Formulation and process optimizations of nano-cosmeceuticals containing purified swiftlet nest

Abstract

The good effects of swiftlet nest (SN) on the skin have been widely investigated. However, the importance of the SN in the area of nano-cosmeceutical research is still limited as it is commonly consumed, rather than being applied to the skin. Formulation and optimization processes are two important issues in the manufacturing of cosmetics. In this work, Response surface methodology (RSM) was utilized to investigate the influence of the nano-cosmeceuticals composition; purified SN (1–5% w/w), and Tween 80 (3–6% w/w) as well as the preparation method; time of homogenization (10–30 min), on the physicochemical properties of nano-cosmeceuticals. The response variables were particle size and zeta potential which were very important characteristics in nano-cosmeceuticals. Formulation and optimization of three independent variables were carried out to obtain an optimum nano-cosmeceuticals with the lowest particle size and high stability formulation. The optimized nano-emulsion containing purified SN with particle size of 136.35 nm and zeta potential of −40.2 mV was successfully formulated. This was obtained experimentally and was closer to the predicted values of 136.22 nm and −40.07 mV, respectively. The optimized formulation remained stable for three months after the centrifugation test as well as storage at different temperatures of 4 °C, 25 °C and 45 °C.

---

Introduction

The term cosmetic, which is used on a daily basis, only places emphasis on outward appearance. Currently, medical skin care products which are termed dermaceuticals or cosmeceuticals are in high demand. Furthermore, cosmeceutical skin care is committed to maintaining and enhancing the long-term function of the skin. Nano-cosmeceuticals are the strongest generation of skin care products using nano-sized systems for the delivery of active ingredients to the skin cells for better penetration. They are subjected to special manufacturing procedures and contain higher concentrations of active substances compared to classical care cosmetics.

In cosmeceuticals, emulsions are widely used as delivery system for certain active ingredients. Emulsions are metastable colloids made out of two immiscible fluids, one being dispersed in the other, in the presence of surfactants. Emulsion droplets exhibit all the classical behaviours of metastable colloids: Brownian motion, reversible phase transitions due to droplet interactions that may be strongly modified and irreversible transitions that generally involve their destruction.¹ Emulsion consists of oil droplets dispersed in an aqueous phase is called oil-in-water or o/w emulsion and water-in-oil or w/o emulsion that consists of water droplets dispersed in an oil phase.² The delivery system used in this nano-cosmeceuticals is nano-emulsion. Nano-emulsion is defined as emulsion systems having particle sizes ranging from 20–500 nm.³

Recently, research about flora and fauna has been undertaken which might prove beneficial to the human body. Some of them are being used as active ingredients in new cosmetics. One of the fauna that was found to have a good value was a swiftlet nest (SN). This is made from the saliva of an insectivorous bird named the swiftlet, which mainly inhabit limestone caves.⁴ SN can be produced by several different swiftlet species in the genus of Aerodramus and Collocalia. The nests are mainly built by male swiftlet and made almost entirely from the saliva secreted by the swiftlet's two sublingual glands.⁵ There have been a number of studies conducted on the benefits of SN in food, medicine and cosmetics. The main component of the SN is glycoprotein. The existence of glycoprotein is capable of promoting cell division, and it has demonstrated the presence of an epidermal growth factor-like protein.^6,7 SN has a good effect on the skin and is attractive to women due to the properties that make the skin delicate and radiant. Although the importance of the SN has provided a high potential in the area of new nano-cosmeceuticals, research in this area is still limited as it is commonly consumed, rather than being applied to the skin. Hence, in this research, new formulations of nano-cosmeceuticals containing purified SN were developed and formulated with using nanotechnology to allow better penetration, in order to treat various skin conditions, such as redness, dehydration or lack of firmness, wrinkles and dark spots, or to brighten the skin.

Formulation and optimization processes are two important issues in manufacturing. The process of optimizing parameters not only increases the utility of the technologist, but also the quality of the product as well. Hence, in this study, response surface methodology was selected as the optimization technique for the formulation of nano-cosmeceuticals, as it is a practical and simple method of analyzing, improving and optimizing over the feasible domain of parameter settings. RSM is the most popular optimization method used in recent years.^8–10 Several studies have been investigated based on the application of the RSM in industrial processes. This method is a collection of statistical techniques in which a response of interest is influenced by several variables. The objective is to optimize this response by determining the relationship between the response and independent variables.¹¹

Therefore, the aim of this work was to find the optimum formulation of nano-cosmeceuticals by using RSM and to simultaneously evaluate the main effects and interaction effects between the factors, including the composition of purified SN percentage, Tween 80, and the time of homogenization on the responses, particle size and zeta potential of the formulation.

