Improving Thermal Stability and Hydrophobicity of Rutile-TiO2 Nanoparticles for Oil-Impregnated Paper Application

Thermal stress and moisture absorption can cause a synergetic negative impact on kraft paper. Among various approaches for improving the dielectric properties of kraft paper, nanotechnology has had promising results. However, the hydrophilicity of most metal oxide nanoparticles renders nanomodified kraft paper more vulnerable to thermal stress and moisture, thereby inducing degradation. In nanomodified kraft paper research, the use of TiO2 nanoparticles has yielded the most promising results. The major shortfall, however, is the hydrophilicity of TiO2. This work investigated surface modifications of rutile-TiO2 nanoparticles (NPs) for improved hydrophobicity and thermal stability. Rutile-TiO2 NPs is a nontoxic metal oxide that can withstand high temperature and is stable in chemical reactions. Two cases of surfactants were used—alkyl ketene dimer (AKD) and alkenyl succinic anhydride (ASA). The intention was to increase heat resistance and reduce the surface free energy of the rutile-TiO2 NPs. The impacts of the surface modifiers on the rutile-TiO2 NPs were characterised using FT-IR, muffle furnace, analytical weight balance, and TGA. It was discovered that new functional groups were formed on the modified NPs examined through FT-IR spectra. This indicates new chemical bonds, introduced through the surface modification. The unmodified rutile-TiO2 NPs absorbed moisture, increasing their mass by 3.88%, compared with the modified nanoparticles, which released moisture instead. TGA analysis revealed that AKD- and ASA-modified rutile-TiO2 needed higher temperatures than the unmodified rutile-TiO2 to markedly decompose. AKD, however, gave better performance than ASA in that regard. As an example, those modified with 5% AKD sustained a 45% higher temperature than the pure TiO2 nanoparticles. Furthermore, in both cases of the surfactants, the higher the percent of surfactant content was, the more thermally stable the nanoparticles became. This work demonstrates the possibility of fabricating rutile-TiO2 NPs to give improved hydrophobicity and thermal stability for possible dielectric applications such as in kraft paper for power transformer insulation.


