Stable Nanofluids for Convective Heat Transfer Applications

Heat transfer forms an important part of all practical applications and the heat transfer fluid forms an integral part of any heat transfer set up. Most such set ups use either air (gaseous fluid) or a liquid as heat transfer fluid. Some of the commonly used heat transfer fluids in industrial processes like transportation, power generation, chemical processes, heating and cooling processes and microelectronics include water, ethylene glycol (EG), propylene glycol, mixtures of glycols with water, and organic oils. The properties of the heat transfer fluids strongly influence the size (area of the heat transfer) of the heat exchanger. The poor heat transfer properties of these fluids are a limitation in improving heat transfer performance and miniaturization of heat exchangers.

One of the emerging areas of research with tremendous promise for industrial applications is to improve heat transfer efficiency of standard heat transfer fluids by dispersing solid particles in the base medium. Early research focused on dispersing micron and millimeter sized particles to enhance the heat transfer properties of the fluid, but the idea had to be abandoned because of issues like high pressure drop in flow based applications, sedimentation, and clogging of flow channels. It has since then been shown by several investigators that the addition of solid nanosized particles/fibers, preferably those possessing higher thermal conductivity such as carbon, metal, metal oxides into heat transfer fluids improves the overall thermal conductivity of the fluid [1,2,3,4,5,6,7]. Choi [1] was the first to officially coin the term “nanofluids” for dispersions of nanoparticles in heat transfer fluids. Since then, there have been several research attempts [6,7,8] that have tried to capture the enhancement in thermal conductivity of these dispersions. In addition to the enhancements demonstrated by the nanofluids, they also exhibits better long-term stability, flow characteristics, and rheological properties as compared to the dispersions made with larger particles.

Thermal conductivity is one of the most important properties of a heat transfer fluid. Extensive efforts have been put to study heat transfer properties of titania (TiO₂)-water, alumina-water, and copper oxide (CuO)-water based systems, apart from a few reports concentrating on exotic materials like carbon nanotubes (CNT) and graphene in oil and water based systems.

The temperature oscillation method [9], the steady state parallel plate method [10], and the transient hot wire method [2] have commonly been used by investigators to determine the thermal conductivity of these dispersions. However, the transient hot wire technique has been used by most investigators due to its simplicity and ease of use [11,12,13,14].

Eastman [6] has shown that thioglycolic acid stabilized metallic Cu nanoparticles gave an enhancement of 40% in the thermal conductivity with only 0.3% volume of solid loading. Karthikeyan [15] reported 31% enhancement with 1 vol. % CuO nanoparticles in water. Masuda [16] have demonstrated that 13 nm alumina nanoparticles in water (1.3 to 4.3-vol. %) enhance the thermal conductivity of water by around 10–29%. Hwang et al. [17] studied the effect of particle loading in alumina-water based systems. They reported that with increase in particle concentration from 0.3-vol. % to 1-vol. %, the enhancement in thermal conductivity goes up from 1.3% to 4%. Alumina particles in transformer oil were studied by Choi et al. [14]. They reported an enhancement in thermal conductivity of 20% for a particle loading of 4 vol. %. With 3 vol. %, TiO₂ nanoparticles in water was reported an enhancement of around 7.4% by Turgut et al. [18].

Ethylene glycol, which is a common industrially used heat transfer fluid, had been studied by several investigators. Oh et al. [19] showed that 4 vol. % of alumina in ethylene glycol showed an enhancement in thermal conductivity of 9.7%. Xie et al. [8] showed an enhancement of 23% with 5 vol. % alumina particles in ethylene glycol. Ethylene glycol based carbon nanotubes dispersion was reported an enhancement of 7.7% and 17% at 25 °C and 50 °C, respectively, [20].

While thermal conductivity plays an important role in determining the heat transfer characteristics of a fluid, it is not the most important parameter when it comes to flow based heat transfer. Most potential industrial applications for nanofluids lie in flow based processes. Chun et al. [21] studied the convective thermal properties of alumina nanoparticles in transformer oil in a double pipe heat exchanger and reported considerable enhancement in heat transfer coefficients under laminar flow conditions. The authors attributed this enhancement to the possible migration of the nanoparticles to the boundary layer region near the wall. Pak and Cho [22] studied the heat transfer enhancement in a circular tube using alumina and TiO₂ nanoparticles in convective condition. They reported that the Nusselt number (Nu) increases with increase in particle concentration and Reynolds number. Heris et al. [23] dispersed CuO and alumina nanoparticles in water to study the effect of the type of nanoparticles on the enhancement in convective heat transfer properties of the base fluid. The authors conducted their tests in a constant wall temperature heat exchanger. They reported that at low particle concentrations, both nanoparticles produced similar enhancements in the convective heat transfer properties of the base fluid but at higher concentrations, alumina nanofluids showed better heat transfer enhancement over the base fluid as compared to CuO based nanofluids. Heris et al. [24] studied the effect of particle concentration on enhancement in heat transfer coefficients in alumina nanofluids under constant wall temperature convective heat transfer condition. They concluded that nanoparticles played an important role in enhancing the heat transfer coefficients of the base fluids as increasing the particle concentration increased the enhancement observed.

