Sol–gel synthesis of transition-metal ion conjugated alumina-rich mullite nanocomposites with potential mechanical, dielectric and photoluminescence properties

Abstract

Nanocrystalline mullite have been synthesized from non-stoichiometric alkoxide precursors via sol–gel route with Co²⁺, Ni²⁺ and Cu²⁺ as dopant metal ions. Transition-metal aluminate spinel phases, formed from the reaction between dopant metal ions and dissolved alumina species, introduced prominent colors to the composites after sintering. Interesting colors combined with suitable densification lead these composites to have potential use as ceramic pigments. A comparative Vickers and Knoop hardness have been evaluated in terms of dislocation movement along grain boundaries with highest hardness and Young’s modulus values of ∼8.7 GPa and ∼207 GPa for copper and cobalt incorporated mullite, respectively. Greater porosity of pure mullite results in an unconventionally high dielectric constant of ∼91 whereas larger interfacial polarization is responsible for the varying dielectric response of transition-metal incorporated mullite composites. Formation of oxygen like defects in the composites cause prominent PL bands with highest PL intensity for dopant cobalt ions in mullite matrix.

---

1. Introduction

Over last few years, the only stable aluminosilicate phase in high-temperature solid-state reaction between alumina and silica, namely mullite, has been used not only in traditional ceramics but also as a functional ceramic.^1,2 Mullite has become a good choice of ceramic material for electronic, optical and high temperature structural applications.³ Owing to its favorable properties such as high melting point, high temperature strength, excellent chemical and thermal stability, low thermal conductivity, thermal expansion coefficient and dielectric constant, and high transmittance in mid-infrared region, mullite has been applied to emerging engineering applications such as components in reinforced composites, thermal barrier, ceramic capacitors, electronic packaging materials, window material in mid-infrared range, insulating material in spark plugs etc.^4–6

Mullite can be represented by the general formula Al^VI₂[Al^IV_2+2xSi_2−2x]O_10−x⊗_x where IV and VI represent four- and six-fold coordination of aluminium cation, respectively, ⊗ denotes an oxygen vacancy and x indicates the number of missing oxygen atoms per average unit cell with its value ranging between 0 and 1. The end member with x = 0 corresponds to Al₂SiO₅ polymorphs such as sillimanite, andalusite and kyanite (mole ratio of Al/Si = 2/1) which are perfectly crystalline states whereas for x = 1, the other end member leads to a silica-free phase known as iota-alumina (ι-Al₂O₃).⁷ Although any mullite composition between sillimanite and ι-Al₂O₃ is possible theoretically, mullite phases commonly fall into the range 0.18 ≤ x ≤ 0.88. Among them frequently referenced mullite comes with four separate species differing in mole ratio between alumina and silica, namely 3 : 2 mullite (x = 0.25), 2 : 1 mullite (x = 0.4), 4 : 1 mullite (x = 0.67) and 9 : 1 mullite (x = 0.842).^7–9

The properties and microstructures of mullite are closely related to the synthesis routes employed and the type of precursor phases.¹⁰ A large number of synthesis procedures of mullite have been reported so far among which thermal decomposition of natural aluminosilicates, sol–gel synthesis, hydrothermal procedures, molten salt synthesis, sintering of Al₂O₃ and SiO₂ powders, chemical vapor deposition and spray pyrolysis are most frequently undertaken.¹¹ Sol–gel synthesis is a well established promising technique to fabricate high purity mullite composites resulting from the precursor, with high degree of homogeneity and homogeneous distribution of components due to nanoscale mixing.¹² Different silicon and aluminium sources have been used for mullite precursor synthesis. The most common precursor is a metal inorganic salt (acetate, chloride, nitrate, sulfate etc.) or an organometallic species such as metal alkoxide. The metallic alkoxides are a very favourable class of precursors used to obtain high performance mullite with desirable physico-chemical properties. During the sol–gel synthesis of mullite precursor from metal alkoxides, the alkoxides are hydrolysed first followed by a polymerization process and finally the precursor turns into well crystalline mullite upon high-temperature sintering via nucleation and growth mechanism.

Different transition metals have strong mineralizing effect on the mullitization reaction with an decrease of the surface area and the pore volume.^13,14 These transition-metal ions distort the local ligand symmetry (Jahn–Teller distortion) and they interact with the silica layer and destabilize the aluminosilicate matrix which results in accelerated phase transformation. Moreover, the transition-metal ions in the form of oxides provide a liquid phase at higher temperature which in turn increase the rate of dissolution of alumina into silica flux thus favouring the mullitization reaction sequences.^15,16

Although the effect of transition-metal ions on the formation and microstructure of mullite has been very well established,^10,17–19 their effect on the mechanical and photoluminescence properties of mullite are rarely studied. Elastic modulus and hardness are two essential parameters for structural materials. It is evident that formation of dislocation networks within matrix grains and strengthened grain boundaries contribute to the mechanical strength of a composite.²⁰ Appropriate doping and formation of other crystalline phases within the matrix may also contribute to the higher microhardness and elastic modulus of mullite which can be applied for structural ceramics. PL emission bands of alumina are reported to be closely related to oxygen related defects or F⁺ centres (oxygen vacancies with one-electron centers) and hence Al-rich or silicon-deficient nanocrystalline mullite may also exhibit photoluminescence bands due to formation of oxygen vacancies.^2,21–23

Mullite without any mineralizer exhibits a low dielectric constant (∼7) but its composites with alumina, cordierite or transition metals show promising dielectric characteristics.^2,24,25 The enhancement of dielectric constant for these types of composites can be attributed to the increased polarization due to the presence of other crystalline phases and mobile charge carriers.

The degenerate d-orbitals of transition-metal ions interact with the electron cloud of ligands to create a non-degeneracy of the d-orbitals. The visible light absorption band of any transition-metal ion corresponds to the energy required to excite an electron from the t_2g level to the e_g level. Therefore, transition-metal ion incorporated ceramic matrices can be used as stable ceramic pigments due to the ability of transition-metal compounds to show vivid colours and stability of ceramics at high temperatures.^26,27 Moreover, transition-metal aluminate phases (spinel phases) can be formed during sintering of the mullite precursors with an appropriate amount of dopant ions since divalent metal ions not incorporated into the mullite structure due to their large ionic radii, react with alumina to form aluminate phases.^27,28 CoAl₂O₄, NiAl₂O₄ and CuAl₂O₄ posses interesting electronic and optical properties for which they find suitable applications in ceramic pigments.^29–32 The main limitation of these aluminates in pigment application is their chemical reactivity at higher temperatures which can be reduced by using inert aluminosilicate (mullite) as encapsulant.²⁷

During the course of our work nanocrystalline mullite phases have been synthesized via sol–gel route from a non-stoichiometric precursor with x = 0.29 (61.54 mol% alumina) which corresponds to a little high alumina content in the mullite species. With increasing alumina content a greater number of oxygen vacancies are formed in the mullite structure to maintain charge neutrality due to which silicon deficient mullite shows some interesting mechanical, electrical and photoluminescence properties. Thus, the effect of transition-metal ions (Co²⁺, Ni²⁺ and Cu²⁺) on different physico-chemical properties of mullite and their effect on mechanical strength, dielectric and photoluminescence properties were studied in this paper.

