crystallization (Schmidt et al., 1999). The

121

Avrami equation determines the rate constant (k) and exponent (n) that, respectively,

122

determine how fast the material crystallizes and in how many dimensions. Powders stored at

123

increasing temperatures and humidities crystallize at faster rates (Soutari et al., 2012).

124

Polarized light microscopy (PLM) is a useful technique in distinguishing crystalline from

125

non-crystalline material (Hartel, 2001) and can be used to directly measure the rate of crystal

126

growth. Mazzobre at al. (2003) used polarized light videomicroscopy (PLV), in conjunction

127

with DSC, to determine the isothermal crystallization kinetics of lactose and lactose-trehalose

128

mixtures. The PLV method proved a useful way of directly observing individual crystal

129

growth and morphological aspects that were undetectable by DSC.

cr

us

an

M

d

te

Ac ce p

130

ip t

111

131

The distribution of fat globules has a significant effect on functional properties of dairy

132

powders (Pisecky, 1997). Techniques such as CLSM and cryo-SEM are useful in highlighting

133

microstructural differences which help explain why dairy powders have various functional

134

properties. CLSM is widely used for studying the microstructure of cheese, milk powder and

135

chocolate (Auty et al., 2001). McKenna (1997) examined whole milk powder using CLSM.

Page 7 of 33

Maher Pg. 7 Kelly et al. (2014) used CLSM to analyse the distribution of fat droplets in powder samples

137

where different fat blends were stabilized by sodium caseinate. Samples were dual-labelled

138

with Nile Red / Fast Green FCF to show the distribution of fat and protein, respectively,

139

throughout the powder particles. Air vacuoles were also visible in powders in images

140

produced by CLSM. SEM techniques have been developed to examine the outer and inner

141

structures of foods (Soottitantawat et al., 2003). SEM has been used in studies of

142

microencapsulation properties of powder structures stabilized by sodium caseinate (Fäldt and

143

Bergenståhl, 1996; Hogan et al., 2001a, b). Cryo-SEM is a useful technique in analyzing

144

nano-sized particles embedded in food matrices (Dudkiewicz et al., 2011).

an

us

cr

ip t

136

145

The objectives of the present study were (1) to investigate the effects of reducing the FGS of

147

emulsions prior to spray drying on water-induced lactose crystallization during storage; and

148

(2) to characterize powder microstructure using correlative microscopy techniques (PLM,

149

CLSM and cryo-SEM) to help explain differences of the physical properties of the powders.

te

d

M

146

150

2. Materials and Methods

152

2.1 Materials

153

Two powders, containing 57.7% w/w lactose, 12.3% w/w sodium caseinate and 27.7% w/w

154

sunflower oil, were used (Maher et al., (2014). The powders differed in their mean FGS as

155

they were both homogenized (60 °C with a single pass) at different pressures prior to spray

156

drying. Conventional emulsions were produced using in-line two-stage homogenization

157

(model NS20006H, GEA Niro, Soavi, Parma, Italy) at first- and second-stage pressures of

158

13.79 MPa and 3.45 MPa, respectively, giving a FGS of ~ 1100 nm. Nanoemulsions were

159

produced by microfluidization with a Microfluidizer (model M-110EH, Microfluidics,

160

Newton, MA, USA) at 100 MPa, giving a FGS of ~ 155 nm. All emulsions were spray dried

Ac ce p

151

Page 8 of 33

Maher Pg. 8 at the same conditions (inlet/outlet dryer temperatures of 180 °C/80 °C) with a single stage

162

pilot-scale spray dryer (Anhydro F1 Lab, Copenhagen, Denmark) equipped with a two-fluid

163

nozzle atomization system (Type 1/8 JAC 316ss) and counter-flow drying. Resulting powders

164

were labelled C and N for conventional emulsion powders (FGS ~ 1100 nm) and

165

nanoemulsions powders (155 nm), respectively.

ip t

161

cr

166 2.2 Methods

168

2.2.1. Dynamic vapour sorption

169

Water sorption isotherms of powders were obtained from the DVS Data Analysis Suite in

170

Microsoft Excel using a DVS-1 analyser (Surface Measurement Systems, London, UK)

171

equipped with a Cahn microbalance. Powder samples (~ 30 mg) were placed in the sample

172

pan with an empty pan used as a reference. Samples were humidified at 55% relative

173

humidity (RH) on the DVS for two days (48 h) for C and N at 25ºC. Mixing dry nitrogen gas

174

(200 mL/min) with saturated water vapour in the correct proportion using mass flow

175

controllers was used to get accurate RH readings. Samples were measured from three

176

replicate spray drying trials.

an

M

d

te

Ac ce p

177

us

167

178

2.2.2. Polarized light microscopy

179

Samples were examined using an Olympus BX51 light microscope (Olympus BX-51,

180

Olympus Corporation, Tokyo, Japan) with a 10x dry objective lens using polarized light.

181

Digital images (TIFF, 8-bit) were taken using a Jenoptik C14 Imagic camera, and captured.

182

Powders were analysed pre- and post- lactose crystallization following storage over a

183

saturated solution of Mg(NO3)2 (55% RH) for 4 days at 20 °C. Crystalline regions appeared

184

as bright areas on the micrographs.

