Transformer oil based multi-walled carbon nanotube–hexylamine coolant with optimized electrical, thermal and rheological enhancements

Abstract

In a one-pot microwave-assisted method, multi-walled carbon nanotubes (MWNT) are functionalized with hexylamine (HA). Based on FT-IR, TGA-DTG, Raman, EDS, CHNS/O and TEM results, the functionalization of MWNT with HA was confirmed . The effect of microwave-assisted functionalized MWNT with HA charged transformer oil at different concentrations was experimentally investigated for electrical, thermal and rheological properties. According to the breakdown voltage, flash point, density, electrical conductivity, thermal parameters and viscosity results of the synthesized transformer oil-based coolants they could be an appropriate alternative for different transformers operating at a nominal voltage less than 170 kV.

---

1. Introduction

Different common oils such as engine oil, transformer oil (TO), turbine oil etc. illustrate weak thermal properties, so many of their applications cannot be entirely realized.^1,2 In order to synthesize TO with good voltage insulation and power apparatus cooling and/or to increase the thermal and dielectric properties as well, numerous researchers have employed different kinds of methods. It is well known that TO can be considered as a dielectric liquid with both insulating and coolant properties. As mentioned above, the low thermal conductivity of TO plays a key role in decreasing the transformer performance.³

In the field of transformer technology, a higher thermal conductivity of TO means a higher heat transfer rate and smaller equipment size, implying a longer lifetime and performance.² In order to reach the abovementioned purpose, an innovative idea was to improve dielectric coolants.⁴ These type of fluids are synthesized by adding different nanostructures to TO, with the aim of increasing the insulating and thermal properties of oil.⁴ A number of studies demonstrated that about 5% growth in the thermal conductivity of TO can result in considerable cost saving to the electrical power generation and transmission distribution infrastructure.⁵ Despite the idea about the fluids including solid particles with higher thermal conductivities than base fluids being reported a century ago, the innovative concept for synthesizing coolants was introduced by Dr Choi and his team in 1995.⁶

Recently, researchers have studied the performance of transformer oil-based coolant nanofluids of different nanoparticles such as magnetite nanoparticles, Al₂O₃.^7–9 The results commonly suggested that a nanoparticle-based transformer oil coolant can be utilized to increase the cooling of a power transformer’s core. It is obvious that the thermal conductivity of the nanostructures may play an important role in the suspension’s thermal conductivity. In the field of coolants, a higher extent of thermal conductivity of the nanoparticles means higher thermal conductivity of the coolants as well as the higher heat transfer coefficient.^10–12

Among different nanostructures, carbon nanotubes (CNT) with unique thermo-physical, electrical, and mechanical properties can be selected as promising materials.^13–16 On the other hand, various applications of CNTs cannot be fully realized due to weak dispersion in solvents and feeble interactions with other materials, which result from strong intertube van der Waals interactions.^17,18 To increase the interactivity of CNTs, surface functionalization was suggested as an efficient and common approach. Therefore, different functionalities such as carboxylic groups,^2,19 aminoacids^17,20,21 etc. have been used for various applications. Aravind et al.¹⁹ functionalized CNTs with carboxylic groups to enhance their water dispersibility. Also, Zeinali et al.² experimentally investigated the effects of carboxylated CNT on TO. Eight effective parameters were investigated and their results reported a drop in the breakdown voltage of TO when the carboxylated CNTs were applied.²

It is obvious that the acidity of transformer oil is a destructive property. According to a previous study,⁶ the acidity of TO increases with an increase of water content in the oil and it becomes more soluble to the oil, which ultimately deteriorates its insulation property.^22,23 So, despite the use of carboxylic groups on CNT surfaces it can also be used to increase dispersibility in different media, thus it cannot be a good choice for TO as an additive. In order to realize excellent dispersion of CNTs in different media, a plethora of carboxylic groups must be decorated on the surface of the CNTs.²⁴ This process first produces numerous defects on the CNT surface and then destroys the main structure creating promising properties with the shortening of their lengths.²⁴

Choi et al.²⁵ synthesized and investigated the dispersions of Al₂O₃ and AlN nanoparticles in TO with an oleic acid surfactant. In order to synthesize a new kind of transformer oil-based coolant, they applied ceramic nanoparticles to enhance the thermal conductivity of TO and the electric insulation property. An 8% improvement in the thermal conductivity and 20% in the overall heat transfer coefficient were observed at an AlN volume concentration of 0.5%. In addition, the viscosity of the transformer oil-based coolant plays a key role in the performance of other required instruments such as the pump and it has a direct connection with required power through the cooling systems.

Chen et al.²⁶ studied the influence of different basefluids such as silicone oil, glycerol, and water and they observed Newtonian behavior at various nanoparticle concentrations and temperatures. Ahmadi et al.²⁷ investigated the influence of CNT on engine oil coolants and their viscosities at various concentrations. They observed about 13% and 3.3% increases in the flash point and pour point of coolants from 0.1 weight% of CNT over base oil.

In the current study, multi-walled carbon nanotubes (MWNTs) were first functionalized with hexylamine (HA) in one-pot by applying a green procedure under microwave irradiation. In order to realize good dispersibility in TO, a decrease in defects and the removal of the acidic property of TO, the proposed method had no acid treatment phase, which is the general step in previous studies. The functionalization phase was investigated via different characterization methods. In the application phase, the HA functionalized MWNT (MWNT–HA) was added to the basefluid and its properties such as, breakdown voltage, density, flash point, pour point, and transformer performance were investigated at two weight concentrations.

2. Experimental section

2.1. Materials

Pristine MWNTs of diameter less than 30 nm, length 5–15 µm and purity more than 95% were obtained from Shenzhen Nano-Tech Port Co. Pristine MWNTs have the properties of low electrical conductivity and suitable thermal conductivity, which could be selected as one of the best choices for this study as per information obtained from relevant companies. All the chemical materials were of analytical grade such as methanol, NaNO₂, N,N-dimethylformamide (DMF), and H₂SO₄ were obtained from Merck Inc. In addition, hexylamine (HA) was purchased from Sigma-Aldrich Co. and the mineral TO was obtained from Nynax, Stockholm, Sweden.

2.2. Functionalization of the MWNTs and preparation of the coolant

Functionalization of the MWNTs with diamine groups as an innovative method was reported by the present researchers³⁰ and Ellison et al.^28,29 Furthermore, Bahr et al.³⁰ and Price et al.³¹ applied NaNO₂ for preparing a semi-stable diazonium ion, which resulted in a radical reaction with CNTs. Accordingly, the above-mentioned method was followed with a slight modification under microwave irradiation was applied for functionalization of the MWNTs with HA. Pristine MWNTs (200 mg), HA (20 ml) and NaNO₂ (200 mg) were poured into a vessel and sonicated for 2 hours at 50 °C to form a uniform suspension. About 0.5 ml of H₂SO₄ was simultaneously added drop by drop into the aforementioned suspension to accomplish the diazonium reaction. The suspension was then transferred into a Teflon vessel and was exposed to microwave radiation for 30 min at 120 °C and 700 W power. The solution was cooled to room temperature, centrifuged with THF and methanol, and easily washed by vacuum filtration with a PTFE membrane. After washing several times with DMF, THF and methanol and removing any unreacted materials by checking the pH of the filtrate, it was vacuum dried for 48 h at 40 °C.

