Abstract
This study is a numerical study on the coagulation as well as the sedimentation effect of nanofluids using the Brownian dynamics method. Three cases are simulated, focusing on the effects of the sizes, volume fraction, and ζ potentials of nanoparticles on the formation of coagulation and sedimentation of nanofluids. The rms fluctuation of the particle number concentration, as well as the flatness factor of it, is employed to study the formation and variation of the coagulation process. The results indicate a superposition of coagulation and sedimentation effect of small nanoparticles. Moreover, it is stable of nanofluids with the volume fraction of particles below the limit of "resolution" of the fluids. In addition, the effect of ζ potentials is against the formation of coagulation and positive to the stability of nanofluids.
Introduction
The nanofluid is characterized by the fluid with nanometersized solid particles dispersed in solution [1], which can increase the heat transfer coefficient [26], enhance the critical heat flux in boiling heat transfer [79], reduce the wall friction force [10], improve the optical characteristics [11], etc. Nanosized particles are utilized because of its better stability than the suspension of microsized particles. For a badly stable suspension, sedimentation or coagulation (agglomeration) may occur. It compromises the abovementioned advantages of the nanosuspension.
As is well known [12], the occurrences of coagulation and sedimentation are the two main factors for the instability of nanofluid. The phenomenon of coagulation is characterized by the formation of particle clusters, i.e., particles are in contact with each other and the cohesion takes place. Then, the clusters grow up. Many researchers investigated the coagulation effect of particles by the Brownian dynamics simulation, focusing on the formation of gelation [13], coagulation rates [14], particle network [15], etc. For example, Hütter [14] identified the characteristic coagulation time scales in colloidal suspensions, and measured their dependencies on the solid content and potential interaction parameters. He also deduced different clustercluster bonding mechanisms in the presence of an energy barrier, etc. Besides, the sedimentation always occurs after a big particle cluster is established, i.e., the particles within the cluster sediment flow downward because of the increased effect of the gravity of the cluster over the buoyancy force of it, and reduced the effect of Brownian motion to the big cluster. Many researches were devoted to the sedimentation [1618] using the Brownian dynamics simulation too. For example, Soppe and Jannsen [17] studied the sediment formation of colloidal particle by a process of irreversible singleparticle accretion. They used the algorithm of Ermak and McCammon, incorporating the interparticle forces and hydrodynamic interaction on the twoparticle level, and analyzed the effect of twoparticle hydrodynamic interactions on the sediment structure, etc. They found that the process of sediment formation by colloidal particle is the result of a delicate balance of sediment field strength, DLVO interactions, and hydrodynamic interactions.
However, there is an important issue about which few researches have been concerned: the interaction between the coagulation and sedimentation for the instability of nanofluids. For example, the processing of coagulation causes the particle clusters to grow up, and then the large clusters are more prone to sedimentation than that of small clusters because of the intensive gravity effect. In other words, the coagulation effect is able to augment the sedimentation effect. Thus, this study is intended to carry out some research on this issue, exploring the complex interaction as well as the close relation between the coagulation and the sedimentation phenomena.
Numerical model
Governing equation
In this study, the Brownian dynamics technique is employed to investigate the motion of nanoparticles. The governing equation is the socalled Langevin equation, which is formulated as follows [19]:
where the superscript 0 indicates that the variable is corresponding to the beginning of the time step Δt; r_{i }is the ith component of the position vector of particle; D_{ij }is the element of the diffusion tensor indexed by (i, j); F_{j }is the force experienced by the jth particle; k_{B} is the Boltzmann constant; and T denotes the temperature. The displacement R_{i}(Δt) is a random displacement of a Gaussian distribution with a zero expectation and a () covariance (). In this study, the RotnePrager tensor [20] is utilized as approximations to the hydrodynamic interaction:
where η is the viscosity, a is the particle radius, δ_{ij }is the Kronecker delta, is the vector from the center of particle i to the center of particle j, and is the unit tensor.
Moreover, three forces, i.e., the attractive Van der Walls force, f_{v}, the repulsive electrostatic force by the electric double layer, f_{e}, and the gravity force f_{g} are considered as the forces experienced by any particle, which are formulated as follows [12]:
where A, d, ε, κ, ζ, ρ_{f}, ρ, and g are the Hamaker constant, the particle diameter, the electric permittivity of the fluid, the inverse of the doublelayer thickness, the zeta potential of the suspension, the density of the fluid, the density of the particle, and the gravity acceleration, respectively.