185

Page 9 of 33

Maher Pg. 9 2.2.3. Confocal laser scanning microscopy

187

Visualization of powder particles was carried out using a Leica TCS SP5 confocal laser

188

scanning microscope (CLSM; Leica Microsystems CMS GmbH, Wetzlar, Germany). Powder

189

samples were placed onto a glass slide and labelled using a mixture of Fast Green and Nile

190

Red or Nile Red alone (Auty et al., 2001). The dye mixture comprised Fast Green (aq. 0.01 g

191

/ 0.1 L) and Nile Red dissolved in polyethylene glycol mol. wt. 400 (0.1g / 0.1 L) mixed in a

192

ratio that both allowed diffusion of the dye molecules into the powder particles whilst

193

maintaining particle integrity and preventing solubilisation. Based on trial and error, pre- and

194

post-lactose crystallized powders were labelled, respectively, with 1:40 or 1:20 ratios of Fast

195

Green to Nile Red. Dual excitation using 488nm/633nm was utilized for all Nile Red and Fast

196

Green mixtures, while single excitation at 488 nm was used for the Nile Red images.

197

Representative images of each sample were taken using 63x oil immersion objective with

198

numerical aperture 1.4. Z-stacks were obtained in order to generate a three-dimensional

199

structure of the particle and to identify surface fat staining. Red and, Green pseudo-coloured

200

images (8-bit), 512 x 512 pixels in size, were acquired using a zoom factor of 1. A minimum

201

of four z-stacks were taken per sample, with representative micrographs being shown.

202

Samples were measured pre and post lactose crystallization (55% RH for 3 days).

cr

us

an

M

d

te

Ac ce p

203

ip t

186

204

2.2.4. Cryo-scanning electron microscopy (Cryo-SEM)

205

Cryo-SEM was used to examine the microstructure of powders. Powders were sprinkled onto

206

the surface of OCT embedding compound (Agar Scientific Ltd, UK) on a slotted aluminium

207

stub, then plunge-frozen into liquid nitrogen slush at   

    

208

to the cold stage using the Alto 2500 cryo-transfer device (Gatan Ltd., Oxford, UK). A cold

209

scalpel was used to fracture samples. The sample was cooled to      

210

at                                 

Page 10 of 33

Maher Pg. 10 imaging at !"#$ '=?Q  

    [ 

212

Zeiss SMT Ltd., Cambridge, UK), operating at 2 kV and a working distance of 6 mm, was

213

used to obtain secondary electron images. Representative micrographs were taken in order to

214

visualize surface characteristics, including surface fat and lactose crystals, where present.

215

Samples were analyzed pre- and post- lactose crystallization (55% RH for 3 days).

ip t

211

cr

216 2.2.5. Statistical analysis

218

Spray drying experiments were carried out in triplicate. One-way analysis of variance

219

(ANOVA) was used with Minitab 15 (Minitab Ltd., Coventry, UK) to determine significant

220

differences between lactose crystallization kinetics of both sets of powders by Fisher’s one-

221

way multiple comparison test. Results were deemed statistically significant if P < 0.05.

M

an

us

217

222 3. Results and discussion

224

3.1 Water sorption kinetics

225

Water sorption kinetics of powders C and N were measured by DVS at 55% RH and 25 °C.

226

These conditions were chosen since Burnett et al. (2004) showed that below 30% RH lactose

227

crystallization will not occur and above 58% RH lactose crystallization is nearly

228

instantaneous.

te

Ac ce p

229

d

223

230

Figure 1a shows the mean water uptake in powders from 3 replicate samples on a solids-non-

231

fat (SNF) basis over a 2880 min (48 h) period. It is shown that there was no difference in

232

initial water sorption behaviour of the powders. The mean maximum sorption rate of water in

233

the first 60 min was 7.4 g water / 100 g SNF for C, and 7.9 g water / 100 g SNF for N.

234

Maximum water sorption occurred in approximately 140 min, upon which there was a

235

decrease in mass and therefore evidence for onset of lactose crystallization (Burnett et al.,

Page 11 of 33

Maher Pg. 11 2004; Burnett et al., 2006). Powders C and N shown in Figure 1a crystallized at different

237

rates, with lactose crystallising faster in N than in C. The mass of water decreased until full

238

crystallization was achieved towards the end of the run. The mean crystallization rate of

239

lactose in each powder showed 0.9 g water / 100 g SNF lost per hour for C and 1.9 g water /

240

100 g SNF lost per hour for N (measured between 700 and 760 min). The onset (m0) and end

241

points (m\) of lactose crystallization were found by calculating the intersection point of

242

tangent lines to the respective baselines using the experimental data in Figure 1a. The mean

243

onset points for C and N were 483 min and 424 min, respectively, showing that N started to

244

crystallize before C. N also had a mean end point of 1098 min compared to a mean end point

245

of 1240 min for C. This clearly showed that lactose crystallized faster in powders from

246

nanoemulsions (674 min in total) than in powders from conventional emulsions (757 min in

247

total). This result was similar to that of Maher at al. (2014) who found using Differential

248

Scanning Calorimetry (DSC) that lactose in powders from nanoemulsions crystallized at

249

significantly (P < 0.05) lower temperatures than in powders from conventional emulsions.