To synthesize the MWNT–HA based transformer oil coolant of different weight concentrations, the HA treated MWNT (MWNT–HA) was sonicated with known amounts of TO for 30 min. Due to the good dispersibility of HA in the oil phase, the MWNT–HA could achieve an appropriate dispersion in TO media. The easily-miscible hexylamine molecules are responsible for better dispersion of MWNT–HA. The synthesized transformer oil-based coolants of different weight concentrations, such as 0.001% and 0.005% were prepared. Some essential properties of TO such as viscosity, density, flash point, pour point, breakdown voltage, electrical conductivity, and thermal conductivity are shown in Table S1 (ESI†).

2.3. Experimental apparatus

A schematic diagram of the heat transfer apparatus is shown in Fig. 1. The experimental transformer was planned and constructed based on an actual transformer. There is a good correlation between the size and materials of the experimental set-up and the oil-25 kVA transformer.² The experimental transformer was oil based and included a reservoir with the dimensions 203 × 100 × 221 mm³.

Fig. 1 Schematic of experimental transformer.

A wire cylindrical heating element was employed to heat the oil and coolants. 4 PT-100 thermocouples with the accuracy of ±0.1 °C were installed in the reservoir of fluid to measure the average temperature of the fluid. Moreover, four thermocouples were fixed at 4 different sides of the transformer’s wall to calculate the average temperature of the walls. Meanwhile, at an actual distance of 5 cm of the bigger wall, a 55 W blower was employed vertically to cool the big wall and introduce forced convection heat transfer.

The ammeter, thermocouples, and the voltmeter had highest precision data of 0.001 A, 0.1 °C, and 0.1 V, respectively. The uncertainty of the experiments were calculated using the Holman method³² and the maximum uncertainty in calculating the heat transfer coefficient and Nusselt number were less than 2.1 and 5.4%, respectively.

Raman spectroscopy (Renishaw confocal spectrometer at 514 nm), thermogravimetric analysis, TGA, (TGA-167 50 Shimadzu), and transmission electron microscopy, TEM, (HT7700, High-Contrast/High-Resolution Digital TEM) were used to analyze samples. Regarding TGA, the mass loss of samples was measured at a heating rate of 10 °C min⁻¹ in air. The preparation of TEM samples consisted of several steps, such as the sonication of MWNTs in ethanol, dropping the sample on a lacey carbon grid followed by drying under vacuum. Electrical conductivity of the samples was measured using an electrical conductivity meter by using two coaxial electrodes, which are the inner electrode and outer electrode. To conduct the measurement, a specific amount of sample was injected between the two electrodes. When the voltage (V) was applied to the electrodes, there would be a current (I) flowing through the sample. Then, the electrical conductivity of the sample was measured using eqn (1)³³

where, σ, L and S are the electrical conductivity, spacing between the electrodes and the effective area of the electrodes respectively.

2.4. Data processing

In order to investigate the influence of MWNT–HA on the thermal properties of TO in a transformer, the natural & forced heat transfer coefficient (h), Nusselt number (Nu), breakdown voltage, flash point, pour point, density, electrical and thermal conductivities and viscosity should be studied as the key parameters.

Parameters (h & Nu) were obtained from eqn (2)–(4):

First, the rate of input power was calculated by eqn (2).

Q = VI (2)

Here, it is obvious that thermal energy transfer is between the hot wall and the cold walls. Thus, the average heat transfer coefficient (h) is determined by eqn (3)

h = Q/A(T_h − T_c) (3)

where, Q, T_c and T_h are heat flux, wall temperature and fluid bulk temperature respectively. Meanwhile, A is the heat transfer area which was considered as 0.13 m². T_c and T_h were considered as the average temperatures obtained from the thermocouples placed on the inner body of the chamber.

The amount of Nu can be calculated from eqn (4)

Nu = hL/k (4)

where, L (0.05 m) is the distance between the cold and hot walls. Also, k is the thermal conductivity of MWNT–HA/TO coolants at the specific temperature and concentration.

Meanwhile, the breakdown voltage, flash point, pours point, density, electrical conductivity, thermal conductivity and viscosity were obtained experimentally. A Brookfield LVDV-III rheometer was used to measure the amounts of viscosity. A KD₂ thermal analyzer (Decagon Devices, Inc., USA) was used to determine the thermal conductivity of coolants and TO at various temperatures. Also, the thermal conductivity reported in this study is the average of 4 replicated measurements with the error value less than 0.0016.

3. Results and discussion

3.1. Functionality

In order to attain a one-pot and rapid functionalization method, less defective structure and selecting functional groups without acidity property, the microwave procedure based on a radical diazonium reaction was applied. Based on recent studies by the present authors,³⁴ the mentioned radical diazonium ions were in a semi-stable form, which easily reacts with MWNTs. Since HA has no acidity property, there was no change in the acidity of TO after addition of MWNT–HA. In addition, the functionalization process was confirmed by some characterization instruments such as derivative thermogravimetric analysis (DTG), Raman and FT-IR spectroscopy, and transmission electron microscopy (TEM).

As discussed above, the covalent functionalization has been performed by the formation of a semi-stable diazonium ion which initiated a radical reaction between the nanotubes and HA chains. FT-IR spectroscopy has been used as one of the best ways to obtain evidence, which is illustrated in Fig. 2 panel (a). In contrast to the pristine sample, the FT-IR spectrum of MWNT–HA demonstrates a couple of peaks at the ranges of 2850–3000 cm⁻¹, indicating the presence of C–H stretching vibrations.³⁵ Also, the peaks at 1396 cm⁻¹ and 1460 cm⁻¹, arise from the bending vibrations of the CH₂ group.³⁵ Formation of the aforementioned peaks after functionalization along with the lack of amine peaks can depict the diazonium reaction between the MWNTs and HA.^29,34 More importantly, the peak at 1492 cm⁻¹ was associated with the stretching vibrations of C=C, which is due to disruption of aromatic π-electrons on the MWNTs surface.

Fig. 2 (a) FT-IR spectra, (b) Raman spectra, (c) thermogravimetric analysis (TGA) and derivative thermogravimetric (DTG) curves of the pristine and MWNT–HA, (d–f) TEM images of MWNT–HA.

