betaFeedback

WorksPhysics and Astronomy

Flux scaling in Rayleigh Bénard convection: a local boundary layer analysis

Abstract

We study the effect of shear due to the large scale flow (LSF) on the heat flux in Rayleigh B'enard convection for a range of near-plate Rayleigh numbers 8×107≤Raw≤5×10148\times 10^7 \leq Ra_w \leq 5\times 10^{14}, by studying its effect on the local boundary layers (BLs) on either sides of the plumes, which are much thinner than the global shear BL created by the LSF velocity VFV_F. Considering these local BLs forced externally by the LSF, we obtain a fifth order algebraic equation for the local boundary layer thicknesses. Solving these equations numerically using ReRe relations for aspect ratios Γ=1Γ=1 and 0.5, we obtain the variation of the local BL thicknesses with the longitudinal distance for various RawRa_w. We find that the average shear acting on the edges of these local BLs (u∣‾z=δ\overline{u|}_{z=δ}) increases as u∣‾z=δ∼Raw1/3\overline{u|}_{z=δ} \sim Ra_w^{1/3} for 8×107≤Raw≤10128\times 10^7\leq Ra_w \leq 10^{12} at Γ=1Γ=1, and as u∣‾z=δ∼Raw0.38\overline{u|}_{z=δ} \sim Ra_w^{0.38} for 1×1011≤Raw≤5×10141\times 10^{11}\leq Ra_w \leq 5\times 10^{14} at Γ=0.5Γ=0.5. We then estimate the average local thermal BL thickness to find the global Nusselt number NuNu.We find that Nu∼RawmNu\sim Ra_w^m, where m≈0.327m\approx 0.327 for 8×107≤Raw≤1×10128\times 10^7 \leq Ra_w \leq 1\times 10^{12} at Γ=1Γ=1, and m=0.33m=0.33 for 1×1011≤Raw≤5×10141\times10^{11}\leq Ra_w \leq 5\times10^{14} at Γ=0.5Γ=0.5. Inspite of the increasing shear on these BLs with increasing RawRa_w, we then surprisingly obtain the classical 1/3 scaling of flux since the shear forcing acting on those BLs remains sub-dominant compared to the NCBL velocities (VblV_{bl}) within these BLs, upto Raw≤5×1014Ra_w\leq 5\times10^{14}.

Key words:

Introduction

Rayleigh-Bénard convection, a type of natural convection occurring between two horizontal plates maintained at different temperatures, has been extensively studied due to its complex and rich fluid dynamics. A key focus in this area is understanding how the Nusselt number NuNu, which measures the efficiency of convective heat transfer, scales with the Rayleigh number RawRa_w. Specifically, the scaling relationship Nu∼RawmNu \sim Ra_w^m has been a topic of contention, with different values of the exponent mm reported in the literature [3, 25, 8, 29, 11, 13].

The classical scaling exponent is 1/3 where the near-wall Rayleigh number Raw=gβΔTwH3/ναRa_w = g\beta\Delta T_w H^3/\nu\alpha. Here, β\beta is the coefficient of thermal expansion, gg the acceleration due to gravity, ν\nu the kinematic viscosity, α\alpha the thermal diffusivity, HH the fluid layer height and ΔTw\Delta T_w the temperature difference between the hot plate temperature ThT_h and the bulk fluid temperature TbT_b; ΔT=2ΔTw\Delta T = 2\Delta T_w, where ΔT=Th−Tc\Delta T = T_h - T_c with TcT_c being cold plate temperature, then Rayleigh number Ra=2RawRa = 2Ra_w.

For Raw<1013Ra_w < 10^{13}, there seems to be a broad acceptance that the scaling exponent has a value of 2/7<m<1/32/7 < m < 1/3 [29, 15, 10]. Also, most studies find the Reynolds number based on the large-scale flow (LSF) velocity VFV_F scaling as Re∼Ra0.43Re \sim Ra^{0.43} [15, 21, 2, 11, 9], where Re=VFH/νRe = V_FH/\nu. With increased RawRa_w, the strength of the LSF increases. This stronger LSF, when interacting with the near-wall region, also increases velocities within the local boundary layers (BLs) close to the hot plate from their natural convection values given by VblV_{bl} [32]. The Reynolds number based on VblV_{bl} is

Rebl=VblH/ν=1.88Raw1/3Pr−0.98,(1)Re_{bl} = V_{bl}H/\nu= 1.88Ra_w^{1/3}Pr^{-0.98}, \tag*{(1)}

[22]. As RawRa_w increases, the thermal boundary layer thickness δT\delta_T decreases rapidly. Consequently, the dimensionless heat transfer coefficient, NuNu, which is proportional to H/2δTH/2\delta_T, increases.

At even higher values of RawRa_w, 1010≤Raw≤101510^{10} \leq Ra_w \leq10^{15}, Iyer et al. [13] demonstrates that the flux scaling approaches the classical 1/3 scaling at aspect ratio Γ=L/H=0.1\Gamma= L/H = 0.1, where LL is the length of the convection cell. However, contradictory inferences report a transition to the so-called ultimate regime, which is observed as a scaling of Nu∼Raw0.38Nu \sim Ra_w^{0.38}, at Ra≈1014Ra \approx10^{14} [11, 8]. He et al. [11] attributed this transition in flux scaling to a change in the scaling of Reynolds number ReRe on RawRa_w from Re∼Raw0.43Re \sim Ra_w^{0.43} to Re∼Raw0.5Re \sim Ra_w^{0.5}, claiming a different boundary layer dynamics at higher values of RawRa_w.

The aspect ratio Γ\Gamma, depending on which the large-scale dynamics significantly varies, also plays a role in deciding the flux scalings [1]. Recently, Samuel et al. [28] studied RB convection in Γ=4\Gamma= 4 and 8 channels for Raw≤1011Ra_w \leq10^{11} relevant to geophysical applications, and showcases that the BLs are fluctuation dominated with fluctuations being much higher in magnitudes that the mean flow. In the absence of persistent LSF in high aspect ratio channels, these results raise questions over the claimed shear-induced transition in flux scalings.

Majority of flux scalings were discovered by spatially and temporally averaging thermal boundary layer thicknesses, and the transition in flux scaling was subsequently claimed using these averaged values of BL thicknesses [7, 1]. In reality, there are local boundary layers on the plates that turns into plumes due to gravitational instability [18]. This instability is modified with increasing shear effects [3], resulting in an increased spacing between plumes (λs\lambda_s) from its no-shear value (λ0\lambda_0) [31]. The plume spacing λs\lambda_s, in the presence of small shear, normalised by the viscous-shear length Zsh=ν/VFZ_{\mathrm{sh}} = \nu/V_{\mathrm{F}}, is given by

λs=λ0+ZshRe3DRaw,(2)\lambda_{\mathrm{s}} = \lambda_0 + \frac{Z_{\mathrm{sh}}Re^3}{DRa_{\mathrm{w}}}, \tag*{(2)}

where, D(Pr)=0.004Pr3D(Pr) = 0.004Pr^3 for Pr>5Pr > 5 and D(Pr)=52.7Pr−2.8D(Pr) = 52.7Pr^{-2.8} for Pr<5Pr < 5. These local BLs, which feed plumes from the sides, are significantly thinner compared to the the global BL thickness, obtained by spatio-temporally averaging the temperature within BLs and plumes. Rather than the global BL, a near-wall temperature drop (ΔTw\Delta T_{\mathrm{w}}) that occurs on thinner, local BLs present on either side of plumes are the ones that decides the flux scalings in Rayleigh-Bénard convection. Moreover, averaging the BL and plume regions together essentially considers differently characterized regions of strain rates as one [35, 34].