250

Maher at al. (2014) explained that this was due to nanoemulsions having a smaller FGS

251

which means that there was more surface area for protein to adsorb to, leaving less protein in

252

the continuous lactose phase.

cr

us

an

M

d

te

Ac ce p

253

ip t

236

254

More detailed quantitative analysis of lactose crystallization was achieved by converting the

255

experimental data to the fraction crystallized over time, and subsequent results were analyzed

256

statistically. This was calculated by comparing the mass of water absorbed by the powder at a

257

given time (mt) to the mass of absorbed water at the end of crystallization (m\) and the mass

258

of water absorbed at the onset of crystallization (m0). Then:

259

Crystallization T t

m0  mt m0  mf

(1)

Page 12 of 33

Maher Pg. 12 The resulting isotherm is shown in Figure 1b and is the mean of 3 replicate powders of each

261

type (C and N). The time has been normalized so that the onset of lactose crystallization is

262

time zero and the y-axis has been normalized to reflect the crystallization fraction. Before

263

lactose crystallization, the crystalline fraction was assumed to be zero and it was assumed to

264

be one after crystallization. Various different equations have been used in literature to model

265

crystallization kinetics, the most common of which is the Avrami model (1939):

266

Tt

267

where k is the rate constant (min-1) and n is the Avrami exponent. The rate constant (k)

268

quantifies how fast the crystals grow indicating the stability of the system (Gaisford et al.,

269

2009) and the Avrami exponent (n) defines the mechanism of crystallization. The Avrami

270

rate constant mainly depends on the crystallization temperature (Kawamura, 1979) and

271

typically follows an Arrhenius-type temperature dependency. It takes both nucleation and

272

crystal growth into account (Sharples, 1966). An Avrami exponent value of 1 represents

273

nucleation and crystal growth from surfaces, a value of 2 represents crystal growth from

274

edges, and a value of 3 indicates crystal growth in three dimensions. Non-integer n values can

275

also be obtained and they represent mixed crystallization mechanisms (Cahn, 1956; Soutari et

276

al., 2012). The sigmoidal shape of the curve in Figure 1b is typical of those produced using

277

the Avrami equation. The larger the Avrami exponent (n) the more sigmoidal the

278

crystallization curve (Avrami, 1940). Transformation rates are low at the beginning and end

279

of the transformation, but rapid in between. Initially there is a slow rate due to the time

280

required to form a significant number of nuclei of the new phase. The nuclei grow rapidly

281

into particles, during the intermediate phase, and consume the old phase while nuclei

282

continue to form in the remaining parent phase. When the transformation is nearly complete,

283

there is less material to transform to nuclei and the production of new particles becomes slow.

284

Also, the already existing particles touch each other forming a boundary where growth stops.

cr

ip t

260

(2)

Ac ce p

te

d

M

an

us

1  exp( kt n )

Page 13 of 33

Maher Pg. 13 This is not accounted for in the Avrami model as it assumes a constant crystal growth rate,

286

which is why the model breaks down as complete crystallization approaches (Soutari et al.,

287

2012).

288

The experimental data was also modelled using a modified version of the Avrami equation

289

from Yang et al. (2010):

290

Tt

291

The same equation was derived by Tobin et al. (1974). Crystallization isotherms were fitted

292

to Eqs. (2) and (3) and the results were compared to the experimental data in Table 1. The

293

Avrami model was found to better describe the experimental crystallization process than the

294

Yang model. The mean Avrami rate constants of 1.1 x 10-8 min-1 (C) and 1.2 x 10-7 min-1 (N)

295

are significantly different (P < 0.05) showing that lactose in powders from nanoemulsions

296

crystallized faster than in conventional emulsions (Table 1). The Avrami integers are very

297

close to 3 indicating the crystals are growing in three dimensions (Soutari et al., 2012).

298

Half-times of lactose crystallization are useful in predicting the storage stability of lactose

299

containing products (Jouppila et al., 1997). The crystallization half-time is calculated using

300

the rate constant (k) and exponent (n) in the following equation:

cr

1 1  kt n

(3)

Ac ce p

te

d

M

an

us

1

ip t

285

1/ n

§ ln 2 · ¨ ¸ © k ¹

301

t1/ 2

302

The t1/2 values, calculated using the Avrami k and n values, were very similar to those found

303

experimentally (Table 1). High coefficients of determination (r2) and similar t1/2 values

304

indicate that the Avrami equation models the experimental data to a high degree of accuracy.

305

Figure 1b shows that Avrami crystallization values are very similar to experimental values.

306

Values found using the Yang equation were less accurate than those from the Avrami

307

equation (Table 1). The coefficient of determination was lower and the t1/2 was much

308

different to the experimental value. The exponents (n) of 4.39 (C) and 4.31 (N) were greater

(4)

Page 14 of 33

Maher Pg. 14 and rate constants (k) were lower than Avrami constants. Similar results were reported by Al-

310

Mulla (2007) who found that the Yang (or Tobin) model did not fit as well as the Avrami

311

model because it gave higher n values and lower r2 values for the crystallization of polymers.

312

The rate constants (k) obtained in this study (1.1 – 12 x 10-8 min-1) were much lower than

313

those obtained by Schmidt et al. (1999), who measured the crystallization kinetics of

314

amorphous lactose at 25 °C in the range 52.5 – 57.5% RH (5.6 – 15.5 x 10-3 min-1). Jouppila

315

et al. (1997) found that the Avrami model did not work for skimmed milk powder (SMP) at

316

53.8% RH, but did work at humidities ]^^!_!`    ${?  [!

