Users Online: 156
Home Print this page Email this page Small font size Default font size Increase font size
Home About us Editorial board Search Ahead of print Current issue Archives Submit article Instructions Subscribe Contacts Login 

 Table of Contents  
Year : 2018  |  Volume : 3  |  Issue : 1  |  Page : 13-21

Numerical and experimental analysis of the double-diffusive convection in a linearly stratified medium

1 Department of Mechanical Engineering, Al Imam Mohammad Ibn Saud Islamic University, Riyadh, Kingdom of Saudi Arabia
2 Department of Mechanical, Riyadh College of Technology, Riyadh, Kingdom of Saudi Arabia

Date of Submission03-Jul-2018
Date of Acceptance10-Sep-2018
Date of Web Publication19-Oct-2018

Correspondence Address:
Dr. Mohammed Almeshaal
Department of Mechanical Engineering, Al Imam Mohammad Ibn Saud Islamic University, Riyadh
Kingdom of Saudi Arabia
Login to access the Email id

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/ijas.ijas_7_18

Rights and Permissions

Introduction: In this paper, we numerically and experimentally studied the onset and the mechanisms of spatiotemporal evolution of the double convective diffusion in a saline medium linearly stratified and heated from below at a constant temperature.
Methodology: The numerical simulations were based on the solution of the Navier–Stokes,energy, and concentration equations in stream-vorticity formulation. The numerical method used is the Hermitian compact one. In the experimental studies, a transparent cavity with free surface was filled with linearly stratified solution. The solution was heated from below at a constant temperature. The convective motions were visualized using shadowgraph to obtain more information on density exchange through the interfaces. Furthermore, particle image velocimetry (PIV) is used to measure the time evolution of the velocity field.
Results: Flow visualizations by Shadowgraph and PIV allowed the comparison between experimental and numerical results.
Conclusion: It was found that a multicellular flow oscillating in space and time mechanisms was observed in both experimental and numerical studies.

Keywords: Double diffusion, flow visualization, numerical simulations, solar pond

How to cite this article:
Almeshaal M, Choubani K. Numerical and experimental analysis of the double-diffusive convection in a linearly stratified medium. Imam J Appl Sci 2018;3:13-21

How to cite this URL:
Almeshaal M, Choubani K. Numerical and experimental analysis of the double-diffusive convection in a linearly stratified medium. Imam J Appl Sci [serial online] 2018 [cited 2018 Dec 14];3:13-21. Available from:

  Introduction Top

The double-diffusive convection in a stratified systems heated from below has many important applications, such as convection in oceans, atmosphere, lakes, solar ponds,[1],[2],[3],[4],[5],[6],[7],[8],[9] and multi-component diffusion in crystal growth, coating, and casting processes.[10],[11] The phenomenon of double diffusion considered in this work is a horizontal fluid layer linearly stratified in density and heated from below. The theoretical modeling existing up-to-date needs considerable improvement. For example, for the situation in which the magnitude of the applied bottom heat is relatively low (solar heating), it is observed that single, warm, and well-mixed layer forms near the bottom. This region is separated from the overlaying fluid by a boundary layer.[12],[13]

The first elaborated models were one-dimensional (1-D); they were based on the input empirical correlations, and they offered little information on the flow structure in the system. The mechanism responsible for the heat and species transfer across the boundary layer, required for continuous growth of the mixed layer, is not well understood.[14] Hence, models based on 2-D governing equations were used to explain the behavior of a mixture of a fluid and a chemical species whose initial density distribution decreases linearly from the bottom to the upper surface, where the fluid layer is subjected to sudden heating at the bottom surface.

Ungan and Bergman,[15] considered a similar problem. However, they imposed a symmetrical condition and considered only the system which, numerically, restricts the physical stability of the problem. However, in unstable problems and due to the nonlinearities, the symmetry can be broken. In the fluid dynamics area, one can find a good number of asymmetrical solutions, such as the problem under consideration. The use of symmetry condition is questionable. The transient evolution of the phenomenon depends on the initial perturbation imposed on the whole system.