Several models have been proposed by different authors to explain the scalings of Nusselt number in Rayleigh-Bénard convection. The formulation of models by various authors exhibits a notable contrast in their approach, as they assume distinct natures of BLs. Initially, Howard [12] considered a marginally stable BLs assuming that the LSF has no impact on the stability of the BLs close to the plates. However, this neglected the interaction between the thermal plumes and the LSF [31]. Therefore, the revisited model of Malkus and Howard recently developed by Creyssels & Martinand [6], considering the LSF as an external input, obtained NuNu values similar to those in experiments and simulations for Ra<1010Ra < 10^{10}. Kraichnan [14] proposed a model with nature of the BL analogous to that in a fully developed shear flow and the ultimate scaling of 1/2 at extremely high RaRa. Castaing et al. [3] developed a model in which the domain was divided into near-wall BLs, turbulent bulk, and intermediate mixing region, and proposed a flux scaling with exponent of RawRa_{\mathrm{w}} equal to 2/7, in agreement with the observed values in the experiments for Ra<1013Ra < 10^{13}. Shraiman & Siggia [36] proposed a model presuming the fully turbulent nature of boundary layers leading to 2/7 scaling. While supporting the transition in flux scalings at Raw∼1014Ra_{\mathrm{w}} \sim10^{14}, Grossmann & Lohse [8] argued that the nature of the boundary layers changes from laminar at low RaRa to fully turbulent at extremely high RaRa. Skrbek & Urban [37] re-analysed experimental data of Roche et al. [25] and He et al. [11] at very high RaRa and attributed the so-called transition to ultimate regime to non-Oberbeck–Boussinesq effects. Recently, a theory on BLs by Lindborg [16] hypothesised the near-wall boundary layers of a semi-turbulent nature, with a presence of thin viscous wall layer of about five Kolmogorov scales at RawRa_{\mathrm{w}} as high as 101510^{15}, and concluded that Nu∼Raw1/3Nu \sim Ra_{\mathrm{w}}^{1/3} as the scaling of the ultimate regime.

Summing up the investigations in the literature, we propose a BL model which describes the near-wall dynamics in Rayleigh Bénard convection for 8×107≤Raw≤5×10148 \times10^7 \leq Ra_{\mathrm{w}} \leq5 \times10^{14}. We consider the two-dimensional laminar natural convection type BLs, which are thinner than the global BL. These are local BLs on the sides of the plumes, forced with the mean Prandtl-Blasius velocity profile generated by the LSF. The LSF velocity scalings are provided as an external input employed from the literature studies [5, 25]. A fifth order scaling equation for the local BLs forced by the shear due to the LSF is obtained using the order of magnitude balance of integral BL equations, which was then solved numerically. Calculating averaged values of the thermal BLT δT\delta T, over the half mean plume spacing (λs/2\lambda_{\mathrm{s}}/2) (2), the variation of the Nusselt number NuNu with RawRa_{\mathrm{w}} is then computed. The scaling of NuNu follows the power law Nu∼Raw0.327Nu \sim Ra_{\mathrm{w}}^{0.327} for Raw≤1012Ra_{\mathrm{w}} \leq10^{12} at Γ=1\Gamma= 1 while Nu∼Raw0.33Nu \sim Ra_{\mathrm{w}}^{0.33} for 1×1011≤Raw≤5×10141 \times10^{11} \leq Ra_{\mathrm{w}} \leq5 \times10^{14} at Γ=0.5\Gamma= 0.5. We demonstrate the reasoning behind the classical 1/3 scaling of the Nusselt number up to Raw=1015Ra_{\mathrm{w}} = 10^{15} by relating it to the relatively high strength of buoyancy effects compared to shear effects inside the BLs.

Analysis of the local BLs forced by large scale flow (LSF)

Figure 1(a) shows the top view of the actual local flow field in a horizontal plane close to the hot plate, obtained by PIV, after applying the plume separation criterion by Shevkar et al. [35]. The figure shows line plumes (coloured regions) aligned by an LSF oriented from bottom left corner to top left corner of the figure. The white regions in the figure are the local BLs in between the plumes, that are externally forced by the LSF. The schematic of the side view, in the vertical plane through B–B in Figure 1(a) is shown in the top figure in 1(b). The direction of LSF in this figure is perpendicular to the plane of the figure. The corresponding schematic of the top view, showing the plumes aligned by the LSF and the local BLs in-between those plumes. A side view of one of such local BL, in vertical plane through A–A in Figure 1(b) is shown in Figure 1(c). These local BLs are an order thinner than the global BL created by the LSF, which span the length of the convection cell (Gunasegarane & Puthenveettil [9]; Shevkar et al. [31, 35]). Since the major part of the temperature drop near the plate occurs across these BLs, these thinner local BLs are more likely to be the ones which decide the flux scaling near the hot surface and not the global BLs as has been assumed so far. Due to their smaller thickness as well as since they are embedded within the global BL, these local BLs are less likely to be affected by the turbulent bulk. These local boundary layers on the hot plate become unstable to turn upwards and form the plumes. Since plumes are the outcome of the instability of these thin local BLs, on their either sides, it is natural to expect that these BLs, before they become unstable, to be laminar. Further, successful scaling laws for the mean spacing between such line plumes (Theerthan & Arakeri [39, 40]; Puthenveettil & Arakeri [21]), their total lengths (Puthenveettil et al. [22]), their mean dynamics ((Gunasegarane & Puthenveettil [9])) and most recently the velocities within local BLs (Shevkar et al. [32]) have all been obtained assuming steady 2D laminar natural convection boundary layers (Rotem & Classen [27]; Pera & Gebhart [19]) feeding these plumes. More importantly, the observed vertical distributions of fluctuations of velocities and temperature near the hot plate have been predicted well by a model that assumes laminar BLs giving rise to laminar plumes (Theerthan & Arakeri [39]). This means that the observed fluctuations in velocity and temperature near the hot-plate are a result of the spatial averaging of a spatially non-uniform field consisting of many local laminar BLs giving rise to many plumes with the bulk fluid in between them, as well as by the lateral motion of such plumes (Shevkar & Puthenveettil [33]). Due to all these reasons we consider these local BLs to be 2D laminar natural convection boundary layers on a horizontal surface, forced by an external shear due to a LSF of strength VFV_F, as shown in Figure 1. Since these local BLs are affected by the shear due to the LSF, in other words, we assume that the local boundary layers that occurs on either sides of the plumes on the plate in turbulent convection, which we use to find the flux scaling, to be of laminar mixed convection nature.

Figure 1. (a) Schematic of the local boundary layers between line plumes being forced by an external shear due to the large-scale flow for Pr>1Pr > 1. The top schematic shows the side view in a vertical plane through B–B while the bottom schematic shows the corresponding top view; (b) Top view of the horizontal velocity field obtained using PIV, overlaid over the corresponding negative horizontal divergence field, in a horizontal plane at a height of hm=1.6 mmh_m = 1.6\,\mathrm{mm} from the hot plate, at Raw=1.31×107Ra_w = 1.31 \times10^7 [31, 35, 32]. The white regions show the local BLs and the colored regions show the plumes. (c) Side view in a vertical plane through A–A in figure (a); —, horizontal velocity profile within the local BL due to forcing by the external shear of the LSF on the natural convection BL; − · − · −, horizontal velocity profile within a NCBL; − − −, horizontal velocity profile within a PBBL.