2. Experimental

2.1. Materials

All reagents used in this synthesis procedure are of analytical grade and >99% pure. The chemicals used in this preparation are aluminium isopropoxide (Al(O-i-Pr)₃) (AIP) (Loba Chemie, Mumbai, India), tetraethoxysilane (TEOS) (Si(OC₂H₅)₄) (Merck, Honenbrunn, Germany), absolute ethanol (Merck, Worli, Mumbai, India), cobalt chloride hexahydrate (CoCl₂·6H₂O) (Merck, Worli, Mumbai, India), nickel chloride hexahydrate (NiCl₂·6H₂O) (Merck, Worli, Mumbai, India), copper chloride hexahydrate (CuCl₂·6H₂O) (Merck, Worli, Mumbai, India).

2.2. Sol–gel synthesis of nanocrystalline mullite from non-stoichiometric precursors

A sol–gel synthesis of non-stoichiometric mullite precursor (61.54 mol% alumina and mole ratio of Al/Si was 3.2) was performed using aluminium isopropoxide (AIP) and tetraethoxysilane (TEOS) as source of alumina and silica respectively (Fig. 1).

Fig. 1 A schematic diagram for the preparation of transition-metal ions conjugated mullite nanocomposites.

The composition of the aluminosilicate precursor was kept slightly in the alumina rich zone intentionally to minimize glassy formation during sintering. Initially, AIP and TEOS with mass ratio 3.15 were simultaneously added to absolute ethanol and magnetically stirred overnight for the formation of precursor sol. Alkoxide solution was then hydrolysed by adding of 20 mL of deionized water to form the precursor sol under stirring at a fixed temperature of 60 °C. Precursors with dopant transition-metal ions were prepared by the addition of 20 mL aqueous solution of cobalt chloride hexahydrate, nickel chloride hexahydrate and copper chloride dehydrate to the alkoxide solution during the hydrolysis step so that the resulting molarity of metal ions in the precursor volume was 0.03, 0.06, 0.14 and 0.23 M, respectively. The pure and doped mullite gels obtained via hydrolysis of sol were dried at 80 °C in a hot-air oven and the dried gel was sintered at 1300 °C in a muffle furnace under air atmosphere with heating rate of 600 °C h⁻¹ and 2 h soaking to obtain transition-metal ions conjugated mullite nanocomposites for further characterization. The designations of pure and doped mullite composite samples are given in Table 1.

Table 1 Description of prepared samples along with their designations

Molarity of dopant salt solution	Final molarity in resulting precursor volume	Co^2+ doped mullite precursors	Ni^2+ doped mullite precursors	Cu^2+ doped mullite precursors
0	0	M00	M00	M00
0.1	0.03	M11	M21	M31
0.2	0.06	M12	M22	M32
0.5	0.14	M13	M23	M33
0.8	0.23	M14	M24	M34

---

2.3. Characterization

Crystallization, phase formation and microstructural analysis of pure and transition-metal ion doped mullite were performed by X-ray diffraction (XRD) (Bruker AXS INC., Madison, WI), Fourier Transform Infrared Spectroscopy (FTIR-8400S, Shimadzu) and FESEM (INSPECT F50, Netherlands).

The phase formation of mullite and mullite composite was studied by the X-ray diffraction using Cu-Kα radiation – 1.5409 Å, 2θ = 10–70°, scan speed 0.1 s per step, increment – 0.02, operating voltage – 40 kV, and operating current – 40 mA.

FTIR spectroscopic studies of mullite powders sintered at different temperatures were done by the KBr pellet method in the wavenumber range 500–1500 cm⁻¹.

Field emission scanning electron micrographs of the samples were taken after being coated with gold via plasma spraying at 0.1 mbar pressure. The samples were kept inside a vacuum chamber with pressure about 5 × 10⁻³ Pa at distance 15 mm away from the detector with spot size 3.0 mm. The emission current and operating voltage were kept at 170 μA and 20 kV, respectively.

Archimedes’ principle was used to measure the relative density of all samples using xylene as buoyant medium. The absolute density of the samples was determined employing the relation:

where W_a is weight of the sample in air, W_b is the weight of the sample in the buoyant medium, ρ is the density of the sample and ρ_b is the density of the buoyant liquid (0.86 g cm⁻³ at 25 °C for xylene). Assuming theoretical densities of mullite, cobalt oxide, nickel oxide and copper oxide as 3.16, 6.11, 6.67 and 6.31 g cm⁻³, the relative density and apparent porosity of the samples were measured.

Vickers (H_V) and Knoop hardness (H_K) of the polished sample pellets were measured using Leco’s hardness tester which contains a nanoindenter (Leco, LV700). In case of Vickers hardness test the indenter is a square based pyramid with angle (ψ) between the two opposite sides being 136° whereas for Knoop indenter a lozenge-based pyramid was used with angle (θ) between two opposite faces being 172°5′ and the angle (φ) between the other two being 130°. Hardness of a material is described as the ratio between the indentation load and parameter representing area of residual impression which depends on the shape of the indenter. Ten impressions were made on the surface of the samples using a load of 1 kg f during 10 s. After removal of the indenter, by measuring the dimension of indentations via an optical microscope with 20× zooming, the Vickers and Knoop hardness number of the composite were calculated according to the following equations:

where H_V and H_K are Vickers and Knoop hardness numbers, respectively, are expressed in MPa, P is the applied load in N, d is the indentation diagonal for Vickers and L is the longer indentation diagonal for Knoop indentation in mm, A_TAC represents the true area of contact and A_PAC represents the projected area of contact for Vickers and Knoop indentation, respectively.³³

The dielectric properties of the composite were investigated by measuring the frequency dependent capacitance and tangent losses using a digital LCR meter (Agilent, E4980A) with circular Ag electrodes under room temperature and pressure with 1 V applied signal and frequency ranging between 20 Hz and 2 MHz. The dielectric constant (ε) and the ac conductivity (σ_ac) of the pure and doped mullite composite were calculated according to the following equations, respectively,

σ_ac = 2πfεε₀ tan δ (5)

where, C, d, A and tan δ are the capacitance, thickness, area and tangent loss of the samples, respectively, f is the frequency of applied ac voltage in Hz and ε₀ is the absolute permittivity of free space with value 8.854 × 10⁻¹² F m⁻¹.