Kazmierczak and Poulikakos,[16] investigated numerically the transient double diffusion phenomena in a stably stratified fluid layer (water and salt) heated from below. The complete form of the governing equations, subjected to the Boussinesq approximation, was obtained and solved. A constant heat flux condition was imposed at the bottom wall. It is found that this kind of boundary condition generates a single well-mixed flow region, which grows with time. The growth of the bottom region and the increase of the local and average bottom wall temperature were investigated. Furthermore, authors analyzed the effect of the geometric aspect ratio (H/L) of the system as well as the effect of the ratio of the thermal to the solutal Rayleigh number (RaT/Ras) on the evolving temperature, concentration, and flow fields. Kazmierczak and Poulikakos,[16] concluded that the geometric aspect ratio of the system has a little impact on the phenomenon but the increase of the stability parameter RaT/Ras from 3 to 10 initiates significant 2D effects on the temperature and the concentration fields.

On the other sides, several theoretical and experimental studies have been conducted on solar ponds and their potential applications in heating, power generation, and desalination. The main aim of these investigations was to understand the thermal behavior of these ponds as both solar energy collectors and heat stores under different operating conditions.[17],[18] Details on principle, types, and characteristics of solar ponds can be found in the review by Kalogirou.[19] The stability of salt-gradient solar ponds was studied by Ben Mansour et al.[20] using a 2-D, transient, variable property diffusive model under the climatic conditions of Tunis city, Tunisia. It was found that the presence of heat extraction with its cooling effect tends to stabilize the pond. Giestas et al.[21] presented a numerical study based on the Navier-Stokes equations to investigate effect of solar radiation absorption on the stability of solar pond. In another study, Giestas et al.[22] analyzed effects of nonconstant diffusivities on the stability of the pond. Suarez et al.[17] analyzed numerically the problem of 2D-coupled transient double diffusion-convection for salt gradient solar ponds. They incorporated latent heat (evaporation) and sensible heat losses, solar shortwave, and longwave radiation fluxes at water-air interface.

In order to obtain an improved understanding of dynamic processes in the Solar Pond Gradient Zone (SPGZ), numerical and experimental studies have been conducted. The goal of the present study is to investigate the onset and the mechanisms of spatiotemporal evolutions of the double-diffusive convection in a saline medium linearly stratified and heated from below at a constant temperature. These conditions simulate the behavior of the fluid flow with heat and mass transfer within an SPGZ. Visualizations by Shadowgraph and PIV allowed comparisons with numerical simulations results.

  Mathematical Formulation Top

The system of interest is shown schematically in [Figure 1]. The density of a mixture of water and salt decreases linearly from the bottom to the top. The sidewalls of the system are impermeable to heat and concentration.
Figure 1: System of interest

Click here to view

Initially, the fluid is isothermal and at rest. Suddenly, a constant temperature is applied at the bottom wall and starts warming up the system. Because of the scenario described above, heat, species, and fluid transports are initiated in the system.

By neglecting thermocapillary effect and Dufour and Soret secondary effects, the equations governing these transport phenomena in unsteady state conditions are as follows:

Continuity equation:

Momentum equation:

Energy equation:

Salinity equation:

In accordance with the usual Boussinesq approximation, the fluid density and thermos-physical properties are assumed to be constant everywhere except in the buoyancy force term of the momentum equation (Equation 2), where the density follows the linear state equation:

The boundary conditions are:

The initial conditions are:

Before attempting to solve the system of Equations 1-4, it is convenient to replace these equations in dimensionless form and to write them with respect to the Cartesian system of [Figure 1].

The dimensionless parameters needed to perform this task are defined as:

To overcome the problem encountered by calculation of the pressure P and to reduce the number of equations, stream function Ψ and vorticity function Ω are used to replace the primitive variables U, V, and P. The two dimensionless functions Ψ and Ω, are introduced as:

Note that by definition, the stream function satisfies the continuity equation (1). Furthermore, the two momentum equations can be combined to eliminate the pressure gradient and obtain a unique vorticity equation.

After the mentioned manipulations, the dimensionless governing equations and boundary and initial conditions become as:

  Numerical Simulation Top

The governing equations (12–19) were solved numerically. Variety of methods is available in literature, and CFD software makes it possible to simulate a large range of industrial problems. In our case, the set of coupling equations (12–15) are solved using a difference finite scheme based on a compact Hermitian method where the function and its first and second derivatives are considered as unknowns. The method is fourth order accurate, O(h 4), for Ψ and second-order accurate, O(h 2), i, in Ω, T, and S.

The Alternate Direction Implicit technique is used to integrate the parabolic equations. Since the scheme is well documented in the literature and has been widely used for natural convection and recirculation zones, there is no need for elaboration. The procedure has the advantage that the resulting tridiagonal matrix instead of a matrix with five occupied diagonals can easily be solved by factorization algorithm.