Introduction
Temperature rise in power transformers can cause thermal-induced degradation of kraft paper insulation. The temperature rise in power transformers may result from electrical faults, overloads, etc. [1,2].
Exposed hydroxyl sites on kraft paper (adsorption sites of cellulose) make it hygroscopic. The exposed site (OH) with a large dipole moment attracts water through the Coulomb force. The presence of water, therefore, modifies the dielectric properties of kraft paper [3][4][5][6][7]. Such a compromised kraft paper suffers from accelerated decomposition. Associated problems include increased dielectric loss (loss factor) and increased susceptibility to partial discharge-induced degradation. Furthermore, the mechanical properties of kraft paper are degraded.
In a power transformer, the effects of pyrolysis and hydrolysis on kraft paper insulation cannot be treated independently, as the presence of one can influence the other. For instance, temperature rise in transformers can lead to moisture presence due to oxidation and can also distort the bonding structure of kraft paper. On the other hand, the presence of moisture in the transformer mostly due to ingress from the atmosphere or natural gassing of the transformer oil affects the hydrogen bond of kraft paper [8][9][10][11][12]. The hydrogen bond defines the cross linkage of cellulose chains on which the dielectric properties of the insulation depend [13,14]. In essence, therefore, the rate of thermal stability or withstand of kraft paper insulation is proportional to its moisture content [15].
To improve on kraft paper's thermal stability, normally the paper is either chemically or physically modified. Manufacturers of transformer insulation introduced different thermally upgraded kraft papers around the late 1950s. The National Electrical Manufacturers Association (NEMA) recognised these improvements in 1962 through the standard TR-1-1962 [16]. For the chemical modification of kraft paper, the water-absorbing group (hydroxyl group) of the cellulose which is vulnerable to thermal effect is replaced with a more stable functional group, cyanide ethylation patented in the early 1960s, as well as acetylation. As for physical modification (thermal stabilisers), the rate of insulation degradation is retarded through the reaction of the additives with degrading factors. Amine compounds are the most frequently used additives for physical modification [2,[16][17][18].
According to Jusneret et al. [2] and Liang et al. [18], both chemical and physical approaches to thermal modification of cellulose increase the risk to health and safety of personnel, as well as environmental contamination. There is a possibility of discharge of highly toxic degradation by-products during thermal ageing, coupled with the displacement of hydrogen bonds, resulting in a decrease in mechanical strength. These setbacks are mainly due to cyanoethylation of kraft paper and the discharge of ammonia due to amino compounds. In the literature, therefore, there is always emphasis on the need to continuously search for less toxic and environmentally friendly kraft paper, with thermal endurance solutions. In that regard, this work was conceived to explore possibilities of improving the thermal stability of kraft paper through less toxic means.
In power transformers, more than 98% of the moisture is contained within kraft paper. To mitigate this, kraft paper is dried to reduce the moisture content to 0.5% (by dry weight), then impregnated with an oil to fill the air voids within the structure. When a transformer is in use, moisture is contained in the solid insulation, dissolved in oil, and also in a form of liquid in the core or bottom of the transformer [4,8,[19][20][21][22][23][24]. The IEEE C57.140 Guide for Evaluation and Reconditioning of Liquid Immersed Power Transformer recommends drying wet insulation (solid and liquid). Some of these power transformer drying (maintenance) methods include factory drying, vacuum dry-out method, off-line processing approach, and the on-line transformer dryer [15,23]. According to a CIGRE working group A2.18 [24], migration of moisture from kraft paper to oil for extraction depends on temperature, and this process could take months. It follows, therefore, that moisture management in power transformers requires mitigating the rate of moisture absorption of kraft paper.
In this paper, the possibility of improving the thermal stability of a kraft paper with less toxic material, as well as controlling its vulnerability to the surrounding moisture, is explored. In the literature, nanotechnology has been recognised as a promising technique for improving the dielectric properties of kraft paper. However, regarding the hydrophilicity and thermal stability challenges of kraft paper, there has not been much research on using the technology as a possible solution. In power transformer technology, the application of nanotechnology is increasingly drawing research attention as a promising tool in addressing major threats to the life of transformer insulation [25]. Kraft paper modified with nanoparticles has been reported to contain more hydrogen bonds, as more traps within cellulose structure are filled with the particles; this improves the bonding strength and the dielectric properties [26].
According to Liao et al. [27], oil-impregnated kraft paper modified with nano-TiO 2 prevents the injection of space charge distortion; a 50% decrease was reported for the modified paper with TiO 2 NPs. Liu et al. [28] improved surface discharge inception voltage on kraft paper modified with nanoparticles. The inception voltage was reported to have improved from 2.85 kV for an unmodified paper to 3.5 kV for the modified one. Hollertz [29] reported a 66% improvement in the tensile strength of paper after modification with nanoparticles. Among the nanoparticles used for modification of kraft paper, metal oxides are commonly used, and these include TiO 2 , ZnO, SiO 2 , and Al 2 O 3 [7,26,[28][29][30].
Of the options, nano-TiO 2 has been reported to be superior. Its compatibility and ability to improve the performance of kraft paper are notable. TiO 2 is reported to be more stable in reactions, nontoxic, low cost, and can withstand higher temperatures. The rutile phase of nano-TiO 2 has high electrical resistivity, thermal stability, and good mechanical strength. In contrast to the other common phases of the TiO 2 NPs (brookite and anatase), the rutile is chemically stable and can withstand very high temperatures and pressure without transforming or easily decomposing. Rutile-TiO 2 has a relatively high dielectric constant, making it suitable for some dielectric applications [31][32][33][34][35][36][37]. Although TiO 2 NPs are an attractive nanofiller for kraft paper, it has a major weakness in being hydrophilic [7,38,39]. It is worth noting that in the literature, most studies on modifying kraft paper with TiO 2 have been with the anatase phase and not the rutile. Therefore, in this work, the investigated nanoparticles were rutile-TiO 2 . This paper presents a fabrication procedure that produced rutile-TiO 2 NPs with improved hydrophobicity and thermal stability.