317

x 10-10 min-1) with a corresponding increase in crystallization half-time (40 h) compared to

318

the powders in this study, even at the higher RH of 66.2%. Similar Avrami indexes were

319

obtained for SMP (~ 3), indicating crystal growth in three dimensions. Overall, this shows

320

that protein had a significant effect on lactose crystallization kinetics. When protein was not

321

present, lactose crystallized quickly (Schmidt et al., 1999). The 11% sodium caseinate in this

322

study caused delayed lactose crystallization, and increasing the protein content to 35% (for

323

SMP) further delayed lactose crystallization (Jouppila et al., 1997). It is postulated that

324

nanoemulsions have less non-adsorbed protein than conventional emulsions due to their

325

smaller FGS giving increased surface area for protein adsorption. Therefore, this study

326

suggests that not only is the total amount of protein important, but the amount of non-

327

adsorbed protein that has the most significant effect on lactose crystallization.

328

To the author’s best knowledge no other studies have compared crystallization of lactose in

329

powders with significantly different FGS. The best comparison can be made with McCarthy

330

et al. (2013) who showed that lactose crystallization occurred at lower ware contents with

331

decreasing protein content (i.e., less protein in the continuous phase) for infant formula

332

powders. Similar results were found by Thomas et al. (2004) who found that increasing E-

Ac ce p

te

d

M

an

us

cr

ip t

309

Page 15 of 33

Maher Pg. 15 333

lactoglobulin contents increased the humidity at which lactose crystallized in E-

334

lactoglobulin/lactose mixtures.

335 3.2 Microscopy

337

3.2.1 Polarized light microscopy

338

Figure 2 shows PLM images of powders C and N when stored in a desiccator over a saturated

339

salt solution of Mg(NO3)2 at 55% RH for 0, 1, 2, 3, and 4 days. The micrographs are

340

consistent with results obtained from DVS, showing that lactose crystals formed more

341

quickly in powder N than powder C over 4 days. Both powders at day 0 showed no evidence

342

of lactose crystals. This was in contrast to other studies (Haque and Roos, 2005; McCarthy et

343

al., 2013; Roetman, 1979) that showed lactose crystals present in fresh powders. At day 1, no

344

crystals were seen in powder C but some small isolated crystals were seen in powder N

345

(Figure 2d). At day 2, some crystals had formed in powder C (Figure 2e) with increased

346

growth of crystals on the edges of powder particles in powder N (Figure 2f). Surface crystals

347

are evident in powder C at day 3 (Figure 2g) but there are noticeably more crystals present in

348

powder N (Figure 2h). By day 4, both powders C and N showed crystallization throughout

349

the particles.

cr

us

an

M

d

te

Ac ce p

350

ip t

336

351

3.2.2. Confocal laser scanning microscopy

352

Confocal laser scanning microscopy results are shown in Figure 3. Labelling with Nile

353

Red/Fast Green highlighted the fat and protein phases of the powder and showed air vacuoles

354

as dark regions within the particle (Figure 3a, b, arrows). Discrete spherical fat droplets, most

355

of which are 1 - 5 Pm in diameter, were clearly seen within powder C (Figure 3a). By

356

contrast, fat droplets in Powder N were small (Figure 3b); although some individual droplets

357

(~ 0.2 - 1Pm diameter) were visible, there was considerable co-localization of the Nile Red

Page 16 of 33

Maher Pg. 16 and Fast Green signals indicating that most fat droplets were below the resolution limit of the

359

CLSM. The protein phase of both powders is shown in red. The protein is more clearly

360

observed in powder C than powder N due to the fat globules being less numerous and evenly

361

spaced, allowing more gaps for the protein phase to be visualized. Very little surface fat was

362

visible for either powder (Figure 3a, b), in contrast with some agglomerated whole milk

363

powder particles that have shown evidence of surface fat pooling (McKenna, 1997). No

364

evidence of lactose crystals was found in fresh powders (Figure 3a, b).

cr

ip t

358

us

365

Powders stored after 4 days at 55% RH are shown in Figure 3 c, d, e, f. Particles from both

367

powders appeared to have deformed and were more agglomerated than fresh powders. Figure

368

3c shows large free fat droplets (arrows) > 5 Pm in diameter as well as fat droplets and pools

369

within the particles. Powder N shows generally less free fat but had more obvious elongated

370

or needle-shaped crystals seen by negative contrast (Figure 3d, arrows). The protein signal

371

was fairly weak and appeared localized at the surface and outer edges of the particles (Figure

372

3 c, d). This may be due to high level of lactose crystallization preventing the stain from

373

diffusing into the particle core.

M

d

te

Ac ce p

374

an

366

375

To study the fat phase in more detail and reduce possible staining-induced coalescence,

376

powders were stained with Nile Red solubilized in polyethylene glycol, mol. wt. 400 and

377

results are shown in Figure 3e, f. The fat globules within the particles no longer appear

378

spherical as in fresh powders, probably due to lactose crystallizing around them and forcing

379

them into irregular pools (see inset Figure 3e). Powder N appeared to have less free fat and

380

more intact air vacuoles, indicating that powder particles from nanoemulsions are more stable

381

to crystallization-induced fat rupture.

382

Page 17 of 33

Maher Pg. 17 Few studies have looked at powders using CLSM post lactose crystallization. McCarthy et al.

