The chosen geometry is a square cavity differentially heated with a side of 0.05 m. It is formed by two vertical walls, each of which is heated at its center. The left active part has a higher temperature ( T r + Δ T / 2 ) than that of the right active part ( T r − Δ T / 2 ) as shown in Figure 1. The boundary conditions are of Dirichlet type (imposed temperatures). The direction x is normal to the vertical walls, the gravity is carried by the axis y.

In Figure 1: s = 0.025 m, represents the height of each heated zone; A/4 = l = 0.0125 m―the height of the adiabatic zone; A = 0.05 m―the side of the square cavity; Tg and Td―the temperatures imposed respectively on the central halves of the left and right vertical walls. Convective heat transfer is a very complex phenomenon, and it depends on the boundary conditions (dynamic and thermal) of the system under consideration.

The dynamic boundary conditions are:

- to the walls, u = 0 m/s and v = 0 m/s.

The thermal boundary conditions give:

- for heated areas, we have:

for Ra = 1000 (Tg = 295.027 K and Td = 294.973 K),

for Ra = 10,000 (Tg = 295.27 K and Td = 294.73 K),

for Ra = 100,000 (Tg = 297.7 K and Td = 292.3 K),

for Ra = 1,000,000 (Tg = 322 K and Td = 268 K).

- for the adiabatic zone, φ = 0 W/m².

The flow is governed by the continuity equation, the equations of motion and the energy equation under the Boussinesq hypothesis, with: U r = α / A for velocity, A for length, A / U r = A 2 / α for time and θ = ( T − T r ) / Δ T corresponds to the dimensionless temperature.

∂ U ∂ X + ∂ V ∂ Y = 0 (1)

∂ U ∂ τ + U ∂ U ∂ X + V ∂ U ∂ Y = − ∂ P ∂ X + P r ( ∂ 2 U ∂ X 2 + ∂ 2 U ∂ Y 2 ) (2)

∂ V ∂ τ + U ∂ V ∂ X + V ∂ V ∂ Y = − ∂ P ∂ Y + P r ( ∂ 2 V ∂ X 2 + ∂ 2 V ∂ Y 2 ) + R a ⋅ P r ⋅ θ (3)

∂ θ ∂ τ + U ∂ θ ∂ X + V ∂ θ ∂ Y = ∂ 2 θ ∂ X 2 + ∂ 2 θ ∂ Y 2 (4)

We have the following boundary conditions (hydrodynamic and thermal) at the walls:

X = 0 , (Heated part of the left wall), U = V = 0 , θ = 1. (5)

X = 1 , (Heated part of the right wall), U = V = 0 , θ = 0. (6)

X = 0 and X = 1 , (Adiabatic parts of left and right walls), U = V = 0. (7)

∂ θ ∂ X = 0. (8)

Y = 0 and Y = 1 , (Lower and upper walls), U = V = 0. (9)

∂ θ ∂ Y = 0. (10)

The dimensionless numbers used are:

The Prandtl number is physically defined as ratio of momentum diffusivity to thermal diffusivity. It provides a measure of the efficiency of diffusion transport through the velocity boundary layer and the thermal boundary layer.

P r = ν α (11)

The Rayleigh number is physically the ratio of buoyancy and viscosity forces multiplied by the ratio of momentum and thermal diffusivities. Below the critical value of the Rayleigh number, heat transfer is primarily due to conduction. Above this critical value, heat transfer is due to convection.

R a = g β Δ T A 3 ν α (12)

The results produced by our calculation code were compared with some results available in the literature. Let’s consider the same model as the benchmark [7] . Table 1 summarizes the maximum horizontal velocity on the vertical mid-plane of the cavity (together with its location) and the maximum vertical velocity on the horizontal mid-plane of the cavity (together with its location). We have also the maximum value of the local Nusselt number on the boundary at X = 0 (together with its location) and the minimum value of the local Nusselt number on the boundary at X = 0 (together with its location). They are compared with those obtained by different authors G. De Vahl Davis [7] (1983) and P. M. Gresho (1980) [8] for a Rayleigh number equal to 100,000. For this Rayleigh number, we used a mesh 81 × 81.