Using Sizing Agents in Making Cellulose Fibre to Be Hydrophobic
Alkyl ketene dimer (AKD) and alkenyl succinic anhydride (ASA) are usually used to make organic compounds to be hydrophobic. They are the most used water-resistive agents (sizing) for manufacturing paper. Application of the sizing agent reduces the surface free energy of the host material, resulting in better hydrophobicity. The mechanism involves anchoring the hydrophilic head of the agents (AKD and ASA), with the host material and the hydrophobic tail extended out to repel water, as demonstrated in Figure 1. As far back as the 1960s, AKD and ASA have been used to control the wettability of paper. The extent of the interaction of the agents with the host determines the degree of resultant hydrophobicity of the modified paper [40][41][42].
Seppanen [43] used AKD and ASA to control the wettability of a paper and reported that the contact angle (angle with which the water makes with the paper surface) indicated a reduction in surface energy of the paper due to the application of the sizing agents. Hubbe [41] also reviewed studies in the literature on the application of internal sizing agents and revealed that AKD and ASA are the commonly used agents by paper manufacturers. It has also been reported that AKD and ASA increase the heat resistance of the resultant modified polymers and in cases where nano-metal oxides are used [40,41,[43][44][45][46][47][48][49]. In this work, therefore, rutile-TiO 2 NPs were surface conditioned with AKD and ASA to study the effectiveness of the modification on thermal and hydrophilic properties of the resultant nanofiller. In the future extension of the present work, the modified TiO 2 NPs will be used to modify kraft paper for improved hydrophobicity and thermal stability. Bonding of inorganic fillers such as TiO2 NPs with an inorganic sizing agent can be difficult. A coupling agent is, therefore, often used to bond the two. The coupling agent can be represented by the formula R (CH2)n SiX3. The 'X' is the hydrolysable group which forms a bond with TiO2 NPs. 'R' is the functional group; it bonds with the polymer (cellulose) or organic material, which, in this case, are the sizing agents. Figure 2 illustrates the interface bonding of organic (sizing agents) and inorganic (TiO2 NPs) material with silane. The nanofiller reacts with 'X' to produce silanol (forming metal hydroxide), while 'R' produces a covalent bond with organic material [39,[50][51][52][53][54]. The rest of this paper presents the experimental work conducted to synthesise the rutile-TiO2 NPs of specific shapes and dimensions. Some of the resultant nanoparticles were surface modified with AKD and others with ASA sizing agent. Both sets were then characterised using various techniques, and the results were analysed. Bonding of inorganic fillers such as TiO 2 NPs with an inorganic sizing agent can be difficult. A coupling agent is, therefore, often used to bond the two. The coupling agent can be represented by the formula R (CH 2 ) n SiX 3 . The 'X' is the hydrolysable group which forms a bond with TiO 2 NPs. 'R' is the functional group; it bonds with the polymer (cellulose) or organic material, which, in this case, are the sizing agents. Figure 2 illustrates the interface bonding of organic (sizing agents) and inorganic (TiO 2 NPs) material with silane. The nanofiller reacts with 'X' to produce silanol (forming metal hydroxide), while 'R' produces a covalent bond with organic material [39,[50][51][52][53][54]. Bonding of inorganic fillers such as TiO2 NPs with an inorganic sizing agent can be difficult. A coupling agent is, therefore, often used to bond the two. The coupling agent can be represented by the formula R (CH2)n SiX3. The 'X' is the hydrolysable group which forms a bond with TiO2 NPs. 'R' is the functional group; it bonds with the polymer (cellulose) or organic material, which, in this case, are the sizing agents. Figure 2 illustrates the interface bonding of organic (sizing agents) and inorganic (TiO2 NPs) material with silane. The nanofiller reacts with 'X' to produce silanol (forming metal hydroxide), while 'R' produces a covalent bond with organic material [39,[50][51][52][53][54]. The rest of this paper presents the experimental work conducted to synthesise the rutile-TiO2 NPs of specific shapes and dimensions. Some of the resultant nanoparticles were surface modified with AKD and others with ASA sizing agent. Both sets were then characterised using various techniques, and the results were analysed. The rest of this paper presents the experimental work conducted to synthesise the rutile-TiO 2 NPs of specific shapes and dimensions. Some of the resultant nanoparticles were surface modified with AKD and others with ASA sizing agent. Both sets were then characterised using various techniques, and the results were analysed.

Materials
AKD sizing agent emulsion was obtained from Sappi™ Technology Centre (Pretoria, South Africa). The ASA, silane coupling agent, and titanium (IV) butoxide were purchased from Sigma-Aldrich (Pty.) Ltd. (St. Louis, MO, USA). Both chemicals were used as received.

Synthesis of Rutile-TiO 2 Nanoparticles
The sol-gel method was used to synthesise rutile-TiO 2 nanoparticles. The synthesis procedure was iteratively explored to target producing rutile-TiO 2 NPs of less than 20 nm. The resultant optimised procedure is summarised as follows: a total of 30 mL titanium butoxide (Ti(OBu) 4 ) was added dropwise into 30 mL ice-water in a glass beaker under magnetic stirring. The resultant aqueous solution was stirred for 30 min and heated to remove water. A white residue was obtained, sonicated, and then washed several times in distilled water before centrifuging. The product was dried in a laboratory oven at 80 • C for 24 h. 500 mg of the dry powder was annealed in a muffle furnace at 700 • C, ramped for 2 h, and held for 2 h, resulting in the formation of the rutile phase of TiO 2 . The resultant NPs were determined to be 19.7 nm on average. The NPs synthesis procedure is illustrated in Figure 3. The synthesised nanoparticles were surface modified, as presented in the next section.
Energies 2021, 14, x FOR PEER REVIEW 5 of 16 AKD sizing agent emulsion was obtained from Sappi™ Technology Centre (Pretoria, South Africa). The ASA, silane coupling agent, and titanium (IV) butoxide were purchased from Sigma-Aldrich (Pty.) Ltd. (St. Louis, MO, USA). Both chemicals were used as received.

