Fast route for synthesis of stoichiometric ...

Materials and Design 109 (2016) 547–555

Contents lists available at ScienceDirect

Materials and Design journal homepage: www.elsevier.com/locate/matdes

Fast route for synthesis of stoichiometric hydroxyapatite by employing the Taguchi method Basam A.E. Ben-Arfa, Isabel M. Miranda Salvado, Jorge R. Frade, Robert C. Pullar ⁎ Department of Materials and Ceramic Engineering, CICECO – Aveiro Institute of Materials, University of Aveiro, 3810-193 Aveiro, Portugal

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Taguchi methods used for simultaneous effects of 5 processing parameters on phase purity and properties of hydroxyapatite. • The five different parameters were pH, synthesis temperature, synthesis time, drying temperature and calcination temperature. • Logarithmic scales used for linear multivariate regression based on kinetic models, giving meaningful regression coefficients. • Comprehensive analysis of effects of processing parameters on 4 properties: crystallite size, surface area, Ca/P ratio and mol% HAp. • Calcination temp had greatest impact on morphology, crystallite size & surface area; pH had greatest influence on Ca:P ratio.

a r t i c l e

i n f o

Article history: Received 19 May 2016 Received in revised form 14 July 2016 Accepted 17 July 2016 Available online 19 July 2016 Keywords: Hydroxyapatite Nanoparticles Nanosynthesis Biomaterials Biocompatibility Taguchi method

⁎ Corresponding author. E-mail address: [email protected] (R.C. Pullar).

http://dx.doi.org/10.1016/j.matdes.2016.07.083 0264-1275/© 2016 Published by Elsevier Ltd.

a b s t r a c t The Taguchi experimental design method is an elegant and efficient way of deriving optimum conditions for processes from the minimum number of experiments. We correlated various relevant synthesis parameters in the precipitation synthesis of single-phase pure hydroxyapatite (Ca10(PO4)6(OH)2, HAp) nanoparticles, via a rapid wet precipitation method, without any aging time. Taguchi planning was used for a systematic study of the combined effects of five different parameters: pH, synthesis temperature, synthesis time, drying temperature and calcination temperature. Using T\aguchi methods, we were able to evaluate the effects of four variations (levels) in each of these five parameters, with just 16 experiments (an L16 (1024) orthogonal array). We assessed the impact of these parameters on four distinct properties, namely crystallite size, surface area, Ca/P atomic ratio and mol% of HAp. Calcination temperature exerted the greatest impact on hydroxyapatite morphology, corresponding to crystallite size and specific surface area, for which the role of other processing parameters was not significant. On the contrary, the Ca:P ratio was affected mainly by pH. These findings were confirmed by microstructural, structural and spectroscopic characterisation. FTIR spectra, revealing the conditions to retain a pure or prevailing hydroxyapatite phase, and also to indicate favourable conditions for A-type substitutions of carbonate for hydroxide groups, or B-type substitution for phosphate groups. © 2016 Published by Elsevier Ltd.

548

B.A.E. Ben-Arfa et al. / Materials and Design 109 (2016) 547–555

1. Introduction Hydroxyapatite (Ca10(PO4)6(OH)2, HAP) is a major component of bone, and has been used as a synthetic biomaterial for implants and a wide range of biomedical applications. It can also be used as a filter or membrane, creating unique chromatography tools [1,2]. HAp is classified as a bioactive, biocompatible and osteoconductive material, which aids local osteosynthesis when used in implants, owing to its high compatibility with the composition of natural bone mineral [3]. It is an important substitute in the treatment of bone disorders, and is also an optimum restoration material for dental decay. It can also be used as a surface coating to improve the biocompatibility of other implants [4], and is used as a biomaterial with functionalised nanostructure as a carrier for therapeutics [5]. Among several methods for HAp synthesis [2–4,6–8], precipitation from solution is the most widely used, this being a convenient low cost method for obtaining HAp powder [9]. The bioactivity of synthetic HAp is governed by several factors, such as pH and temperature of synthesis, and thermal treatment including drying and calcination [4]. Starting materials also influence the morphology and crystallinity of the product, and the mode of addition of the reactants effects the subsequent growth of HAp when prepared by precipitation [10]. It was reported previously that changing the stirring time and synthesis temperature leads to highly crystalline HAp [11], while changing pH affects the morphology and the particle size of the synthesised HAp [12]. Almost all previous studies have focused on the effect of only one to three parameters at a time, applying various methods [13–16]. This investigation will uniquely focus on the simultaneous influence of five different parameters on the synthesis of HAp, namely: i) pH; ii) synthesis temperature; iii) time of synthesis reaction; and iv) drying temperature; v) calcination temperature. This is achieved through the application of a simple, fast, low cost and highly efficient analysis process, using Taguchi methods [17]. Taguchi methods are statistical methods that were originally developed after the Second World War to improve the quality of manufactured goods, and are finding increasing application in the engineering and biotechnology sectors to improve procedures [18]. Taguchi methods have been used to investigate and optimise various inorganic syntheses and ceramic processes, such as the hydrothermal synthesis of titania nanoparticles [8], the formation of SiAlON and mullite oxide/non-oxide ceramic composites [19], the synthesis of superhydrophobic polyvinylidene fluoride (PVDF)/zinc oxide composites [20], and for a wide variety of different applications such as photocatalytic zinc oxide nanoparticles [21], ceria nanoparticles for catalyst supports [22], sol-gel applied anti-corrosion titania nanocoatings [21], biomaterials [23,24], etc. The Taguchi method is an experimental design method, used to find solutions for complex problems with a large number of variables and levels using relatively few actual experiments [25]. In the Taguchi method, orthogonal arrays are used to efficiently determine the effect of variables and levels to achieve a robust design, greatly lowering the number of trials and the subsequent cost in time, manpower and resources. Based on Taguchi experimental design, the optimised experimental conditions can be determined by comparing a mean of the means [8,18]. In the case of our study, we studied the five parameters listed above, with four levels of variation for each one. Applying fractional factorial design and the Taguchi statistical method, the 1024 (=45) separate experiments required to reproduce every possible combination can be reduced to only 16 experiments (an L16 (1024) orthogonal array), as shown in Table 1. Each of the experiments gives a unique combination of factors, which can be analysed to give the optimum synthesis conditions for HAp. In addition, Taguchi planning allows one to measure the impact of those processing parameters on four distinct properties, namely crystallite size (nm), specific surface area (m2/g), Ca/P atomic ratio and mol% of HAp phase present, and analyse these results to determine the optimum conditions.