It is noted in Equation (4) that the results for r_{ij } d = 0 is meaningless when the contact between the two particles occur, and they will adhere to each other or rebound back. Thus, we treat the condition with as the situation when the two particles are separated, so that Equation (4) works. Otherwise, it results in coalescence between the two particles. Once the coalescence between colliding particles takes place, the clusters start growing up.
Simulation conditions
In this study, three cases with different diameters of particles (Case 1), different volume fractions (Case 2), and different zeta potentials (Case 3) are simulated, respectively (Table 1). For these cases, the parameters of the material, as well as other relevant parameters, are illustrated in Table 2.
Table 1. Three cases with different diameters of particles, volume fractions, and zeta potentials
Table 2. Parameters used in this simulation
In this simulation, the boundary conditions in the x and y directions (Table 2) in the horizontal plane are both periodic, whereas the top and bottom walls of the simulation domain in the zdirection are treated as adhesive walls to which the particles adhere immediately once they come into contact with them. It is reasonable to conclude thus, since the agglomerated particle clusters always adhere to the bottom walls or the top interfaces.
Initially, a random distribution is given to the particles. As time advances, the possible movements of particles are computed through solution of the governing equations.
Simulation results
Case 1: effect of particle sizes
This section deals with the effect of particle sizes on the coagulation and sedimentation. Figure 1 shows the simulation results at t = 0, 5, 10, and 50 μs for Case 1 due to the effects of different sizes of particles. Figure 1a, b, c, d shows the results of d = 10 nm at t = 0, 5, 10, and 50 μs, respectively. Similarly, Figure 1e, f, g, h shows the results for d = 25 nm, whereas Figure 1i, j, k, l shows them for d = 50 nm. It is seen that the coagulation takes place the most intensively and rapidly for the smallest size of particles (Figure 1a, b, c, d), moderately for the intermediate size of particles (Figure 1e, f, g, h), and weakly and slowly for the largest size of particles (Figure 1e, f, g, h, i, j, k, l). More importantly, the results of the intermediate sizes are due to coagulation but with weak sedimentation, whereas the results of the smallest sizes are due to both the effects of coagulation and sedimentation. For the largest size of the particles, it is neither due to coagulation nor sedimentation. It looks complicated. As is known, the larger particles bear the major effect of gravity, and they are the most prone to sedimentation. However, it is only true of the single particle without coagulation. With the superposition of the coagulation effect, it can amplify or augment the trend of sedimentation through coagulation. Owing to the increasing agglomeration of the particles, the gravity effect may play an important and even a dominant role, which causes possible sedimentation of the whole agglomeration (the upper part of the agglomeration in Figure 1d is due to the adhesive boundary on the upper wall). In other words, there exists a balance between the sedimentation effect of the largesized individual particles and the sedimentation effect of smallsized aggregated particles. The former is caused solely by the gravity effect, whereas the latter is caused by the superposition of the coagulation and the gravity effects, i.e., the amplification and augmentation of the gravity effects of the aggregated particles due to coagulation.
Figure 1. Snapshots of simulation results. (Case 1 for d = 10 nm (ad), d = 25 nm (eh), d = 50 nm (il) at t = 0, 5, 10, and 50 μs respectively).
It is necessary to mention that Figure 1i, j, k, l does not indicate the stability of the nanofluids. Alternatively, it indicates a relatively stable characteristic compared to Figure 1a, b, c, d. After the evolution over a long time, possible coagulation or sedimentation may also occur.
In order quantify the degree of coagulation, we need to define some functions. Let us divide the simulation domain L_{x }× L_{y }× L_{z }by N_{x }× N_{y }× N_{z }cubic meshes by the cell volume (δ_{x }× δ_{y }× δ_{z}). The mean number concentration of particles is the mean number of particles within each mesh volume (δ_{x }× δ_{y }× δ_{z}). Then, the rms value of the particle concentration R_{1} and the flatness factor of the number concentration R_{4} are formulated as follows:
The rms of concentration means the fluctuation of the number concentration of particles, and it is closely related to the formation of particle clusters due to coagulation. The flatness factor means the intensity of fluctuation of the number concentration, thereby indicating the intensity of coagulation. Thus, these two functions are helpful in enabling the quantification of particle coagulation.