Raman characterization is a strong measurement for analyzing structure and sp² and sp³ hybridized carbon atoms in carbon-based materials and functionalization by following alterations in holes.^36–39 The Raman spectra of the pristine MWNT, and MWNT–HA are presented in Fig. 2 panel (b). While the pristine MWNTs are weak in terms of D band intensity, the fairly strong D band in the MWNT–HA sample can be seen at 1343 cm⁻¹. The ratio of the intensities of the D-band to that of the G-band (I_D/I_G) was considered to be the amount of disordered carbon (sp³-hybridized carbon) relative to graphitic carbon (sp²-hybridized carbon).³⁹ In functionalization studies of MWNT, the higher intensity ratio of I_D/I_G indicates a higher disruption of aromatic π–π electrons, implying partial damage of the graphitic carbon produced by covalent functionalization.^39,40 The I_D/I_G ratio of MWNT–HA (0.91) is relatively higher than that of pristine MWNTs (0.49), which confirmed the successful functionalization via a diazonium reaction under microwave irradiation. A significant increase in I_D/I_G can also confirm that the present method is completely successful for functionalization of MWNT without acid-treatment phase.

As further evidence, thermogravimetric analysis was conducted to investigate functionalization of MWNTs with HA. TGA is a technique of thermal analysis in which alterations in the structure of materials are measured as a function of temperature. Also, TGA presents results about chemical functionalization of MWNTs with different functional groups.^41,42

Fig. 2 panel (c) presents the TGA and derivative thermogravimetric (DTG) curves of the pristine MWNTs and MWNT–HA. It can be seen that the TGA results of the pristine sample illustrate no mass loss up to 500 °C. However, there is an obvious weight loss in the temperature range of 50–150 °C in the HA-MWNT curve. This mass loss was attributed to the functionality of HA as an unstable organic part on the surface of the MWNTs. Also, it is obvious that the DTG curve of the pristine MWNTs shows no phase of degradation up to 500 °C and a noticeable step of mass loss in the temperature range of 500–730 °C is attributed to the bulk degradation temperatures of the main graphitic structures. In addition to the first step of mass loss, the first step of degradation in the DTG curve of the MWNT–HA in the temperature range of 50–150 °C is associated with the bulk degradation temperatures of HA as an unstable organic loaded on the surface of the MWNTs.

The degree of functionalization of modified MWNTs is generally calculated by means of thermogravimetric analysis (TGA). The functionalization degree of MWNT–HA was calculated in order to assess the functionalization yield of the proposed method. Eqn (5) is used to measure the functionalization degree from TGA results:⁴³

While pristine MWNTs lost only 1.01% of mass up to 500 °C, the TGA curve of the HA functionalized MWNT displays a considerable mass loss of 11.1% in the range of 50–200 °C corresponding to the decomposition of the attached HA moiety. On the basis of TGA weight loss, we estimated that the amount of functional groups per gram (mmol g⁻¹) of MWNT–HA is 1.4661.

Overall, TGA or DTG, FT-IR and Raman results are in good agreement with each other regarding functionalization of MWNTs with HA. Utilizing a microwave method and semi-stable diazonium ions simultaneously provided a very effective route to attack the main graphitic structures of MWNTs.

3.2. Morphological analysis

TEM analysis was carried out to study the morphological structure of MWNTs. Fig. 2 panels (d–f) show the TEM images of the MWNT–HA. As a first indication of functionalization, there are some MWNTs with open tips, which is a result of the radical reaction of the diazonium ions with MWNTs. In addition, the functionalization procedure alters the graphitic (sp²) carbon network to sp³ hybridized states of carbon, which is in a good agreement with more defects on the MWNT surface after treatment. Thus, more surface roughness in the MWNT–HA images is logical. Even though TEM images illustrate no indication of functional groups, the more surface deterioration of treated MWNTs with HA can be considered as evidence to support this.¹³ These morphological changes are in agreement with the DTG, FT-IR and Raman results.

3.3. Elemental analysis

To confirm the diazonium coupling, elemental analyses were performed by the energy dispersive X-ray spectroscopy (EDS) and combustible elemental analysis (CHNS/O). EDS spectra of pristine MWNTs and MWNT–HA (Fig. 3) show a trace amount of Ni, O and Cd in their chemical structures. The trace amounts of Cd and Ni are mainly due to trapped catalysts during the synthesis of the pure MWNTs. The EDS results are presented in Table 1 to ascertain accurately the exact amount of each element. EDS analysis shows an increase of oxygen content and the presence of a very little amount of nitrogen in the MWNT–HA. Note that the MWNT–HA showed high purity except a very small amount of nitrogen and oxygen. It is noteworthy that our attached functional group has just C and H elements. To confirm the results obtained by EDS and to obtain information on the hydrogen content, combustible elemental analysis was performed. Combustible elemental analysis (CHNS/O) was performed using a Perkin Elmer 2400 series II and was used to measure carbon, oxygen, nitrogen and hydrogen elemental content. The instrument was used in CHN operating mode to convert the sample elements to simple gases (CO₂, H₂O and N₂). The PE 2400 analyzer automatically performed combustion, reduction, homogenization of product gases, separation and detection. The results of the CHNS/O analysis are shown in Table 2. The results are in very good agreement with those obtained by the EDS results. As more evidence, the CHNS/O results show that H content substantially increased in MWNT–HA, which further confirms that hexane has been successfully grafted onto MWNTs. Also, no trace of nitrogen was observed in the MWNT–HA, which confirm a successful diazonium coupling.

Fig. 3 EDS spectra of pristine MWNTs and MWNT–HA.

Table 1 EDS results for pristine MWNTs and MWNT–HA

Element symbol	Pristine MWNTs	MWNT–HA
C	93.1 ± 0.7	90.2 ± 0.9
O	2.9 ± 1.1	3.2 ± 0.5
N	0.0	0.1 ± 0.9

---

Table 2 Elemental composition of pristine MWNTs and MWNT–HA

Sample	Chemical composition (wt%)
	C	N	O	H
Pristine MWNT	90.23	0.00	2.54	1.09
MWNT–HA	89.85	0.09	2.41	3.28

---

3.4. Stability analysis

Fig. 4 panel (a) illustrates the UV-vis spectra of the MWNTs for weight concentrations of 0.001 and 0.005. UV-vis spectroscopy is commonly applied for investigation of the stability of coolants including solid nanoparticles and is able to measure the sedimentation time. According to the Beer–Lambert law, the absorbance of a solution is directly proportional to the concentration of the absorbing species such as particles in the solution. As a raw spectrum of MWNT–HA/TO coolant, the sharp peak at 275 nm is attributed to the presence of MWNTs, and the intensity of the peak increases with weight concentration. Quantitative analysis of the dispersion state and the long-term stability of the MWNT and MWNT–HA/TO coolants was performed using UV-vis spectroscopy, as shown in Fig. 4 panel (b). Thus, the absorbance at a wavelength of 275 nm was measured over 30 days for both weight concentrations. It can be seen that the relative concentration of MWNT–HA decreases insignificantly over time. As a result, a maximum sediment of 26.1% was obtained for the highest weight concentration of 0.005, which confirmed the suitable dispersibility of MWNT–HA in TO.