Integral relations

The xx-momentum BL equation, u∂u/∂x+w∂u/∂z=−(1/ρ)∂p/∂x+ν∂2u/∂z2u\partial u/\partial x+w\partial u/\partial z=-(1/\rho)\partial p/\partial x+\nu\partial^{2}u/\partial z^{2}, integrated across the velocity BL thickness δ(x)\delta(x) for such BLs is,

∫0δu∂u∂x dz+∫0δw∂u∂z dz=−1ρ∫0δ∂p∂x dz+ν∫0δ∂2u∂z2 dz(3)\int_{0}^{\delta}u\frac{\partial u}{\partial x}\,dz+\int_{0}^{\delta}w\frac{\partial u}{\partial z}\,dz=-\frac{1}{\rho}\int_{0}^{\delta}\frac{\partial p}{\partial x}\,dz+\nu\int_{0}^{\delta}\frac{\partial^{2}u}{\partial z^{2}}\,dz \tag*{(3)}

where p(x,z)p(x,z) is the pressure within the boundary layer and uu & ww the horizontal and vertical velocity components, respectively. Rewriting the second term in (2.1) as ∫0δ(∂(wu)/∂z−u∂w/∂z) dz\int_{0}^{\delta}(\partial(wu)/\partial z-u\partial w/\partial z)\,dz, integrating, replacing ∂w/∂z=−∂u/∂x\partial w/\partial z=-\partial u/\partial x from continuity, noticing that ∫0δ(u∂u/∂x) dz=∫0δ(∂u2/∂x) dz/2\int_{0}^{\delta}(u\partial u/\partial x)\,dz=\int_{0}^{\delta}(\partial u^{2}/\partial x)\,dz/2, and since ∂u/∂z∣z=δ≃0\partial u/\partial z|_{z=\delta}\simeq0 for small

external shear, we rewrite (3) as

∫0δ∂2u∂x2 dz+(wu)∣z=δ=−1ρ∫0δ∂p∂x dz−ν∂u∂z∣z=0.(4)\int_{0}^{\delta} \frac{\partial^{2}u}{\partial x^{2}}\,dz + (wu)\big|_{z=\delta} = -\frac{1}{\rho}\int_{0}^{\delta}\frac{\partial p}{\partial x}\,dz - \nu\left.\frac{\partial u}{\partial z}\right|_{z=0}. \tag*{(4)}

In the second term in (4), the vertical velocity at the velocity boundary layer edge can be obtained by integrating the continuity equation, ∂u/∂x+∂w/∂z=0\partial u/\partial x + \partial w/\partial z = 0, and applying the Leibnitz rule as

w∣z=δ=−∂∂x∫0δu dz+u∣z=δ∂δ∂x.(5)w\big|_{z=\delta}=-\frac{\partial}{\partial x}\int_{0}^{\delta}u\,dz+u\big|_{z=\delta}\frac{\partial\delta}{\partial x}. \tag*{(5)}

To obtain the term u∣z=δu\big|_{z=\delta} which is the external forcing of these local BLs, that occur in (4) and (5), we assume that these local boundary layers are embedded within a PBBL spanning the entire length of the hot plate and which is driven by the large-scale flow strength of VFV_F, acting at a distance of δpb\delta_{pb} from the hot plate; the schematic of such an arrangement is shown in figure 1(c). Using the Von-Karman velocity profile

u(z)/VF=2z/δpb−(z/δpb)2,(6)u(z)/V_F=2z/\delta_{pb}-(z/\delta_{pb})^{2}, \tag*{(6)}

for the PBBL, the dimensionless shear velocity acting on the edges of the local natural convection boundary layers, at a height of z=δz=\delta from the hot plate, is then,

A(x)=u∣z=δVF=2δδpb−(δδpb)2,(7)A(x)=\frac{u\big|_{z=\delta}}{V_F}=2\frac{\delta}{\delta_{pb}}-\left(\frac{\delta}{\delta_{pb}}\right)^2, \tag*{(7)}

where the mean PBBL thickness, independent of xx, is

δpb=0.922H/Re(8)\delta_{pb}=0.922H/\sqrt{Re} \tag*{(8)}

([2, 38]).

Applying Leibnitz rule, to the first term in (4), substituting (5) in (4), simplifying, and replacing u∣z=δu\big|_{z=\delta} using (7), we obtain,

∂∂x∫0δu2dz−A(x)VF∂∂x∫0δu dz+1ρ∫0δ∂p∂xdz+ν∂u∂z∣z=0=0.(9)\frac{\partial}{\partial x}\int_{0}^{\delta}u^{2}dz-A(x)V_F\frac{\partial}{\partial x}\int_{0}^{\delta}u\,dz+\frac{1}{\rho}\int_{0}^{\delta}\frac{\partial p}{\partial x}dz+\nu\left.\frac{\partial u}{\partial z}\right|_{z=0}=0. \tag*{(9)}

To replace the unknown pressure in (9), we integrate the z-momentum BL equation, ∂p/∂z=ρgβ(T−TB)\partial p/\partial z=\rho g\beta(T-T_B), across the velocity BL thickness, to obtain,

p=−ρgβ∫0δ(T−TB) dz,(10)p=-\rho g\beta\int_{0}^{\delta}(T-T_B)\,dz, \tag*{(10)}

where T(z)T(z) is the temperature distribution within the local BL, and TBT_B the fluid temperature above the local BLs. Substituting (10) in (9), we obtain the integral momentum balance equation for the local natural convection boundary layers, forced externally by the large-scale flow, as

∂∂x∫0δu2dz−AVF∂∂x∫0δu dz−gβ∫0δ∂∂x∫0δ(T−TB) dz dz+ν∂u∂z∣z=0=0.(11)\frac{\partial}{\partial x}\int_{0}^{\delta}u^{2}dz-AV_F\frac{\partial}{\partial x}\int_{0}^{\delta}u\,dz-g\beta\int_{0}^{\delta}\frac{\partial}{\partial x}\int_{0}^{\delta}(T-T_B)\,dz\,dz+\nu\left.\frac{\partial u}{\partial z}\right|_{z=0}=0. \tag*{(11)}

The local BL energy equation, u∂T/∂x+w∂T/∂z=α∂2T/∂z2u\partial T/\partial x+w\partial T/\partial z=\alpha\partial^{2}T/\partial z^{2}, integrated across the local thermal BL thickness δT(x)\delta_T(x), is

∫0δTu∂T∂x dz+∫0δTw∂T∂z dz=α∫0δT∂2T∂z2 dz.(12)\int_{0}^{\delta_T}u\frac{\partial T}{\partial x}\,dz+\int_{0}^{\delta_T}w\frac{\partial T}{\partial z}\,dz=\alpha\int_{0}^{\delta_T}\frac{\partial^{2}T}{\partial z^{2}}\,dz. \tag*{(12)}

The first term in (12) can be written as ∫0δT(∂(uT)/∂x)dz−∫0δTT(∂u/∂x)dz\int_{0}^{\delta_T} \left(\partial(uT)/\partial x\right) dz-\int_{0}^{\delta_T}T\left(\partial u/\partial x\right)dz. Applying Leibnitz rule to the first term in this equation, the first term in (2.10 becomes,

∂∂x∫0δTuTdz−(uT)∣z=δTdδTdx−∫0δTT∂u∂xdz.(13)\frac{\partial}{\partial x}\int_{0}^{\delta_T}uT dz-(uT)|_{z=\delta_T}\frac{d\delta_T}{dx}-\int_{0}^{\delta_T}T\frac{\partial u}{\partial x}dz. \tag*{(13)}

Similarly, we rewrite the second term in (12) as ∫0δT(∂(wT)/∂z)dz−∫0δTT(∂w/∂z)dz\int_{0}^{\delta_T}\left(\partial(wT)/\partial z\right)dz-\int_{0}^{\delta_T}T\left(\partial w/\partial z\right)dz. Integrating the first term of this equation, and using continuity equation on the second term of this equation, the second term in (12) becomes,

(wT)∣z=δT+∫0δTT∂u∂xdz.(14)(wT)|_{z=\delta_T}+\int_{0}^{\delta_T}T\frac{\partial u}{\partial x}dz. \tag*{(14)}

Replacing w∣z=δTw|_{z=\delta_T} in (14) by −∂/∂x∫0δTudz+u∣z=δT(dδT/dx)-\partial/\partial x\int_{0}^{\delta_T}u dz+u|_{z=\delta_T}(d\delta_T/dx), obtained by integrating the continuity equation across the thermal BL thickness, similar to (5), and applying Leibnitz rule, the second term in (12) becomes,

−∂∂x∫0δTuTBdz+u∣z=δTTBdδTdx+∫0δTT∂u∂x.(15)-\frac{\partial}{\partial x}\int_{0}^{\delta_T}uT_B dz+u|_{z=\delta_T}T_B\frac{d\delta_T}{dx}+\int_{0}^{\delta_T}T\frac{\partial u}{\partial x}. \tag*{(15)}