Table 1 L16 orthogonal array of experiments, accompanied by the results of crystallite size (D), specific surface area (A), Ca:P ratio and % Hap. The five parameters studied were: pH; synthesis temperature Ts/K; stirring time ts/min; drying temperature Td/K and calcination temperature Tc/K.

E1 E2 E3 E4 E5 E6 E7 E8 E9 E10 E11 E12 E13 E14 E15 E16

pH

Ts (K)

ts (min)

Td (K)

Tc (K)

D (nm) (nm)

A (m2/g)

Ca:P

Hap (%)

7.5 7.5 7.5 7.5 8.5 8.5 8.5 8.5 9.5 9.5 9.5 9.5 5.5 5.5 5.5 5.5

318 333 348 363 318 333 348 363 318 333 348 363 318 333 348 363

30 60 90 120 60 30 120 90 90 120 30 60 120 90 60 30

333 343 353 363 353 363 333 343 363 353 343 333 343 333 363 353

673 973 1273 1473 1473 1273 973 673 973 673 1473 1273 1273 1473 673 973

0 19 9 1 103 54 27 9 57 1 42 62 81 73 1 21

105 38 19 4 4 13 18 83 19 92 4 8 4 4 69 19

1.43 1.76 1.72 1.73 1.55 1.72 1.72 1.57 1.88 1.69 1.69 1.67 1.87 1.74 1.97 1.87

100 100 100 77 21 84 75 100 51 100 47 66 86 52 100 48

2. Experimental 2.1. Experimental design Taguchi planning was selected because its performance has been widely demonstrated in many technical and scientific fields. Five variable parameters (factors) with 4 levels were chosen for the synthesis of hydroxyapatite, as shown in Table 1. To reduce the total number of experiments, fractional factorial design was employed. An L16 orthogonal array of experiments was chosen to evaluate the effect of pH, synthesis temperature, stirring time, drying temperature and calcination temperature on the crystallite size, specific surface area, Ca/P molar ratio, and the percentage of HAp phase of the nano particles (NPs). This method is an efficient way to assess simultaneous effects of several different parameters, also at different levels, without undue experimental effort, cost and time. The conventional approach to vary every single parameter independently would imply enormous efforts, as demonstrated on comparing the actual number of experiments for 5 independent variables at 4 different levels (16) to the corresponding number of experiments for all the combinations of variables and their levels (45 = 1024). This enormous economy in cost and time allows simultaneous inspection of several relevant properties, as demonstrated in the actual study of effects of synthesis conditions (pH, temperature and time), and processing steps (drying and calcination temperatures) on crystallite size, specific surface area, Ca:P ratio and hydroxyapatite content of the resulting materials. This planning also allows further insight on mechanisms involved by complementary structural, microstructural and spectroscopic characterisation, even if this information is only qualitative. 2.2. Synthesis procedure Nanostructured HAp may be synthesised via precipitation in aqueous media, such as in reference [26]. However, in this work a new, easy and very fast route was implemented to synthesis HAp. Analytical grade calcium acetate monohydrate (Ca(CH3CO2)2·H2O) and orthophosphoric acid (H3PO4) were used as starting precursors for calcium and phosphorous respectively, maintaining the stoichiometric ratio of reactants at a Ca:P molar ratio of 1.67. Ammonia (NH3) was used as a buffer to keep the pH near constant at a designated value (between 5.5 and 9.5). All reagents were purchased from Sigma Aldrich, Germany, and were ACS grade. The reactants were added to distilled water at a fixed stirring rate for all experiments. After the necessary synthesis time, the product was dried using a rotary evaporator for 1 h. Then all samples were subjected to heat treatment at predefined temperature


549

ranges. The same procedure was repeated with different experimental conditions. Details of the experiments are presented in Table 1.

2.3. Characterisation of hydroxyapatite For all characterisation techniques, samples were used as a powder, sieved by a 45 μm mesh. Microstructure of the samples was observed by scanning electron microscope (SEM, Hitachi S-4100), on samples coated with carbon. Energy-dispersive X-ray spectroscopy (EDS, Rontec) was used to check for the presence of impurities, as well as perform qualitative elemental composition analysis. Infrared transmittance spectra of HAp powders were obtained using a Fourier Transform Infrared Spectrometer (FTIR, Brucker Tensor 27) in the range of 350–4000 cm−1, with 128 scans and a resolution of 4 cm−1. Pellets for FTIR were prepared by mixing a 1:150 (by weight) proportion of the sample with KBr and pressing this to obtain discs. Crystalline phase content of the samples was determined by X-ray diffraction (XRD, PANalytical XPERT-PRO Diffractometer system), using Cu Kα radiation (Kα = 1.54059) with 2θ varying from 10 to 80° in steps of 0.026 s−1. The diffraction patterns were compared with JCPDS standards. The crystallite size for all HAp end products was calculated from the corrected broadening of the XRD peaks of the whole pattern, using the Williamson-Hall method [27], according to the following formula:

βhkl cosθ ¼

Kλ þ 4ε sinθ D

ð1Þ

where D, θ, λ, ε, and β are the crystallite size, diffraction angle, wavelength (1.540598 Ǻ), lattice strain, and the peak's full width at half maximum (FWHM), respectively. Thus, for the main peaks of each pattern, the crystallite size was obtained from the intersection of the slope with the vertical (y) axis, after plotting (βhkl cosθ) of the above equation on the y-axis and (4εsinθ) on the x-axis, as shown in Fig. 1. The relative percentage of the HAp phase in wt% for all samples was calculated using PANalytical X-pert pro software. The specific surface area values were obtained from the adsorption isotherms using a 5-point Brunauer-Emmett-Teller (BET) method on a Micromeritics Gemini 2380 surface area analyser, with N2 used as the adsorbate. Standard pre-treatment outgassing conditions were 105 °C under vacuum for 12 h. The phosphorous content in the samples was determined by a Shimadzu UV-3100 spectrophotometer, using a phosphorus detection reagent, while calcium content was determined by atomic absorption spectroscopy (GBC Avanta PM).

Fig. 2. XRD difractograms of as-synthesised HAp.

3. Experimental results 3.1. XRD measurements X-ray diffractograms (Fig. 2) of the as-dried powders (before calcination) for all 16 experiments showed the precipitation of HAp as the sole crystalline phase. Subsequent calcination of the samples led to well defined peaks, with higher intensities of single phase HAp at the lower calcination temperature (400 °C), while the appearance of tri-calcium phosphate (TCP) as a second crystalline phase together with HAp was often observed at higher calcination temperatures, as presented in Fig. 3. The fraction of tri-calcium phosphate formed at the highest calcination temperatures (1000 and 1200 °C) also tends to increase with decreasing Ca:P ratio (Table 1), in close agreement with differences between Ca:P ratio in the hydroxyapatite and tri-calcium phosphate phases. However, XRD failed to show evidence of any other Ca-rich crystalline phases, to account for the co-existence of hydroxyapatite and tri-calcium phosphate with differences in the Ca:P ratio. Ca-rich segregation was expected mainly when the overall Ca:P ratio exceeded the ideal stoichiometry (1.667) of hydroxyapatite. Indirect evidence of Ca-based segregation may be inferred from FTIR bands at about 881 cm−1 and 1385 cm−1, which could be ascribed to minor traces of calcium carbonate, as discussed below. However, if it is present, it is below levels detectable by XRD. XRD was also the basis for quantification of the HAp content, by the Williamson-Hall method (Eq. (1)), as shown in Table 1, and this was taken as one of the main guidelines for the optimisation of processing conditions, as discussed below. 3.2. FTIR spectroscopy

Fig. 1. Average crystallite size estimation using the Williamson-Hall method.

FTIR transmittance spectra is often used to identify the presence of characteristic groups in HAp, with emphasis on phosphate, hydroxyl and carbonate groups, and to assess the impact of processing conditions on these groups (Fig. 4). Bands at about 571 and 604 cm−1 can be ascribed to a triply-degenerated bending mode of phosphate groups (ν4), and are poorly suited to reveal the most relevant structural effects. Thus, it is better to emphasise more sensitive bands which may be more suitable signatures of relevant structural changes, such as the bands ascribed to CO2– 3 groups acting as substitutes for hydroxyl groups (1546 cm−1) and phosphate groups (1460 cm−1), in close agreement with the relevant literature [28]. Note that both of these bands tend to disappear after calcination at 1000–1200 °C, in agreement with literature findings [29]. In addition, FTIR spectra also show the extinction of

550


Fig. 3. XRD difractograms for synthesised HAp at different calcination temperatures: a) 400 °C, b) 700 °C, c) 1000 °C and d) 1200 °C).

an additional band at about 881 cm−1, ascribed to CO2– 3 adsorbed on the surface or to traces of CaCO3, which is expected to decompose at temperatures in the order of 1000 °C or higher. On comparing the FTIR spectra of samples calcined at different temperatures (Fig. 4), one clearly sees that intermediate calcination temperatures are best suited to retain Type-B substitutions in the hydroxyapatite phase. This is especially true for samples calcined at

700 °C (Fig. 4b), which show an enhanced substitution of CO2– 3 for in B-type HAp, based on the band at 1460 cm−1, agreeing with PO3− 4 other reports in the literature [3,7,10,30]. Though these spectra also – −1 ), this band show type-A substitution of CO2– 3 for OH (1546 cm shows poorer resolution in most samples calcined at 700 °C, which is consistent with expected stability differences between type-A and type-B carbonated hydroxyapatites [29].

Fig. 4. FTIR spectra of calcined samples showing the impact of heat treatment at different temperatures: a) 400 °C, b) 700 °C, c) 1000 °C and d) 1200 °C.