Methodology

Materials

Swiftlet nest (SN) was collected from a local SN farmer in Selangor, Malaysia. Polyoxyethylene sorbitan mono-oleate (Tween 80) was purchased from Fluka Chemie GmbH, USA. Potassium sorbate and jojoba oil were obtained from Making Cosmetics Inc., while olive oil was purchased from Borges, Nasmark, Malaysia. Essential oil was purchased from Wellness, Original Ingredient, Malaysia. Deionized water was produced in the laboratory (Halal Products Research Institute, UPM, Malaysia). All chemicals used were of analytical, food or cosmetic grade classes.

Purification of swiftlet nest

Raw SN from swiftlet species of Aerodramus fuciphagus was soaked in deionized water for 24 hours and the water was changed for every 2 hours. Any leftover particles and feather were removed. The cleaned SN was freeze dried for three days. Water was removed from SN after it is frozen at −80 °C and placed under vacuum. The freeze dried SN was ground in a blender and then SN was stored in chiller prior to use.

Preparation of nano-cosmeceuticals containing swiftlet nest

Nano-cosmeceuticals were formulated using a mixture of jojoba and olive oil as dispersed oil phase and deionized water containing SN and Tween 80 as the continuous aqueous phase. The aqueous phase, consisting of water (q.s), was heated to a temperature of 60–65 °C and then the purified SN was added, followed by adding Tween 80. At the same time, the oily phase consisting of jojoba oil and olive oil phase was heated to the same temperature. The oil phase was then added to the water phase drop by drop with continuous stirring. Then, the mixture was homogenized using a homogenizer (IKA®T18 Basic ULTRA-TURRAX, Germany) at 13 500 rpm. The speed of the homogenizer was reduced to 9500 rpm while adding potassium sorbate solution (0.3%) drop wise as a preservative. The speed of the mixer was further reduced to 6500 rpm for complete homogenization while adding 0.1% essential oil; until the emulsion reached room temperature. The final products were put into sample bottles.

Experimental design

A three-factor central composite design (CCD) was utilized to study the effect of Tween 80 amount (3–6% w/w, A), SN amount (1–5% w/w, B), and time of homogenization (10–30 min, C) on the two response variables: particle size and zeta potential. Hence, based on the CCD, a total of 20 experiments was run using Design Expert software (version 6.0.6, Stat ease Inc., Minneapolis, USA). The experimental runs involved 8 factorial points, 6 axial points and 6 replicates of centre points at 5 levels of each variable in two blocks. The data in block 1 contains design points of eight factorial points, plus four centre points were collected first and analyzed. Then, the experiment was followed by other formulation in block 2. The use of blocked design with allows the estimation of individual and interaction factor effects independently of block effects. Blocks are assumed to have no impact on nature and shape of response surface.¹² But, for each block, the experiments were carried out in a randomized order to minimize the effect of unexplained variability on the actual response due to extraneous factors.¹⁴ To determine the repeatability of the method, the centre point was repeated six times. The proper choice of design is very important in any response surface investigation. The choice of CCD as the experimental design is because it is more precise for estimating factor effects, the interaction effect between factors can be evaluated and permits optimization in the full factor space.¹³ The experimental data were analyzed by response surface regression procedure and the results were statistically analyzed by the corresponding analyses of variances. An appropriate polynomial model was chosen based on the statistical significance of the model (p < 0.05) and the lack of fit value of the model provided by Design-Expert software was not significant. The matrix of CCD is shown in Table 1.