Synthesis of Rutile-TiO2 Nanoparticles
The sol-gel method was used to synthesise rutile-TiO2 nanoparticles. The synthesis procedure was iteratively explored to target producing rutile-TiO2 NPs of less than 20 nm. The resultant optimised procedure is summarised as follows: a total of 30 mL titanium butoxide (Ti(OBu)4) was added dropwise into 30 mL ice-water in a glass beaker under magnetic stirring. The resultant aqueous solution was stirred for 30 minutes and heated to remove water. A white residue was obtained, sonicated, and then washed several times in distilled water before centrifuging. The product was dried in a laboratory oven at 80 °C for 24 hours. 500 mg of the dry powder was annealed in a muffle furnace at 700 °C, ramped for 2 hours, and held for 2 hours, resulting in the formation of the rutile phase of TiO2. The resultant NPs were determined to be 19.7 nm on average. The NPs synthesis procedure is illustrated in Figure 3. The synthesised nanoparticles were surface modified, as presented in the next section.

Surface Modification of the Rutile-TiO2 Nanoparticles
Three portions of 10 g each of TiO2 nanoparticles were dispersed in 32 mL of distilled water and sonicated for 30 minutes. Then, 5 vol/vol%, 3 vol/vol%, and 1 vol/vol% of AKD were added, making samples X, Y, and Z. The same procedure was repeated for ASA, making samples A, B, C. After 35 minutes of stirring, 1.5 vol/vol% silane was added to each mixture to connect the dissimilar compounds together and was stirred for 1 hour. For ASA, alum was added to the solution to further fuse and retain the ASA with the TiO2. Both mixtures were heated for 20 minutes at 60 °C with continued stirring. The obtained samples were centrifuged and rinsed with distilled water to remove the impurities and dried at 80 °C. Figure 4 shows the block illustration of the modification model. The reaction mechanism model is illustrated in Figure 5. The hydrolysis of the silane (Si-OH) makes TiO2 NPs react with the hydrolysable group (Si(OH)3), while the organofunctional group reacts with sizing agents (AKD and ASA).

Surface Modification of the Rutile-TiO 2 Nanoparticles
Three portions of 10 g each of TiO 2 nanoparticles were dispersed in 32 mL of distilled water and sonicated for 30 min. Then, 5 vol/vol%, 3 vol/vol%, and 1 vol/vol% of AKD were added, making samples X, Y, and Z. The same procedure was repeated for ASA, making samples A, B, C. After 35 min of stirring, 1.5 vol/vol% silane was added to each mixture to connect the dissimilar compounds together and was stirred for 1 h. For ASA, alum was added to the solution to further fuse and retain the ASA with the TiO 2 . Both mixtures were heated for 20 min at 60 • C with continued stirring. The obtained samples were centrifuged and rinsed with distilled water to remove the impurities and dried at 80 • C. Figure 4 shows the block illustration of the modification model. The reaction mechanism model is illustrated in Figure 5. The hydrolysis of the silane (Si-OH) makes

Characterisation of the Surface-Modified Rutile-TiO2 Nanoparticles
Phase identification of the sample was conducted using X-ray diffraction (Bruker MeasSrv (D2-205530)/D2205530 diffractometer), as well as with Raman spectroscopy using Bruker Infinity 1 spectrometer (Billerica, MS, USA) fitted with a 50× objective lens for imaging. The nanoparticles shape was identified using transmission electron microscopy (TEM). The Fourier-transform infrared spectroscopy (FT-IR) spectra of the samples were obtained with an attenuated total reflectance (ATR) in the wavelength range of 420-4000 cm −1 using PerkinElmer [55]. Analytical weight balance was used to measure changes in mass of the sample due to moisture absorption and heat. A thermogravimetric analyser (PerkinElmer STA 4000 analyser; Waltham, MA, USA) was used to study the thermal stability of the samples. The various results of each analysis are presented in the next section.

X-ray Diffraction Results
X-ray diffraction (XRD) was used to determine the crystallinity and phase of the synthesised TiO2 NPs. The XRD pattern shows distinct diffraction peaks which match the standard rutile-TiO2 reference peaks. All peaks are in good agreement with the standard spectrum (JCPDS no.: 88-1175). As shown in Figure 6, the diffraction depicts sharpness of the peaks, which indicates highly crystalline nanoparticles [56][57][58].