The photoluminescence of the composite was measured using a Cary Eclipse fluorescence spectrophotometer, Agilent Technologies, taking 0.1 mg mL⁻¹ solution of the samples in absolute ethanol medium. The fluorescence microscopic images of selected samples were taken under Zeiss Axiocam MRc with 50× zooming.

3. Results and discussion

3.1. X-Ray diffraction analysis

The comparative powder X-ray powder diffraction data of pure and Co(II), Ni(II) and Cu(II) ion incorporated mullite composites are shown in Fig. 2.

Fig. 2 Comparative powder X-ray diffraction patterns of pure and (a) Co²⁺, (b) Ni²⁺ and (c) Cu²⁺ doped mullite nanocomposites sintered at 1300 °C.

Well crystalline mullite phase is reflected in the diffraction pattern along with reflections of alumina for both pure and doped mullite composites. For 0.03 M solution of all dopant transition metals, very little change in mullitization occurred, as confirmed by the small difference between heights of characteristic mullite peaks compared to the pure mullite. In comparison to other metal ions the difference in peak heights for 0.03 M nickel chloride doping (Fig. 2b) was much lower, which is indicative of better mullitization for nickel ion doping. This is due to a higher extent of interaction of Ni²⁺ with aluminosilicate matrix. Co²⁺, Ni²⁺ and Cu²⁺ ions due to their lower valency and larger ionic radius cannot be readily incorporated into the mullite structure but their oxides have been found to form Co, Ni and Cu aluminates (spinel phases) due to solid-state diffusion of these oxides within alumina. The spinel formation rate follows the sequence CuAl₂O₄ > CoAl₂O₄ > NiAl₂O₄ which depends on the diffusivity of metal ions towards AlO₆ octahedra.³⁰ Increasing concentration of transition metals has a detrimental effect on mullitization since alumina is consumed to form spinel phases (aluminate phases) at higher doping concentration of metal ions resulting in a lower dissolution rate of Al₂O₃ into molten SiO₂ flux. The intensity of the peaks corresponding to crystalline alumina that are observed in the diffractogram of doped mullite composites, decreased significantly as the aluminate phases were found to grow from the respective oxides at higher doping. In the case of cobalt and copper ion, spinel phases started to appear from 0.03 M salt concentration (Fig. 2a and c) and grew distinctly at higher doping concentration whereas for nickel ion doping, reflections corresponding to nickel aluminate phase started to grow at 0.06 M nickel chloride concentration (Fig. 2b) due to low spinel forming capability of nickel ion. In air atmosphere, Co²⁺, Ni²⁺ and Cu²⁺ from their respective salts, form CoO, NiO and CuO upon high-temperature sintering followed by the formation of aluminate phases from these oxides upon reacting with alumina. At higher temperature, around 1050 °C, CuO (Cu(II)) may transform into Cu₂O (Cu(I)) which also contribute to the formation of aluminate phases besides Cu(II).³⁴ Due to higher diffusivity of Cu⁺ during the incubation period at 1300 °C, copper silicate phases along with aluminate phases were also observed in the X-ray diffraction pattern (Fig. 2c) when the salt concentration reaches 0.23 M, which indicates that Cu⁺ reacted not only with AlO₆ octahedra but also with SiO₄ tetrahedra. On the other hand, owing to lower diffusivity of Co(II) and Ni(II) as compared to Cu(I), they react only with AlO₆ octahedra and not with SiO₄ tetrahedra.

The average particle size was calculated from the Debye–Scherrer equation, D = 0.9λ/β cos θ, where D is average particle diameter, λ is wavelength of Cu-Kα radiation, and β is full width of half maxima in radians for the characteristic mullite peak at 2θ = 16° (where θ is Bragg’s angle) corresponding to the (110) plane. The dimensions of the particles as calculated from the Debye–Scherrer equation are in the nanometer regime for pure and transition metal–mullite composites (Table 2).

Table 2 Particle size from Debye-Scherrer’s equation, relative density, apparent porosity, dielectric constant, hardness and Young’s modulus for pure and transition-metal ions doped mullite composites

Sample	Average particle size (nm) (from XRD)	Bulk density (%)	Apparent porosity (%)	Dielectric constant at 1 kHz	Vicker’s hardness (GPa)	Knoop hardness (GPa)	Young’s modulus (GPa)
M00	46	62.59	37.41	91.3	1.6	1.7	46.7
M11	43	66.77	33.23	98.1	3.5	3.6	76.5
M12	48	80.84	19.16	73.1	5.2	5.6	135.7
M13	51	83.47	16.53	80.2	5.9	6.3	153.6
M14	59	84.88	15.12	68.2	6.9	7.5	207.3
M21	43	81.09	18.91	108.6	2.6	2.7	63.3
M22	43	82.1	17.9	175.3	5.4	5.8	142.4
M23	55	96.23	3.77	64.7	5.9	6.2	153.2
M24	57	98.89	1.11	67.7	6.5	6.9	167.2
M31	58	64.56	35.44	50.1	3.3	3.4	72.2
M32	52	74.9	25.1	56.3	6.8	7.1	170.5
M33	65	76.24	23.76	77.2	7.5	7.8	177.2
M34	56	82.91	17.09	64.9	8.7	8.7	174.6

---

3.2. FTIR analysis

FTIR spectra of Co²⁺, Ni²⁺ and Cu²⁺ doped mullite composites are given in Fig. 3.

Fig. 3 Fourier-transform infrared spectra of pure and (a) Co²⁺, (b) Ni²⁺ and (c) Cu²⁺ doped mullite nanocomposites sintered at 1300 °C.