Substituting (13) and (15) in (12) and simplifying, since the last term in (12) becomes −α∂T/∂z∣z=0-\alpha\left.\partial T/\partial z\right|_{z=0} since ∂T/∂z∣z=δ≃0\left.\partial T/\partial z\right|_{z=\delta}\simeq0, we obtain the integral energy equation for the local NCBLs forced by the LSF as

∂∂x∫0δTu(T−TB)dz=−α∂T∂z∣z=0.(16)\frac{\partial}{\partial x}\int_{0}^{\delta_T}u(T-T_B)dz=-\alpha\left.\frac{\partial T}{\partial z}\right|_{z=0}. \tag*{(16)}

Scaling relation for the local BL thickness

We now convert the integral equations (11) and (16) into scaling relations, using the following relevant characteristic scales near the hot/cold plate in turbulent convection. We take the characteristic scale of temperature difference as ΔTw\Delta T_w, and the local characteristic vertical distances within the velocity BL and the thermal BL as the velocity and thermal BL thicknesses, δ(x)\delta(x) and δT(x)\delta_T(x), respectively. The characteristic horizontal velocity within the BLs, when the BLs are forced externally by the shear due to the LSF, is not known apriori; we take it as UcU_c, whose value we find later. Substituting these characteristic scales in the order of magnitude balances of (11) and (16), we obtain, respectively,

Uc2δx−A(x)VFUcδx−gβΔTwδ2x+νUcδ∼0,and(17)\frac{U_c^2\delta}{x}-\frac{A(x)V_FU_c\delta}{x}-\frac{g\beta\Delta T_w\delta^2}{x}+\frac{\nu U_c}{\delta}\sim0,\quad\text{and} \tag*{(17)}
UcΔTwδTx∼αΔTwδT.(18)\frac{U_c\Delta T_w\delta_T}{x}\sim\frac{\alpha\Delta T_w}{\delta_T}. \tag*{(18)}

For small shear forcing of the local BLs, given by A<1A<1, which we show later to be the case for Raw<1015Ra_w<10^{15}, we now assume,

δδT=C2Prn,(19)\frac{\delta}{\delta_T}=C_2Pr^n, \tag*{(19)}

where, C2C_2 and nn are positive constants. Equation (19) implies that we assume δ\delta and δT\delta_T to have the same dependence on RawRa_w and ReRe in the presence of small shear forcing, so that their ratio becomes independent of RawRa_w and ReRe. Since C2C_2 and nn are positive, δT<δ\delta_T<\delta for Pr>1Pr>1, with the reduction in δT\delta_T with respect to δ\delta being a function of PrPr. This reasonable assumption has given a good prediction for UcU_c in the presence of shear, earlier in Shevkar et al. [32].

Replacing δT\delta_T in (18) with δ\delta from (19), we obtain the expression for UcU_c as

Uc∼αxδ2C22Pr2n.(20)U_c \sim\frac{\alpha x}{\delta^2} C_2^2 Pr^{2n}. \tag*{(20)}

Substituting (20) in (17), we obtain a scaling relation for the local velocity BL thickness δ(x)\delta(x), when these BLs are forced externally by the shear due to the LSF, as

RaxC22(δx)5+A(x)Pr2nRex(δx)2−C22Pr5nE∼0,(21)\frac{Ra_x}{C_2^2}\left(\frac{\delta}{x}\right)^5 + A(x)Pr^{2n}Re_x\left(\frac{\delta}{x}\right)^2 - C_2^2Pr^{5n}E \sim0, \tag*{(21)}

where, Rax=gβΔTwx3/ναRa_x = g\beta\Delta T_w x^3/\nu\alpha is the local Rayleigh number based on xx. A(x)A(x) is defined in (7), the dimensionless shear forcing at the edge of the local BLs due to LSF, Rex=VFx/νRe_x = V_Fx/\nu the local shear Reynolds number based on xx, and E(Pr)=(C22Pr2n−1+1)/C22Pr3nE(Pr) = (C_2^2Pr^{2n-1} + 1)/C_2^2Pr^{3n}.

Results and discussion

Local boundary layer thicknesses and dimensionless shear

Rewriting (21) in terms of the external shear Reynolds number (Re=VFH/νRe = V_FH/\nu) and the near-plate Rayleigh number (RawRa_w), and rearranging, we obtain,

RawC22(δH)5+APr2nRexH(δH)2−C22Pr5nE(xH)2∼0.(22)\frac{Ra_w}{C_2^2}\left(\frac{\delta}{H}\right)^5 + APr^{2n}Re\frac{x}{H}\left(\frac{\delta}{H}\right)^2 - C_2^2Pr^{5n}E\left(\frac{x}{H}\right)^2 \sim0. \tag*{(22)}

Equation (22) describes the variation of the local BLT δ(x)\delta(x) with the longitudinal distance xx for a given ReRe and RawRa_w. This δ(x)\delta(x), in turn, decides the dimensionless external shear A(x)A(x) through (7), for the appropriate dependence of ReRe on RawRa_w, since δpb\delta_{pb} in (7) is a function of ReRe as given by (8). The expression of ReRe varies based again on the aspect ratio Γ\Gamma, as given later in (23) and (24). The above equation has the appropriate behaviour in the limiting cases. When Re→0Re \to0, (22) shows that δ/x∼Rax−1/5\delta/x \sim Ra_x^{-1/5}, the expected variation in NCBL [26]. When Raw→0Ra_w \to0, δ/x∼Re−0.5\delta/x \sim Re^{-0.5}, where Reδx=u∣z=δx/νRe_{\delta x} = u|_{z=\delta x}/\nu, the expected variation in PBBL. We now numerically solve (22) and (7) simultaneously at Pr=1Pr = 1 to obtain the dimensionless velocity BL thickness δ(x)/H\delta(x)/H and the dimensionless external shear A(x)A(x) (7), as a function of xx, using two Reynolds number relations proposed for two different Γ\Gamma, for the corresponding ranges of near-plate Rayleigh numbers. We use

Re=1.345Raw3/7Pr−0.76,(23)Re = 1.345Ra_w^{3/7}Pr^{-0.76}, \tag*{(23)}

for 8×107≤Raw≤1×10128 \times10^7 \le Ra_w \le1 \times10^{12} at Γ=1\Gamma= 1 [5], and

Re=0.18Raw0.48Pr−0.75,(24)Re = 0.18Ra_w^{0.48}Pr^{-0.75}, \tag*{(24)}

for 1×1011≤Raw≤5×10141 \times10^{11} \le Ra_w \le5 \times10^{14} at Γ=0.5\Gamma= 0.5 [25]. We take only the real and finite solutions of (22) and (7) in the present analysis.

Figures 2(a) shows the variations of the dimensionless local BLT δ/H\delta/H as a function of the dimensionless horizontal location x∗=x/(λs/2)x^* = x/(\lambda_s/2), where λs/2\lambda_s/2 is half the plume spacing with shear, as obtained by using (23) for different RawRa_w in (2); figure 2(b) shows the same, obtained by using (24). The insets in figures 2(a) and 2(b) show the same plots as in the corresponding main figures in log-log scale. Power law fits of the form δ/H=Bx∗m\delta/H = Bx^{*m} for these RawRa_w are also shown in each of the inset figures. At any x∗x^*, with increase in RawRa_w the values of δ\delta decrease. Since the variation of mm with RawRa_w is small (see figure 3(b)), this decrease can be quantified by the variation of the prefactors BB, which is shown in figure 3(a). As shown in figure 3(b), at any x∗x^*, δ/H\delta/H decreases with RawRa_w as Raw−1/3Ra_w^{-1/3}, in the same way as the behaviour in NCBLs the absence of shear (see Puthenveettil et al. [22]). It then appears that the effect of shear does not seem to be enough to offset the decrease in δ\delta with increase in RawRa_w seen in NCBLs.