Fig. 4 (a) UV-vis spectrum of MWNT–HA/TO coolant at different concentrations, (b) the colloidal stability of MWNT–HA in TO as a function of time and weight concentration and (c) the UV-vis spectrum of the MWNT–HA for weight concentrations of 0.001 and 0.005 mechanically shaken for 5 min every day.

To check whether the decrease in signal intensity over time is just due to reversible particle settling or gradual irreversible particle aggregation, Fig. 4 panel (c) was plotted. The UV-vis spectrum of MWNT–HA for weight concentrations of 0.001 and 0.005 obtained by performing UV measurements for 30 days on suspensions mechanically shaken every day. The sonication was performed for 5 min at room temperature with a power rating of 390 W. The measurement was again carried out at the peak wavelength of each material to trace the alteration in intensity which can be further used to describe the suspension stability at both weight fractions of 0.001% and 0.005%. It can be seen that both colloidal mixtures show a lower than 1% decrease in the relative concentration as time progressed, indicating that the colloidal suspensions have remained stable for 1 month without irreversible particle aggregation.

The particle size distribution change was analyzed using the dynamic light scattering (DLS) method to check the aggregate size over time. For measurements, the samples were transferred into the folded capillary cell (polycarbonate with gold plated electrodes) for investigation of particle size distribution using a Zetasizer Nano (Malvern Instruments Ltd., United Kingdom) at 25 °C.

Fig. 5 shows graphs of the particle size distributions in the nanofluids measured using the DLS method. These graphs present the size distribution of the particles. Fig. 5 panels (a–f) show the DLS results of MWNT–HA. Although the difference is insignificant, the overall distribution patterns show that the distribution is found to move to a slightly larger particle size when the distribution of the final day is compared to those of previous days, i.e., the particles are insignificantly aggregated. Also, a comparison of particle size distribution of pristine MWNTs (Fig. 5g) and MWNT–HA (Fig. 5a–f) presented a big difference. While the pristine MWNTs completely degraded after 24 h to reach a particle size distribution of 3033 nm, the MWNT–HA material shows a particle size distribution around 400 nm for 35 day, indicating a stable colloidal system in TO media. Therefore, the existence of MWNT–HA in TO provides a great benefit as MWNT–HA provides a stable colloidal system which can enhance the heat transfer properties.

Fig. 5 Average particle size distribution after (a) 1 day, (b) 7 days, (c) 14 days, (d) 21 days, (e) 28 days, (f) 35 days for MWNT–HA and (g) 1 day for pristine MWNTs.

Table 3 demonstrates the particle size distributions and zeta potential for pristine MWNT/TO and MWNT–HA/TO colloids over a period of 35 days. Firstly, MWNT–HA/TO shows no big aggregation and coagulation at both concentrations. It can be seen that the particle size distribution for MWNT–HA/TO nanofluids shows a gradual increase in the overall hydrodynamic size, which substantiates the formation of small aggregation, which is in agreement with the UV-vis results. On the other hand, pristine MWNTs mostly settled after 24 h. The above results confirm the critical role of HA molecules which act as a covalent stabilizer to prevent rapid colloidal instability associated with the increase in graphitic domain within the carbon-based structure.

Table 3 Zeta potential, average particle size distribution, mobility and polydispersity index (PDI) of pristine MWNT and MWNT–HA in TO media

Label in Fig. 5	Sample	Time (day)	Average particle size distributions (nm)	Polydispersity index (PDI)	Zeta potential (mV)	Mobility (µmcm V^−1 s^−1)
a	MWNT–HA	1	367.4	0.297	−52.3	−4.097
b	MWNT–HA	7	407.3	0.222	−49.4	−3.870
c	MWNT–HA	14	401.5	0.259	−46.9	−3.678
d	MWNT–HA	21	411.0	0.239	−38.6	−3.024
e	MWNT–HA	28	400.9	0.245	−46.6	−3.655
f	MWNT–HA	35	389.3	0.251	−50.4	−3.951
g	Pristine MWNTs	1	3033.0	0.655	−23.6	−1.853

---

According to stabilization theory, the electrostatic repulsions between particles increases if the zeta potential has a high absolute value which then leads to good stability of the suspensions. Particles with a high surface charge tend not to agglomerate, since contact is opposed. Typically accepted zeta-potential values are summarized in Table S2.†⁴⁴

Table 3 also shows the zeta-potential and the polydispersity index (PDI) for pristine MWNTs and MWNT–HA/TO at their natural pH. The zeta-potential and polydispersity index (PDI) are commonly utilized as an index of the magnitude of electrostatic interaction between colloidal particles and thus can be considered as a measure of the colloidal stability of the solution. According to Table S2,† the zeta potential must be as large as possible (positively or negatively) to make a common repulsive force between the particles.⁴⁵ It can be seen that after functionalization with HA, MWNT–HA shows a more negatively charged zeta-potential and is around −45 mV over a period of 35 days. The zeta potential results of MWNT–HA in TO media suggest an appropriate stability over a period of 35 days at 25 °C. Indeed, the zeta-potential gradually show some fluctuations over a period of 35 days, despite remaining mostly stable with time.⁴⁵

3.5. Electrical and thermo-physical analysis

3.5.1. Electrical properties. As a key parameter, the electrical resistivity of MWNT–HA/TO coolants as well as pure TO was investigated in the temperature range of 20–50 °C. Electrical conductivity and resistivity behavior of samples for different weight concentrations are illustrated in Fig. 6 panels (a) and (b), respectively. The results of electrical conductivity and resistivity of the pure TO and synthesized coolants demonstrated different changes by MWNT–HA loading into the pure TO. Alteration of the electrical conductivity and resistivity was in the range of 7–23% with changing concentration. It can be seen that the prepared samples present larger electrical conductivity enhancement at a higher weight concentration of MWNT–HA. Thus, the intrinsic electrical transfer capacity of the MWNTs should be the main reason for electrical conductivity enhancement or decreasing thermal resistance. Also, the lack of acidic agent retards the corrosion phenomenon and enhances the electrical conductivity of the prepared coolants, which as a result increases the lifetime of the system.

Fig. 6 (a) Electrical conductivity and (b) electrical resistivity of MWNT–HA/TO coolant at different weight concentration and temperatures.