Characteristic bands of mullite were evident at 572 cm⁻¹ (octahedral Al–O stretch), 730 cm⁻¹ (tetrahedral Al or Si bend), 830 cm⁻¹ (tetrahedral Al–O stretch), 950 cm⁻¹ (Si–O stretch), 1130 cm⁻¹ (tetrahedral SiO₄ stretch) and 1165 cm⁻¹ (Si–O stretch) for pure and transition-metal ion doped mullite samples.^8,35,36 In the case of Ni²⁺ and Cu²⁺ doping, an additional peak for Al–O stretching vibration was observed at 595 cm⁻¹ (Fig. 3b and c). Reflections of octahedral and tetrahedral Al–O stretching vibrations, tetrahedral Si–O stretching in the FTIR spectra is indicative of the presence of mullite in the composites whereas the decreased transmission intensity of Al–O stretching vibrations due to transition-metal ion doping clearly reflects the consumption of alumina towards the formation of spinel phases.

3.3. FESEM analysis

Microstructures of mullite and transition-metal ions conjugated mullite composites are shown in Fig. 4.

Fig. 4 Field emission scanning electron micrographs of (a) pure mullite (M₀₀), cobalt-doped mullite (b) M₁₁, (c) M₁₂, (d) M₁₃ and (e) M₁₄, nickel-doped mullite (f) M₂₁, (g) M₂₂, (h) M₂₃ and (i) M₂₄ and copper-doped mullite (j) M₃₁, (k) M₃₂, (l) M₃₃ and (m) M₃₄.

The nanometer dimension of equiaxed mullite grains was obvious from the FESEM micrograph (Fig. 4a) of pure mullite with average particle diameter ∼100 nm along with clearly visible porosity. The micrographs of cobalt- and nickel-ion doped mullite with lowest doping concentration (Fig. 4b and f) are similar to that of pure mullite with increased grain size, which indicates no significant morphological changes to the mullite structure due to doping. Non-uniform microstructure with equiaxed grains distributed among abnormal elongated grains was observed for 0.06 M doping concentration of cobalt chloride salt which reflects abnormal grain growth of mullite.³⁷ Improved microstructures were observed for much higher cobalt ion concentration with some elongated grains of average width ∼120–160 nm distributed among equiaxed grains of average diameter ∼100–150 nm that identify the biphasic nature of the precursor gels. In case of dopant nickel ion, no drastic morphological changes are visible up to 0.06 M doping concentration whereas in case of 0.14 and 0.23 M doping concentration, conglomerated sheet like and concoid structure of nickel aluminate were reflected, respectively, in the micrographs (Fig. 4h and i). Elongated grains of average width ∼100 nm decorated with smaller mullite grains of average diameter ∼80 nm were obtained (figures j and k) for 0.03 and 0.06 M doping concentration of copper chloride solution whereas for 0.14 M concentration some small copper aluminate grains (Fig. 4l) were also observed along with mullite grains. In case of highest doping with copper ion, sheet like structures (Fig. 4m) of copper aluminate were formed. No alumina structures and no metal aluminate structures at lower doping concentrations of the dopant solutions were observed due to their lower extent of formation. The increase in grain size and densification behavior for higher doping concentration in all the composites were also reflected in the FESEM microstructures.

3.4. Densification behavior

Densification behavior of transition metal–mullite nanocomposites are shown in Fig. 5.

Fig. 5 Densification behavior and variation of apparent porosity of mullite doped with (a) Co²⁺, (b) Ni²⁺ and (c) Cu²⁺.

The highly porous nature (∼37% apparent porosity) was evident in undoped mullite system due to higher gelation rate, higher evaporation rate of alcohol solvent and lower phase segregation. In case of all dopant transition-metal ions, the relative density of the doped mullite composites increased with higher extent of doping accompanied by a gradual decrease in apparent porosity. The highest relative density of ∼99% was obtained for the mullite sample doped with 0.23 M dopant solution of nickel chloride and for highest doping concentration of cobalt and copper solution, the density level was found to be ∼85 and ∼83%, respectively (Table 2). This higher relative density upon increased doping, along with dwindling apparent porosity, may be caused due to increased consolidation at 1300 °C, formation of interstitial aluminate phases in transition-metal ions doped mullite composite, higher cation mobility due to formation of cation vacancy resulting from the substitution of Al³⁺ by transition-metal ions, and finally due to higher rate of body diffusion for the contribution of liquid phases by transition-metal ions in terms of their corresponding oxides.^38,39

3.5. Color of the transition metal–mullite composites

Formation of transition-metal complexes by Co, Ni and Cu due to the incorporation of oxygen results in a shift of the absorption band towards the visible region (wavelength between 400 to 800 nm) of the electromagnetic spectrum and thus accounts for a wide range of applications as ceramic pigments. CoAl₂O₄, NiAl₂O₄ and CuAl₂O₄ due to their interesting physical and chemical properties including their excellent colours are widely used as ceramic pigments.²⁷ The formation of these aluminates from transition-metal ion doped mullite precursors after sintering at 1300 °C introduced sustainable color in the samples which make these mullite based transition-metal aluminates suitable for pigment applications.²⁶ No absorption in the visible region occurred for pure mullite samples which results in the white color of the pellet whereas deep blue, pale blue and brown colour appeared for Co, Ni and Cu doped mullite composites, respectively, due to formation of CoAl₂O₄, NiAl₂O₄ and CuAl₂O₄ (Fig. 6).

Fig. 6 Digital images for the pellets of pure and transition-metal ions conjugated mullite composites after sintering at 1300 °C.

The appearance of color was found to be intensified for increasing doping concentration of all three transition-metal ions which may be due to the increased formation of aluminate phases.

3.6. Mechanical properties

3.6.1. Comparison between Vickers and Knoop hardness. To compare between Vickers and Knoop hardness values, 9.8 N stresses were applied on the polished specimen surfaces of all samples during 10 s in order to negate indenter size effects. The difference between the hardness values for differently shaped Vickers and Knoop indenters may be caused due to the different extent of elastic recovery resulting from mismatch between the plastic zone beneath the indentation and the surrounding elastic deformed region after removal of the indenters.³³ Lawn and Howes⁴⁰ proposed that change in the indentation diagonal length was negligible for Vickers indentation though elastic recovery occurs along the depth of the indentation, whereas for Knoop indentation Marshall et al.⁴¹ observed that length of the minor diagonal is often less than the theoretical length as calculated via geometric consideration of the indenter, which is attributed to elastic recovery of the material. Ratio of short (s) and long indentation diagonal lengths (L) of Knoop impression was expressed as a function of the ratio between Knoop hardness (H_K) and Young’s modulus (Y) by Marshall et al. according to following equation.