Some difficulties were encountered in implementing the vortices boundary conditions, but different approximations were tested, and the Padé approximation was employed to overcome the numerical instability.[23]

The set of equations is solved using an unsteady numerical procedure [Figure 2]. At each step of time:
Figure 2: Algorithm Resolution of unsteady numerical procedure in (Ψ,Ω,T,S) formulation: IROUT: Outer iteration number, IRINMAX: Max inner iteration number, IROUMAX: Max outer iteration number, IRINMAX: Max inner iteration number, △τ: Step time, (△X, △Y): Step space in horizontal and vertical position, (NX,NY): Grid nodes, TESTΨ: Convergence test of Ψ

Click here to view

The convergence of the stream function equation is achieved using internal iterations. The convergence criterion are defined by the following relation:

For the vorticity, temperature and salinity functions, the convergence criteria are:

χ is equal to 10−6 for the stream function Ψ and 10−4 for the Ω, T and S. [Figure 2] shows the flow diagram, i.e., algorithm of the solution.

  Numerical Results Top

The numerical simulations were conducted for the configuration [Figure 1] using the following parameters:

  • H = 2.10−2 m, L = 10−1 m,
  • A salt solution with △C0= 10% of mass,
  • A difference temperature: △T0=(40–20)= 20°C,
  • The values of RaT, Pr, Le, N, and Bi were, respectively:
  • RaT= 6.1x106, Pr = 4.68, Bi = 0.2, Le = 73, N = 7.

Several grid sizes were tested to ensure that results are independent of the mesh size. A grid of 101 × 21 nodes was selected and used.

Results of the computations are presented in the form of contours plots of stream function and density distribution at different times [Figure 3]. It was found that:
Figure 3: Spatiotemporal evolution of (a) density and (b) streamlines (numerical results)

Click here to view

  • At the beginning (t = 20 s), small vortices appeared in the bottom of the cavity while density remained stable
  • After some time (t = 30 s), the number of vortices increases in the vertical direction [Figure 3]b inducing the acceleration of the diffusion of concentration. At the same time, density started to oscillate over the bottom surface [Figure 3]a
  • As the time further increased, the recirculation zone extends to the horizontal direction, and density oscillation starts to be dumped in the vertical direction.

[Figure 4] shows the dimensionless density distribution for three positions of y (upward:19/20, 1/2, and 1/20).
Figure 4: Iso-density distribution with different position of y (numerical results)

Click here to view

As the time increases, the appearance of the oscillatory movements near the bottom surface (y = 1/20) starts.

One can notice that the heat and species are exchanged by molecular diffusion in the beginning of the process. Then, vortices appear and convective movement takes place inside the liquid, and a double diffusion starts. The convective rolls are ejected from the bottom to upper surface but are quickly dumped in the vicinity of the hot plate. As the time increases, these rolls coalesce and give rise to a well-mixed layer of a limited thickness.

  Experimental Setup Top

All experiments were carried out in a rectangular Plexiglas tank with free surface and with inner dimensions of 20.10−2 m × 2.10−2 m. The sides of the tank, except for a window of (10.10−2 m × 2.10−2 m), necessary for flow visualization, were insulated. The tank was fitted in the bottom with an aluminum heat exchanger thermally insulated, through which circulates water coming from an adjustable-thermostat to provide heating at a constant temperature over the base. The instruction-temperature is obtained with a precision of ±0.2°C. A photograph of the entire apparatus is shown in [Figure 5]a, and a schematic sketch is depicted in [Figure 5]b.
Figure 5: (a) Experimental setup, (b) Sketch of experimental setup

Click here to view

The desired linear salinity gradients were set up using a modified two-tank technique of Oster.[24] The resulting concentration was varying from Cb (high concentration) at the bottom to zero at the surface. The temperature profile of the stratified solution was measured by copper-constantan thermocouples. The diameter of each thermocouple is about 0.2 mm. Temperature measurements were obtained with a precision of ±0.8%. Thermocouples were positioned at 4 mm intervals along the mid vertical axis, except for the lowest one which had been set at 2 mm above the bottom of the cell to depict possible 3-D regimes. All the thermocouples were connected to an Agilent data logger. The measurements of density were made using a single-point conductivity probe of constant 0.954. Interface location and layer formation were monitored with a shadowgraph system. It uses the fact that the index of refraction of fluids is typically dependent on temperature.