Wen and Ding [25] studied the convective heat transfer properties of alumina-water nanofluids in laminar region under constant heat flux condition. They observed considerable increase in the heat transfer coefficients for the nanofluids. They also observed higher enhancement near the entrance region and attributed that to the migration of the nanoparticles to the high shear region near the wall and disturbing the boundary layer. Xuan and Li [26] investigated the thermal enhancement and hydrodynamic properties of nanofluids under turbulent flow conditions.

There are several theoretical models that study the mechanisms for the enhancement of thermal conductivity of fluids due to the presence of nanoparticles. Classical models of the Maxwell-Garnett type and Hamilton-Crosser model do not take in to consideration the effect of Brownian motion of the nanoparticles. They are based on effective medium approximation and take into consideration the layering of liquid molecules around the solid surface at the solid–liquid interface [8]. Xuan and Li [26] and Jang and Choi [27] developed models based on the Brownian motion of the nanoparticles dispersed in fluid. But, no single model is able to capture all the properties and features of the heat transfer characteristics of nanofluids.

In spite of all the results and potentials demonstrated by nanofluids in several scientific research papers and technical presentations over the last couple of decades, the concept has not been widely commercialized yet. The key issues with nanofluids are: (i) a robust, cost-effective and scalable method to produce nanofluids to industrial scale has not yet been developed, (ii) stability in industrial applications is not yet established, and (iii) meaningful data in flow based heat transfer process do not exist. The present work aims at addressing all the above issues. A top down approach is employed to produce stable metal oxide nanoparticles in base fluids by high energy wet-milling of corresponding metal oxides powders of micron size. The milling process is optimized for type of metal oxide particles, dispersants, and base fluids of industrial interest. A concentrated suspension (10 vol. %) of nano alumina particles in base fluid (water, EG, their mixtures, and a commercial coolant) and a suitable dispersant (selected by use of molecular dynamics simulation) was prepared. This concentrated suspension was further diluted to desire concentration to produce a stable nanofluid in large quantities for heat transfer enhancement studies. The nanofluids thus obtained were tested for overall heat transfer coefficient on a custom made double pipe heat exchangers. In this work effect of various parameters like size, concentration of nanoparticles, type of based fluid for nanofluids were investigated for heat transfer enhancement compared to base fluid.

The experimental set up fabricated for the study is shown in Fig. 1. The set up consists of a double pipe heat exchanger for studying the heat transfer characteristics of nanofluids. The test section consists of two concentric 520 mm long tubes made of stainless steel. The internal diameter of the inner tube is 6 mm with 1 mm wall thickness. The outer tube, which is also made of stainless steel, has an internal diameter of 25 mm with 1 mm wall thickness. Polyurethane foam has been used as insulation for the heat exchanger. Four K-type thermocouples (T1–T4) were installed at the inlet and outlet of the two fluid streams. These thermocouples were connected to a data acquisition system (Agilent 34972A2) for automated logging of temperature-time data. This set-up has been used to generate all the reported data, unless otherwise specified.

The nanofluid (test fluid) was flown through the inner tube and the hot water used to heat the test fluid was flown through the outer tube. The temperature of the outer stream was kept constant by use of a custom-made constant temperature water bath of 50 L capacity. Flow rates were maintained at the desired levels by use of peristaltic pumps for the test fluid and centrifugal pumps for the outer heating fluid.

Alumina (A16-SG grade) was used to produce nanofluids using four industrially used cooling liquids as base fluids. Testing for the concept of enhancement in heat transfer coefficient in nanofluids was carried out for nano oxide particles dispersed in double distilled (DD) water, EG, mixtures of DD water, and EG. A commercially available internal combustion (IC) engine coolant containing additives such as sodium silicate, disodium phosphate, sodium molybdate, sodium borate, and dextrin was used to disperse nanoparticles to produce nanofluids of industrial interest. Nanofluids preparation is of critical importance when the nanoparticles dispersion in a coolant was used for potential industrial applications like IC engine cooling. Simply mixing nanoparticles with the base fluid will not result in stable dispersions. In order to prepare nanofluids suitable for industrial use, proper mixing technique and stabilization of the particles is required.