On the other hand, Gong et al.⁴² explained that two hardness values differing in magnitude is indicative of a difference between degree of elastic recovery in both cases neglecting different plastic zones for the indenters. They further determined a correlation between s/L and the hardness ratio H_V/H_K according to the following equation

Later, Chicot et al.³³ showed that the eqn (7) given by Gong et al. is not valid for all ranges of hardness values. Eqn (7) is not valid for 0–7.9 GPa hardness values since in this range, according to Chicot et al. the Vickers hardness value (H_V) is less than the Knoop hardness value (H_K) which make the measured value of s/L higher than the theoretical one (0.1406). Due to elastic recovery, the shortening of the minor diagonal of Knoop indentation and hence lowering of s/L value are obvious after removal of the indenter and thus the experimental value of s/L must be less than the theoretical value of s/L which equals to 0.1406 as calculated from the geometric point of view. In this context the higher value of measured s/L after removal of the Knoop indenter compared to the theoretical one has no physical significance since it corresponds to elastic collapse. To remove this ambiguity, Chicot et al. modified eqn (6) and (7) to be applicable for all ranges of hardness values. They suggested that Vickers and Knoop hardness values should be calculated in the same scenario for which they consider true area of contact instead of projected area of contact for Knoop hardness measurement to relate it with Vickers hardness measurement from eqn (2). In order to recalculate the Knoop hardness value using the true area of contact, Chicot et al. suggested an equation which is a modified form of eqn (3) and given as follows:

Eqn (8) and (3) are very similar, only differing in the numerical constant (14.299 in eqn (3) and 12.873 in eqn (8)) for which it can be written from these two equations that

Chicot et al. also suggested to modify eqn (6) and (7) correctly according to eqn (10) and (11):

and

During the course of our work we followed Chicot et al.’s formalism since all the values of hardness for all samples, except copper-ion doped mullite at its highest doping concentration, fall below the critical hardness value of 7.9 GPa (Table 2). All hardness data were in agreement with the generalization of hardness and thus Vickers hardness is obtained to be less than that of Knoop hardness value for less than 7.9 GPa hardness. In case of 0.23 M dopant copper solution, both the hardness values was found to be 8.7 GPa which may be attributed to the fact that this hardness value is not too far from the critical value (7.9 GPa) of hardness, so that the difference between the two hardness values is not very prominent. Plots between (H_K)_TAC and H_V as well as between s/L and H_V/(H_K)_TAC are shown separately for different metal ion doping (Fig. 7a and b).

Fig. 7 Plots of (a) (H_K)_TAC vs. H_V and (b) s/L vs. H_V/(H_K)_TAC.

Both the plots show a straight line nature only differing in slope (positive slope for (H_K)_TAC vs. H_V plot and negative slope for s/L vs. H_V/(H_K)_TAC plot) and this straight line nature of the graphs is in agreement with Chicot et al. and suggests their modified equations to be valid for mullite ceramics and mullite matrix composites. Young’s modulus of pure mullite and transition-metal ion mullite composites were calculated from eqn (10) and their variation was discussed in the following section.

3.6.2. Doping concentration dependent mechanical properties. Fig. 8a shows the variation of Vickers and Knoop hardness with increasing doping concentration of transition-metal ions.

Fig. 8 Variation of (a) hardness and (b) Young’s modulus with doping concentration of transition-metal ions.

The Vickers and Knoop hardness of pure mullite was found to be 1.6 and 1.7 GPa, respectively, which may be due to the higher resulting porosity of pure mullite (Table 2). The increase in both Vickers and Knoop hardness values with increasing doping concentration of transition-metal ions were observed and highest value (∼8.7 GPa) for both the hardness values were obtained in the case of 0.23 M copper chloride solution (Table 2) which is higher than other mullite composites reported in literature.^4,24,43 The higher densification, larger grain size and the formation of dislocation networks in matrix grains as well as strengthened grain boundaries are responsible for higher mechanical strength of composites.⁴⁴ Among several strengthening mechanisms, chiefly grain boundary strengthening is responsible for the enhanced strength of the transition-metal ion doped mullite composites. Mechanical properties of materials can be enhanced by prohibiting dislocation motions in the grains. The transition-metal ions cause lattice distortions in mullite that hinder dislocation motion which results in higher mechanical strength of mullite upon doping with transition metals. Grain size has a significant influence on the mechanical properties of polycrystalline materials. Due to different crystallographic orientations of grains, grain boundaries arise which act as an impediment to dislocation motions in the grains; because the lattice structure of adjacent grains are different in different crystallographic orientations, more energy is required for a dislocation to change its direction and move to a neighboring grain. Smaller grain size and interstitial formation of aluminate phases for transition-metal ion doping introduce a larger number of grain boundaries, which oppose dislocation motions, resulting in higher values of hardness.²⁰ Variation of Young’s modulus with doping concentration of dopant transition-metal salts are shown in Fig. 8b. Due to the nanometer dimensions of mullite and spinel grains, increase in Young’s modulus of the mullite composite with higher doping with transition-metal ions can be explained by the inverse Hall–Petch effect. According to this effect, yield strength increases with increasing grain size which signifies a higher extent of elasticity of the material, resulting in higher Young’s modulus. The small decrease in the value of Young’s modulus for 0.23 M doping concentration of copper chloride may be caused due to the formation of copper silicate phases (Fig. 8b).

3.7. Dielectric behavior

3.7.1. Doping concentration dependent dielectric properties for different transition-metal ions. Fig. 9 shows the variation of dielectric constants and tangent losses at 1 kHz frequency for cobalt, nickel and copper ion doped mullite composites.

Fig. 9 Doping concentration dependent dielectric constant and tangent losses of mullite composite doped with (a) Co²⁺, (b) Ni²⁺ and (c) Cu²⁺ at 1 kHz frequency.

Highest dielectric constant (∼175) was obtained in case of nickel doping at its highest concentration of doping (M₂₄) (Table 2) which may be explained in terms of Jahn–Teller distortion. The unpaired electron in the t_2g orbital of Co²⁺ (d⁷ electron system) and three electrons in two degenerate e_g orbitals of Cu²⁺ (d⁹ electron system), lead to Jahn–Teller distortion in the mullite composite gel matrix, which in turn lowers the energy.^13,16 However, Ni²⁺ due to its symmetrically filled e_g orbitals and t_2g orbital show no distortion in the precursor gel for which the energy of nickel ion doped mullite composite should be greater than that of cobalt and copper doped mullite composites.¹⁶ This higher energy of nickel ion doped mullite composites promotes higher entropy of the induced dipoles upon the application of external electric field resulting in higher dielectric constant. Moreover, it is obvious that in electronic and ionic polarization processes, the force due to an external electric field for which movement of internal individual dipoles occurs is balanced by elastic binding forces. The Young’s modulus of nickel doped composites being lower than that of the other transition-metal ions (Co²⁺ and Cu²⁺) conjugated mullite composites (Table 2), it can be supposed that elastic binding force which is resistant to polarization processes is lower, that leads to higher polarization of the internal dipoles, and hence higher dielectric constant in nickel mullite composites.