For the thermal and dynamic modeling of our problem, we first discretized the domain of study by a staggered cartesian mesh; the control volume of the component u is staggered in the direction x relative to the main control volume, that of the component v is staggered in the direction of y. Moreover, the staggered mesh technique consists in taking the vector and scalar variables in different nodes. This mesh is adapted to our choice of the distribution of the calculation points. Those of the temperature and the pressure are at the center of the meshes while those of the components u and v of the velocity are located on the faces of the control volume. We use the finite volume method which allows to

U max : the maximum horizontal velocity, V max : the maximum vertical velocity, N u max : the maximum value of the local Nusselt number and N u min : the minimum value of the local Nusselt number.

model both the flow of a fluid and the transfer of heat. This method relies on a discretization technique, which converts the differential equations to the partial derivatives into nonlinear algebraic equations, which can then be solved numerically. It was described in 1971 by Patankar and Spalding and published in 1980 by Patankar [9] . Close to the active walls, the gradient of temperature and velocity is strong. For a better representation to the required accuracy of the reality of thermal transfers especially near the walls, the dimensions of the mesh are variable as shown in Figure 2 (41 × 41 for Ra = 1000, 10,000 to 81 × 81 (mesh finer)

for Ra = 100,000, 1,000,000).

There is a change in the horizontal velocity curves from 41 × 41, 61 × 61 to 81 × 81. Hence considering the accuracy of the results required and computational time involved, a 41 × 41 grid size is chosen for computations with the two lower Rayleigh numbers. An 81 × 81 grid size is chosen also for the two upper Rayleigh numbers. The theoretical analysis of Rayleigh [10] was done at the beginning of the 20th century. Natural convective heat transfer is became an interesting field of study phenomenon [11] .

In the literature, some papers describe natural convection in rectangular enclosures [12] [13] [14] [15] [16] . However, in other studies, authors investigate the natural convection heat transfer in square cavities [17] . The study of heat transfer by natural convection in enclosures is an active field of research [18] and the fluid studied in the cavity may be air, water [19] [20] , etc.

The differential Equations (1)-(4) governing the physical situation are translated into algebraic equations using the finite volume scheme. The system of dimensionless algebraic equations with boundary conditions associated (5)-(10) is iteratively solved. For accurate numerical simulation, a mesh size of 41 × 41 was selected for computations with Rayleigh numbers Ra = (1000, 10,000) and a mesh size of 81 × 81 for Ra = (100,000, 1,000,000). We have developed a numerical code with FORTRAN 90. The governing equations are discretized with the Central Difference Scheme (CDS). This scheme (order 2) generates a linear interpolation to the problem boundary. An iterative process is employed to find the velocity and temperature fields. The process is repeated until the following convergence criterion for velocity and temperature is met. The calculation stopped when the follows inequalities were satisfied:

Max [ ∑ i , j | ϕ i , j n + 1 − ϕ i , j n | ∑ i , j | ϕ i , j n + 1 | ] ≤ 10 − 5 . (13)

The same convergence criterion is imposed in terms of relative error for velocity and temperature. In the above expression n is any time level and ϕ = U, V, θ. In this study, we used the finite volume method with quadrilateral control volumes and a staggered mesh. The latter is the subdivision of the field of study into longitudinal and transverse grids whose intersection represents a node, where the variables P and θ are located while the components U and V of the velocity vector are in the middle of the segments connecting two nodes adjacent. After discretization of the differential transport equations we obtain a system of nonlinear algebraic equations, these equations describe the discrete properties of the fluid at the nodes in the solution domain. For temporal discretization, we have used a numerical method of solving numerical differential equations, based on a multi-step method. The temporal integration was carried out on a mesh staggered according to an Adams-Bashforth scheme (order 2).

d Y d τ = f ( τ , Y ) , with τ = τ 0 , Y ( τ 0 ) = Y 0 . (14)

{ Y 0 ( given ) , Y 1   approached   ( o n e - s t e p   m e t h o d ) , Y n + 1 = Y n + Δ τ 2 ( 3 f ( τ n , Y n ) − f ( τ n − 1 , Y n − 1 ) ) . (15)

In the first case, we have the following solutions for different thermally active locations. Figure 3 shows the profile for the different locations of the thermally active walls. It is observed that the fluid particle moves with greater velocity for the middle active locations and the velocity is less for the top/bottom active locations. Figure 3 shows that the variation of the vertical velocity reaches a maximum in the vicinity of the left heated wall and decreases as one moves away in order to be canceled towards x = 0.025 m. It descends by forming a kind of central symmetry and reaches a minimum in the vicinity of the right heated wall. The evolution of the vertical velocity in the cavity along the ordinate shows that the amplitude of this velocity is greater when the height is between the active

parts than on those which are inactive (adiabatic). In Figure 4, we have represented the temperature profiles along the ordinate. Temperature decreases almost linearly with abscissa. Temperatures increase in the vicinity of the heated left part and decrease near the active right part.