B.A.E. Ben-Arfa et al. / Materials and Design 109 (2016) 547–555 Table 2 Correlation matrix for the effects of preparation parameters on crystallite size (D), specific surface area (A), Ca:P ratio (Ca), and hydroxyapatite % (Ha).

pH Ts ts Td Tc pH ∗ Ts pH ∗ ts pH ∗ Td pH ∗ Tc Ts ∗ ts Ts ∗ Td Ts ∗ Tc ts ∗ Td ts ∗ Tc Td ∗ Tc

D (nm)

A (m2/g)

Ca:P

Hap %

0.098 −0.171 0.066 0.049 0.931 −0.037 −0.021 0.141 0.857 −0.052 −0.145 0.641 0.075 0.592 0.867

0.057 −0.075 −0.052 −0.070 −0.888 −0.040 0.060 −0.014 −0.761 −0.022 −0.070 −0.738 −0.032 −0.573 −0.823

−0.443 0.111 0.183 0.461 0.006 −0.219 0.012 −0.077 −0.174 0.077 0.321 0.036 0.267 0.156 0.092

−0.114 0.097 0.221 −0.008 −0.604 0.045 0.177 −0.116 −0.602 0.287 0.118 −0.393 0.217 −0.155 −0.533

The FTIR spectra also show a band at 632 cm−1 ascribed to hydroxyl OH groups. Interestingly, the intensity of this absorption band is the highest at 1000 °C (Fig. 4c), i.e., when A-type substitution of CO2– 3 for OH−has been extinguished. The 632 cm−1 band is also a marker of the ability to retain the HAp phase, against its decomposition to tricalcium phosphate, at 1000 °C. Note the correlation between the intensity of this band and the actual HAp content in E3, E6, E12 and E13 (Table 2). Thus, the presence of OH– groups (band 632 cm−1) and simultaneous presence of the main bands ascribed to phosphate groups are consistent with the formation of HAp, as known from relevant literature [7,10,31]. The 632 cm−1 band is also indicative of the ability to retain the prevailing HAp at even higher temperatures, as observed mainly for E4. phosphate groups are the doubly Relevant bands ascribed to PO3− 4 degenerated bending mode ν2 (473 cm−1), symmetric stretching mode ν1 (960 cm− 1), and triply degenerated asymmetric stretching mode v3 (1092 cm− 1). Again, one observes sharper peaks for the ν1 stretching band (960 cm−1) for samples calcined at 1000 °C, i.e. after 3− suppression of B-type substitution of CO2– 3 for PO4 , while retaining −1 HAp as the prevailing phase. The 960 cm band is also the highest in E3. A similar conclusion can be drawn for the bending mode ν2 (473 cm−1) in samples calcined at 1000 °C. A small absorption peak is also seen at about 1385 cm−1. Though this is not clearly identified, it is often most visible in samples without phase purity, i.e., when partial decomposition of the HAp phase is more likely to yield a CaO-rich phase, to account for differences in Ca:P between TCP and HAp. Thus, this may be a spurious peak caused by interactions of traces of a CaO-rich phase with atmospheric gases or other species resulting from decomposition of reactants (ammonia, acetates, etc.) –

3.3. SEM images and EDS analysis Processing conditions are also known to exert an important effect on morphology of synthesised HAp particles [32], and their subsequent changes with heat treatments. Thus, the morphology of the powders were also investigated (Fig. 5). This clearly shows agglomeration in the as-synthesised powders, with typical sizes N 1 μm, increasing with synthesis temperature and/or time. SEM also suggests that subsequent calcination contributes to agglomeration (see E1, E2 and E3 in Fig. 5af); this is also confirmed by a decrease in specific surface area (Table 1). In addition, electron micrographs confirm that crystallite sizes are in the nanoscale range, as also determined from XRD (Table 1). It can also be observed that E4 and E5 heated at 1200 °C (Fig. 5g-h) have different microstructures, as clusters with larger sizes can be seen in E5 (Fig. 5h), apparently without significant differences in specific surface area (Table 1). This may be correlated to a higher degree of

551

transformation of hydroxyapatite to tri-calcium phosphate in sample E5, probably promoted by its relatively low Ca:P ratio, induced by a higher pH. Processing conditions may also determine the effective solubility of precursors, the composition of resulting synthesised powders, and their homogeneity. However, very sharply defined peaks for Ca and P were shown by EDS (Fig. 5i), confirming that these are the only detectable elements in all the samples, and without evidence of heterogeneities of Ca:P ratio or traces of contamination. 4. Statistical analysis of results 4.1. Correlation guidelines Table 1 shows the L16 orthogonal array of the five factors (pH, Ts, ts, Td and Tc), and their combined impacts on crystallite size, specific surface area, Ca/P atomic ratio and the HAp phase percent in each sample. The corresponding correlation matrix (Table 2) shows the strongest statistical relevance is for the effects of calcination temperature, except for its negligible effect on Ca:P ratio. Indeed, one would not expect significant effects of calcination temperature on Ca:P, except possibly for slight losses of phosphorus at the highest calcination temperature. The other processing parameters only show relatively weak statistical relevance for one of the measured properties, namely effects of pH and drying temperature on Ca:P ratio. The role of pH is indeed expected if one considers that pH determines equilibrium conditions for metal oxides or oxyhydroxides, and concentrations in the residual liquid phases. However, this cannot explain the apparent effects of drying temperature on Ca:P. The correlation matrix was also extended for different combinations of pairs of independent variables. However, this reflects mainly the role of the independent variable with highest correlation, as found for correlations between crystallite size and different combinations of calcination temperature with other variables. The signal to noise ratio (Fig. 6A through 6D) is consistent with the prevailing effects identified in the correlation matrix (Table 2); this is also emphasised by the differences between the minimum and maximum values of signal to noise ratio (Table 3), as well as the sum of squares and corresponding Pareto distributions. However, this is still insufficient to provide meaningful physical interpretation for the effects of processing parameters. Thus, we re-examined the overall dependence in detail using a suitable regression analysis. 4.2. Multivariate regression We re-examined the effects of processing parameters by applying a multivariate formula to describe the dependence of a generic property Y as a function of pH, synthesis temperature, synthesis time, drying temperature and calcination temperature. However, the regression model was based on exponential or power law dependences (Eq. (2)), rather than a simple linear regression: γ