3.5.2. Thermal conductivity. TO can be considered as an oil with a high electrical insulation property and heat transfer agent. Similar to other critical factors, the thermal conductivity and the rate of heat transfer of TO play a key role in the selection of the transformer fluid. By looking at the transformer oil’s normal operating temperature (253 K to 363 K), the effect of temperature on the thermal conductivity of TO also plays a key role in the performance of transformers.⁴⁶ In contrast to other basefluids, the thermal conductivity of the transformer oil decreases with increasing temperature.⁴⁶ It is a big disadvantage in applying transformer oil, in particular, at a high range of temperatures, resulting in low performance of the transformer.² It is important to note that the extent of thermal conductivity reduction with temperature increase is different for different transformer oils. A decrease of molecular association with an increase in temperature makes a considerable difference to TO’s thermal conductivity because the molecular chain is so long and the molecular weight is so great. Thus, it is obvious that TO with higher thermal conductivity can be more useful. Previous studies illustrated that various parameters such as thermal conductivity of working fluid and nanostructures, size, concentration, temperature, pH and shape of nanostructure influence the thermal conductivity of coolants.^47,48 Also, some of the thermal phenomena such as Brownian motion, thermophoresis, and diffusiophoresis can influence the thermal conductivity of coolants.^2,49

Fig. 7 shows the thermal conductivity plot of MWNT–HA/TO coolant as a function of temperature and concentration. Two different weight concentrations of 0.001% and 0.005% are considered and the variation of thermal conductivity with concentration and temperature are studied. To avoid a significant increase in effective viscosity and electrical behavior, MWNT–HA at low concentrations is considered in the present study. Fig. 7 clearly demonstrates that the thermal conductivity of MWNT–HA/TO coolant is higher than that of pure TO. It can also be seen that the thermal conductivity of the TO decreases as the temperature increases, where the MWNT–HA/TO coolants at both concentrations show an upward trends. In a similar study, Beheshti et al.² showed that the thermal conductivity of MWNT-COOH/TO coolants increased up to 60 °C and it decreased considerably at higher temperatures. This drop may be attributed to the settlement of MWNT, which can result from the disruption of carboxyl groups on the MWNT surface at high temperature. In contrast with a previous study,¹⁰ the problem of accumulation at 60 °C is completely eradicated by the functionalization of MWNT with HA. On the other hand, some researchers in the field of heat transfer concluded that the nanoparticles are governed by the continuous irregular reciprocating motion and phenomena such as Brownian motion, thermophoresis and diffusiophoresis which could influence thermal conductivity of the coolant.⁴ Above all, it is a big success that the trend of thermal conductivity illustrates a rising trend over the temperature range of 30–80 °C.

Fig. 7 Thermal conductivity plot of pure TO and MWNT–HA/TO coolant at different weight concentrations.

On the other hand, the obtained results confirm that temperature plays a key role in increasing the thermal conductivity of MWNT–HA/TO coolant, which cannot be neglected. For coolants loaded with MWNTs, the vital role in heat transfer enhancement is played by the nanoparticles.^50,51 Liquid molecules generate layers around the MWNTs, thereby altering the local ordering of the liquid layer at the interface region. Thus, the liquid layer at the interface would exhibit a higher thermal conductivity compared to the base-fluid. Differences in thermal conductivity enhancement may be attributable to differences of the thermal boundary resistance around the nanoparticles occurring for different base fluids.⁵² The higher slope in thermal conductivity of MWNT–HA/TO coolants with higher concentration is due to a higher rate of formation of the surface nanolayers.¹⁰ In addition, the role of Brownian motion of particles in nanofluids may be an important parameter in determining the thermal conductivity enhancement and is also an important factor, in particular for cases with significant change in viscosity with temperature, which is certainly the case for TO.⁵² According to recent studies,^19,53,54 increasing the local ordering of the liquid layers at the interface region is one of the main reasons for thermal conductivity enhancement. Thus, the resistance for heat removal is decreased by thinning of the thermal boundary layer via MWNT loading in basefluids and preparing nanolayers with higher thermal conductivity than the bulk liquid.

A summary of experimental studies on the thermal conductivity of TO-based nanofluids is listed in Table S3.† It is worth mentioning that most of the recent studies focused on suspensions with high nanoparticle concentration, which increase the possibility of sediment. Compared with recent studies,² the present work reaches higher enhancement in thermal conductivity under similar experimental conditions. For example, Patel et al.⁵⁵ obtained 10, 11.5, 14 and 17% enhancements in the thermal conductivities of TO-based nanofluids by the addition of Al₂O₃ at 20, 30, 40 and 50 °C, respectively. Such enhancements are obtained for 3% Al₂O₃ loading, while we work at very low weight fractions of 0.001 and 0.005. In fact, the strategy of the present study is the utilization of low concentration of highly-soluble MWNT–HA in TO to avoid sediment, since the bulk fluid in the transformer is motionless.

3.5.3. Convective heat transfer coefficient. In order to investigate the convective heat transfer coefficient, several experiments at different input powers of 49.76, 60.25, 69.95, 81.1, 90.4, 100.36, 110.42, 120.6, 130.19, 140.16 and 149.83 were applied. Also, the natural and forced heat transfer coefficients were studied at both weight concentrations of MWNT–HA at different powers.

Natural convection heat transfer occurred by the buoyancy force, which results from the density deviations formed by temperature variations in the layers of fluid. This type of heat transfer generates in the absence of an external source for the movement of the bulk fluid. According to a previous study,⁵⁶ when a fluid is in contact with a surface with higher temperature, the molecules of the fluid separate and form a layer of lower density. As a result, the layers of fluid with lower density are displaced by the cooler layers of fluid, thus the cooler layers of fluid sink. It is noteworthy that the nanostructure concentrations and thermal conductivity can also influence natural convection in coolants.²

The natural convection heat transfer coefficient and Nusselt number of pure TO and MWNT–HA/TO coolants versus input power are illustrated in Fig. 8 panels (a) and (b), respectively. It can be seen that similar upward trends are obvious in the presence of the pure TO and MWNT–HA altered coolants in both figures. It is obvious that the heat transfer coefficient is a function of temperature. According to the results, the input power increases with an increase of the average temperature of the bulk fluid in the transformer, which is implying a higher heat transfer coefficient. As the thermal conductivity of the coolants at different temperatures is applied for calculation, the evaluated Nusselt number results show some weak fluctuations.

Fig. 8 (a) Natural convection heat transfer coefficient, (b) Nusselt number (natural), (c) forced convection heat transfer coefficient and (d) Nusselt number (forced) of pure TO and MWNT–HA/TO coolant.

Forced convection is a mechanism of heat transfer in which fluid motion is created via an external source such as a fan and pump. The amount of forced convection heat transfer coefficient and Nusselt number versus input power are respectively shown in Fig. 8 panels (c) and (d) for pure oil and MWNT–HA/TO coolants. The natural convection heat transfer coefficient and Nusselt number increase with an increase of input power and concentration, the forced convection heat transfer also shows similar results of enhancement with an increase of power and concentration. Surprisingly, the amount of thermal conductivity, heat transfer coefficient and/or Nusselt number show significant enhancements in the presence of MWNT–HA, which is attributed to the good dispersibility of MWNT in TO as one of the reasons.