At any RawRa_w, δ/H\delta/H increase with increase in the longitudinal distance along the BL. The inset of Figure 2(a) shows that for Raw=8×107Ra_w=8\times10^7, δ/H\delta/H approximately scale as x∗0.366x^{*0.366}, which changes to δ/H∼x∗0.364\delta/H\sim x^{*0.364} at Raw=1×1012Ra_w=1\times10^{12}. Similarly, in the inset of Figure 2(b), δ/H∼x∗0.3875\delta/H\sim x^{*0.3875} at Raw=1×1011Ra_w=1\times10^{11} which changes to δ/H∼x∗0.3803\delta/H\sim x^{*0.3803} at Raw=5×1014Ra_w=5\times10^{14}. The variation of the exponents (mm) with RawRa_w, along with the corresponding exponent of 2/52/5 for NCBL, is shown in Figure 3(b). The corresponding scaling in NCBLs, namely, δ/H∼x∗2/5\delta/H \sim x^{*2/5}, is also shown by black dashed lines in the insets of Figure 2(a) and Figure 2(b). Figure 3(b) shows that, with increase in RawRa_w, mm decreases in both the ranges of RawRa_w, deviating more from the value of m=2/5m = 2/5 in NCBLs. The change in the longitudinal dependence of δ/H\delta/H from that in NCBLs, imply that the shear due to the LSF changed the nature of the local BLs from a natural convection type to a mixed convection type. Then, with increase in RawRa_w, the local BLs become more and more affected by the shear of the LSF, to become more and more of the nature of mixed convection type. As shown in Figure 3(b) for Γ=0.5\Gamma= 0.5 case, where ReRe is given by (24), the values of mm take a jump to around 0.385, which is closer than the values of mm for Γ=1\Gamma= 1 to the value of m=2/5m = 2/5 in NCBLs. We expect these larger values of mm for Γ=0.5\Gamma= 0.5, which implies a more natural convection nature of the local BLs, is due to the weaker LSF strengths at this lower Γ\Gamma. We show later that the actual shear forcing of these local BLs, namely u∣z=δu|_{z=\delta}, that we obtain from (23) and (7) is indeed lower when ReRe is given by (24) than (23). Further, eventhough the strengths of the LSF increase with RawRa_w (see 3.2 and 3.3), at any RawRa_w, their values also decrease to zero as we approach the plates (see 2.4 and Figure 1(c)). In such a vertically varying LSF velocity field close to the plates, a decreasing thickness of the local BLs with RawRa_w, as could be seen to occur in Figure 2(a) and Figure 2(b), may even reduce the actual shear forcing of these local BLs with increase in RawRa_w. To understand these issues, we now look at the variation of the actual shear forcing at the edge of these local BLs with RawRa_w.

Variation of $\delta/H$ with $x^*=x/(\lambda_s/2)$ for fourteen different $Ra_w$

Figure 2. Variation of δ/H\delta/H with x∗=x/(λs/2)x^*=x/(\lambda_s/2) for fourteen different RawRa_w. (a), variation of δ/H\delta/H obtained using (23); □, Raw=8×107Ra_w=8\times10^7; ◇, Raw=2×108Ra_w=2\times10^8; ▽, Raw=5×109Ra_w=5\times10^9; △, Raw=2×109Ra_w=2\times10^9; ▷, Raw=2×1010Ra_w=2\times10^{10}; ◁, Raw=1×1011Ra_w=1\times10^{11} and ○, Raw=1×1012Ra_w=1\times10^{12}. Log-log plot of figure (a) is plotted as an inset. Inset of figure (a): − − −, ∼x2/5\sim x^{2/5}; —, ∼x0.366\sim x^{0.366}; —, ∼x0.365\sim x^{0.365}; —, ∼x0.364\sim x^{0.364}. (b), variation of δ/H\delta/H obtained using (24); ◇, Raw=1×1011Ra_w=1\times10^{11}; ▽, Raw=1×1012Ra_w=1\times10^{12}; △, Raw=1×1013Ra_w=1\times10^{13}; ▷, Raw=5×1013Ra_w=5\times10^{13}; ◁, Raw=9×1013Ra_w=9\times10^{13}; ○, Raw=2×1014Ra_w=2\times10^{14} and □, Raw=5×1014Ra_w=5\times10^{14}. Log-log plot of figure (b) is plotted as an inset. Inset of figure (b): − − −, ∼x2/5\sim x^{2/5}; —, ∼x0.3875\sim x^{0.3875}; —, ∼x0.383\sim x^{0.383}; —, ∼x0.3803\sim x^{0.3803}.

Variation of prefactors and exponents with $Ra_w$

Figure 3. Variation of (a) prefactors in the relation δ/H=Bx∗m\delta/H=Bx^{*m} with RawRa_w for the data shown in Figure 2. —, 2.5Raw−0.332.5Ra_w^{-0.33}; (b) Variation of exponents mm in the same; − − −, 0.37Raw−0.000620.37Ra_w^{-0.00062}; − − −, 0.409Raw−0.002110.409Ra_w^{-0.00211}; − · − · −, m=0.4m=0.4 for NCBLs.

Figures 4(a) and 4(b) show the variation of A(x∗)A(x^*) (7), the dimensionless shear at the edge of the BLs, with the dimensionless horizontal location x∗x^*, for different RawRa_w; Figure 4(a) shows the case of Γ=1\Gamma= 1 using (23) while Figure 4(b) shows the case of Γ=0.5\Gamma= 0.5 using (24). For any RawRa_w, the dimensionless shear increases with increase in x∗x^*. The dependence of A(x∗)A(x^*) on x∗x^* for different RawRa_w can be better understood from the inset plots in Figure 4(a) and 4(b), which show the same plots as in the main figures in log-log scale. The inset in Figure 4(a) shows that, at Γ=1\Gamma= 1, A(x∗)∼x∗0.33A(x^*) \sim x^{*0.33} at Raw=8×107Ra_w = 8 \times10^7 which changes to A∼x∗0.352A \sim x^{*0.352} at Raw=1×1012Ra_w = 1 \times10^{12}. For Γ=0.5\Gamma= 0.5 case, the inset of Figure 4(b) shows that the longitudinal dependence of A(x∗)A(x^*) remains around x∗0.367x^{*0.367}, with marginal increase to x∗0.368x^{*0.368} at the higher RawRa_w. Then, with increasing RawRa_w, the exponent of x∗x^* in the longitudinal dependence of A(x∗)A(x^*) increases, this increase itself reduces with increasing RawRa_w to become marginal at the highest RawRa_w.

Variation of $A(x^*)$ with $x^*$ for fourteen different $Ra_w$

Figure 4. Variation of A(x∗)A(x^*) with x∗x^* for fourteen different RawRa_w. In figure (a); □\square, Raw=8×107Ra_w = 8 \times10^7; ⋄\diamond, Raw=2×108Ra_w = 2 \times10^8; ▽\triangledown, Raw=5×109Ra_w = 5 \times10^9; △\triangle, Raw=2×109Ra_w = 2 \times10^9; ▹\triangleright, Raw=2×1010Ra_w = 2 \times10^{10}; ◃\triangleleft, Raw=1×1011Ra_w = 1 \times10^{11} and ∘\circ, Raw=1×1012Ra_w = 1 \times10^{12}. Log-log plot of figure (a) is plotted as an inset. Inset of figure (a): —, ∼x∗0.33\sim x^{*0.33}; —, ∼x∗0.339\sim x^{*0.339}; —, ∼x∗0.352\sim x^{*0.352}. In figure (b); ⋄\diamond, Raw=1×1011Ra_w = 1 \times10^{11}; ▽\triangledown, Raw=1×1012Ra_w = 1 \times10^{12}; △\triangle, Raw=1×1013Ra_w = 1 \times10^{13}; ▹\triangleright, Raw=5×1013Ra_w = 5 \times10^{13}; ◃\triangleleft, Raw=9×1013Ra_w = 9 \times10^{13}; ∘\circ, Raw=2×1014Ra_w = 2 \times10^{14} and □\square, Raw=5×1014Ra_w = 5 \times10^{14}. Log-log plot of figure (b) is plotted as an inset. Inset of figure (b): —, ∼x∗0.3668\sim x^{*0.3668}; —, ∼x∗0.3665\sim x^{*0.3665}; —, ∼x∗0.3679\sim x^{*0.3679}.