Y∝ expð−αpHÞ expð−β=T s Þt s expð−δ=T d Þ expð−ε=T c Þ

ð2Þ

This is consistent with generic trends expected for kinetics of chemical reactions or transformations, including dependences on time, processing temperatures, and/or pH dependent activities of relevant reactants, as found for [H+] = exp.(− 2.30 pH). The model behaviour described by Eq. (2) is also suitable for ready linearisation, as follows: ln ðY Þ ¼ −αpH−β=T s þ γ ln ðt s Þ−δ=T d −ε=T c þ θ

ð3Þ

This transformation of independent variables provides unique conditions to retain the simplicity of linear regression, while providing close relations between regression coefficients and meaningful physical parameters. For example, the pH coefficient α may be indicative of the

552


Fig. 5. Selective SEM images for some experiments showing the microstructure differences between as-synthesised and calcined samples for E1–E4. Any squares on the SEM images represent areas used for EDS measurements. The EDS spectra of image a) are shown in i) as an example.

B.A.E. Ben-Arfa et al. / Materials and Design 109 (2016) 547–555 -10

40

pH

Tc

Ts

Td

pH

Tc

Ts

Td

ts

ts 30

(S/N)A

(S/N)D

-20

553

-30

20

-40

10

1

2

3

4

1

2

3

level 45

4

level

pH

Tc

Ts

Td

6

ts

(S/N)Hap

(S/N)Ca:P

40

pH

Tc

Ts

Td

ts

5

35

4

30 1

2

3

4

1

2

level

3

4

level

Fig. 6. Signal to noise ratio of the dependence of crystallite size, specific surface area, % of hydroxyapatite, and Ca:P ratio on pH, synthesis temperature and time, drying temperature and calcination temperature.

dependence on soluble species in aqueous media (e.g. ln(aCa2+) ∝ pH or 2− dependence of H2PO− 4 : HPO4 on pH), the time parameter γ may be indicative of the controlling reaction mechanism, and reciprocal temperature parameters provide information on activation energies (e.g. ε = Ea/ R) or enthalpy changes. On the contrary, optimised computer codes such as Latin hypercube design [33] often fail to find applicability in experimental planning, except possibly for rare cases of multi-parametric characterisation [34]. Eq. (3) was thus the basis for the regression analysis of the dependence of crystallite size, specific surface area and % Hap on processing parameters, by least square fitting, yielding the following regression coefficients: ln ðD=nmÞ ¼ 0:073pH−1197=T s þ 0:712 ln ðt s Þ−1594=T d −5254=T c þ 12:75

ð4Þ

Table 3 Sum of squares, corresponding Pareto distributions and difference between maximum and minimum results of signal to noise ratio for the effects of pH, synthesis temperature, synthesis time, drying temperature and calcination temperature on average crystallite size, specific surface area, Ca:P ratio and hydroxyapatite %.

D

A

Ca:P

HAp

Sum of squares Contribution (%) Δ(S/N) Sum of squares contribution (%) Δ(S/N) Sum of squares Contribution (%) Δ(S/N) sum of squares Contribution (%) Δ(S/N)

pH

Ts

ts

Td

Tc

4.9 0.3% 1.39 19.7 1.1% 2.87 3.2 39.3% 1.11 115.6 19.9% 7.58

38.9 2.5% 2.23 19.3 1.1% 0.73 0.8 9.4% 0.33 92.5 16.0% 1.17

4.9 0.3% 0.99 19.1 1.1% 2.74 0.5 6.2% 0.16 92.5 16.0% 6.66

15.1 1.0% 2.03 17.7 1.0% 2.66 1.9 23.4% 0.59 91.1 15.7% 5.66

1513.8 96.0% 7.70 1646.4 95.6% 14.29 1.8 21.7% 0.66 187.7 32.4% 7.30

ln A= m2 g−1 ¼ 0:058pH−179=T s −0:171 ln ðt s Þ−409=T d þ 3594=T c þ 1:194

ð5Þ

ln ðHap=%Þ ¼ −0:035pH−553=T s þ 0:162 ln ðt s Þ þ 207=T d þ 725=T c þ 4:15

ð6Þ

These fitting parameters were used to compute the expected contribution when a given independent variable varies from its lowest to highest level, while retaining the remaining variables at the lowest level (Table 4). Partial linear regression was used for the effects of pH, and synthesis temperature and time, on the Ca:P ratio; this was based on considering the independence of the Ca:P ratio from calcination temperature, and also on the drying temperature, as discussed above. Thus, the simplified model for the dependence of the Ca:P ratio on processing parameters, reduced to dependence on pH, synthesis temperature and synthesis time, was as follows: ln ðCa : P Þ ¼ −0:022pH−73=T s þ 0:032 ln ðt s Þ þ 0:795

ð7Þ

Table 4 contributions of independent variables to changes in crystallite size (D), specific surface area (A), Ca:P ratio (Ca), and hydroxyapatite % (HAp), on varying a specific parameter from its minimum to maximum level while keeping the remaining parameters at their average values (e.g. ΔD = 102 nm on increasing calcination temperature from 400 °C to 1200 °C).