Characterisation of the Surface-Modified Rutile-TiO2 Nanoparticles
Phase identification of the sample was conducted using X-ray diffraction (Bruker MeasSrv (D2-205530)/D2205530 diffractometer), as well as with Raman spectroscopy using Bruker Infinity 1 spectrometer (Billerica, MS, USA) fitted with a 50× objective lens for imaging. The nanoparticles shape was identified using transmission electron microscopy (TEM). The Fourier-transform infrared spectroscopy (FT-IR) spectra of the samples were obtained with an attenuated total reflectance (ATR) in the wavelength range of 420-4000 cm −1 using PerkinElmer [55]. Analytical weight balance was used to measure changes in mass of the sample due to moisture absorption and heat. A thermogravimetric analyser (PerkinElmer STA 4000 analyser; Waltham, MA, USA) was used to study the thermal stability of the samples. The various results of each analysis are presented in the next section.

X-ray Diffraction Results
X-ray diffraction (XRD) was used to determine the crystallinity and phase of the synthesised TiO2 NPs. The XRD pattern shows distinct diffraction peaks which match the standard rutile-TiO2 reference peaks. All peaks are in good agreement with the standard spectrum (JCPDS no.: 88-1175). As shown in Figure 6, the diffraction depicts sharpness of the peaks, which indicates highly crystalline nanoparticles [56][57][58].

Characterisation of the Surface-Modified Rutile-TiO 2 Nanoparticles
Phase identification of the sample was conducted using X-ray diffraction (Bruker MeasSrv (D2-205530)/D2205530 diffractometer), as well as with Raman spectroscopy using Bruker Infinity 1 spectrometer (Billerica, MS, USA) fitted with a 50× objective lens for imaging. The nanoparticles shape was identified using transmission electron microscopy (TEM). The Fourier-transform infrared spectroscopy (FT-IR) spectra of the samples were obtained with an attenuated total reflectance (ATR) in the wavelength range of 420-4000 cm −1 using PerkinElmer [55]. Analytical weight balance was used to measure changes in mass of the sample due to moisture absorption and heat. A thermogravimetric analyser (PerkinElmer STA 4000 analyser; Waltham, MA, USA) was used to study the thermal stability of the samples. The various results of each analysis are presented in the next section.

X-ray Diffraction Results
X-ray diffraction (XRD) was used to determine the crystallinity and phase of the synthesised TiO 2 NPs. The XRD pattern shows distinct diffraction peaks which match the standard rutile-TiO 2 reference peaks. All peaks are in good agreement with the standard spectrum (JCPDS no.: 88-1175). As shown in Figure 6, the diffraction depicts sharpness of the peaks, which indicates highly crystalline nanoparticles [56][57][58]. The average crystallite size of the obtained rutile-TiO2 nanoparticles was calculated to be 19.7 nm. The size was determined by the use of the Scherrer equation (Equation 1) [56] as given in Equation (1).

Raman Spectroscopy Results
Raman spectroscopy was used to characterise the obtained rutile-TiO2 NPs. Four random spots were scanned on the specimen. The spectrum in Figure 7 shows the Raman peak shifts for the rutile-TiO2 NPs. The rutile-TiO2 has four characteristic phonon modes at 143 ± 2 cm −1 (B1g), 445 ± 5 cm −1 (Eg), 610 ± 1 cm −1 (A1g), and 235 ± 5 cm −1 . From Figure 7, it is evident that rutile exhibits dominant peaks at 448.16 cm −1 and 611.04 cm −1 . These peaks are attributed to Eg and A1g, active modes of rutile-TiO2 correspondently. The third-most active optical phonon mode of rutile from the graph is the 144.06 cm −1 which is attributed to B1g. The peak at 825.16 cm −1 is the weakest and poorly observed. This peak is assigned B2g. A visible broadband at 235.76 cm −1 , as observed on the graph, does not correspond with any fundamental modes of the rutile phase. This may be attributed to the disorderinduced scattering or second-order effect (SOE) as a consequence of multi-phonon processes [59][60][61]. The average crystallite size of the obtained rutile-TiO 2 nanoparticles was calculated to be 19.7 nm. The size was determined by the use of the Scherrer equation (Equation (1)) [56] as given in Equation (1).

Raman Spectroscopy Results
Raman spectroscopy was used to characterise the obtained rutile-TiO 2 NPs. Four random spots were scanned on the specimen. The spectrum in Figure 7 shows the Raman peak shifts for the rutile-TiO 2 NPs. The rutile-TiO 2 has four characteristic phonon modes at 143 ± 2 cm −1 (B 1g ), 445 ± 5 cm −1 (E g ), 610 ± 1 cm −1 (A 1g ), and 235 ± 5 cm −1 . From Figure 7, it is evident that rutile exhibits dominant peaks at 448.16 cm −1 and 611.04 cm −1 . These peaks are attributed to E g and A 1g , active modes of rutile-TiO 2 correspondently. The third-most active optical phonon mode of rutile from the graph is the 144.06 cm −1 which is attributed to B 1g . The peak at 825.16 cm −1 is the weakest and poorly observed. This peak is assigned B 2g . A visible broadband at 235.76 cm −1 , as observed on the graph, does not correspond with any fundamental modes of the rutile phase. This may be attributed to the disorder-induced scattering or second-order effect (SOE) as a consequence of multi-phonon processes [59][60][61]. The average crystallite size of the obtained rutile-TiO2 nanoparticles was calculated to be 19.7 nm. The size was determined by the use of the Scherrer equation (Equation 1) [56] as given in Equation (1).