In the present work, the nanofluids were prepared in-situ by wet milling of the alumina powder in the base fluid in presence of a suitable dispersant at a very high particle loading (30–40%) in a planetary mill (Fritsch P5, Germany) with yttrium stabilized zirconium dioxide as grinding media (Zirconox, Jyoti Ceramics, Nashik, India) and further diluting the concentrated suspension with the corresponding base fluid (water, EG, their mixtures, and commercially available engine coolants) to the desired concentration. The grinding media to powder ratio has been selected to be an optimum 25:1 for the present experiments. The particle size distribution of the primary slurry as well as the final ground product was measured by using laser scattering particle size analyzer (HORIBA LA910 and HORIBA LA 950S2). The dispersant for the process has been selected by studying the interaction energies of the tested dispersants with the surface of alumina using molecular modeling. The dispersant with the lowest interaction energy (most favored thermodynamically to adsorb on the surface) is selected as the dispersant for the process.

The ground slurry thus obtained, was used to prepare nanofluids in large quantities (0.5–25 L). The concentrations of nanoparticles varied from 0.5 vol. % to 1.25 vol. % in the final product. The primary preground particles were found to be around 527 nm in size. The final size distribution for the alumina nanoparticles in the four-tested base fluids is shown in Fig. 2 and Table 1, respectively. The median particle size was found to be around 80–95 nm for all base fluids. Transmission electron microscopy (TEM) image of alumina-water nanofluid is shown in Fig. 3. Alumina particles of 50 nm size and crystal of citrate salt were observed under TEM. It can be seen that the alumina particles are fully covered with the citrate crystals. This confirms that alumina particles like citrate molecules on its surface.

The size distributions show that not only is the developed process independent of the base fluid but also highly reproducible.

The adsorption of surfactants at the solid–liquid interface results in certain desirable changes in macroscopic properties of the particles such as wettability and colloidal stability. The differences in the macroscopic properties created as a consequence of the intervention of the surfactant, in turn drives the particular end-application, for example dispersion, flotation, flocculation, changes in rheology, detergency etc. Molecular modeling computations capture the different molecular interactions of surfactant molecule with the particle surface at sufficiently high degree of complexity as well as with adequate scientific rigor. These interactions quantified in terms of interaction energies, computed purely on theoretical basis, can thus be related to the macroscopic properties such as dispersion capability of a given surfactant for the particular particle in a given solvent. Accordingly, we have computed interaction energies of several surfactants at particle-solvent interface to find out the most suitable dispersant for the given system.

Universal force field (UFF) [28,29,30,31] as implemented in Material Studio (MS) [32] has been used to model the inorganic surface-dispersant interactions. We have successfully demonstrated though our earlier work that UFF can be used to model such systems with reasonable accuracy [33,34,35,36]. The geometry of dispersant molecules was optimized using UFF as implemented in Forcite module of MS. A surface cell was created from the unit cell of the inorganic crystal at its cleavage plane and optimized with the help of surface builder module in MS. The optimized reagent molecule was docked on the mineral surface. The initial geometry of surface-reagent complex was created physically on the screen with the help of molecular graphics tools, taking into consideration the possible interactions of reagent functional groups with surface atoms. The reagent molecule was then allowed to relax completely on the surface using Forcite geometry optimization. Several initial conformations (∼20) were assessed so as to locate the minimum energy conformation of the inorganic surface-reagent complex. The partial charges on the atoms were calculated using charge equilibration method [37]. The intramolecular van der Waal interactions were calculated only between atoms which are located at distances greater than fourth nearest neighbors. A modified Ewald summation method [38] was used for calculating the nonbonded coulomb interactions while for van der Waal interactions a direct cut off at a distance less than r/2 (where r is the length of the simulation cell) was employed. Smart minimizer as implemented in Forcite module of MS was used for geometry optimization. The optimization was considered to be converged when a gradient of 0.0001 Kcal/mole is reached. Structure of the complex obtained through static energy minimization method represents only a local minimum energy structure; it was further optimized to find a global minimum energy structure through molecular dynamics (MD) simulations. In order to get realistic interaction energies, we have modeled several conformations of surfactant on the particle surface and optimized it to get a global minimum. Figure 4 depicts one such optimized complex for a given surfactant-particle-solvent system. MD calculations were run using constant energy microcanonical ensemble method (NVE) at 300 K with time step of 1 fs. Total run length was ∼300 ps. During the simulations, the temperature was controlled by velocity scaling method. During simulations, atom-based cut-off method was employed for calculating both Van der Waals and electrostatic forces. The interaction energies were computed as