Figure 2a, b shows the R_{1} and R_{4} for Case 1. It is seen from Figure 2a, b that the coagulation of the small particles is the fastest. They become almost totally coagulated immediately even at the beginning. Comparatively, the coagulation of the larger particles takes place slowly and increasing steadily. However, the final level of coagulation of the larger particles is greater than that of smaller particles. In addition, the degree as well as the rapidity of the coagulation of the intermediate particles is intermediate between that of the smaller and the larger particles.
Figure 2. The flatness factor of the concentration distribution of nanoparticles. (The R_{1}(a) and R_{4}(b) for Case 1).
Case 2: effects of volume fractions
In this section, the effect of volume fraction, i.e., the concentration of particles, is studied. As aforementioned, the smaller particles are more prone to coagulate than the larger particles, under the same condition of the volume fractions. However, the process of coagulation is also closely related to the number of particles contained in it.
For example, Figure 3a, b, c, d shows one of the results of Case 2 where only n_{p }= 400 particles are simulated. Compared to Figure 1a, b, c, d, it is seen that the coagulation does not takes place at t = 50 μs. It says that the coagulation is to be regarded appropriately as a process with an excess of particle content. When the particle numbers go beyond the superior limit of the resolvent, then the coagulation will certainly take place. Thus, when the n_{p }= 4200 particles are simulated, more intensive coagulations are observed correspondingly (Figure 3e, f, g, h).
Figure 3. Snapshots of simulation results. (Case 2 for n_{p }= 400 (ad) and n_{p }= 4200 (eh) at t = 0, 5, 10, and 50 μs respectively).
In addition, Figure 4a, b shows the R_{1} and R_{4} of Case 2. It is seen from Figure 4 that the R_{1} and R_{4} for n_{p }= 400 are always relatively of small value, indicating a stable status almost without coagulation and sedimentation, although R_{4} is slightly fluctuated when t < 0.06. Moreover, compared to n_{p }= 1200, it is seen that the concentration fluctuation R_{1} and flatness factor of concentration R_{4} for n_{p }= 400 are relatively of lower values. It validates the conclusions derived from the observation of Figure 3a, b, c, d, in comparison with Figure 1e, f, g, h.
Figure 4. The flatness factor of the concentration distribution of nanoparticles. (The R_{1}(a) and R_{4}(b) for Case 2).
With the increased number of particles, it is seen that the R_{1} and R_{4} are increased too (n_{p }= 1200 and 2100, respectively, Figure 4). However, when the particle number is extremely large, all the spaces are almost stuffed with particles, leading to a homogeneous distribution and a low fluctuation in the number concentration (n_{p }= 4200, in Figure 4).
Case 3: effects of ζ potentials
The previous sections showed the results with ζ = 0 eV. As seen from Equation (5), no repulsive effect has been considered between the particles since f_{e} = 0. Thus, this section will focus on the effect of the repulsive effect by varying the ζ potentials.
Comparing with Figure 1e, f, g, h, it is seen that the degree of coagulation with ζ = 0.01 eV is attenuated (Figure 5a, b, c, d), and it almost disappears with ζ = 0.05 eV (Figure 5e, f, g, h). It indicates that the repulsive effect induced by the ζ potentials is beneficial to the stability of nanofluids, since it acts against the coagulation process.
Figure 5. Snapshots of simulation results. (Case 3 for ζ = 0.01 eV (ad) and ζ = 0.05 eV (eh) at t = 0, 5, 10, and 50 μs, respectively).
This conclusion is also validated by the variations of R_{1} and R_{4} (Figure 6a, b, respectively). It is seen that the fluctuation of the number concentration of particles with ζ = 0.05 eV increases much more slowly than the cases with smaller ζ. Although the R_{1} and R_{4} for ζ = 0.05 eV are still increasing with time, which indicates that the coagulation process may still take place after a fairly long time, their effects against the formation of coagulation are very clear. In other words, it is positively beneficial for the stability of nanofluids.
Figure 6. The flatness factor of the concentration distribution of nanoparticles. (The R_{1}(a) and R_{4}(b) for Case 3).
Conclusions
The findings of this study are briefly summed up as follows:
1. A complicated superposition of the coagulation and sedimentation effects for small particle is observed. The mechanisms of sedimentation for the larger and the smaller particles are different. The former is caused mainly by the great gravity effect of any individual particle, whereas the latter is mainly due to the coagulation process, and the superposition of coagulation causes the sedimentation of the whole agglomeration of particles.
2. There exists a superior limit of the fluid for particle content. When the volume fraction is below the limit, it is hard for the coagulation to occur. In contrast, the coagulation will certainly take place when the concentration of nanoparticles is beyond the capacity of "resolution" of the fluids.
3. The effect of ζ potentials is beneficial for the stability of nanofluid, since it resists the formation of coagulation. In other words, increase in the value of ζ potentials is helpful to make the nanofluid more stable.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
All authors contributed equally.
Acknowledgements
This study is supported by the National High Technology Research and Development of China 863 Program (2007AA05Z254), for which the authors are grateful.
References