In dielectrics some interactive forces between adjacent molecules are supposed to exist that try to prevent their orientation according to the external field. This short range interaction force can be suggested to reduce the dielectric constant of the composites. At 1 kHz frequency, in the case of cobalt ion doping, the maximum dielectric constant (∼98) was obtained for M₁₁ and for copper ion, the maximum value of dielectric constant (∼34) was achieved for sample M₃₃ with a sudden decrease in dielectric constant for 0.03 M doping at the same frequency. The Maxwell–Wagner–Sillars (MWS) interfacial polarization mechanism describes polarizations at the interfaces of alumina, mullite and aluminate phases which may be dominant at lower frequencies such as 1 kHz. The dielectric constant of pure mullite (∼91) was observed to be higher than that of other reported ones which may be due to the resulting larger porosity. Formation of elongated grains distributed among equiaxed spherical grains can reduce the interfacial area per unit volume but enhance the resistive interaction force between grains due to larger intersection probability of elongated grains within spherical particles.⁴⁵ Since there are no morphological changes occurring for M₁₁, the dielectric constant increased relative to pure mullite due to larger grains having larger interfacial area per unit volume, leading to greater interfacial polarizations. A drop in dielectric constant values was observed for M₁₂ since some elongated grains have formed within equiaxed grains resulting in lower interfacial area and larger interaction probabilities that inhibit polarization. Again an increase in the value of ε occurred for larger interfacial polarization resulting from grain growth followed by final decreased value for agglomeration of grains. The variation of dielectric constant for copper doped mullite composites follow a similar pattern to that of cobalt doped samples. The only exception is a drop of ε value for M₃₁ which is probably due to formation of interconnected elongated grains. For further doping, the dielectric constant increased up to 0.14 M doping concentration followed by a decrease for highest doping concentration which can be explained in the same way to that of cobalt doped samples. In case of nickel doped samples, no abrupt variation of dielectric constant was observed since no morphological changes occur for nickel ion doping. With doping concentration increased up to 0.06 M, a continuous increase in ε value occurred due to interfacial polarizations at the interfaces of mullite alumina and nickel aluminates. Further doping with nickel ion results in a decrease in ε value and this remains almost invariant with higher doping due to exaggerated grain growth.

The tangent loss (tan δ) is indicative of energy dissipation in a dielectric system that is proportional to the imaginary part of dielectric constant.^25,46 Generally loss in the mullite composites happens due to the migration of mobile cations and deformation losses resulting from the network constructed by silica.⁴⁷ It was reported earlier from the Cole–Cole plot of dielectric permittivity that Debye-like relaxation is the most dominant relaxation type which can occur in mullite ceramics.⁴⁸ Thus, non-linear increase of tangent losses which was observed for all three transition-metal ions doped mullite composite (Fig. 9a–c) may be due to different Debye-like relaxations.⁴⁹ The increased tangent losses can be caused due to higher ionic conductivity through oxygen defects formed due to transition-metal ion doping whereas lowering of tangent losses occurred as a result of morphological changes in the composites.²⁵

Fig. 10 shows the variation of dielectric constant and ac conductivity of mullite with increased doping concentrations of Co²⁺, Ni²⁺, Cu²⁺ at 100 Hz, 10 kHz at 1 MHz frequencies.

Fig. 10 Concentration dependent dielectric constant for (a) Co²⁺, (b) Ni²⁺ and (c) Cu²⁺ and ac conductivity for (a) Co²⁺, (b) Ni²⁺ and (c) Cu²⁺ doped mullite composites.

At higher frequencies the variation of dielectric constants are not very prominent due to restricted response of induced dipoles in the composites. Ac conductivities of the composites show the same nature as that of dielectric constant.

3.7.2. Frequency dependent dielectric properties. Fig. 11a, d and g show the frequency dependent dielectric constants of Co²⁺, Ni²⁺, Cu²⁺ ion doped mullite nanocomposites respectively.

Fig. 11 Frequency response of (a) dielectric constant, (b) tangent losses and (c) ac conductivity for pure and Co²⁺ doped mullite composites, (d) dielectric constant, (e) tangent losses and (f) ac conductivity for pure and Ni²⁺ doped mullite composites and (d) dielectric constant, (e) tangent losses and (f) ac conductivity for pure and Cu²⁺ doped mullite composites.

The obvious decrease of dielectric constant with increased frequency is readily attributed to the restricted movement of different polarizations with frequency. The polarization process in transition-metal ion doped mullite nanocomposites can be viewed as similar to that of conduction processes. Among various conduction mechanisms such as Schottky, Poole–Frenkel, ionic hopping and space charge limited current mechanisms, ionic hopping between the oxygen vacancies in mullite structure is responsible for the conduction processes in mullite ceramic composites.^50–52 Doping with transition metals tends to increase this conductivity by several orders of magnitude.⁵³ The electrical conduction mechanism can be interpreted by the electron-hopping model proposed by Heikes and Johnston.⁵⁴ The interwell hopping is supposed to contribute to the dielectric relaxation of transition metal–mullite nanocomposites at lower frequencies whereas the dielectric constants of all samples remain almost constant at high frequency range because beyond a certain frequency of the applied electric field, intrawell hopping becomes predominant and the charge carriers in the composites do not have enough time for long range hopping before the reversal of the external field. As a result, only intrawell hopping exists at the high frequency range since the average intrawell hopping distance, of the order of one lattice spacing, is less than that for interwell hopping which is in the range of a few nanometres. Therefore, the average polarisation decreases as the frequency is increased resulting in a lowering of dielectric constant at higher signal frequency.