where E_combined is the total interaction energy calculated using Eq. (2). E_{complex,solvent} is the total energy of the optimized surface-reagent complex in the presence of solvent, E_surface, E_reagent, and E_solvent are the total energies of free surface, reagent and solvent molecules, computed separately. ΔE_{solvent-reagent} is the interaction energy computed for the interaction of solvent and reagent molecule and ΔE_{surface-solvent} is the contribution due to interaction of solvent molecules with the surface. These energies are subtracted from ΔE_combined to get the final interaction energy [ΔE_solvent] of reagent molecule with the inorganic surface [Eq. (2)]. It is worth noting that the more negative magnitude of interaction energy indicates more favorable interactions between the dispersant and inorganic surface. The magnitude of this quantity is thus an excellent measure of the relative intensity/efficiency of interaction amongst various dispersants. For example, in the presence of water based on the magnitude of computed interaction energies, carboxylic acid dispersants were ranked as follows, are shown in Table 2 [39].

Since the present work is concerned with flow-based applications, it is prudent to make note of the effect nanoparticles have on viscosity of the dispersion. In the present work, viscosity is measured using Brookfield viscometer (Model: LVDV – I, spindle S18). Since viscosity of liquids is known to decrease with increase in temperature, the viscosity in the present work is measured at room temperature (∼23 °C) as the expected temperatures in actual applications are expected to be much higher.

Figure 5 shows the changes in viscosity of water that is observed on addition of nanoparticles at three different concentrations. It has been found that 4 wt. % (∼1 vol. %) alumina causes an increase in viscosity of about 30%. On the other hand, alumina particles in EG caused negligible change in viscosity.

It must be noted that the EG-alumina nanofluids contain around 5 vol. % water that is added to solubilize the dispersant. Hence, there is a decrease in viscosity of the nanofluids with respect to pure EG. This decrease in viscosity of EG on addition of water is a commonly known phenomenon and is attributed to molecular aggregation between water and EG molecules as EG has a high heat of hydration (∼7.1 kJ/mol) and the ability of water molecules in the hydrogen bonded aggregate to exchange positions freely with other water molecules in other aggregates. The saturation of hydrogen bonds in water molecules does not happen in the case of EG unlike in the case of primary alcohols like methanol and ethanol because of the ability of EG to form more than one independent collective hydrogen bond [40].

Marked change in viscosity is observed when alumina nanoparticles are added to a 1:1 (volume) mixture of EG and water. The viscosity jumped to around 7 cP and 12.7 cP from 2.9 cP on addition of 2 wt. % and 4 wt. % of alumina, respectively, to the mixture. The reason for this marked increase in currently under investigation and will be communicated through a later publication. It is believed that citrate ion which has a higher heat of hydration (∼22.4 kJ/mol) than EG at low concentration forms hydrogen bonds with the water molecules easier than EG molecules. This leaves the system to have a large number of free EG molecules, which cannot join water, and the viscosity increases toward pure EG. The addition of particles further enhances the increase.

In the case of commercial coolant because of the addition of a small amount of water during milling the viscosity drops below that of the pure coolant for low particle concentrations (2 wt. %). As more particles are added, the viscosity goes up and matches that of the pure coolant at 6 wt. % particles loading. Although commercially available coolants are mixtures of ethylene glycol and water, they also contain additives that act as viscosity modifiers, which arrest the drastic increase of viscosity as in the case of pure EG-water mixture. It is also seen from the preliminary results that at the concentrations tested, nanofluids behave as Newtonian fluids. The values of viscosity obtained for different nanofluids at various concentrations and under different shear rates are summarized in Table 3.

The dispersions produced by the technique proposed in this paper, are found to be stable for over 30 months even after undergoing several heating and cooling cycles. The nanofluids based on commercial coolants have been tested for over 3000 heating and cooling cycles with no detrimental effect on stability. This proposed technique can produce up to 10 L of 2 wt. % alumina particles in a single batch using one planetary mill. The heat transfer properties of nanofluids thus produced have been tested as described in the previous sections. It has been found that the enhancement in overall heat transfer coefficient is a strong function of the base fluid used, concentration of nanoparticles added, flow rate of the test fluid, and the material of the heat transfer surface.