Table 1 The experimental data obtained for the two responses based on the CCD matrix^a

Formulation number	Type	A	B	C	Particle size (nm)	Zeta potential (mV)
a A: Tween 80 amount; B: swiftlet nest amount; C: time of homogenization.
1	Fact	3.61	1.81	14	140.4	−48.3
2	Fact	3.61	4.19	14	329.4	−42.8
3	Fact	3.61	1.81	26	116.6	−51
4	Fact	5.39	4.19	26	110.3	−44.2
5	Fact	5.39	1.81	14	98.2	−31
6	Fact	5.39	1.81	26	105.1	−34.4
7	Centre	4.50	3.00	20	123.5	−35.7
8	Fact	5.39	4.19	14	280	−46.9
9	Centre	4.50	3.00	20	123.1	−36.9
10	Centre	4.50	3.00	20	125.9	−36.3
11	Fact	3.61	4.19	26	127.5	−39.8
12	Centre	4.50	3.00	20	123.1	−36.2
13	Axial	6.00	3.00	20	91.5	−36.1
14	Axial	4.50	3.00	30	97.6	−44.9
15	Centre	4.50	3.00	20	118.4	−35.6
16	Axial	4.50	5.00	20	253.8	−43
17	Centre	4.50	3.00	20	118.1	−35.6
18	Axial	3.00	3.00	20	135	−47.8
19	Axial	4.50	3.00	10	254.2	−47.2
20	Axial	4.50	1.00	20	95.3	−37.4

---

Statistical analysis

Numerical optimization was carried out to determine the optimum condition of the independent variables when predicting the variation of ingredient compositions as well as preparation conditions using response optimizer in the Design Expert software. The desired goal for each variable was chosen. The optimal composition and conditions of nano-cosmeceuticals formulations were chosen based on the condition of attaining minimum particle size and the highest absolute value of zeta potential. The responses were then analyzed jointly by conferring to them either the same importance or weight for simultaneous optimization of the multiple responses. The Design Expert software overlaying all the responses to obtained the optimal condition. The optimal condition that depended on the independent variables was also obtained using the predicted equations determined by RSM. By using the polynomial regression equation, the response surface behaviour was explored for the response function (Y_i). The generalized response surface model is shown below:

Y_i = A₀ + β_aA + β_bB + β_cC + β_aaA² + β_bbB² + β_ccC² + β_abAB + β_acAC + β_bcBC

where Y_i is the predicted response (particle size/zeta potential); A₀ is constant; and β_i, β_ii and β_ij are the linear, quadratic and interaction coefficients, respectively.¹³ The statistical significance of the term in the regression equations was determined by utilizing analysis of variance (ANOVA) for each response and the significance test level was set at 5% (p > 0.05). Response surfaces and 3-D contour plots of the fitted polynomial regression equations were generated to better visualize the interaction effect of the independent variables on responses.

Particle size analysis

The particle size distribution is one of the most important physical characteristics of a nano-cosmeceutical. The sample was measured by a diffusion method using a dynamic light scattering (DLS) particle analyzer (Malvern Nano ZS90, Malvern, UK). The measurement was performed at a scattering angle of 173° and a temperature of 25 °C. The nano-cosmeceuticals were diluted with deionized water to the required concentration. Then, the diluted emulsions were placed in the disposable folded capillary cell. The count rate was maintained between 100 and 300 kcps. The measurements were performed in triplicate and the values were reported as mean values. The particle size was measured based on the intensity weighted distribution (Z-average) within range of 20–500 nm³.

Zeta potential analysis

The surface charge of the inventions was tested using Malvern Nano ZS90, Malvern, UK. The measurement was performed at a scattering angle of 173° and a temperature of 25 °C. The nano-cosmeceuticals were diluted with deionized water to the required concentration. Then, the diluted emulsions were placed in the disposable folded capillary cell. The count rate was maintained between 100 and 300 kcps. Particles with a zeta potential value more positive than +30 mV or more negative than −30 mV were considered stable. All the measurements were repeated three times.

Stability study

Centrifugal tests were carried out immediately after preparation. Each sample (10 g) was subjected to 15 min centrifugation at 4500 rpm, at room temperature (25.0 ± 0.5 °C). The thermal stability of nano-cosmeceuticals was evaluated at different temperatures. Nano-cosmeceutical samples (10 g) were placed at room temperature (25.0 ± 0.5 °C), 45 °C and 4 °C. Observations were made on a monthly basis for three months.

Results and discussion

Particle size and zeta potential analysis

The mean values of particle size and zeta potential of the formulation are shown in Table 1. Both responses were significantly different between assays with very low probability value (P model, F < 0.0001), which is a basic requirement for further RSM.¹⁵ Based on the resultant data, nano-cosmeceuticals show particle sizes below 400 nm and zeta potential values less than −30 mV by restraining the range of SN amount, Tween 80 amount and time of homogenization at levels of 1–5%, 3–6% and 10–30 min, respectively.