384

(2013) used CLSM and found free fat partially covering lactose crystals post crystallization

385

of a model infant formula after it had leaked from the interior of the powder. McKenna

386

(1997) showed that for whole milk powder, when stained for fat only, that the entire powder

387

particle could be seen with lactose crystals clearly evident from negative contrast. The lactose

388

crystals in that study were rhomboidal D-crystals due to the lactose crystallizing in the milk

389

prior to spray drying. Needle-shaped crystals were seen post lactose crystallization in the

390

current study. Jouppila et al. (1997) and Bushill et al. (1965) reported that lactose in skimmed

391

milk powder crystallized at 55% RH in a 5:3 molar ratio of anhydrous D-lactose: E-lactose,

392

therefore our results are assumed to be the same.

an

us

cr

ip t

383

M

393 3.2.3 Cryo-scanning electron microscopy

395

It is important to characterize the outer topography of particles as it can give information on

396

flow properties of spray dried powders (Soottitantawat et al., 2003). Cryo-SEM images are

397

presented in Figures 4 and 5. The surface morphology of fresh powders from C and N

398

powders is shown in Figures 4a and b, respectively. The surfaces of all powders appeared to

399

be very smooth with little indentation and powder rupture, similar to data of Hogan et al

400

(2001b) who also used sodium caseinate as an emulsifier. Sodium caseinate is a very efficient

401

emulsifier and dominates the powder surface, even in small concentrations, due to a higher

402

surface activity of protein in water during the drying process (Fäldt and Bergenståhl, 1996).

403

Very little solvent extractable free fat was measured previously by Maher et al. (2014), with

404

powder N (0.4%) having a significantly lower FF content than powder C (2.3%). There was

405

little evidence of free fat on the surface of these powders. Fractured particles are shown in

406

Figure 4c, d, e, f. The advantage of cryo-SEM is that the air vacuoles are easily visible and

407

well preserved. Both powders showed high levels of occluded air in the form of rounded

Ac ce p

te

d

394

Page 18 of 33

Maher Pg. 18 408

vacuoles (Figure 4c, d, arrows), although powder N appeared to have slightly larger vacuoles

409

and thinner cross-walls than powder N. Vacuoles were distributed evenly throughout the

410

powder particles and were mostly in the size range 5 – 50 Pm.

ip t

411 Conventional emulsion powders (C) have an uneven distribution of oil droplets spread

413

throughout the powder particles, whereas nanoemulsion powders (N) have an even

414

distribution of much smaller oil droplets. The size of the oil droplets was ~ 1100 nm for C

415

and ~ 150 nm for N. The smaller oil droplets in nanoemulsions were more embedded in the

416

matrix and therefore less likely to be extracted upon addition of solvent. No evidence of

417

lactose crystals was found in any of the fresh powders. Higher magnifications revealed the

418

powder wall structure at the nano-scale (Figure 4e, f, g, h). In powder C spherical fat

419

droplets, mostly 200 nm to 5 Pm in diameter, are seen embedded within the wall matrix

420

(Figure 4 e, f, arrows). Fat droplets were smaller in powder N (< 200 nm diameter) and could

421

only be seen at very high magnifications (Figure 4h, arrow).

us

an

M

d

te

422

cr

412

The powders were also examined by cryo-SEM post lactose crystallization (Figure 5). At low

424

magnification, agglomeration (or melding) of powder particles was evident (Figure 5a, b).

425

Higher magnification revealed the presence of radiating clusters of needle-like crystals on

426

powder surfaces (Figure 5c, d). Surfaces of lactose-crystallized powders were more uneven

427

with various indentations and holes. The irregular surface, as a result of lactose

428

crystallization, was also observed by Murrieta-Pazos et al. (2011). Faldt and Bergenståhl

429

(1996) found using electron spectroscopy for chemical analysis (ESCA) that a protein film

430

dominates the powder surface even after lactose crystallization. Solid bridges formed

431

between particles after lactose crystallization making them agglomerate and cluster, which is

432

in contrast to the more discrete individual particles from the fresh powder. No large crystals

Ac ce p

423

Page 19 of 33

Maher Pg. 19 were observed on the powder surface. Faldt and Bergenståhl (1996) postulated that fat

434

globules reduce the size of lactose crystals formed after humidification. Fractured particles

435

showed crystals embedded within the wall and also projecting into vacuoles (Figure 5e, f,

436

arrows). There was little evidence of intact spherical fat droplets in either powder, suggesting

437

that they had been disrupted by lactose crystals. This observation is consistent with CLSM

438

results which showed irregular fat pools dispersed within the powder particles (Figure 3e)

439

and highlights the advantage of a correlative microscopy approach. Cryo-SEM had the

440

advantages of good structural preservation and high resolution but could not easily

441

distinguish between fat, protein or amorphous lactose. CLSM revealed the distribution of fat

442

in powders more clearly, while PLM revealed the onset and progress of crystallization over

443

time and confirmed DVS data which indicated that powders prepared from nanoemulsions

444

crystallized at a faster rate. Combining cryo-SEM with CLSM and PLM thus provided a

445

powerful set of tools for characterising the microstructures of dairy powders.

446

te

d

M

an

us

cr

ip t

433

Conclusions

448

Powders prepared from nanoemulsions crystallized significantly (P < 0.05) more quickly

449

than those prepared from conventional emulsions, as measured using DVS. Rates were

450

calculated with the Avrami equation and crystals were found to form in three dimensions.

451

Evidence of crystal growth was observed with PLM and showed the same trend as results

452

from DVS. CLSM and cryo-SEM show differences in powder structure with powders from

453

nanoemulsions having smaller fat globules that are evenly distributed throughout powder

454

particles. It is suggested that reduced protein in the continuous phase with lactose contributed

455

to observed increase in crystallization rate in powders prepared from nanoemulsions.