The enhancement of convective heat transfer can be attributed to a reduction of the thermal boundary layer thickness. Aravind et al.^50,57 showed that carbon nanomaterials such as MWNTs have a tendency to decrease the thermal boundary layer thickness, which lowers the difference between temperatures of the bulk fluid and wall in the transformer and subsequently enhances the convective heat transfer coefficient. To clarify this, the heat transfer coefficient can be approximately modeled as k/δ_t, where δ_t represents the thermal boundary layer thickness. So, to increase the heat transfer coefficient, either k should be increased or δ_t should be decreased or both. As a future comparison, Table S4† shows a comparison of the natural convection and forced convection heat transfer coefficients of nanofluids with different nanoparticle loading. The convective heat transfer coefficient of the MWNT–HA/TO nanofluid was compared with those of other nanoparticle-based TO nanofluids reported in the literature. Obviously, the present samples had a relatively high natural convection and forced convection heat transfer coefficients in comparison to other nanoparticles loading in TO at similar weight fraction, in particular, as compared with MWNT-COOH/TO nanofluids² under similar experimental conditions. Our results show natural convection and forced convection heat transfer coefficient enhancements of 23 and 28% at weight fraction of 0.005, which demonstrate a significant increase at a very low concentration.

3.5.4. Breakdown voltage. Breakdown voltage is a dielectric strength of TO and plays an essential role in the electrical performance of transformers. Breakdown voltage is obtained by detecting the voltage required for the generation of a spark between two electrodes with a precise gap in the oil. A small extent of breakdown voltage demonstrates the presence of moisture content in the TO. According to previous studies,^2,58 different impurity and percentage of moisture content can easily disturb the breakdown voltage results. On the other hand, functionalization of MWNTs is the only route for increasing the colloidal stability of MWNTs in TO media. Thus, it is essential to apply functional groups with hydrophobic properties for functionalization of MWNTs. HA as a functional group can meet both criteria of having good dispersion and hydrophobic properties. In addition, to prevent acidity, a novel one-pot method without acid-treatment is employed.

Breakdown voltage of pure TO with and without sonication and MWNT–HA/TO at both weight concentrations of 0.001 and 0.005 are plotted in Fig. 9 and the mean breakdown voltages with standard deviation are listed in Table 4. In order to measure the dielectric breakdown voltage with different weight fractions of MWNT–HA in the colloidal suspension, the experiment was repeated 60 times for each of the weight concentrations according to the ASTM D-92 standard using the dielectric breakdown measurement device, Mugger’s automatic laboratory oil tester. All experiments were performed using mushroom electrodes set at a gap of 1 mm and at room temperature. The dielectric breakdown voltage of all samples was measured 1 day after preparing. The most influential factor affecting the performance of the dielectric strength of the transformer oil is the degradation caused by water and other contaminants,⁵⁹ which with the sonication process could also introduce moisture to the oil. Fig. 9 panels (a) and (b) show the dielectric breakdown voltage of pure oil with and without 30 min sonication (time for performing the diazonium reaction) to investigate the effect of the sonication process on TO performance. Results suggest that the sonication has an insignificant effect on the dielectric breakdown voltage (∼0.4 kV), which is obtained by doing just 30 min sonication using the Bioruptor (4 s “ON”, 4 s “OFF”).

Fig. 9 The dielectric breakdown voltage of pure oil (a) with and (b) without 30 min sonication and (c) MWNT–HA/TO at (c) 0.001 and (d) 0.005.

Table 4 Breakdown voltage (kV) of pure TO and MWNT–HA/TO at different concentrations

Samples	Mean breakdown voltage (kV)	S.D. (kV)
Pure TO	58.03166667	1.137174
Sonicated pure TO	57.65333333	1.311039
MWNT–HA/TO (0.001%)	54.09833333	1.60521
MWNT–HA/TO (0.005%)	52.06666667	2.47283

---

Fig. 9 panels (c) and (d) show the experimental results with the various weight concentrations of MWNT–HA. Generally, increasing the MWNT–HA concentration leads to a mild decrease of the dielectric breakdown voltage. Again, the mean breakdown voltages and standard deviations for the above-mentioned samples are given in Table 4. It was found that breakdown voltages level-off with the rising test number, no upward or downward trend was obtained in experiments, which owes to the effective energy control of the test equipment. With the increase of weight concentration of MWNT–HA in TO from 0.001 to 0.005%, the dielectric strength decreases from 54.098 kV to 52.066 kV. The standard deviations of both samples are larger than that of transformer oil, indicating that the breakdown voltage of the samples is less predictable. The MWNT–HA loading in TO media leads to a small drop in the breakdown voltage, which is mainly caused by the MWNTs.

3.5.5. Density of fluids. The density of MWNT–HA/TO coolants as well as pure TO at various temperatures is illustrated in Fig. 10. Density plays a key role in evaluating the natural convection heat transfer. As mentioned above, natural convection is a function of the buoyancy force or density variation. As shown in Fig. 10, there is a diverse connection between the temperature and density of the fluids. The density of the coolants increases with an increase of MWNT–HA concentration.

Fig. 10 Density of the MWNT–HA/TO coolants at different concentrations and temperatures.

3.5.6. Flash point. The flash point is considered as the temperature at which TO produces sufficient vapor for creating a flammable mixture with air. The flash point is a very essential temperature for defining the probability of fire hazards in a transformer in the presence of TO. Therefore, a higher flash point can be considered as a desirable property for TO.⁶⁰

A seta semi-automatic cleveland open cup flash point tester was used to measure the flash point. The flash points of the pure TO with and without sonication and also MWNT–HA/TO at two different weight fractions were measured using the basis of American Society for Testing and Materials (ASTM) D-92.⁶¹ The trend of changes of flash point as a function of MWNT–HA concentration is shown in Fig. 11. Firstly, the flash point was tested for the pure oil with and without sonication to investigate the effect of sonication. Variation of the flash point in the presence of sonication and lack of sonication is shown in Fig. 11, indicating an increase of around one degree in the volatility of the TO when performing sonication for 30 min. Also, the variation of the flash point as a function of MWNT–HA concentration is indicated with an increase in the volatility of the TO with MWNT addition. It is obvious that a higher flash point temperature is required for safer handling of TO. The rate of increase in the flash point of the MWNT–HA/TO coolant at the 0.001 weight fraction with respect to the TO is 4.69%, and the highest amount of increase is related to the 0.005 weigh fraction sample, which is 7.86%.

Fig. 11 Flash point of pure TO and MWNT–HA/TO coolants.

3.5.7. Viscosity. Viscosity is known as the resistance of fluid to flow. High-resistance transformer oil produces an obstacle for the convection circulation in the transformers. Obviously, low viscosity is an essential property for an excellent TO. Another critical property for TO is the decrease of viscosity with the increase of temperature at a lower rate. The viscosity of pure TO and MWNT–HA/TO coolant at different concentrations and a high shear rate of 235 s⁻¹ were investigated experimentally, as shown in Fig. 12.

Fig. 12 (a) The viscosity of pure TO and MWNT–HA/TO coolants at different weight concentrations and temperatures at a shear rate of 235 s⁻¹, (b) viscosity enhancement MWNT–HA/TO coolants at different weight concentrations in comparison with pure TO.