Raman Spectroscopy Results
Raman spectroscopy was used to characterise the obtained rutile-TiO2 NPs. Four random spots were scanned on the specimen. The spectrum in Figure 7 shows the Raman peak shifts for the rutile-TiO2 NPs. The rutile-TiO2 has four characteristic phonon modes at 143 ± 2 cm −1 (B1g), 445 ± 5 cm −1 (Eg), 610 ± 1 cm −1 (A1g), and 235 ± 5 cm −1 . From Figure 7, it is evident that rutile exhibits dominant peaks at 448.16 cm −1 and 611.04 cm −1 . These peaks are attributed to Eg and A1g, active modes of rutile-TiO2 correspondently. The third-most active optical phonon mode of rutile from the graph is the 144.06 cm −1 which is attributed to B1g. The peak at 825.16 cm −1 is the weakest and poorly observed. This peak is assigned B2g. A visible broadband at 235.76 cm −1 , as observed on the graph, does not correspond with any fundamental modes of the rutile phase. This may be attributed to the disorderinduced scattering or second-order effect (SOE) as a consequence of multi-phonon processes [59][60][61].

Transmission Electron Microscopy (TEM) Results
Transmission electron microscopy (TEM) was used to examine the samples' morphology and size. As shown in Figure 8, the particles are agglomerated; therefore, particle size distribution cannot be ascertained from the images obtained. However, as earlier reported in the XRD analysis using the Scherrer equation, the average size of nanoparticles was calculated as 19.7 nm. The agglomeration obtained in the particles could be due to the high temperature used during calcination (which exists between the transition of the two polymorphous of TiO 2 ). Additionally, for small particles, Van der Waals attraction force that is present between the NPs leads to agglomeration [60][61][62][63]. From the TEM images, hexagonal-like shape is evident. To reduce agglomeration, high shear force mixing, ball milling, or ultrasonication has to be employed for mixing with the host polymer such as the case of kraft paper [64].

Transmission Electron Microscopy (TEM) Results
Transmission electron microscopy (TEM) was used to examine the samples' morphology and size. As shown in Figure 8, the particles are agglomerated; therefore, particle size distribution cannot be ascertained from the images obtained. However, as earlier reported in the XRD analysis using the Scherrer equation, the average size of nanoparticles was calculated as 19.7 nm. The agglomeration obtained in the particles could be due to the high temperature used during calcination (which exists between the transition of the two polymorphous of TiO2). Additionally, for small particles, Van der Waals attraction force that is present between the NPs leads to agglomeration [60][61][62][63]. From the TEM images, hexagonal-like shape is evident. To reduce agglomeration, high shear force mixing, ball milling, or ultrasonication has to be employed for mixing with the host polymer such as the case of kraft paper [64].

Fourier-Transform Infrared Spectroscopy (FT-IR)
The FT-IR was recorded at room temperature using a PerkinElmer Fourier-transform-attenuated total reflectance-infrared spectrometer (FT-UATR-IR-2) in the range from 420-4000 cm −1 . As shown in Figure 9a, the spectrum of the obtained sample (rutile-TiO2) was examined and compared with what is in the literature. The appearance of a band in the region 531 cm −1 of the TiO2 NPs spectrum indicates the presence of metal to oxygen bonding which corresponds to O-Ti-O bonding [64][65][66][67][68][69]. Compared with what is available in the literature, the characteristic of -OH group (3200-3650 cm −1 ) is not obvious in the TiO2 spectrum [69][70][71][72][73]. This could be due to the annealing temperature and time, leading to the removal of hydrolysable surface group, thereby fabricating ligand-free TiO2 NPs [74][75][76].
The FT-IR of the modified TiO2 NPs was conducted to examine the spectra of each sample. Figure 9b shows the FT-IR spectra of pure AKD. From the plot, a weak and broad peak is evident between 3041 and 3565 cm −1 which is characteristic of (OH) vibrations in the AKD spectrum [77]. Comparing Figure 9b,c of the modified TiO2 NPs, the modified TiO2 NPs show the presence of AKD features on its structure. The double bond carboncarbon stretching vibration-absorption peak at 1708 cm −1 relates to the absorption band characteristic peak 1846 and 1633 cm −1 in the spectrum of AKD which usually correspond to C=O and C=C. However, there was an absence of OH frequency absorption band in the modified TiO2 NPs, suggesting the dehydration of water molecules in the complexes. The appearance of double bond carbon stretching at 1708 cm −1 is very weak (almost not existing). This indicates that the functional group is very symmetrical but does not imply that the functional group is absent.
The adsorption bands at 2849 and 2916 cm −1 of the modified TiO2 NPs are from -CH2symmetric and asymmetric stretching vibration [48]. The spectra of the modified TiO2 NPs with AKD, therefore, shows that new absorption bands are present which is evidence of