Fig. 11b, e and h show the frequency dependence of tangent losses (tan δ) for Co(II), Ni(II) and Cu(II) doped mullite composites. The loss value was higher at lower frequencies and gradually decreased at higher frequency up to a lower saturated value. The tangent loss being representative of energy dissipation in dielectrics, is proportional to the imaginary part of the dielectric constant, and depends on the randomness of grain orientation and grain boundaries. The higher value of losses at lower frequencies can be attributed to the favourable movement of dipoles at lower field reversal time and lower value of losses at higher frequency is due to the limited dielectric response. Moreover, at lower frequencies, due to lower conductivity, electrons require more energy for conduction through grains. Therefore, energy dissipation is high at a low frequency range resulting in higher tangent losses. At high frequency, higher conductivity resulting in exchange of electrons between two conducting grains becomes favourable, resulting a decrease in loss.⁴⁷

The ac conductivities of cobalt, nickel and copper ion doped mullite composites are shown in Fig. 11c, f and i respectively. An increase of conductivity of all samples with increased frequency of the applied field occurred obeying the frequency dependent part of Johnscher’s universal power law which is as follows:

σ(ω) = σ_dc + σ₀ω^S

where σ_dc is the dc conductivity which is a frequency independent part and σ₀ is the frequency dependent part with S ranging between 0 and 1. The linear increase of conductivities can be correlated directly with the Maxwell–Wagner–Sillars (MWS) interfacial space charge polarization effect and dielectric relaxation at higher frequencies. The mobility of transition-metal ions and aluminium cations is facilitated in the amorphous phases produced by doping.²⁵ Due to this higher mobility of charges, cations acquire an easy path to move, resulting in larger conductivity. Moreover, the complex structural rearrangements in mullite composites, ion hopping through oxygen defects and charge conduction via grain boundaries are responsible for higher conductivity of the mullite composites.⁴⁷

3.8. Photoluminescence properties

The photoluminescence spectra of Co²⁺, Ni²⁺ and Cu²⁺ doped mullite nanocomposite were obtained with an excitation wavelength of 255 nm and shown in Fig. 12.

Fig. 12 Photoluminescence spectra of pure and (a) Co²⁺, (b) Ni²⁺ and (c) Cu²⁺ doped mullite nanocomposites with excitation wavelength at 255 nm and fluorescence microscopic images of (d) M₀₀, (e) M₁₃, (f) M₂₃ and (g) M₃₃ with blue light excitation.

Strong PL bands were observed at 310, 347, 436 and 460 nm with weak bands at 362, 379, 407 and 424 nm. The radiative centers leading to the PL bands are supposed to be generated from Al–O bonds in mullite bonding structures.²² The dimensions of the mullite and spinel grains are too high to show quantum confinement effect,⁵⁵ excluding the possibility of PL emission due to a quantum confinement, and the observed PL emission bands may be attributed to the oxygen related defects in the transition metal–mullite nanocomposites. The PL emissions are expected to result from the radiative recombination of photo-excited holes with electrons occupying the oxygen vacancies. The production of oxygen vacancies in the mullite structure is obvious when Si⁴⁺ lattice sites are replaced by Al³⁺ and the vacancy formation equation is as follows:

2Si⁴⁺ + O²⁻ → 2Al³⁺ + ⊗

where ⊗ represents an oxygen vacancy.¹⁵ Therefore it is obvious that a large number of oxygen vacancies can be formed in alumina rich mullite structures since large numbers of substitutions occur that result in prominent oxygen defects. The mullite synthesized during our work being silica deficient, it can be ascribed to form definite oxygen vacancies for which even in case pure mullite sample, PL emission bands were visible.

Transition-metal ions in their oxide states meanwhile react with alumina, which can introduce defect structures in mullite which is responsible for PL emissions. The defect reaction can be represented such as the following:

where M = Co, Ni or Cu, and V_O designates vacancy.

In solid solution of transition-metal oxides and alumina, transition-metal oxides dissolve substitutionally and the dopant metal ion will acquire one negative charge which is compensated by the formation of a positive effective charge or oxygen vacancy. In this defect reactions neutral metal and oxygen lattices sites are also formed in order to maintain charge neutrality.

The higher doping concentration of transition-metal ions thus can be attributed to the increase in oxygen vacancy density in the composite structures. This increase in the density of localized defect states in the energy band gap can cause an increase in intensity of PL bands (UV region) and is confirmed by the prominent enhancement of PL intensity around 310 nm at higher doping concentrations.

Fluorescence microscopic images are shown in Fig. 12d–g for M₀₀, M₁₃, M₂₃ and M₃₃ samples with excitation in blue light. The fluorescence corresponding to pure mullite sample is not very prominent which can be readily predicted from the lower PL intensity. The highest PL intensity of cobalt ion doped mullite composite was reflected from the fluorescence microscopic image of M₁₃ possessing highest intensity followed by that of M₃₃ and M₂₃.

4. Conclusion

Multifunctional ceramic composites of transition-metal ion conjugated mullite have been developed. The higher densification (∼99% relative density for nickel ion doping), moderate hardness and Young’s modulus due to different ion doping make these composites not only useful in structural application, but also as ceramic pigments in paint systems due to the formation of colored aluminate spinel phases. The highest value of hardness (∼8.7 GPa) and Young’s modulus (∼207 GPa) was found for highest doping concentration of copper and cobalt ions, respectively. Higher dielectric constants of pure (∼91) and metal ion doped mullite composites (∼175 for 0.06 M nickel chloride doping) have revealed a wide scope for these composites in useful electrical applications such as ceramic capacitors. Prominent fluorescent properties in both UV and visible region resulting from oxygen defect structures with highest fluorescence intensity obtained for cobalt doped mullite composites make such materials also suitable for optical applications. Thus, the developed ceramic composites with a unique combination of physico-chemical, dielectric, mechanical and photoluminescence properties may have promising structural, electrical and optical applications in everyday life.

Acknowledgements

Authors are grateful to DST (INSPIRE PROGRAM (No. DST/INSPIRE Fellowship/2013/797, Code: IF130870)), Government of India for their financial support.