456

Elongated lactose crystals (probably a mixture 5:3 molar ratio of D and E-lactose), initially

457

at the surface but progressively occurring within the particles, were observed post

Ac ce p

447

Page 20 of 33

Maher Pg. 20 humidification at 55% RH. To the authors’ knowledge, there have been no studies on

459

powders prepared from nanoemulsions using CLSM, SEM, or cryo-SEM that show the

460

distribution of fat globules within the particles in situ. This paper provides a useful insight

461

into how reducing the FGS of emulsions pre spray drying impacts upon the physical

462

properties of powders.

ip t

458

cr

463 Acknowledgements

465

The authors would like to acknowledge the National Food Imaging Centre, Teagasc Food

466

Research Centre, Moorepark, Fermoy, Co. Cork, Ireland for use of the microscopes. The

467

project was funded by the Irish Department of Agriculture, Food and the Marine under the

468

Food Institutional Research Measure (FIRM).

M

469

479

d

Industrial relevance

te

Relatively little is known about the physical and chemical properties of nano-sized emulsions when spray dried. Drying of milk-based food ingredients is a major industry and high-shear processing is increasingly used to reduce fat droplet size to improve stability. This paper highlights one unintended consequence of reducing fat droplet size, namely increasing the tendency for lactose to crystallize. This should be factored in to any future formulations that involve high shear processing of milk derived ingredient s for subsequent spray drying.

Ac ce p

470 471 472 473 474 475 476 477 478

an

us

464

Page 21 of 33

Maher Pg. 21 479 480

Table 1 Summary of kinetic parameters obtained by fitting the crystallization isotherms to the

481

Avrami model and Yang model.

ip t

482 483

n

t1/2

r2

k

(min-1)

-

(min)

-

(min-1)

C

1.1 x 10-8 a

3.06 a

363 a

0.995

7.3 x 10-12 a

N

1.2 x 10-7 b

2.76 b

287 b

0.985

2.3 x 10-9 a

485

a-b

486

way ANOVA.

490 491 492 493 494

r2

t1/2

-

(min)

-

(min)

4.39 a

323 a

0.976

368 a

4.31 a

262 b

0.970

285 b

d

M

Values within a column not sharing a common letter differ significantly (P < 0.05) by one-

te

489

t1/2

Ac ce p

488

Experimental

n

an

k

Powder

487

Yang

us

Avrami

cr

484

495 496 497 498 499

Page 22 of 33

Maher Pg. 22 500 Figure captions

502

Figure 1. a) Dynamic vapour sorption graph of control (C) and nanoemulsion (N) powders

503

held at 55% RH for 24 h showing lactose crystallization. b) Crystallization isotherms for C

504

and N at 25 °C for experimental data and those calculated using the Avrami equation.

506

Figure 2. Polarized light (crossed polars) micrographs of control (a,c,e,g,i) and nanoemulsion

507

(b,d,f,h,j) powders. after storage for 0 (a,b), 1 (c,d), 2 (e,f), 3 (g,h) and 4 (i,j) days at 55%

508

relative humidity. Arrows indicate small crystals, scale bar = 100 Pm.

an

us

cr

505

ip t

501

509

Figure 3. Confocal scanning laser micrographs of showing internal microstructures of control

511

(a,c,e) and nanoemulsion (b,d,f) powders after storage for 0 (a,b) and 4 (c,d,e,f) days at 55%

512

relative humidity. Figure 4 a – d are fluorescently labelled with Nile red/fast green to show

513

fat (green) and protein (red); Figure 4 e & f are labelled with Nile Red to show fat phase.

514

Arrows indicate air vacuoles (a, b), discrete fat droplets (c) and solid lactose crystals (d). Fig

515

3e inset shows irregularly shaped coalesced fat regions. Scale bars = 100 Pm.

d

te

Ac ce p

516

M

510

517

Figure 4. Cryo-scanning electron micrographs of control (a,c,e,g) and nanoemulsion (b,d,f,h)

518

powders prior to crystallization. Images show powder particle surface morphology (a,b) and

519

interior features of fractured particles (c – h). Arrows indicate air vacuoles (c,d) and fat

520

droplets (e,g,h). Scale bars = 20 Pm (a,b), 10 Pm (c,d), 2 Pm (e,f) and 500 nm (g,h).

521 522

Figure 5. Cryo-scanning electron micrographs of control (a,c,e) and nanoemulsion (b,d,f)

523

powders after storage for 4 days at 55% relative humidity showing crystals (arrowed) on the

Page 23 of 33

Maher Pg. 23 powder surfaces (a,b,c,d) and also interior of fractured particles. Scale bars = 20 Pm (a,b), 5

525

Air vacuole Pm (c,d) and 2 Pm (e,f).

526

Figures

Ac ce p

te

d

M

an

us

cr

ip t

524

527 528

Figure 1.

529 530

Page 24 of 33

Maher Pg. 24 531 532 533

a

b

c

d

e

f

ip t

534 535

cr

536

538

us

537

an

539 540

te

544 545 546 547 548

g

549 550 551 552

i

Ac ce p

543

d

542

M

541

h

j

553 554 555

Page 25 of 33

Maher Pg. 25 556

558 559

Figure 2.

a

b

c

d

ip t

557

560

cr

561 562

us

563

an

564

566

M

565

567

d

568

te

569

571 572 573 574 575 576

Ac ce p

570

e

f

577 578 579 580

Page 26 of 33

Maher Pg. 26 581 582 583

Figure 3.

a

b

c

d

ip t

584 585

cr

586

589

an

588

590

M

591 592

597 598 599 600 601

Ac ce p

595 596

f

te

e

d

593 594

us

587

g

h

602 603 604

Figure 4.