Similar to other oil-based nanofluids, the rheological behavior of MWNT–HA/TO (Fig. 12a) shows two typical characteristics: (i) an enhancement of viscosity with increasing weight fraction of MWNT–HA, and (ii) a decrease in viscosity with increasing temperature, which is due to weakening of the intermolecular forces of the fluid itself. Fig. 12b illustrates the amount of enhancement in the viscosity with MWNT–HA loading. It shows the enhancement in the MWNT–HA/oil viscosity is less than 10% for both weight fractions and all temperatures. It can be concluded that the enhancement in the MWNT–HA/oil viscosity is insignificant with an increase in weight concentration, which can be attributed to the low weight fraction of the MWNT–HA in the oil. In addition, in agreement with Ko et al.,⁶² the viscosity decreases with an increase in temperature.

The viscosity of pure TO and MWNT–HA/TO coolant as a function of the shear rate for different concentrations are presented in Fig. S1 (ESI†). The effective viscosities of pure transformer oil and treated coolants were measured for the shear rate range of 35–235 s⁻¹. Firstly, MWNT–HA/TO and pure TO have a linear shear stress/shear rate relationship, which can be categorized as a Newtonian fluid.^2,63 It may be seen for all temperatures as the viscosity of the TO is independent of the shear strain rate, indicating Newtonian behavior. Also, after the addition of MWNT–HA, the samples showed Newtonian behavior.

We believe that since TO exhibits Newtonian behavior, it dominates the rheological properties and the whole mixture behaves like a Newtonian fluid with low concentrations of MWNT–HA. Due to some fluctuations and the significant error in measurement, we were not able to identify the exact amount of viscosity for a shear rate of around zero. Based on some recent research,^2,63,64 TO is categorized as a Newtonian fluid.

4. Conclusion

MWNTs are functionalized with hexylamine (HA) via a diazonium reaction under microwave radiation by a rapid and one-pot technique to obtain highly dispersed MWNTs in TO media without acidic properties. The procedure was fast and simple and resulted in a suitable degree of hydrophobic functionalization as well as solubility in TO media. Based on the results of FT-IR, TGA-DTG, Raman, EDS, CHNS/O and TEM the functionalization of MWNT with HA was confirmed. The results of natural and forced convection heat transfer coefficients, breakdown voltage, flash point, density, electrical and thermal conductivities and viscosity verified the synthesized MWNT–HA based transformer oil is a promising alternative oil for use in transformers. Also, the promising achievements in the thermal and rheological properties support this claim. We obtained 10% enhancement in the thermal conductivity of TO-based nanofluids at a very low weight concentration of 0.005. Results also showed that the natural convection and forced convection heat transfer coefficient enhancements of 23 and 28% at a weight fraction of 0.005, which demonstrate a significant increase at a very low concentration. Observation of enhancement in natural- and forced-convection heat transfer coefficients and thermal conductivity of working fluids with MWNT–HA loading in TO, a feeble decrease in the breakdown voltage and lack of remarkable change in rheological properties resulted in a higher performance transformer.

Nomenclature

Nu	Nusselt number
T c	Average temperature of walls, °C
H	Heat transfer coefficient, W m^−2 K^−1
K	Thermal conductivity, W m^−1 K^−1
L	Length, m
T h	Average temperature of oil, °C
Q	Input power, W
A	Heat transfer area
V	Voltage, V
I	Current, A

Acknowledgements

The authors gratefully acknowledge the Bright Sparks Unit of the University of Malaya, the UMRG Grant RP012B-13AET and the High Impact Research Grant UM.C/625/1/HIR/MOHE/ENG/45, Faculty of Engineering, University of Malaya, Malaysia for support to conduct this research.