In a second case, we have the following results for different thermally active locations. Figure 5 shows the profile for the different locations of the thermally

active walls. It is observed that the fluid particle moves with greater velocity for the middle active locations and the velocity is less for the top/bottom active locations. With a larger temperature variation and therefore a higher Rayleigh number, we observe in Figure 5 that the amplitudes of velocities increased in the vicinity of the heated left wall by being canceled towards x = 0.025 m and decreasing near the active right wall. The evolution of the vertical velocity in the cavity along the ordinate has become greater than in the preceding case, with the velocity amplitude being greater when the height approaches the center of the active part (y = 0.025 m). In Figure 6, temperature decreases with abscissa and the convection is accentuated because we have increased the variation of temperatures between active parts. Temperatures increase again in the vicinity of the heated left part and decrease near the active right part.

In the third case, we have the following findings for different thermally active locations. Figure 7 shows the profile for the different locations of the thermally active walls. It is observed that the fluid particle moves with greater velocity for the middle active locations and the velocity is less for the top/bottom active locations. For an even larger Rayleigh number, the flow regime remains laminar with greater velocity amplitudes in Figure 7. Temperatures have decreased again with abscissa in Figure 8.

In the fourth case, we have the following results for different thermally active locations. Figure 9 shows the profile for the different locations of the thermally

active walls. It is observed that the fluid particle moves with greater velocity for the middle active locations and the velocity is less for the top/bottom active locations. The flow regime always remains laminar with higher velocity and temperature variations (Figure 9 and Figure 10).

Figures 11-14 show the impact of variation in the Rayleigh number. With an increase in the Rayleigh number from 1000 to 1,000,000, we have some noticeable changes. There are fluctuations in the velocity profiles and in particular for

the maximum values u_max and v_max. The velocity components (Figure 11 and Figure 12) increase progressively in amplitude and the maximum of v_max moves to the vertical walls with the increase of the Rayleigh number. The centro-symmetry’s property is retained throughout the range of the Rayleigh number studied. We observe a symmetry of the curves of velocity with respect to the

center of the cavity. For our model, the variation of the Rayleigh number represents a variation proportional to the temperature difference between the right and left active parts. In the zone 0.0125 m ≤ y ≤ 0.0375 m, the temperature increases progressively (Figure 13). This is where the heated parts are located. Fluid particles move with greater velocity due to an increase in buoyancy force for large values of the Rayleigh number. On the other hand, in the zone 0.0125 m ≤ x ≤ 0.375 m, the temperature remains almost constant (Figure 14). For a small Rayleigh number, of the order of 1000, we have noticed the dominance of the heat transfer mode by conduction. Beyond this value, convection dominates and appears more clearly for Ra = 1,000,000. Temperature and velocity are influenced by the increasing the Rayleigh number, which increases them at the same time. The local temperature profile certifies a centro-symmetry property of the local temperature field. The determination of these parameters allowed us to analyze the behavior of the fluid inside the cavity.

The results are displayed graphically in terms of velocity vectors and isotherms. In Figure 15, we show the velocity vectors corresponding to the end of the simulation (final time). The circulation of the fluid inside the cavity is shown by the velocity vectors. In Figure 15(a), we can observe the existence of a stagnant zone in the center of the cavity, which means that the heat exchange takes place

in an intense way at the corners of the cavity. In Figure 15(b), the stagnant zone at the center is always present inside the cavity. In the case of a low Rayleigh number, the fluid rotates around the cavity midpoint. In Figure 15(c) & Figure 15(d), we noticed the existence of two cells of recirculation inside the cavity, one to the upper left side and the other to the lower right side. The two remaining corners are less active. The flow leaving an active wall region is free to move either upwards or downwards. The formation of the two recirculation cells shows that the solution has a symmetrical flow. Fluid circulation therefore fills the entire cavity even at a very high Rayleigh number.

In Figure 16(a), the isotherms are almost parallel to the active parts. It is indicating that only the conduction mode of heat transfer prevails in this stage. For a low value of the Rayleigh number equal to 1000, the dominance of the conduction heat transfer mode was noted. In Figure 16(a) & Figure 16(b), isotherms show a monocellular flow represented by a sort of vortex rotating in the sense of a clock. The flow of the fluid takes place in the form of a large recirculation loop.

As it approaches the hot wall, the particle gains heat (its kinetic energy), heats up (becomes lighter) and rises along the hot wall with a trajectory that will be further modified by the presence of the upper horizontal wall. Along the latter, the fluid particle cools as it approaches the cold wall: we have the formation of a circulation cell. The warm air rises and the cold air descends. In Figure 16(c) & Figure 16(d), we notice that the isotherms are close to one another and condense near the active situations, where the temperature gradients are high, whereas they are negligible for the rest of the walls of the cavity. This reflects the existence of the thermally strong boundary layers close to these zones. We show that there is the phenomenon of convection in the vicinity of the active places, in particular for a high Rayleigh number, of the order of 100,000. Isotherms are crowded around the active locations on the left and right sides of the enclosure. The heat transfer is high because this situation gives the minimum distance between points on the hot and cold surfaces.