Interestingly, at any x∗x^*, A(x∗)A(x^*) decreases with increase in RawRa_w. By plotting the prefactors of the approximately parallel lines in the insets of Figure 4(a) and 4(b), we can find this dependence of A(x∗)A(x^*) on RawRa_w at any x∗x^*, which is shown in Figure 5. Here, the dependence of A(x∗)A(x^*) on x∗x^* is considered to remain approximately the same at the different RawRa_w, since even when RawRa_w changes by about 6 orders, the exponent of x∗x^* changes only by 6%, as could be seen from Figure 3(b). Figure 5 shows that the dimensionless shear A(x∗)A(x^*), at any horizontal location decreases with increase in RawRa_w as A(x∗)∼Raw−0.098A(x^*) \sim Ra_w^{-0.098} for Raw≤1012Ra_w \le10^{12} at Γ=1\Gamma= 1 and A(x∗)∼Raw−0.089A(x^*) \sim Ra_w^{-0.089} for Raw≥1×1011Ra_w \ge1 \times10^{11} at Γ=0.5\Gamma= 0.5. However, at the same time, the external LSF velocities VFV_F increase with increase in RawRa_w as, VF∼(ν/H)Raw3/7V_F \sim(\nu/H)Ra_w^{3/7} (23) for Raw≤1012Ra_w \le10^{12} at Γ=1\Gamma= 1, and as VF∼(ν/H)Raw0.48V_F \sim(\nu/H)Ra_w^{0.48} (24) for Raw≥1×1012Ra_w \ge1 \times10^{12} at Γ=0.5\Gamma= 0.5. Since A(x∗)=u∣z=δ/VFA(x^*) = u|_{z=\delta}/V_F from (7), u∣z=δu|_{z=\delta} then scales as (ν/H)Raw0.331(\nu/H)Ra_w^{0.331} for Raw≤1012Ra_w \le10^{12} at Γ=1\Gamma= 1, and as (ν/H)Raw0.398(\nu/H)Ra_w^{0.398} for Raw≥1×1012Ra_w \ge1 \times10^{12} at Γ=0.5\Gamma= 0.5, as shown in the inset of Figure 5.

Variation of the local dimensionless shear at the edge of the BL with $Ra_w$

Figure 5. Variation of the local dimensionless shear at the edge of the BL with RawRa_w; —, ∼Raw−0.098\sim Ra_w^{-0.098} and - - -, ∼Raw−0.089\sim Ra_w^{-0.089}. Inset shows the variation of the local shear velocity at the edge of BL with RawRa_w; —, ∼Raw0.331\sim Ra_w^{0.331} and - - -, ∼Raw0.398\sim Ra_w^{0.398}.

Hence, at any x∗x^*, eventhough the magnitudes of the dimensionless shear (A) acting on the local BLs (u∣z=δu|_{z=\delta}) decrease with increase in RawRa_w, the magnitudes of the actual dimensional shear acting at the edge of these local BLs do increase with increase in RawRa_w. In agreement with this increase in u∣z=δu|_{z=\delta} with RawRa_w, as we saw earlier in Figure 2(a), 2(b) and 3(b), the longitudinal dependence of the dimensionless local BLT δ(x)/H\delta(x)/H deviates away from the corresponding NCBL scaling, with increase in RawRa_w, implying a shift in the nature of these BLs towards mixed convection.

Average shear on the BLs and flux scaling

We now estimate the dimensionless average thermal boundary layer thickness δT/H‾\overline{\delta_T/H} at each RawRa_w by numerically averaging the various values of δT/H\delta_T/H over the horizontal dimensionless plume spacing in the presence of shear (λs/(2H)\lambda_s/(2H)), as

δT‾H=2Hλs∫0λs/(2H)δTH dx,(25)\frac{\overline{\delta_T}}{H}=\frac{2H}{\lambda_s}\int_0^{\lambda_s/(2H)}\frac{\delta_T}{H}\,dx, \tag*{(25)}

where, λs\lambda_s is given by (2). Using (2.17), for Pr=1Pr=1, (25) can be rewritten as,

δT‾H=2Hλs∫0λs/(2H)1C2δH dx.(26)\frac{\overline{\delta_T}}{H}=\frac{2H}{\lambda_s}\int_0^{\lambda_s/(2H)}\frac{1}{C_2}\frac{\delta}{H}\,dx. \tag*{(26)}

We evaluate (26) using the δ\delta obtained from the numerical solution of (22) and (7), and λs\lambda_s from (2), with ReRe obtained from (23) and (24) for the corresponding range of RawRa_w, with the value of C2=0.72C_2=0.72, as suggested by Shevkar et al. [32]. Using these values of δT‾/H\overline{\delta_T}/H obtained from (26), for each of the Reynolds number relations (23) and (24), for the different RawRa_w corresponding to the range of (23) and (24), we then obtain Nu=H/(2δT‾)Nu=H/(2\overline{\delta_T}), for Γ=1\Gamma=1 and Γ=0.5\Gamma=0.5, respectively; the obtained variations of NuNu with RawRa_w are shown in Figure 6(a). The inset in Figure 6(a) shows the corresponding variation of Nu/Raw1/3Nu/Ra_w^{1/3} with RawRa_w; Nu/Raw1/3∼Raw−0.0059Nu/Ra_w^{1/3}\sim Ra_w^{-0.0059} for Γ=1\Gamma=1 and Nu/Raw1/3∼Raw−0.0016Nu/Ra_w^{1/3}\sim Ra_w^{-0.0016} for Γ=0.5\Gamma=0.5. Figure 6(a) shows the resulting NuNu vs RawRa_w scalings, namely, Nu∼Raw0.327Nu\sim Ra_w^{0.327} for 8×107≤Raw≤1×10128\times10^7\le Ra_w\le1\times10^{12} at Γ=1\Gamma=1, and Nu∼Raw0.332Nu\sim Ra_w^{0.332} for 1×1011≤Raw≤5×10141\times10^{11}\le Ra_w\le5\times10^{14} at Γ=0.5\Gamma=0.5. Then, the present model where the local BLs in between the plumes, forced externally by the shear due to the LSF, determine the flux, predicts the classical 1/3 scaling of flux for the whole range of 8×107≤Raw≤5×10148\times10^7\le Ra_w\le5\times10^{14}; no transition to any ultimate regime is observed at Raw=2.5×1014Ra_w=2.5\times10^{14} as suggested by He et al. [11].

Plots of $Nu$ versus $Ra_w$ and mean local dimensionless shear versus $Ra_w$

Figure 6. (a) Variation of NuNu with RawRa_w; —, 0.1842Raw0.3270.1842Ra_w^{0.327} and − − −, 0.225Raw0.330.225Ra_w^{0.33}. · · ·, 0.137Raw1/30.137Ra_w^{1/3}. Inset figure shows the variation of Nu/Raw1/3Nu/Ra_w^{1/3} with RawRa_w; —, ∼Raw−0.0059\sim Ra_w^{-0.0059} and − − −, ∼Raw−0.0016\sim Ra_w^{-0.0016}. (b) Variation of mean local dimensionless shear with RawRa_w; —, 0.734Raw−0.1020.734Ra_w^{-0.102} and − − −, 0.428Raw−0.0890.428Ra_w^{-0.089}. Top-right inset shows the variation of u‾∣z=δ\overline{u}|_{z=\delta} with RawRa_w; —, ∼Raw1/3\sim Ra_w^{1/3} and − − −, ∼Raw0.38\sim Ra_w^{0.38}. Bottom-left inset shows the variation of u‾∣z=δ/Vbl\overline{u}|_{z=\delta}/V_{bl} with RawRa_w. —, 0.58Raw−0.00670.58Ra_w^{-0.0067} and − − −, 0.039Raw0.0590.039Ra_w^{0.059}.