The liquid was seeded with particles tracer. The illumination of the particles was provided by a plane-light-sheet, generated by expanding the beam of a 20 mW He-Ne laser by means of a cylindrical lens. This sheet can move parallel to the walls of the cell to depict a probably 3-D regimes. Images were recorded by a CCD camera of resolution 578 × 512 pixels and 25 FPS. Images will be stored on a computer hard disk. Fluid velocities are determined by the displacements of the seeded particles at a given time separation, with an error of 10−7 m/s.

  Experimental Results Top

Experimental conditions associated with the linearly salt-stratified solution heated from below at a constant temperature are fixed by: A = 5, △T0= 20°C, Salt solution = (10% of mass).

Shadowgraph visualization [Figure 6] shows that once the heating begins, a vertical oscillatory disturbance occurs in the bottom layer. This disturbance grows with time. This oscillatory motion seems to be the main mechanism responsible for the growth of the first mixed layer which strongly influences overall system performance. Indeed, when the mixed layer grows to a critical height, the interface behaves like a wall. The heat seems to be localized in the bottom layer. Hence, upper layer will be heated from the below at a constant temperature using a fictive base plate formed by a moving density interface. Thereby, the interface is so energetic to be shared by oscillatory motion, but it will be pushed up, and the bottom layer rises. As the bottom layer grows, the same mechanism is reproduced in the upper edge of the interface; the later motion will be prompted. The motion appeared in the upper edge of the interface does not persist for a long time. This leaves the interface eroded essentially from the lower front. Since the convection in the lower layer is stronger than that in the upper layer, the differential turbulence levels across the interface cause its migration upward while manifesting net entrainment from the upper layer.[3]
Figure 6: Shadowgraph visualization

Click here to view

Observations of the entrainment process during the heating of a nonlinear salinity gradient from below, lead us to distinguish two behaviors:

  1. An oscillatory regime corresponding to the deformation of the interface
  2. A steady two-layer regime where the interface is affected by an oscillatory motion in both lower and upper edges. In this regime, the density stratification acts as a barrier to motions that try to penetrate through it.

PIV visualization shows the appearance of eddies in the homogeneous layers [Figure 7]. These eddies are responsible for the convective mixing, and they induce a significant shearing on the edges of the interface.
Figure 7: Particle image velocimetry visualization of the vortices at the bottom of the stratified layers

Click here to view

It is of interest to mention that these observations based on the experimental tests come to confirm the above numerical results. Similar flow structures have been observed using both studies. For instance, numerical and experimental results showed that vortices appeared in the bottom then were travelled up, but not exceeding a fixed distance. During this process, density oscillated in space and time and then coalesces to form the first mixed layer. Time variations of the main variables of this problem obtained numerically have been confirmed by experiments.

  Conclusion Top

This study presented numerical and experimental results of the double-diffusive convection in a linearly stratified medium of a salt solution heated from below at a constant temperature. Initially, the salt solution was at rest, isothermal, and its density decreased linearly with distance from the bottom wall.

Numerical results, as well as experimental, showed that vortices appeared in the bottom then were travelled up, but not exceeding a fixed distance. During this process, density oscillated in space and time and then coalesces to form the first mixed layer.

Experimentally, we distinguish three steps of onset of the first mixed layer:

  1. The onset, of aperiodic oscillations moving in space and time
  2. A well-mixed layer is born, and an oscillation movement appears at the free surface later, the average wavelength is of the same magnitude of the stratified layer thickness. These oscillations become more structured with a given period and continue to propagate in the horizontal direction. During this phase, a well-mixed layer appears with an upper interface thin layer. At the same time, a similar oscillation movement appears at the free surface and moves toward the bottom interface where it merges: a new interface was born
  3. This new interface becomes thinner and quasi-plane, consequently, the stratification is nonlinear, and it was governed by the immigration of the single interface upward. This interface oscillated in time and space until the final homogenization of the whole system.

These results agreed with precedent ones but provided more details on the mechanism of double diffusion.


Greek symbols




Nondimensional numbers

Nondimensional quantities

Financial support and sponsorship


Conflicts of interest

There are no conflicts of interest.