ΔD (nm) ΔA (m2/g) Δ(Ca:P) ΔHAp (%)

pH

Ts

ts

Td

Tc

6 4 −0.153 −10

9 1 0.049 15

19 −4 0.075 16

8 2 – −4

102 −91 – −44

554


The calcination temperature (Tc) has the greatest effect on crystallite size, with an activation energy in the order of 44 kJ/mol (Eq. (4)), and without significant effects from other parameters; this is consistent with the relevant literature [13], and indicates that changes occurring during calcination mask the original size and morphology of the as-synthesised powders. In fact, synthesis temperature and other conditions of synthesis may determine the morphology of hydroxyapatite crystals, particularly using solid precursors of Ca (e.g. [35]). However, the actual crystallite size ranges were evaluated by structural characterisation of the HAp phase, which implies uncertainties, mainly because the highest crystallite sizes were in the order of 100 nm, at the limit of accuracy of the Williamson-Hall method used. The prevailing effect of calcination temperature was also confirmed for changes in specific surface area, with activation energies in the order of 30 kJ/mol. The effect of synthesis time on particle size also explains the negative impact on specific surface area. On the contrary, increasing pH exerts a positive effect on specific surface area, possibly by promoting formation of needle-like HAp particles with high aspect ratio [36], which is likely to hinder subsequent morphological changes on calcining. The Ca:P ratio is slightly higher than the ideal stoichiometry (Ca:P = 10:6) in several samples (Table 1), and multivariate fitting suggests that pH is most effective in controlling this excess of Ca (Table 4). The effect of pH can be understood by taking into account that the Ca:P ratio of soluble species in aqueous media increases with pH [37], implying the opposite trend for the Ca:P ratio in precipitated phases. In addition, the impact of synthesis time indicates complex dynamic changes, as reported for the CaO/P2O5/H2O-based system in classical literature [38]; this includes time dependent changes in pH. Multivariate fitting also confirms the prevailing negative effect of high calcination temperatures on HAp, as widely recognised in the literature. These results are in a good agreement with results obtained by Chafik et al. [39], Costescu et al. [16] and Irma Bogdanoviciene et al. [40], based on synthesis of HAp by wet precipitation or sol-gel, and reporting that calcination temperatures must be lower than 1000 °C to obtain single-phase pure HAp. Also, these authors reported the presence of the CaO phase as a result of HAp decomposition at temperatures higher than 1000 °C. These multivariate results are summarised in Fig. 7. However, somewhat different results were reported by others [41, 42], indicating that single phase HAp nanoparticles could be obtained only at temperatures up to about 700 °C, with onset of TCP and calcium oxide (CaO) at temperatures above 800 °C. This may provide clues to understand the effects of other processing parameters (i.e. pH and synthesis time) on the ability to retain the prevailing HAp phase in calcined samples. In fact, higher reactivity is expected for samples with high specific surface area, implying also facile high temperature decomposition

Fig. 7. comparison between predictions obtained by multivariate regression and actual experimental results.

of HAp. Thus, the contribution of synthesis time may be related to its negative impact on specific surface area. An inverse correlation between specific surface area and HAp content also explains the effects of pH, with a positive contribution to specific surface area, and negative impact on HAp content. However, this may still be debatable. For example, Kweh et al. suggested that pH ~9 provides an appropriate environment for the formation of stoichiometric pure HAp powders [43]. 5. Conclusions It can be concluded from these experiments that the Taguchi method is a powerful means to minimise the number of experiments required to achieve a fuller understanding of the effects of processing factors on HAp powders synthesised in aqueous media. Suitable models were found to describe the simultaneous impacts of pH, synthesis temperature and time, drying temperature and subsequent calcination temperature on: i) crystallite size, ii) specific surface area, iii) Ca:P ratio and iv) % HAp content. This provides a clearer understanding of the prevailing factors, additional relevant contributions, and also of the indirect effects of synthesis parameters, which are related to correlations between different characteristics of the resulting powders. This includes reciprocal correlations between the ability to retain HAp and specific surface area or Ca:P ratio. This understanding of relevant effects and their approximate quantification allows for the prospect of optimisation of these biomaterials by adjusting the actual % HAp content independently of the calcium-tophosphorus (Ca:P) ratio, and probably also independently of crystallite size and specific surface area. This also enables a flexibility in designing multiphase biomaterials with adjusted distributions of HAp and other phases (e.g. tricalcium or tetracalcium phosphates [44]), while retaining suitable morphology, or seeking functionalisation with preferential groups. Acknowledgements R.C. Pullar was supported by FCT Grant SFRH/BPD/97115/2013 for this work. This work was developed in the scope of the project CICECO − Aveiro Institute of Materials (Ref. FCT UID/CTM/50011/ 2013), financed by national funds through the FCT/MEC and when applicable co-financed by FEDER under the PT2020 Partnership Agreement. References [1] P. Gagnon, R. Frost, T. Ogawa, CHT™ Ceramic Hydroxyapatite — A New Dimension in Chromatography of Biological Molecules, 2009. [2] O. Takagi, N. Kuramoto, M. Ozawa, S. Suzuki, Adsorption/desorption of acidic and basic proteins on needle-like hydroxyapatite filter prepared by slip casting, Ceram. Int. 30 (2004) 139–143, http://dx.doi.org/10.1016/S0272–8842(03)00061–0. [3] U. Anjaneyulu, D. Pattanayak, U. Vijayalakshmi, The facile and phase pure evaluations of nano hydroxyapatite powder by sol-gel method, Int. J. ChemTech Res. 7 (2014) 1516–1520 (http://www.embase.com/search/results?subaction= viewrecord&from=export&id=L602149120). [4] M.H. Santos, M. De Oliveira, L.P.D.F. Souza, H.S. Mansur, W.L. Vasconcelos, Synthesis control and characterization of hydroxyapatite prepared by wet precipitation process, Mater. Res. 7 (2004) 625–630, http://dx.doi.org/10.1590/S151614392004000400017. [5] R. Subbiah, M. Veerapandian, K.S. Yun, Nanoparticles: functionalization and multifunctional applications in biomedical sciences, Curr. Med. Chem. 17 (2010) 4559–4577, http://dx.doi.org/10.2174/092986710794183024. [6] A. Yasukawa, T. Kunimoto, K. Kamiuchi, K. Kandori, T. Ishikawa, Preparation of lead hydroxyapatite particles using acetamide, J. Mater. Chem. 9 (1999) 1825–1830, http://dx.doi.org/10.1039/a900758j. [7] H. Zhang, Y. Yan, Y. Wang, S. Li, Morphology and formation mechanism of hydroxyapatite whiskers from moderately acid solution, Mater. Res. 6 (2002) 111–115. [8] S. Naghibi, M.A. Faghihi Sani, H.R. Madaah Hosseini, Application of the statistical Taguchi method to optimize TiO2 nanoparticles synthesis by the hydrothermal assisted sol-gel technique, Ceram. Int. 40 (2014) 4193–4201, http://dx.doi.org/10. 1016/j.ceramint.2013.08.077. [9] T.V. Safronova, S.A. Korneichuk, V.I. Putlyaev, V.K. Krut’ko, Ceramics based on calcium hydroxyapatite synthesized from calcium acetate, calcium hydroxide, and potassium hydrophosphate, Glass. Ceram. 69 (2012) 1–7, http://dx.doi.org/10.1007/ s10717–012–9409-1.