As we saw earlier, if the local BLs are purely natural convection type, with the whole of ΔTw\Delta T_w occurring across them, when Nu∼Raw1/3Nu\sim Ra_w^{1/3}, the classical scaling we observe for Raw>1011Ra_w>10^{11}. At the same time, in contrast, we showed in § 3.1 that the shear at the edge of these local BLs, u∣z=δu|_{z=\delta}, increase with increase in RawRa_w (see the inset in Figure 5), thereby making these local BLs deviate more from NCBLs. To understand this seemingly contradictory prediction of an increasing shear forcing at the edge of these BLs with increase in RawRa_w, coinciding with a progressive tending towards a classical 1/3 flux scaling, for 8×107≤Raw≤5×10148 \times10^{7} \le Ra_w \le5 \times10^{14}, we now look at the average external shear on these local BLs, as well as the relative magnitude of the average shear forcing with respect to NCBL velocities, predicted by the present model.

We estimate the averaged dimensionless shear A‾\overline{A} at the edge of the local BLs by longitudinally averaging A(x∗)A(x^*), obtained by numerically solving (22) and (7), over the distance λs/(2H)\lambda_s/(2H), using a similar expression as (26), where δ/C2H\delta/C_2H is replaced with A(x)A(x).

Figure 6(b) shows the variation of A‾\overline{A} with RawRa_w. A‾\overline{A} decreases with increase in RawRa_w, similar to the reduction of A(x∗)A(x^*) with RawRa_w at any x∗x^*, noticed earlier in Figure 4(a), Figure 4(b) and Figure 5. Figure 6(b) shows that A‾∼Raw−0.102\overline{A} \sim Ra_w^{-0.102} for Raw≤1012Ra_w \le10^{12} at Γ=1\Gamma= 1 and A‾∼Raw−0.089\overline{A} \sim Ra_w^{-0.089} for Raw≥1×1011Ra_w \ge1 \times10^{11} at Γ=0.5\Gamma= 0.5, the exponents being almost the same as those obtained for A(x)A(x).

As in the case of A(x∗)A(x^*), A‾=u‾∣z=δ/VF\overline{A} = \left.\overline{u}\right|_{z=\delta}/V_F, with VF∼(ν/H)Raw3/7V_F \sim(\nu/H)Ra_w^{3/7} for Raw≤1012Ra_w \le10^{12} at Γ=1\Gamma= 1 and as (ν/H)Raw0.48(\nu/H)Ra_w^{0.48}, for Raw≥1×1011Ra_w \ge1 \times10^{11} at Γ=0.5\Gamma= 0.5, the average shear at the edge of the BLs, u‾∣z=δ∼(ν/H)Raw1/3\left.\overline{u}\right|_{z=\delta} \sim(\nu/H)Ra_w^{1/3} for Raw≤1012Ra_w \le10^{12} at Γ=1\Gamma= 1 and as (ν/H)Raw0.38(\nu/H)Ra_w^{0.38} for Raw≥1011Ra_w \ge10^{11} at Γ=0.5\Gamma= 0.5, as shown in the top right inset in Figure 6(b). Hence, the spatially averaged shear at the edge of the local BLs (u‾∣z=δ\left.\overline{u}\right|_{z=\delta}) increase with RawRa_w, eventhough the dimensionless average shear at the edge of the BLs (A‾\overline{A}) decreases with increase in RawRa_w. Eventhough the shear forcing of the local BLs increases faster with RawRa_w at higher RawRa_w (as Raw0.38Ra_w^{0.38}, see top-right Figure 6(b)) compared to that at lower RawRa_w (as Raw1/3Ra_w^{1/3}, see top-right Figure 6(b)), the magnitude of shear at the higher RawRa_w are smaller, clearly due to the lower value of Γ=0.5\Gamma= 0.5 at the higher values of RawRa_w.

The bottom left inset of Figure 6(b) shows the variation of u‾∣z=δ/Vbl\left.\overline{u}\right|_{z=\delta}/V_{bl} with RawRa_w, where VblV_{bl} (1) is the characteristic natural convection boundary layer velocity [21]. For Raw≤1012Ra_w \le10^{12} at Γ=1\Gamma= 1, u‾∣z=δ/Vbl\left.\overline{u}\right|_{z=\delta}/V_{bl} decreases with increase in RawRa_w as ∼Raw−0.0067\sim Ra_w^{-0.0067}, while it increases with increase in RawRa_w for Raw≥1×1011Ra_w \ge1 \times10^{11} at Γ=0.5\Gamma= 0.5, as ∼Raw0.059\sim Ra_w^{0.059}. This changeover in scaling of u‾∣z=δ/Vbl\left.\overline{u}\right|_{z=\delta}/V_{bl} for Γ=0.5\Gamma= 0.5 occurs due to the change in dependence of ReRe on RawRa_w, from Re∼Raw3/7Re \sim Ra_w^{3/7} (23) for Γ=1\Gamma= 1, to Re∼Raw0.48Re \sim Ra_w^{0.48} (24) for Γ=0.5\Gamma= 0.5. Similar changeovers in the scaling of A(x∗)A(x^*), A‾\overline{A} and NuNu, due to this change in dependence of ReRe on RawRa_w, are observed for Γ=0.5\Gamma= 0.5 in Figure 5, Figure 6(b) and Figure 6(a), respectively. Eventhough u‾∣z=δ/Vbl\left.\overline{u}\right|_{z=\delta}/V_{bl} increase with increasing RawRa_w at Γ=0.5\Gamma= 0.5 for 1×1011≤Raw≤5×10141 \times10^{11} \le Ra_w \le5 \times10^{14}, we note that it is always less than 0.3, and due to the lower Γ\Gamma, much lesser than the corresponding values of u‾∣z=δ/Vbl∼0.45\left.\overline{u}\right|_{z=\delta}/V_{bl} \sim0.45 at Γ=1\Gamma= 1. Hence, eventhough the average shear forcing (u‾∣z=δ\left.\overline{u}\right|_{z=\delta}) increases with increase in RawRa_w (see top-right inset in Figure 6(b)) natural convection velocities are still the dominant velocities within these local BLs, till Raw=5×1014Ra_w = 5 \times10^{14}, for both the aspect ratios, Γ=1\Gamma= 1 and Γ=0.5\Gamma= 0.5.

This dominance of the NC velocities within the local BLs, forced externally by the shear due to the LSF, even at RawRa_w as high as 5×10145 \times10^{14}, explains the observation of the classical 1/3 scaling of flux by the present model for 8×107≤Raw≤5×10148 \times10^{7} \le Ra_w \le5 \times10^{14}, with no transition to the ultimate regime seen at Raw≥1013Ra_w \ge10^{13}. Hence, eventhough the shear forcing of these BLs increase with increase in RawRa_w (see top right inset in Figure 6(b)), the natural convection velocities dominate in these local BLs to give a classical 1/3 scaling of flux. An extrapolation of the trend of variation of u‾∣z=δ/Vbl\left.\overline{u}\right|_{z=\delta}/V_{bl} for Γ=0.5\Gamma= 0.5 seen in the bottom inset of Figure 6(b) to higher RawRa_w shows that u‾∣z=δ/Vbl∼1\left.\overline{u}\right|_{z=\delta}/V_{bl} \sim1 at Raw∼1023Ra_w \sim10^{23}; for higher RawRa_w we could expect the local BLs to become shear dominant, possibly resulting in a transition in flux scaling. However this expectation must be qualified by the observation that the plume spacing, which determines the longitudinal extent of these local BLs, given by (2) is valid only till Raw=5×1016Ra_w = 5 \times10^{16} at Γ=0.5\Gamma= 0.5. In any case (2) is a decreasing function of RawRa_w, and hence it is possible that at some high RawRa_w, the local BL length could become of the order of plume thickness; a new regime of flux scaling could be expected beyond such an RawRa_w.