605

Page 27 of 33

Maher Pg. 27 606 607 608

a

b

c

d

ip t

609 610

cr

611 612

an

614 615

M

616 617

d

618

e

f

te

619

622 623 624 625 626

Ac ce p

620 621

us

613

Figure 5.

627 628 629 630

Page 28 of 33

Maher Pg. 28 631 632 References

634

Al-Mulla, A., (2007). Isothermal crystallization kinetics of poly(ethylene terephthalate) and

635

poly(methyl methacrylate) blends. eXPRESS Polymer Letters 1(6), 334-344.

636

Arvanitoyannis, I., Blanshard, J.M.V., (1994). Rates of Crystallization of Dried Lactose-

637

sucrose Mixtures. Journal Of Food Science 59(1), 197-205.

638

Auty, M.A.E., Twomey, M., Guinee, T.P., Mulvihill, D.M., (2001). Development and

639

application of confocal scanning laser microscopy methods for studying the distribution of fat

640

and protein in selected dairy products. Journal of Dairy Research 68, 417-427.

641

Avrami, M., (1939). Kinetics of phase change. I. General theory. Journal of Chemical Physics

642

7, 1103-1112.

643

Avrami, M., (1940). Kinetics of Phase Change II. Transformation-Time Relations for

644

Random Distribution of Nuclei. Ibid 8, 212-224.

645

Burnett, D., Thielmann, F., Booth, J., (2004). Determining the critical relative humidity for

646

moisture-induced phase transitions. International Journal of Pharmaceutics 287, 123-133.

647

Burnett, D., Thielmann, F., Sokoloski, T., Brum, J., (2006). Investigating the moisture-

648

induced crystallization kinetics of spray-dried lactose. International Journal of Pharmaceutics

649

313, 23-28.

650

Bushill, J.H., Wright, W.B., Fuller, C.H.F., Bell, A.V., (1965). The crystallisation of lactose

651

with particular reference to its occurrence in milk powder. Journal of the Science of Food and

652

Agriculture 16(10), 622-628.

653

Cahn, J.W., (1956). Transformation kinetics during continuous cooling. Acta Metallurgia 4,

654

572-575.

Ac ce p

te

d

M

an

us

cr

ip t

633

Page 29 of 33

Maher Pg. 29 Dudkiewicz, A., Tiede, K., Loeschner, K., Jensen, L.H.S., Wierzbicki, R., Boxall, A.B.A.,

656

Molhave, K., (2011). Characterization of nanomaterials in food by electron microscopy.

657

Trends in Analytical Chemistry 30(1), 28-43.

658

Fäldt, P., Bergenståhl, B., (1996). Changes in Surface Composition of Spray-Dried Food

659

Powders due to Lactose Crystallization. LWT - Food Sci. Technol. 29(5–6), 438-446.

660

Gaisford, S., Verma, A., Saunders, M., Royall, P.G., (2009). Monitoring crystallisation of

661

drugs from fast-dissolving oral films with isothermal calorimetry. International Journal of

662

Pharmaceutics 380, 105-111.

663

Haque, M.K., Roos, Y.H., (2005). Crystallization and X-ray diffraction of spray-dried and

664

freeze-dried amorphous lactose. Carbohydrate Research 340, 293-301.

665

Hartel, R., (2001). Measurements of crystalline structure in foods. In: Crystallization in

666

Foods. Aspen Publishers: Gaithersburg, MD, p 34-90.

667

Hogan, S.A., McNamee, B.F., O'Riordan, E.D., O'Sullivan, M., (2001a). Emulsification and

668

microencapsulation properties of sodium caseinate/carbohydrate blends. International Dairy

669

Journal 11(3), 137-144.

670

Hogan, S.A., McNamee, B.F., O'Riordan, E.D., O'Sullivan, M., (2001b). Microencapsulating

671

Properties of Sodium Caseinate. Journal of Agricultural and Food Chemistry 49, 1934-1938.

672

Iglesias, H.A., Chirife, I., (1978). Delayed crystallization of amorphous sucrose in humidified

673

freeze dried model systems. Journal of Food Technology 13, 137-144.

674

Jouppila, K., Kansikas, J., Roos, Y.H., (1997). Glass transition, water plasticization, and

675

lactose crystallization in skim milk powder. J Dairy Sci 80(12), 3152-3160.

676

Jouppila, K., Roos, Y.H., (1994). Glass transitions and crystallization in milk powders. J

677

Dairy Sci 77(10), 2907-2915.

678

Kawamura, K., (1979). The DSC Thermal Analysis of Crystallization Behaviour in Palm Oil.

679

Ibid 56, 753-758.

Ac ce p

te

d

M

an

us

cr

ip t

655

Page 30 of 33

Maher Pg. 30 Kedward, C.J., MacNaughtan, W., Blanshard, J.M.V., Mitchell, J.R., (1998). Crystallization

681

Kinetics of Lactose and Sucrose On Isothermal Differential Scanning Calorimetry. Journal Of

682

Food Science 63(2), 192-197.

683

Kedward, C.J., MacNaughtan, W., Mitchell, J.R., (2000). Isothermal and non-isothermal

684

crystallization in amorphous sucrose and lactose at low moisture contents. Carbohydrate

685

Research 329(2), 423-430.