Choi SUS: Enhancing thermal conductivity of fluids with nanoparticle.

Wang X, Xu X, Choi SUS: Thermal conductivity of nanoparticle fluid mixture.
J Thermophys Heat Transfer 1999, 13:474480. Publisher Full Text

Lee SP, Choi SUS, Li S, Eastman JA: Measuring thermal conductivity of fluids containing oxide nanoparticles.
Trans ASME 1999, 121:280289. Publisher Full Text

Keblinski P, Eastman JA, Cahill DG: Nanofluids for thermal transport.
Mater Today 2005, 8:3644. Publisher Full Text

Eastman JA, Phillpot SR, Choi SUS, Keblinski P: Thermal transport in nanofluids.
Annu Rev Mater Res 2004, 34:219246. Publisher Full Text

Yu W, France DM, Routbort JL: Review and comparison of nanofluid thermal conductivity and heat transfer enhancements.
Heat Transfer Eng 2008, 29:432460. Publisher Full Text

You MS, Kim JH: Effect of nanoparticles on critical heat flux of water in pool boiling heat transfer.
Appl Phys Lett 2003, 83:33743376. Publisher Full Text

Xue HS, Fan JR, Hong RH, Hu YC: Characteristic boiling curve of carbon nanotube nanofluid as determined by the transient calorimeter technique.
Appl Phys Lett 2007, 90:184107. Publisher Full Text

Kim H, Buongiorno J, Hu LW, McKrell T: Nanoparticle deposition effects on the minimum heat flux point and quench front speed during quenching in waterbased alumina nanofluids.
Int J Heat Mass Transfer 2010, 53:15421553. Publisher Full Text

Hu ZS, Dong JX: Study on antiwear and reducing friction additive of nanometer titanium oxide.
Wear 1998, 216:9296. Publisher Full Text

Zhang ZK, Cui ZL: NanoTechnology and NanoMaterial. Beijing: National Defense Industry Press; 2001.

Dobias B, Qiu X, Rybinski WV: SolidLiquid Dispersions. New York: Marcel Dekker, Inc; 1999.

Whittle M, Dickinson E: Brownian dynamics simulation of gelation in soft sphere systems with irreversible bond formation.

Hütter M: Coagulation rates in concentrated colloidal suspensions studied by Brownian dynamics simulation.

Hütter M: Local structure evolution in particle network formation studied by Brownian dynamics.
J Colloid Interface Sci 2000, 231:337350. PubMed Abstract  Publisher Full Text

Ansell GC, Dickinson E: Sediment formation by Brownian dynamics simulation: effect of colloidal and hydrodynamic interactions on the sediment structure.
J Chem Phys 1986, 85:40794086. Publisher Full Text

Soppe WY, Jannsen GJM: Brownian dynamics simulation of sediment formation of colloidal particles.
J Chem Phys 1992, 96:69967004. Publisher Full Text

Nuesser W, Versmold H: Brownian dynamics simulation of sedimented colloidal suspensions.
Mol Phys 1999, 96:893903. Publisher Full Text

Ermak DL, McCammon JA: Brownian dynamics with hydrodynamic interactions.
J Chem Phys 1978, 69:1352. Publisher Full Text

Rotne J, Prager S: Variational treatment of hydrodynamic interaction in polymers.
J Chem Phys 1969, 50:4831. Publisher Full Text