  References Top

Akrour D, Tribech M, Kalache D. A theoretical and numerical study of thermo-solutal convection: Stability of a salinity gradient solar pond. Thermal Sci 2011;80:15-67.  Back to cited text no. 1
Anderson MD, Mcfadden BG, Coriell RS, Murray TB. Convective instabilities during the solidification of an ideal ternary alloy in a mushy layer. J Fluid Mech 2010;647:309-33.  Back to cited text no. 2
Choubani K, Zitouni S, Charfi K, Safi MJ, Akbarzadeh A. Experimental study of the salt gradient solar pond stability. Sol Energy 2010;84:24-31.  Back to cited text no. 3
Choubani K, Safi MJ. Experimental simulation of the salt gradient solar pond stability. J Renew Sustain Energy 2012;4:063143.  Back to cited text no. 4
Gargett AE, Tanny J, Kelley DE, Fernando HJ, Ozsoy E. The diffusive regime of double-diffusive convection. Prog Oceanog 2003;56:461-81.  Back to cited text no. 5
Giestas MC, Milhazes JP, Pina H. Numerical modeling of solar ponds. Energy Procedia (Elsevier) 2014;57:2416-25.  Back to cited text no. 6
Mishra M, Trevelyan PM, Almarcha C, De Wit A. Influence of double diffusive effects on miscible viscous fingering. Phys Rev Lett 2010;105:204501.  Back to cited text no. 7
Scmid M, Busbridge M, Wüest A. Double-diffusive convection in Lake Kivu. Am Soc Limnol Oceanogr 2010;55:225-38.  Back to cited text no. 8
Thompson FA, Veronis G. Diffusively-driven overturning of a stable density gradient. J Marine Res 2005;63:291-313.  Back to cited text no. 9
Chen CF, Turner JS. Crystallization in a double-diffusive system. J Geophys Res 1980;85:2573-93.  Back to cited text no. 10
Ostrach S, Fluid mechanics of crystal growth-The Freeman Scholar Lecture, ASME, Transactions, Journal of Fluids Engineering 1983;105:5-20.  Back to cited text no. 11
Meyer KA. A numerical model to describe the layer behaviour in salt-gradient solar ponds. ASME J Solar Energy Engng 1983;25:63-75.  Back to cited text no. 12
Bergman TL, Incropera FP, Viskanta RA. Differential model for salt-stratified, double-diffusive systems heated from below. Int J Heat Mass Transfer 1985;28:779-88.  Back to cited text no. 13
Incropera FP. Buoyancy effects in double-diffusive and mixed convection flows. Proc Int Heat Transfer Conf 1986;1:121-30.  Back to cited text no. 14
Ungan A, Bergman TL. A two-dimensional numerical simulation of a linear stratified thermohaline system heated from below. Paper No. 87-WA/HT-19. ASME Winter Annual Meeting, Boston, MA; 1987.  Back to cited text no. 15
Kazmierczak M, Poulikakos D. Transient double diffusion in a stably stratified fluid layer heated from below. Int J Heat Fluid Flow 1990;11:30-9.  Back to cited text no. 16
Suarez F, Tyler SW, Childress AE. A fully coupled, transient double-diffusive convective model for salt-gradient solar ponds. Int J Heat Mass Transfer 2010;53:1718-30.  Back to cited text no. 17
Ben Mansour R, Nguyen CT, Galanis N. Numerical study of transient heat and mass transfer and stability in a salt-gradient solar pond. Int J Therm Sci 2004;43:779-90.  Back to cited text no. 18
Kalogirou S. Seawater desalination using renewable energy sources. Prog Energy Combust Sci 2005;31:242-81.  Back to cited text no. 19
Ben Mansour R, Nguyen CT, Galanis N. Transient heat and mass transfer and long-term stability of a salt-gradient solar pond. Mech Res Commun 2006;33:233-49.  Back to cited text no. 20
Giestas M, Pina H, Joyce A. The influence of radiation absorption on solar pond stability. Int J Heat Mass Transfer 1996;39:3873-85.  Back to cited text no. 21
Giestas M, Pina H, Joyce A. The influence of non-constant diffusivities on solar ponds stability. Int J Heat Mass Transfer 1997;40:4379-91.  Back to cited text no. 22
Safi MJ, Loc TP. Development of thermal stratification in two dimensional cavity a numerical study. Int J Heat Mass Transfer 1994;37:2017-24.  Back to cited text no. 23
Oster G. Density gradients. Sci Am 1965;213:70-6.  Back to cited text no. 24


  [Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7]


Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

  In this article
Mathematical For...
Numerical Simulation
Numerical Results
Experimental Setup
Experimental Results
Article Figures

 Article Access Statistics
    PDF Downloaded52    
    Comments [Add]    

Recommend this journal