B.A.E. Ben-Arfa et al. / Materials and Design 109 (2016) 547–555 [10] M. Manoj, R. Subbiah, D. Mangalaraj, N. Ponpandian, C. Viswanathan, K. Park, Influence of growth parameters on the formation of hydroxyapatite (HAp) nanostructures and their cell viability studies, Nanobiomedicine 2 (2015) 1–11, http://dx. doi.org/10.5772/60116. [11] M. Pretto, A.L. Costa, E. Landi, A. Tampieri, C. Galassi, Dispersing behavior of hydroxyapatite powders produced by wet-chemical synthesis, J. Am. Ceram. Soc. 86 (2003) 1534–1539. [12] A. Sáenz, M.L. Montero, G. Mondragón, V. Rodríguez-Lugo, V.M. Castaño, Effect of pH on the precipitation of hydroxyapatite on silica gels, Mater. Res. Innov. 7 (2003) 68–73, http://dx.doi.org/10.1007/s10019-003-0230-x. [13] S.S.A. Abidi, Q. Murtaza, Synthesis and characterization of nano-hydroxyapatite powder using wet chemical precipitation reaction, J. Mater. Sci. Technol. 30 (2014) 307–310, http://dx.doi.org/10.1016/j.jmst.2013.10.011. [14] K. Agrawal, G. Singh, D. Puri, S. Prakash, Synthesis and characterization of hydroxyapatite powder by sol-gel method for biomedical application, J. Miner. Mater. Charact. Eng. 10 (2011) 727–734. [15] M.H. Mahmud, K.A. Salam, M.A. Gafur, A.A. Rana, M.R. Qadir, S.M. Masum, et al., Chemical characteristics of hydroxyapatite from oyster shell by thermo-chemical process, Int. J. Innov. Res. Sci. Eng. Technol. 04 (2015) 5039–5047, http://dx.doi. org/10.15680/IJIRSET.2015.0407002. [16] A. Costescu, I. Pasuk, F. Ungureanu, A. Dinischiotu, M. Costache, F. Huneau, et al., Physico-chemical properties of Nano-sized hexagonal hydroxyapatite powder synthesized by sol-gel, Dig. J. Nanomater. Bios. 5 (2010) 989–1000. [17] G. Taguchi, S. Chowdhury, Y. Wu, Taguchi's Quality Engineering Handbook, John Wiley, New Jersey, 2005. [18] R. Roy, A Primer on the Taguchi Method, second ed. society of manufacturer engineers/ Van Nostrand Reinhold, Dearborn, Michigan, USA, 2010. [19] A. Jamshidi, A.A. Nourbakhsh, S. Naghibi, K.J.D. Mackenzie, Application of the statistical Taguchi method to optimize X-SiAlON and mullite formation in composite powders prepared by the SRN process, Ceram. Int. 40 (2014) 263–271, http://dx. doi.org/10.1016/j.ceramint.2013.05.133. [20] A.M.A. Mohamed, R. Jafari, M. Farzaneh, An optimization of superhydrophobic polyvinylidene fluoride/zinc oxide materials using Taguchi method, Appl. Surf. Sci. 288 (2014) 229–237, http://dx.doi.org/10.1016/j.apsusc.2013.10.013. [21] Y.T. Chung, M.M. Ba-Abbad, A.W. Mohammad, N.H.H. Hairom, A. Benamor, Synthesis of minimal-size ZnO nanoparticles through sol-gel method: Taguchi design optimisation, Mater. Des. 87 (2015) 780–787, http://dx.doi.org/10.1016/j.matdes.2015.07.040. [22] B. Nematollahi, M. Rezaei, E. Nemati Lay, Synthesis of nanocrystalline CeO2 with high surface area by the Taguchi method and its application in methanation, Chem. Eng. Technol. 38 (2015) 265–273, http://dx.doi.org/10.1002/ceat.201400092. [23] S.-S. Mehdi, M.-T. Khorasani, J. Ahmad, Hydrothermal processing of hydroxyapatite nanoparticles—a Taguchi experimental design approach, J. Cryst. Growth 361 (2012) 73–84. [24] T.T. Ajaal, R.W. Smith, Employing the Taguchi method in optimizing the scaffold production process for artificial bone grafts, J. Mater. Process. Technol. 209 (2009) 1521–1532, http://dx.doi.org/10.1016/j.jmatprotec.2008.04.001. [25] S. Naghibi, A. Jamshidi, O. Torabi, R.E. Kahrizsangi, Application of Taguchi method for characterization of corrosion behavior of TiO2 coating prepared by sol-gel dipping technique, Int. J. Appl. Ceram. Technol. 11 (2014) 901–910, http://dx.doi.org/10. 1111/ijac.12077. [26] R.N. Correia, M.C.F. Magalhaes, P.A.A.P. Marques, A.M.R. Senos, Wet synthesis and characterization of modified hydroxyapatite powders, J. Mater. Sci. Mater. Med. 7 (1996) 501–505, http://dx.doi.org/10.1007/BF00705432. [27] V. Mote, Y. Purushotham, B. Dole, Williamson-Hall analysis in estimation of lattice strain in nanometer-sized ZnO particles, J. Theor. Appl. Phys. 6 (2012) 1–8, http:// dx.doi.org/10.1186/2251-7235-6-6.