Fitting the response surface models

The variation in the particle size and zeta potential were predicted by employing RSM as the responses were the function of the nano-cosmeceuticals composition and preparation variables of the formulations. Table 1 shows the experimental data obtained for the two response variables based on the central composite design matrix.

The experimental data were statistically analyzed. The statistical analysis was used to determine the best fitted model for the three independent variables. The estimated regression coefficients, R², adjusted R², regression (p-value), regression (F-value), lack of fit (p-value) and probability values related to the effect of the three independent variables are shown in Table 2. Negative values of coefficient estimates denote negative influence of independent variables on the responses, while positive values indicate the directly proportional relationship between factors and responses.

Table 2 Analysis of variance (ANOVA) for the model^a

Source	Particle size			Zeta potential
	Coefficient estimate	F-value	p-value	Coefficient estimate	F-value	p-Value
a A0 is constant; A, B, C and D are the linear, regression coefficient for the linear effect; A^2, B^2 and C^2: regression coefficient for the quadratic effect; AB, AC and BC: regression coefficient for the interaction effect of the quadratic polynomial model.
A0	120.95	—	—	−36.13	—	—
A	−14.17	619.91	<0.0001	3.30	226.43	<0.0001
B	47.85	7072.88	<0.0001	−1.35	37.80	0.0002
C	−47.73	7038.39	<0.0001	0.25	1.34	0.2768
A^2	−1.54	7.74	0.0213	−1.90	79.33	<0.0001
B^2	20.13	1320.10	<0.0001	−1.28	36.12	0.0002
C^2	20.61	1383.44	<0.0001	−3.35	246.26	<0.0001
AB	−1.61	4.71	0.0582	−5.30	342.00	<0.0001
AC	7.86	111.87	<0.0001	−0.12	0.19	0.6730
BC	−44.34	3557.43	<0.0001	1.48	26.49	0.0006

---

Response variable	F-value	(p-value)	R^2	Adjusted R^2	Lack of fit (p-value)
Particle size	2328.25	<0.0001	0.9996	0.9991	0.0720
Zeta potential	105.49	<0.0001	0.9906	0.9812	0.0583

---

The response surface analysis demonstrated that the second-order polynomial used for particle size has a higher coefficient of determination (R² = 0.9996) as compared to the zeta potential (R² = 0.9906). The obtained coefficient of determination shows that more than 90% of the response variation of the particle size and zeta potential could be described by RSM models as the function of the main nano-cosmeceuticals and preparation variables. It was observed that the lack of fit gave no indication of significance (p < 0.05) for the model, therefore proving the satisfactory fitness of the response surface model to the significant (p < 0.05) factors effect.

Table 2 shows that only one independent variable (B) exhibited a positive effect on the response of particle size. Thus, the positive effect of SN on response implies that higher SN amount causes higher particle size. For zeta potential, two independent variables (A and C) presented a positive effect. Coefficients with more than one factor, or higher order terms in the regression equation, represent the interaction between terms or the quadratic relationship, respectively, which suggest a non-linear relationship between factors and responses.¹⁶ Both of the responses were affected by the interaction of independent variables, presenting a quadratic relationship. The interaction effects between A and C were favourable only for particle size response. A favourable effect was also noticed for zeta potential response, for the interaction between B and C. However, it was observed that the interaction between A and B had an inverse effect for both responses.

The coefficient significance of the quadratic polynomial models was evaluated by using ANOVA. For any of the terms in the models, a large F-value and a small p-value indicated a more significant effect on the respective response variables.¹⁷ Table 2 also shows the effect of independent variables on the variation of the physicochemical properties of nano-cosmeceuticals. All independent variables affect on the particle size of the nano-cosmeceuticals for the linear terms. The quadratic term of all independent variables also had a significant effect on the particle size of nano-emulsions. For the interaction between Tween 80 amount and time of homogenization (AC) and between SN amount and time of homogenization (BC), the significant effects were observed on the particle size of nano-cosmeceuticals.

The variable which exhibited the largest effect on the zeta potential of the nano-cosmeceuticals for the linear term were Tween 80 amount with F-value of 226.43 and followed by SN amount with F-value of 37.80. The time of homogenization showed insignificant effects (p > 0.05). The quadratic terms of all three factors exhibited significant effects (p < 0.05) on the zeta potential value with larger F-value of 246.26 for C² that indicated more significant effects. However, the interaction between Tween 80 and SN amount showed highest effect on the zeta potential compared to the rest of terms.