Fourier-Transform Infrared Spectroscopy (FT-IR)
The FT-IR was recorded at room temperature using a PerkinElmer Fourier-transformattenuated total reflectance-infrared spectrometer (FT-UATR-IR-2) in the range from 420-4000 cm −1 . As shown in Figure 9a, the spectrum of the obtained sample (rutile-TiO 2 ) was examined and compared with what is in the literature. The appearance of a band in the region 531 cm −1 of the TiO 2 NPs spectrum indicates the presence of metal to oxygen bonding which corresponds to O-Ti-O bonding [64][65][66][67][68][69]. Compared with what is available in the literature, the characteristic of -OH group (3200-3650 cm −1 ) is not obvious in the TiO 2 spectrum [69][70][71][72][73]. This could be due to the annealing temperature and time, leading to the removal of hydrolysable surface group, thereby fabricating ligand-free TiO 2 NPs [74][75][76].
The FT-IR of the modified TiO 2 NPs was conducted to examine the spectra of each sample. Figure 9b shows the FT-IR spectra of pure AKD. From the plot, a weak and broad peak is evident between 3041 and 3565 cm −1 which is characteristic of (OH) vibrations in the AKD spectrum [77]. Comparing Figure 9b,c of the modified TiO 2 NPs, the modified TiO 2 NPs show the presence of AKD features on its structure. The double bond carboncarbon stretching vibration-absorption peak at 1708 cm −1 relates to the absorption band characteristic peak 1846 and 1633 cm −1 in the spectrum of AKD which usually correspond to C=O and C=C. However, there was an absence of OH frequency absorption band in the modified TiO 2 NPs, suggesting the dehydration of water molecules in the complexes. The appearance of double bond carbon stretching at 1708 cm −1 is very weak (almost not existing). This indicates that the functional group is very symmetrical but does not imply that the functional group is absent.
The adsorption bands at 2849 and 2916 cm −1 of the modified TiO 2 NPs are from -CH2symmetric and asymmetric stretching vibration [48]. The spectra of the modified TiO 2 NPs with AKD, therefore, shows that new absorption bands are present which is evidence of the presence of new chemical bonds formed through modification. There is a reaction and connection between TiO 2 NPs and AKD. results in the appearance of a band at 1689 cm −1 . The band corresponds to the stretchin vibration of the C=O bond [78][79][80]. The absence of the bands at 1863 to 1778 cm −1 indicate that the composite is free of unreacted modifier (ASA).The adsorption bands at 2849 an 2919 cm −1 of the modified TiO2 NPs are also from -CH2-symmetric and asymmetr stretching vibration [48]. Figure 9e, therefore, demonstrates that new adsorption band and chemical bonds are formed after modification. The FT-IR spectrum of ASA is shown in Figure 9d. The spectrum peaks for ASA arising from anhydride carbonyl stretching were identified at 1778 cm −1 and 1863 cm −1 which correspond with the literature. The 1778 cm −1 represents the stretching vibration of C=O which conforms with 1785 ± 5 cm −1 . At 916 cm −1 , the stretching vibration band of five-membered cyclic anhydrides (succinic anhydride) is identified. In Figure 9e which characterises the modified TiO 2 NPs with ASA, there is an absence of succinic anhydride grouping (1778 cm −1 and 1863 cm −1 ). This is due to the hydrolysis of the anhydride which results in the appearance of a band at 1689 cm −1 . The band corresponds to the stretching vibration of the C=O bond [78][79][80]. The absence of the bands at 1863 to 1778 cm −1 indicates that the composite is free of unreacted modifier (ASA).The adsorption bands at 2849 and 2919 cm −1 of the modified TiO 2 NPs are also from -CH2-symmetric and asymmetric stretching vibration [48]. Figure 9e, therefore, demonstrates that new adsorption bands and chemical bonds are formed after modification.

Moisture Absorption Characteristics
Each sample of the surface-modified TiO 2 NPs, including the unmodified, was weighed using an analytical weigh balance. Then, 10 mL of distilled water was dropped in each sample and was exposed to air for 3 days. The experiment was triplicated; the average values and the error deviations are summarised in Table 1. The unmodified rutile-TiO2 NPs (sample 'T'), after having been exposed to airabsorbed moisture, had a weight increase of 3.88%. In contrast, the modified samples generally lost weight by releasing water into the atmosphere. These results demonstrate the hydrophilic nature of TiO 2 and the effect of sizing agents on reducing the hydrophilicity behaviour of TiO 2 .