555

[28] F. Ren, Y. Ding, Y. Leng, Infrared spectroscopic characterization of carbonated apatite: a combined experimental and computational study, J. Biomed. Mater. Res. A 102A (2014) 496–505. [29] Z.Z. Zyman, D.V. Rokhmistrov, V.I. Glushko, I.G. Ivanov, Thermal impurity reactions and structural changes in slightly carbonated hydroxyapatite, J. Mater. Sci. Mater. Med. 20 (2009) 1389–1399, http://dx.doi.org/10.1007/s10856-009-3706-4. [30] M.M. Figueiredo, J.A.F. Gamelas, A.G. Martins, Characterization of Bone and BoneBased Graft Materials Using FTIR Spectroscopy, in: T. Theophile (Ed.), Infrared Spectrosc. - Life Biomed. Sci. Intech 2012, pp. 315–338 (http://cdn.intechopen. com/pdfs/36055.pdf). [31] L. Berzina-Cimdina, N. Borodajenko, Research of Calcium Phosphates Using Fourier Transform Infrared Spectroscopy, in: T. Theophile (Ed.), Infrared Spectrosc. Mater. Sci. Eng. Technol. Intech 2012, p. 510. [32] R. Kumar, K.H. Prakash, P. Cheang, K.A. Khor, Temperature driven morphological changes of chemically precipitated hydroxyapatite nanoparticles, Langmuir 20 (2004) 5196–5200. [33] P.Z.G. Qian, Nested Latin hypercube designs, Biometrika 96 (2009) 957–970, http:// dx.doi.org/10.1093/biomet/asp045. [34] F. Wang, Y. Hwang, P.Z. Qian, X. Wang, A statistical-guided approach to precise characterization of nanowire morphology, ACS Nano 4 (2010) 855–862. [35] Y. Kim, C. Ohtsuki, Hydroxyapatite formation from calcium carbonate single crystal under hydrothermal condition: effects of processing temperature, Ceram. Int. 42 (2016) 1886–1890. [36] J. Liu, X. Ye, H. Wang, M. Zhu, B. Wang, H. Yan, The influence of pH and temperature on the morphology of hydroxyapatite synthesized by hydrothermal method, Ceram. Int. 29 (2003) 629–633, http://dx.doi.org/10.1016/S0272-8842(02)00210-9. [37] H. McDowell, T.M. Gregory, W.E. Brown, Solubility of Ca5(PO4)3OH in the system Ca(OH)2-H3PO4-H2O at 5, 15, 25, and 37 °C, J. Res. Natl. Bur. Stand. - A. Phys. Chem. 81 (1977) 273–281. [38] K.S. TenHuisen, P.W. Brown, Variations in solution chemistry during the calcium deficient and stoichiometric hydroxyapatite formation from CaHPO4·2H2O and Ca4(PO4)2O, J. Biomed. Mater. Res. 36 (1997) 233–241. [39] B.C. El Idrissi, K. Yamni, A. Yacoubi, A. Massit, A novel method to synthesize nanocrystalline hydroxyapatite: characterization with X-ray diffraction and infrared spectroscopy, IOSR J. Appl. Chem. 7 (2014) 107–112, http://dx.doi.org/10.9790/ 5736-0753107112. [40] I. Bogdanoviciene, A. Beganskiene, K. Tõnsuaadu, J. Glaser, H.J. Meyer, A. Kareiva, Calcium hydroxyapatite, Ca10(PO4)6(OH)2 ceramics prepared by aqueous sol-gel processing, Mater. Res. Bull. 41 (2006) 1754–1762, http://dx.doi.org/10.1016/j. materresbull.2006.02.016. [41] M.H. Fathi, A. Hanifi, Evaluation and characterization of nanostructure hydroxyapatite powder prepared by simple sol-gel method, Mater. Lett. 61 (2007) 3978–3983, http://dx.doi.org/10.1016/j.matlet.2007.01.028. [42] W. Feng, L. Mu-Sen, L. Yu-Peng, Q. Yong-Xin, A simple sol-gel technique for preparing hydroxyapatite nanopowders, Mater. Lett. 59 (2005) 916–919, http://dx.doi.org/ 10.1016/j.matlet.2004.08.041. [43] S.W.K. Kweh, K.A.P. Khor, P. Cheang, The production and characterization of hydroxyapatite (HA) powders, J. Mater. Process. Technol. 89 (1999) 373–377, http:// dx.doi.org/10.1016/S0924-0136(99)00061-8. [44] H. Kim, R.P. Camata, Y.K. Vohra, W.R. Lacefield, Control of phase composition in hydroxyapatite/tetracalcium phosphate biphasic thin nanoparticles for biomedical applications, J. Mater. Sci. Mater. Med. 16 (2005) 961–966.