Response surface analysis

For the optimization of nano-cosmeceuticals containing SN, response surface analyses were plotted in three dimensional model graphs. The response surface plots for particle size and zeta potential, which are used to interpret the interaction effect of the variables, are presented in Fig. 1–4, respectively. The third factor was kept at a constant level.

Fig. 1 Response surface plots showing the interaction effects of time of homogenization and Tween 80 amount on particle size.

Fig. 2 Response surface plots showing the interaction effects swiftlet nest (SN) amount and time of homogenization on particle size.

Fig. 3 Response surface plots showing the interaction effects of time of homogenization and swiftlet nest (SN) amount on zeta potential.

Fig. 4 Response surface plots showing the interaction effects of swiftlet nest amount and Tween 80 amount on zeta potential.

Fig. 1 and 2 demonstrate that the particle size was decreased with increasing homogenization time. The homogenization time is an important duration factor for the resulting particles. Short periods of homogenization restrain the droplet into nano-particles; otherwise, long periods of time may cause instability of the colloidal particles due to the high input of energy that leads to the aggregation of colloidal particles into larger micro-particles.¹⁸ In order to avoid aggregation, colloidal particles are typically stabilized kinetically by electrostatic repulsions.¹⁹ Repulsive electrostatic forces could be formed when SN, which acts as a protein, is dissolved in an electrolyte solution. These repulsive forces between proteins prevent aggregation and facilitate dissolution.

Fig. 1 shows that increasing Tween 80 content lead to a decrease in particle size. This could be due to the fact that emulsifier plays a vital role in the formation of emulsion as it lowers the interfacial tension, thereby the Laplace pressure, p, is reduced and the stress required for droplet deformation is reduced.¹³

The zeta potential is a stability indicative parameter in colloidal systems like submicron emulsions due to electrostatic repulsion.^19,20 We found that initially, with increasing the time of homogenization, the negative zeta potential value decreases and then increases with further increase in the time of homogenization (Fig. 3). Fig. 3 and 4 demonstrates that by increasing the amount of SN and Tween 80, the negative zeta potential value increased. This condition could be due to inter-particle tend to repel each other as the same charges give rise to higher repulsion which resulted in the decrease rate of coagulation and flocculation. In addition, increasing amount of Tween 80 decreased the negative value of zeta potential with decreasing of SN amount. The reason for this behaviour could be due to van der Waals attraction forces that will eventually aggregate. This phenomenon is not only contributed by the surfactant role but is also due to the SN amount.

Optimization of nano-cosmeceuticals formulation

By using Design-Expert software, the desirability function was probed to acquire an optimized formulation. The optimum formulation of nano-cosmeceuticals containing SN was formulated with the smallest particle size and highest absolute value of zeta potential value. The response surface and contour plot were used to visualize the interaction between the independent variables. By investigating the interaction effect between the independent variables and evaluating the optimization constraints, the optimum nano-cosmeceuticals was prepared with a composition of 2.58% SN, 3.99% Tween 80 and time of homogenization of 17 min.

In order to verify the optimum formulation, the nano-cosmeceuticals using the optimal ingredient and homogenization level were formulated and analyzed and the results were statistically compared to the predicted values of the mathematical model. The predicted response values and the actual obtained response values for the optimized products were within the range and found to be not statistically different at the 95% confidence level. Based on the optimum formulation, the predicted values of particle size and zeta potential were 136.22 nm and −40.07 mV, respectively. The analysis showed that the nano-cosmeceuticals formulation had the particle size value of 136.35 nm and zeta potential value of −40.2 mV.

Stability study

The nano-cosmeceuticals containing purified SN prepared based on the recommended optimum conditions remained stable after centrifugation as well as at room temperature (25.0 ± 0.5 °C), 45 °C and 4 °C, with no separation observed during the three months of storage. The excellent stability could be due to the steric stabilizing effect of the non-ionic emulsifier (Tween 80) in which a bulk steric barrier is formed against particle collision. Thus, this phenomenon prevents the occurrence of flocculation and coalescence. In essence, it is hypothesized that the whole nano-system is in a stable state, which might be due to the rapid absorption of the non-ionic surfactant, Tween 80 onto the droplet interface.