References

1. A. Amiri, M. Shanbedi, H. Yarmand, H. K. Arzani, S. Gharehkhani, E. Montazer, R. Sadri, W. Sarsam, B. Chew and S. Kazi, Energy Convers. Manage., 2015, 105, 355–367.
2. A. Beheshti, M. Shanbedi and S. Zeinali Heris, J. Therm. Anal. Calorim., 2014, 118, 1451–1460.
3. B. X. Du and X. L. Li, IEEE Trans. Dielectr. Electr. Insul., 2014, 22, 851–858.
4. C. Choi, H. S. Yoo and J. M. Oh, Curr. Appl. Phys., 2008, 8, 710–712.
5. J. Davidson and D. Bradshaw, US20040206491 A1, 2003.
6. S. Choi, Am. Soc. Mech. Eng., 1995, 231, 99–106.
7. J. G. Hwang, M. Zahn, F. M. O’Sullivan, L. A. Pettersson, O. Hjortstam and R. Liu, J. Appl. Phys., 2010, 107, 014310.
8. V. Segal, A. Rabinovich, D. Nattrass, K. Raj and A. Nunes, J. Magn. Magn. Mater., 2000, 215, 513–515.
9. T. Cader, S. Bernstein and C. Crowe, US5898353 A, 1999.
10. A. Amiri, R. Sadri, M. Shanbedi, G. Ahmadi, B. Chew, S. Kazi and M. Dahari, Energy Convers. Manage., 2015, 92, 322–330.
11. M. Shanbedi, S. Z. Heris, A. Amiri, S. Adyani, M. Alizadeh and M. Baniadam, Numer. Heat Transfer, Part A, 2014, 66, 947–962.
12. M. Shanbedi, A. Amiri, S. Rashidi, S. Z. Heris and M. Baniadam, Heat Transfer Eng., 2015, 36, 315–324.
13. A. Amiri, R. Sadri, G. Ahmadi, B. Chew, S. Kazi, M. Shanbedi and M. S. Alehashem, RSC Adv., 2015, 5, 35425–35434.
14. M. Shanbedi, D. Jafari, A. Amiri, S. Z. Heris and M. Baniadam, Heat Mass Transfer, 2013, 49, 65–73.
15. M. N. M. Zubir, A. Badarudin, S. Kazi, M. Misran, A. Amiri, R. Sadri and S. Khalid, J. Colloid Interface Sci., 2015, 454, 245–255.
16. J. Saien, H. Bamdadi and S. Daliri, J. Ind. Eng. Chem., 2015, 21, 1152–1159.
17. B. Ghiadi, M. Baniadam, M. Maghrebi and A. Amiri, Russ. J. Phys. Chem. A, 2013, 87, 649–653.
18. A. Amiri, R. Sadri, M. Shanbedi, G. Ahmadi, S. Kazi, B. Chew and M. N. M. Zubir, Energy Convers. Manage., 2015, 101, 767–777.
19. S. J. Aravind, P. Baskar, T. T. Baby, R. K. Sabareesh, S. Das and S. Ramaprabhu, J. Phys. Chem. C, 2011, 115, 16737–16744.
20. H. Zare-Zardini, A. Amiri, M. Shanbedi, A. Taheri-Kafrani, S. N. Kazi, B. T. Chew and A. Razmjou, J. Biomed. Mater. Res., Part A, 2015, 103, 2959–2965.
21. H. Zare‐Zardini, A. Amiri, M. Shanbedi, M. Memarpoor‐Yazdi and A. Asoodeh, Surf. Interface Anal., 2013, 45, 751–755.
22. K. Miners, IEEE Trans. Power Appar. Syst., 1982, 751–756.
23. T. V. Oommen and E. M. Petrie, IEEE Trans. Power Appar. Syst., 1984, PAS-103, 1923–1931.
24. S. Z. Heris, M. Fallahi, M. Shanbedi and A. Amiri, Heat Mass Transfer, 2015, 1-9, 1548–1549,  DOI:10.1007/s00231-015-.
25. C. Choi, H. Yoo and J. Oh, Curr. Appl. Phys., 2008, 8, 710–712.
26. L. Chen, H. Xie, W. Yu and Y. Li, J. Dispersion Sci. Technol., 2011, 32, 550–554.
27. H. Ahmadi, A. Rashidi, A. Nouralishahi and S. S. Mohtasebi, Int. Commun. Heat Mass Transfer, 2013, 46, 142–147.
28. A. Amiri, M. Maghrebi, M. Baniadam and S. Z. Heris, Appl. Surf. Sci., 2011, 257, 10261–10266.
29. M. D. Ellison and P. J. Gasda, J. Phys. Chem. C, 2008, 112, 738–740.
30. J. L. Bahr and J. M. Tour, Chem. Mater., 2001, 13, 3823–3824.
31. B. K. Price and J. M. Tour, J. Am. Chem. Soc., 2006, 128, 12899–12904.
32. J. P. Holman, Experimental methods for engineers, 3rd edn, 1966.
33. M. Dong, L. Shen, H. Wang, H. Wang and J. Miao, J. Nanomater., 2013, 2013, 164.
34. A. Amiri, G. Ahmadi, M. Shanbedi, M. Savari, S. N. Kazi and B. T. Chew, Sci. Rep., 2015, 5, 17503.
35. I. Kalinina, K. Worsley, C. Lugo, S. Mandal, E. Bekyarova and R. C. Haddon, Chem. Mater., 2011, 23, 1246–1253.
36. A. Das, S. Pisana, B. Chakraborty, S. Piscanec, S. K. Saha, U. V. Waghmare, K. S. Novoselov, H. R. Krishnamurthy, A. K. Geim, A. C. Ferrari and A. K. Sood, Nat. Nanotechnol., 2008, 3, 210–215.
37. B. Das, R. Voggu, C. S. Rout and C. N. R. Rao, Chem. Commun., 2008, 5155–5157,  10.1039/b808955h.
38. L. Feng, L. Yang, Z. Huang, J. Luo, M. Li, D. Wang and Y. Chen, Sci. Rep., 2013, 3, 3306.
39. R. Tian, X. Wang, M. Li, H. Hu, R. Chen, F. Liu, H. Zheng and L. Wan, Appl. Surf. Sci., 2008, 255, 3294–3299.
40. W. Lv, D.-M. Tang, Y.-B. He, C.-H. You, Z.-Q. Shi, X.-C. Chen, C.-M. Chen, P.-X. Hou, C. Liu and Q.-H. Yang, ACS Nano, 2009, 3, 3730–3736.
41. Y. Xie, Q. Huang, M. Liu, K. Wang, Q. Wan, F. Deng, L. Lu, X. Zhang and Y. Wei, RSC Adv., 2015, 5, 68430–68438.
42. C.-F. Wang, S.-W. Kuo, C.-H. Lin, H.-G. Chen, C.-S. Liao and P.-R. Hung, RSC Adv., 2014, 4, 36012–36016.
43. A. Amiri, M. Shanbedi, G. Ahmadi, H. Eshghi, B. T. Chew and S. N. Kazi, Colloids Surf., A, 2015, 487, 131–141.
44. A. Ghadimi, R. Saidur and H. S. C. Metselaar, Int. J. Heat Mass Transfer, 2011, 54, 4051–4068.
45. B. White, S. Banerjee, S. O’Brien, N. J. Turro and I. P. Herman, J. Phys. Chem. C, 2007, 111, 13684–13690.
46. H. Wang, S. J. Ma, H. Yu, Q. Zhang, C. M. Guo and P. Wang, Pet. Sci. Technol., 2014, 32, 2143–2150.
47. H. Yarmand, S. Gharehkhani, G. Ahmadi, S. F. S. Shirazi, S. Baradaran, E. Montazer, M. N. M. Zubira, M. S. Alehashem, S. N. Kazi and M. Daharia, Energy Convers. Manage., 2015, 100, 419–428.
48. B. Pal, S. S. Mallick and B. Pal, J. Ind. Eng. Chem., 2015, 25, 99–104.
49. S. F. Seyed Shirazi, S. Gharehkhani, H. Yarmand, A. Badarudin, H. S. Cornelis Metselaar and S. N. Kazi, Mater. Lett., 2015, 152, 192–195.
50. S. S. J. Aravind, P. Baskar, T. T. Baby, R. K. Sabareesh, S. Das and S. Ramaprabhu, J. Phys. Chem. C, 2011, 115, 16737–16744.
51. N. Jha and S. Ramaprabhu, J. Phys. Chem. C, 2008, 112, 9315–9319.
52. L. S. Sundar, M. K. Singh, E. V. Ramana, B. Singh, J. Grácio and A. C. Sousa, Sci. Rep., 2014, 4, 4039.
53. S. Murshed, K. Leong and C. Yang, Appl. Therm. Eng., 2008, 28, 2109–2125.
54. C.-J. Yu, A. Richter, A. Datta, M. Durbin and P. Dutta, Phys. B, 2000, 283, 27–31.
55. H. E. Patel, T. Sundararajan and S. K. Das, J. Nanopart. Res., 2010, 12, 1015–1031.
56. S. M. Ghiaasiaan, Convective heat and mass transfer, Cambridge University Press, 2011.
57. S. S. J. Aravind and S. Ramaprabhu, RSC Adv., 2013, 3, 4199–4206.
58. M. Ikeda, T. Teranishi, M. Honda and T. Yanari, IEEE Trans. Power Appar. Syst., 1981, 921–928.
59. Y. Lv, Y. Zhou, C. Li, Q. Wang and B. Qi, IEEE Electr. Insul. Mag., 2014, 30, 23–32.
60. M. A. Wahab, M. Hamada, A. Zeitoun and G. Ismail, Electr. Power Syst. Res., 1999, 51, 61–70.
61. D. Kong, D. J. am Ende, S. J. Brenek and N. P. Weston, J. Hazard. Mater., 2003, 102, 155–165.
62. G. H. Ko, K. Heo, K. Lee, D. S. Kim, C. Kim, Y. Sohn and M. Choi, Int. J. Heat Mass Transfer, 2007, 50, 4749–4753.
63. D. H. Fontes, G. Ribatski and E. P. Bandarra Filho, Diamond Relat. Mater., 2015, 58, 115–121.
64. M. Singh and L. Kundan, IJTARME, 2013, 2, 125–130.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra17687e

---