Four base fluids have been tested in the present work—water, ethylene glycol, 1:1 (volume) mixture of water and ethylene glycol and a commercially available automobile engine coolant purchased from local market. It has been found that as the thermal conductivity of the base fluid decreases, the enhancement shown by the nanofluids increases. 4 wt. % alumina-water nanofluid at a flow rate of 50 kg/h shows an enhancement of 9% for pure water while alumina-EG nanofluid under the same conditions shows an enhancement of nearly 54%, as shown in Fig. 6 [41,42].

The effect of particle loading on the enhancement of overall heat transfer coefficient is investigated. It is observed that as particle loading is increased, the enhancement in overall heat transfer coefficient also increases. This observation is consistent for all the nanofluids prepared in the present work (Fig. 7). This proves that the presence of nanoparticles is responsible for the enhancement in heat transfer properties of the fluids. Further, a nanofluid with > 4 wt. % particles loading is under investigation. It is expected that there is limit on maximum particle loading in base fluid, as particle loading will increase the viscosity and will have impact on the stability of nanofluids.

Nanofluids flow rates are tested in the laminar regime. These flow rates are selected based on the actual flow rates of a fluid through a single tube of the engine radiator. Three mass flow rates tested here are 33 kg/h, 50 kg/h, and 85 kg/h. Although the flow rates tested are in the laminar regime, it is found that the enhancement increases on increasing the flow rate for all four kinds of nanofluids tested is shown in Fig. 8.

The two heat transfer fluid streams in any process are kept in thermal contact by a heat transfer surface, which also plays the role of preventing the two fluid streams from mixing. The heat transfer surface plays an important role in the defining the efficiency of an energy process. Commonly used heat transfer surfaces include stainless steel, copper and aluminum.

To study the effects of the heat transfer surface, two double pipe heat exchangers are used; (A) stainless steel heat exchanger and (B) copper heat exchanger. The details of the two heat exchangers are summarized in Table 4 and the heat exchange circuit is as described in Fig. 1 for both the heat exchangers.

It is seen that higher the thermal conductivity of the heat transfer surface lower is the enhancement. This is attributed to the masking of the features of the heat transfer fluid by the highly conducting heat transfer surface as shown in Fig. 9. In this case copper (A) several times more conducting than the stainless steel (B) showed lower enhancement as compare to steel surface (Fig. 9).

The technique used in the present study to determine stability was to monitor the PSD of the dispersion. It is intuitive and well documented in literature that unstable dispersions of particles tend to agglomerate and settle down. Agglomeration leads to increase in the particle size. At presented in the Table 5, the PSD of as prepared and a 12 month old sample are almost identical. As presented in Fig. 10, no sedimentation could be observed when sample was visually inspected.

The present work proposes a technique for large-scale production of stable alumina nanofluids in four different base fluids such as double distilled water, ethylene glycol, mixtures of glycol, and water and a commercially available automobile engine coolant. The method, which is based on milling of coarse alumina particles in base fluid in the presence of a dispersant selected by using molecular modeling and simulations, is capable of producing around 10 L of 2 wt. % alumina nanofluids in a batch. The present work also looked at optimizing the time and dosage of addition of the selected dispersant. The set of nanofluids produced by this technique is found to be stable up to 30 months even after subjecting to several heating and cooling cycles. It has been found that with the exception of EG-water mixture, the viscosity of the dispersions do not rise by a significant amount. From the preliminary test results on rheological behavior of the nanofluids, it is indicated that the nanofluids behave like Newtonian fluids.

Convective heat transfer tests have been performed in custom made double pipe heat exchangers. Enhancement of around 10–60% has been observed. It has been seen that enhancement is higher for base fluids with poor heat transfer capabilities, higher particle loading, higher flow rates, and poorly conducting heat transfer surface.

This work demonstrates that nanofluids can be used in commercial applications by properly altering certain parameters.

Nomenclature

A =

area of heat transfer

C_p =

specific heat

E_surface =

interaction energies of free surface

E_reagent =

interaction energies of dispersant molecules

E_solvent =

interaction energy of solvent molecules

E_{complex, solvent} =

interaction energy of surface-reagent complex in presence of solvent

ΔE_{solvent-reagent} =

interaction of solvent molecule with reagent molecule

ΔE_{surface-solvent} =

interaction of surface molecule with solvent

ΔE =

final interaction of reagent molecule with inorganic surface

h =

heat transfer coefficient

$m·$ =

mass flow rate

Tp =

temperature of the fluid streams

$ΔTLMTD$ =

log mean temperature difference

T1–T4 =

four thermocouples used to measure temperature of the fluid streams

U =

overall heat transfer coefficient

Subscripts

cf =

test fluid

HW =

heating water

i =

inner heat transfer area

in =

inlet of a stream of fluid

out =

outlet of a stream of fluid