Conclusions

In this paper, nano-cosmeceuticals containing SN with desirable characteristics was successfully formulated. The study showed that response surface methodology is a beneficial tool for identifying and optimizing the best combination of independent variables SN, Tween 80 amounts and time of homogenization for the preparation of nano-cosmeceuticals formulations. The variation in the average particle size and zeta potential were predicted by employing second order polynomial regression. Generally, the linear effect of SN and Tween 80 had a significant effect (p < 0.05) on the zeta potential while all independent variables had a significant effect on the particle size. The interaction effect between Tween 80 amount and time of homogenization and between SN amount and time of homogenization had the significant effect (p < 0.05) on the particle size. Conversely, the zeta potential was significantly (p < 0.05) affected by the interaction effect between Tween 80 and SN amount and between SN amount and time of homogenization. The quadratics of all three independent variables had the significant (p < 0.05) effects on the two response variables studied. The final goal to obtain a nano-cosmeceuticals formulation containing SN with the lowest particle size and high stability formulation was determined to be 2.58% SN, 3.99% Tween 80, 90.03% deionized water and 3.4% other ingredients, with a time of homogenization of 17 min. Hence, the optimized formulation obtained will be very beneficial and helpful to the SN and nano-cosmeceuticals industries worldwide.

Notes and references

1. J. Bibette, F. Leal Calderon and P. Poulin, Rep. Prog. Phys., 1999, 62, 969–1033.
2. S. Zainol, Master thesis, Universiti Putra Malaysia, 2014.
3. I. Solè, A. Maestro, C. M. Pey, C. González, C. Solans and J. M. Gutiérrez, Colloids Surf., A, 2006, 288, 138–143.
4. F. Ma and D. Liu, Food Res. Int., 2012, 48(2), 559–567.
5. D. L. M. Goh, K. Y. Chua, F. T. Chew, T. K. Seow, K. L. Ou, F. C. Yi and B. W. Lee, J. Allergy Clin. Immunol., 2001, 107(6), 1082–1088.
6. M. H. Ng, K. H. Chan and Y. C. Kong, Biochem. Int., 1986, 13, 521–531.
7. Y. C. Kong, W. M. Keung, T. T. Yip, K. M. Ko, S. W. Tsao and M. H. Ng, Comp. Biochem. Physiol., Part B: Biochem. Mol. Biol., 1987, 87, 221–226.
8. C. Chellamboli and M. Perumalsamy, RSC Adv., 2014, 4, 22129–22140.
9. A. Ryad, K. Lakhdar, K.-S. Majda, A. Samia, A. Mark, A.-D. Corinne and G. Eric, Int. J. Mol. Sci., 2010, 11(11), 4726–4740.
10. D. Baş and İ. H. Boyacı, J. Food Eng., 2007, 78(3), 836–845.
11. S. Raissi and R. Farsani, Proc. World Acad., 2009, 3, 251–255.
12. H. Mirhosseini and C. P. Tan, Food Chem., 2009, 115, 324–333.
13. S. Zainol, M. Basri, H. B. Basri, A. F. Shamsuddin, S. S. Abdul-Gani, R. A. Karjiban and E. Abdul-Malek, Int. J. Mol. Sci., 2012, 13(10), 13049–13064.
14. S. Y. Tang, S. Manickam, T. K. Wei and B. Nashiru, Ultrason. Sonochem., 2012, 19(2), 330–345.
15. P. Ritthiruangdej, W. Srikamnoy and T. Amatayakul, Nat. Sci., 2011, 334, 325–334.
16. S. K. Motwani, S. Chopra, S. Talegaonkar, K. Kohli, F. J. Ahmad and R. K. Khar, Eur. J. Pharm. Biopharm., 2008, 68(3), 513–525.
17. A. M. Joglekar and A. T. May, Cereal Foods World, 1987, 32, 857–868.
18. A. Gardouh, M. Ghorab and S. Abdel-Rahman, ARPN J. Sci. Technol., 2011, 2(10), 996–1006.
19. P. Arosio, Stability and aggregation kinetics of colloidal systems: Application to polymer colloids, proteins and peptides [dissertation], ETH Zurich. DISS. ETH NO.20098., 2011.
20. N. Zhao, Y. Qin, Y. Sun, J. Sun and Z. He, Asian J. Pharm. Sci., 2011, 6(3–4), 151–158.

---