References

1. D. X. Li and W. J. Thomson, J. Am. Ceram. Soc., 1991, 74, 2382.
2. A. Kool, P. Thakur, B. Bagchi, N. A. Hoque and S. Das, Appl. Clay Sci., 2015, 114, 349.
3. I. A. Aksay, D. M. Dabbs and M. Sarikaya, J. Am. Ceram. Soc., 1991, 74, 2343.
4. M. M. S. Sanad, M. M. Rashad, E. A. Abdel-Aal and M. F. El-Shahat, Ceram. Int., 2013, 39, 1547.
5. K. Yoshida, H. Hyuga, N. Kondo and H. Kita, Mater. Sci. Eng., B, 2010, 173, 66.
6. E. Tkalcec, S. Kurajica and H. Ivankovic, J. Eur. Ceram. Soc., 2005, 25, 613.
7. S. Aryal, P. Rulis and W. Y. Ching, J. Am. Ceram. Soc., 2012, 95, 2075.
8. C. H. Ruscher, G. Schrader and M. Gotte, J. Eur. Ceram. Soc., 1996, 16, 169.
9. R. X. Fischer, H. Schneider and D. Voll, J. Eur. Ceram. Soc., 1996, 16, 109.
10. J. Parmentier and S. Vilminot, Chem. Mater., 1997, 9, 1134.
11. H. Schneider and S. Komarneni, Mullite, Wiley-VCH, Weinheim, 2005.
12. L. S. Cividanes, T. M. B. Campos, L. A. Rodrigues, D. D. Brunelli and G. P. Thim, J. Sol-Gel Sci. Technol., 2010, 55, 111.
13. B. Bagchi, S. Das, A. Bhattacharya, R. Basu and P. Nandy, J. Am. Ceram. Soc., 2009, 92, 748.
14. J. Parmentier and S. Vilminot, J. Alloys Compd., 1998, 264, 136.
15. H. Schneider, J. Schreuer and B. Hildmann, J. Eur. Ceram. Soc., 2008, 28, 329.
16. B. Bagchi, S. Das, A. Bhattacharya, R. Basu and P. Nandy, J. Sol-Gel Sci. Technol., 2010, 55, 135.
17. L. B. Kong, H. Huang, T. S. Zhang, Y. B. Gan, J. Ma, F. Boey and R. F. Zhang, Mater. Sci. Eng., A, 2003, 359, 75.
18. L. B. Kong, T. S. Zhang, J. Ma and F. Boey, J. Eur. Ceram. Soc., 2003, 23, 2247.
19. E. E. Kiss and P. S. Putanov, React. Kinet. Catal. Lett., 2003, 79, 325.
20. J. P. Buban, K. Matsunaga, J. Chen, N. Shibata, W. Y. Ching, T. Yamamoto and Y. Ikuhara, Science, 2006, 311, 212.
21. X. S. Peng, L. D. Zhang, G. W. Meng, X. F. Wang, Y. W. Wang, C. Z. Wang and G. S. Wu, J. Phys. Chem. B, 2002, 106, 11163.
22. H. K. Seong, U. Kim, M. H. Kim and H. J. Choi, J. Am. Ceram. Soc., 2007, 90, 1937.
23. O. Khatim, T. H. N. Nguyen, M. Amamra, L. Museur, A. Khodan and A. Kanaev, Acta Mater., 2014, 71, 108.
24. M. A. Camerucci, G. Urretavizcaya, M. S. Castro and A. L. Cavalieri, J. Eur. Ceram. Soc., 2001, 21, 2917.
25. D. Roy, B. Bagchi, S. Das and P. Nandy, Mater. Chem. Phys., 2013, 138, 375.
26. S. Roy, S. Kar, B. Bagchi and S. Das, Appl. Clay Sci., 2015, 107, 205.
27. M. Sales, C. Valentin and J. Alarcon, J. Eur. Ceram. Soc., 1997, 17, 41.
28. D. Mazza, A. Delmastro and S. Ronchetti, J. Eur. Ceram. Soc., 2000, 20, 699.
29. S. Kurajica, E. Tkalcec and J. Schmauch, J. Eur. Ceram. Soc., 2007, 27, 951.
30. P. H. Bolt, F. H. P. M. Habraken and J. W. Geus, J. Solid State Chem., 1998, 135, 59.
31. M. Sales, C. Valentin and J. Alarcon, J. Sol-Gel Sci. Technol., 1997, 8, 871.
32. M. Sales, J. Vila and J. Alarcon, J. Mater. Sci., 1998, 33, 4435.
33. D. Chicot, D. Mercier, F. Roudet, K. Silva, M. H. Staia and J. Lesage, J. Eur. Ceram. Soc., 2007, 27, 1905.
34. T. Martisius and R. Giraitis, J. Mater. Chem., 2003, 13, 121.
35. A. Kool, P. Thakur, B. Bagchi, U. Rajak, T. Das, S. Kar, G. Chakraborty, T. K. Mukhopadhyay and S. Das, J. Asian Ceram. Soc., 2014, 2, 297.
36. P. Padmaja, G. M. Anilkumar, P. Mukundan, G. Aruldhas and K. G. K. Warrier, Int. J. Inorg. Mater., 2001, 3, 693.
37. M. Zhou, K. Sun and J. Wu, J. Mater. Sci. Lett., 2001, 20, 2021.
38. J. H. Li, H. W. Ma and W. H. Huang, J. Hazard. Mater., 2009, 166, 1535.
39. S. M. Naga and A. El-Maghraby, Mater. Charact., 2011, 62, 174.
40. B. R. Lawn and V. R. Howes, J. Mater. Sci., 1981, 16, 2745.
41. D. B. Marshall, T. Noma and A. G. Evans, J. Am. Ceram. Soc., 1982, 65, c175.
42. J. Gong, J. Wang and Z. Guan, Mater. Lett., 2002, 56, 941.
43. M. I. Osendi and C. Baudin, J. Eur. Ceram. Soc., 1996, 16, 217.
44. D. Ghahremani, T. Ebadzadeh and A. Maghsodipour, Ceram. Int., 2015, 41, 1957.
45. P. Thakur, A. Kool, B. Bagchi, S. Das and P. Nandy, Phys. Chem. Chem. Phys., 2015, 17, 1368.
46. M. Jaroszewski, Eur. Phys. J. B, 2015, 88, 262.
47. K. Halder, D. Roy and S. Das, J. Mater. Sci.: Mater. Electron., 2015, 26, 5803.
48. R. H. M. Van de leur and C. A. P. Zevenhoven, J. Mater. Sci., 1991, 26, 4086.
49. P. Thakur, A. Kool, B. Bagchi, N. A. Hoque, S. Das and P. Nandy, Phys. Chem. Chem. Phys., 2015, 17, 13082.
50. R. Kaur, C. Singh, V. Bhikhan, M. Jaroszewski and S. Bindra Narang, Eur. Phys. J. B, 2015, 88, 245.
51. G. Mata-Osoro, J. S. Moya, M. Morales, L. A. Diaz, H. Schneider and C. Pecharromán, Int. J. Mater. Res., 2012, 103, 408.
52. S. P. Chaudhuri, S. K. Patra and A. K. Chakraborty, J. Eur. Ceram. Soc., 1999, 19, 2941.
53. C. Neusel, H. Jelitto and G. A. Schneider, J. Appl. Phys., 2015, 117, 154902.
54. R. Heikes and D. R. Johnston, J. Chem. Phys., 1957, 26, 582.
55. C. Liang, G. Meng, Y. Lei, F. Phillipp and L. Zhang, Adv. Mater., 2001, 13, 1330.

---