Thermal-Stress-Induced Weight Loss of Surface-Modified Rutile-TiO 2 NPs
To determine thermal-stress-induced weight loss characteristics, 300 mg portions of each sample were measured separately and placed in a furnace. The furnace temperature was set to ramp up to 400 • C in 20 min and then stayed constant for 40 min. The samples were then placed in a desiccator with silica gel to avoid moisture absorption as the samples cooled off. Analytical weight balance was used to measure the change in mass of each sample. This experiment was triplicated; Table 2 shows the average value of the weight loss and error deviation of each case. From the table, the surface-modified samples lost more weight than the unmodified sample, and the amount of weight loss is directly proportional to the percent content of the sizing agent applied on each sample. The weight loss of the composites can be attributed to the breakage or disintegration of the bond between organic and inorganic compounds [81]. This indicates that the modified samples contain more than one compound. In essence, the weight loss under heating is predominated by the decomposition of the surfactants. This phenomenon is further confirmed in the thermogravimetric analysis, as presented in the next section.

Thermogravimetric Analysis (TGA) Results
The Thermogravimetric analysis (TGA) of the samples was conducted under nitrogen gas. From the graph of pure TiO 2 NPs (Figure 10a), a rapid loss of moisture occurred between 75 and 90 • C, followed by a significant degradation at about 186 • C which is attributed to the decomposition of the TiO 2 [45,82]. The initial steep gradient (representing a rapid loss of moisture) and the subsequent steep depression on the graph further confirm the results obtained in Table 1 about moisture absorption of the unmodified sample. As for the 5% AKD-modified rutile-TiO 2 NPs, there is drastic thermal decomposition at 270 • C which is about 45% higher than that of pure rutile-TiO 2 NPs. For the reduced AKD content of 3%, the decomposition of modified rutile-TiO 2 NPs occurs at a lower temperature of 261 • C which is about 40% higher than that of the pure rutile-TiO 2 NPs. A further reduction of AKD to 1% resulted in the modified rutile-TiO 2 NPs decomposing drastically at 253 • C which is about 36% higher than that of pure rutile-TiO 2 NPs. For AKD, in all cases of surface medication of rutile-TiO 2 NPs, there is increased thermal stability of at least 36%. In the case of ASA modified NPs, the temperatures at which the modified NPs drastically decomposed were 211 • C, 194 • C, 187 • C for 5%, 3% and 1% ASA content, respectively. It is worth noting that for all cases of ASA modification, the temperature at which drastic decomposition occurred was lower than the least % AKD content modified sample. This shows that AKD modified samples are more thermally stable than the ASA. However, the ASA-modified samples show improvement in thermal stability except for 1% content whose thermal stability is the same as that of the pure rutile-TiO 2 NPs. Table 3 summarises the effects of surfactant content and type on the thermal stability of the resultant modified samples. The decomposition of the modified rutile-TiO 2 NPs samples at higher temperatures could be attributed to improved heat resistance associated with the long-chain organic compounds from surfactants [57,79,[82][83][84]. Both cases of the modified TiO 2 NPs show that the modified samples have better thermal stability than pure rutile-TiO 2 NPs. which drastic decomposition occurred was lower than the least % AKD content modified sample. This shows that AKD modified samples are more thermally stable than the ASA. However, the ASA-modified samples show improvement in thermal stability except for 1% content whose thermal stability is the same as that of the pure rutile-TiO2 NPs. Table  3 summarises the effects of surfactant content and type on the thermal stability of the resultant modified samples.

Conclusions
A protocol to synthesise 19.7 nm rutile-TiO 2 NPs was proposed and presented in this work. It was further demonstrated that the moisture absorption characteristics, as well as thermal stability of the nanoparticles, can be improved through surface modification of the rutile-TiO 2 NPs. Two types of surfactants were investigated: AKD and ASA. In the case of AKD surface-modified rutile-TiO 2 NPs, the FT-IR spectra indicate the presence of new functional groups. Improvement in terms of rate of moisture absorption (within an approximate deviation of ±0.14%) and thermal stability were determined in surface-modified rutile-TiO 2 NPs. For ASA surface-modified rutile-TiO 2 NPs, there was improvement observed in the rate of moisture absorption and thermal stability except for the least % content of ASA sample that behaved similar to the unmodified rutile-TiO 2 NPs. Possible applications of the more thermally stable and hydrophobic modified rutile-TiO 2 can be explored in future works.