686

Kelly, G.M., O’Mahony, J.A., Kelly, A.L., O’Callaghan, D.J., (2014). Physical

687

characteristics of spray-dried dairy powders containing different vegetable oils. Journal of

688

Food Engineering 122(0), 122-129.

689

Lai, H., Schmidt, S., (1990). Lactose crystallisation in skim milk powder observed by

690

hydrodynamic equilibria, scanning electron micrroscopy and 2H nuclear magnetic resonance.

691

J Food Sci 55(4), 994-999.

692

Maher, P.G., Roos, Y.H., Fenelon, M.A., (2014). Physicochemical properties of spray dried

693

nanoemulsions with varying final water and sugar contents. J. Food Eng. 126(0), 113-119.

694

Mahlin, D., Berggren, J., Alderborn, G., Engstrom, S., (2004). Moisture-Induced Surface

695

Crystallization of Spray-Dried Amorphous Lactose Particles Studied by Atomic Force

696

Microscopy. Journal of Pharmaceutical Sciences 93(1), 29-37.

697

Mazzobre, M.a.F., Aguilera, J.M., Buera, M.a.P., (2003). Microscopy and calorimetry as

698

complementary techniques to analyze sugar crystallization from amorphous systems.

699

Carbohydrate Research 338(6), 541-548.

700

McCarthy, N.A., Gee, V.L., Hickey, D.K., Kelly, A.L., O'Mahony, J.A., Fenelon, M.A.,

701

(2013). Effect of protein content on the physical stability and microstructure of a model

702

infant formula. International Dairy Journal 29(1), 53-59.

703

McKenna, A.B., (1997). Examination of whole milk powder by confocal laser scanning

704

microscopy. Journal of Dairy Research 64, 423-432.

Ac ce p

te

d

M

an

us

cr

ip t

680

Page 31 of 33

Maher Pg. 31 Miao, S., Roos, Y.H., (2005). Crystallization Kinetics and X-ray Diffraction of Crystals

706

Formed in Amorphous Lactose, Trehalose, and Lactose/Trehalose Mixtures. Journal Of Food

707

Science 70(5), E350-E358.

708

Murrieta-Pazos, I., Gaiani, C., Galet, L., Cuq, B., Desobry, S., Scher, J., (2011). Comparative

709

study of particle structure evolution during water sorption: Skim and whole milk powders.

710

Colloids and Surfaces B: Biointerfaces 87(1), 1-10.

711

Pisecky, J., (1997). Handbook of milk powder manufacture, Copenhagen.

712

Roetman, K., (1979). Crystalline lactose and the structure of spray-dried milk products as

713

observed by scanning electron microscopy. Netherlands Milk and Dairy Journal 33, 1-11.

714

Roos, Y., Karel, M., (1990). Differential scanning calorimetry study of phase transitions

715

affecting the quality of dehydrated materials. Biotechnology Progress 6(2), 159-163.

716

Roos, Y.H., Karel, M., (1991). Plasticizing Effect of Water on Thermal Behaviour and

717

Crystallization of Amorphous Food Models. J Food Sci 56(1), 775-777.

718

Schmidt, E.A., Law, D., Zhang, G.G.Z., (1999). Nucleation and Crystallization Kinetics of

719

Hydrated Amorphous Lactose above the Glass Transition Temperature. Journal of

720

Pharmaceutical Sciences 88(3), 291-296.

721

Sharples, A., (1966). Overall Kinetics of Crystallization. In: Sharples, A., editor. Introduction

722

to Polymer Crystallization, Edward Arnold Ltd, London, p 44–59.

723

Shimada, Y., Roos, Y., Karel, M., (1991). Oxidation of methyl linoleate encapsulated in

724

amorphous lactose-based food model. Journal of Agricultural and Food Chemistry 39(4),

725

637-641.

726

Soottitantawat,

727

Microencapsulation by spray-drying: Influence of emulsion size on the retention of volatile

728

compounds. J. Food Sci. 68(7), 2256-2262.

Ac ce p

te

d

M

an

us

cr

ip t

705

A.,

Yoshi,

H.,

Furata,

T.,

Ohkawara,

M.,

Linko, P.,

(2003).

Page 32 of 33

Maher Pg. 32 Soutari, N., Buanz, A.B.M., Gul, M.O., Tuleu, C., Gaisford, S., (2012). Quantifying

730

crystallisation rates of amorphous pharmaceuticals with dynamic mechanical analysis

731

(DMA). International Journal of Pharmaceutics 423, 335-340.

732

` {!€!!$ !‚ ƒ$!['„!†  ‡ˆ-Lactoglobulin Interaction During

733

Storage of Model Whey Powders. Journal of Dairy Science 87(5), 1158-1166.

734

Tobin, M.C., (1974). Theory of phase transition kinetics with growth site impingement. I.

735

Homogenous nucleation. Journal of Polymer Science Part B: Polymer Physics 12, 399-406.

736

Urbanovivi, E., Segal, E., (1990). New formal relationships to describe kinetics of

737

crystasllisation. Thermochimica Acta 171, 87-94.

738

Yang, J., Grey, K., Doney, J., (2010). An improved kinetics approach to describe the physical

739

stability of amorphous solid dispersions. International Journal of Pharmaceutics 384(1–2),

740

24-31.

an

M te

744

Ac ce p

743

d

741 742

us

cr

ip t

729

Page 33 of 33