Conclusions

Our study’s key contribution lies in obtaining Nusselt number variation across a wide range of 8×107≤Raw≤5×10148 \times10^{7} \le Ra_w \le5 \times10^{14} through a numerical solution of a fifth-order algebraic scaling equation for boundary layers forced by the LSF in RBC. We propose a boundary layer model (refer Figure 1) which assumes the laminar natural convection BLs forced by the mean Prandtl-Blasius velocity profile generated due to the LSF velocity VFV_F. By integrating the BL equations (§ 2.1) and using the appropriate scaling arguments for BL flow (§ 2.2) and the shear forcing boundary condition at the edge of the BL (2.4) in RBC, we obtained an algebraic equation (3.1). We use the expressions of Reynolds number based on the LSF velocity VFV_F given by Cioni et al. [5] for Γ=1\Gamma= 1 and Raw≤1012Ra_w \le10^{12} (3.2) and Roche et al. [25] for Γ=0.5\Gamma= 0.5 and 1×1011≤Raw≤5×10141 \times10^{11} \le Ra_w \le5 \times10^{14} (3.3). The algebraic equation was solved numerically to obtain the variation of thermal BL thickness with the longitudinal distance at various RawRa_w. At any RawRa_w, the averaged thermal BL thickness was then obtained by integrating δT\delta_T over half the mean plume spacing λs\lambda_s (3.5) and later this averaged thickness was used to obtain NuNu. The results reveal that NuNu exhibits a power-law dependence on RawRa_w, with an exponent of m=0.327m = 0.327 for Raw≤1012Ra_w \le10^{12}, while mm approaches 1/31/3 for 1×1011≤Raw≤5×10141 \times10^{11} \le Ra_w \le5 \times10^{14}.

It has been shown that the dimensionless shear at the edge of the local boundary layer decreases as RawRa_w increases. The relationship between the average dimensionless shear and RawRa_w is given by A‾∼Raw−0.102\overline{A} \sim Ra_w^{-0.102} for 8×107≤Raw≤1×10128 \times10^{7} \le Ra_w \le1 \times10^{12} at Γ=1\Gamma= 1 and A‾∼Raw−0.089\overline{A} \sim Ra_w^{-0.089} for 1×1011≤Raw≤5×10141 \times10^{11} \le Ra_w \le5 \times10^{14} at Γ=0.5\Gamma= 0.5. The average shear velocity acting at the edge of the boundary layer, denoted by u‾∣z=δ\left.\overline{u}\right|_{z=\delta}, is related to the dimensionless velocity scale A‾\overline{A} through the expression A‾=u‾∣z=δ/VF\overline{A} = \left.\overline{u}\right|_{z=\delta}/V_F. For Raw≤1012Ra_w \le10^{12}, the velocity scale VFV_F scales as Raw3/7Ra_w^{3/7}, while for Raw>1×1011Ra_w > 1 \times10^{11}, VFV_F scales as Raw0.48Ra_w^{0.48}. Thus, u‾∣z=δ\left.\overline{u}\right|_{z=\delta} increases with increasing RawRa_w, scaling as u‾∣z=δ∼Raw1/3\left.\overline{u}\right|_{z=\delta} \sim Ra_w^{1/3} for 8×107≤Raw≤10128 \times10^{7} \le Ra_w \le10^{12} at Γ=1\Gamma= 1, and u‾∣z=δ∼Raw0.38\left.\overline{u}\right|_{z=\delta} \sim Ra_w^{0.38} for 1×1011≤Raw≤5×10141 \times10^{11} \le Ra_w \le5 \times10^{14} at Γ=0.5\Gamma= 0.5. Similarly, the local shear at the edge of the boundary layer, u∣z=δ\left.u\right|_{z=\delta}, increases with RawRa_w, scaling as u∣z=δ∼Raw0.331\left.u\right|_{z=\delta} \sim Ra_w^{0.331} for 8×107≤Raw≤10128 \times10^{7} \le Ra_w \le10^{12} at Γ=1\Gamma= 1, and u∣z=δ∼Raw0.398\left.u\right|_{z=\delta} \sim Ra_w^{0.398} for 1×1011≤Raw≤5×10141 \times10^{11} \le Ra_w \le5 \times10^{14} at Γ=0.5\Gamma= 0.5. Eventhough, u‾∣z=δ\left.\overline{u}\right|_{z=\delta} increased with increasing RawRa_w, their normalization with the NCBL velocity VblV_{bl}, u‾∣z=δ/Vbl\left.\overline{u}\right|_{z=\delta}/V_{bl} decreases with RawRa_w for Raw≤1012Ra_w \le10^{12} and then exhibits a turnaround at Raw∼1013Ra_w \sim10^{13}, as shown in the inset of Figure 6(b). However, this turnaround in the scaling of the shear velocity acting at the edge of BLs does not induce a transition in the flux scaling to the so called ultimate regime for Raw≤1015Ra_w \le10^{15}. The most probable reason is the much smaller than one values of u‾∣z=δ/Vbl\left.\overline{u}\right|_{z=\delta}/V_{bl}, the sub-dominant shear forcing acting on the BLs compared to the NCBL velocities within the BLs for the given range of RawRa_w.

Based on the aforementioned theoretical findings for Raw≤1015Ra_w \le10^{15}, it is suggested that the classical scaling of flux was observed due to the sub-dominance of shear forcing on boundary layers relative to natural convection velocities within the boundary layers. Our observations does not indicate any possibility of the ultimate regime until Raw=1015Ra_w = 10^{15}. This is in contrast to the suggestion made by He et al. [11] and consistent with the results of Iyer et al. (2019) and Lindborg [16]. For Raw>1015Ra_w > 10^{15}, the shear effects would be dominant when u‾∣z=δ/Vbl≫1\left.\overline{u}\right|_{z=\delta}/V_{bl} \gg1. If the trend of u‾∣z=δ/Vbl\left.\overline{u}\right|_{z=\delta}/V_{bl} with RawRa_w at Γ=0.5\Gamma= 0.5, shown in the inset of Figure 6(b), continues, it suggests that values of u‾∣z=δ/Vbl≫1\left.\overline{u}\right|_{z=\delta}/V_{bl} \gg1 could potentially cause BLs to become turbulent, leading to a transition in the flux scaling. As per the trend at Γ=0.5\Gamma= 0.5, this transition would then occur at Raw≈1023Ra_w \approx10^{23}, in the range of RawRa_w proposed originally by Kraichnan [14], i.e., 1021≤Raw≤102410^{21} \le Ra_w \le10^{24}. Furthermore, the transition in the flux scaling for Γ∼1\Gamma\sim1 and Pr∼1Pr \sim1 could occur at RawRa_w slightly less than 102310^{23}.

In conclusion, our analysis shows that laminar natural convection-type boundary layers (BLs), when subjected to a sufficiently large driving strength similar to that in Rayleigh-Bénard convection (RBC), are not expected to remain laminar for Raw≫1015Ra_w \gg10^{15}. They may first become Prandtl-Blasius type and then transition to a turbulent state, leading to a change in flux scaling. Finally, the LSF cannot be defined solely by scaling laws but must be treated as a statistical quantity. In addition, some regions of boundary layers get strongly affected by the shear due to the dynamically varying LSF and others are not ([32]). The dynamic, wide-ranging LSF observed close to the plate highlights the critical need to study boundary layers at the local scale, as opposed to the global scale in Rayleigh-Bénard convection. Further investigations for Raw≥1015Ra_w \ge10^{15}, utilizing high-resolution DNS and experimental data, are necessary to study the effect of shear generated by the LSF on local boundary layers at various spatial locations. Such studies are crucial to clarify the complete picture, including the sub-critical transitions.

References

  1. [1]AHLERS, G., BODENSCHATZ, E., HARTMANN, R., HE, X., LOHSE, D., REITER, P., STEVENS, R. J. A. M., VERZICCO, R., WEDI, M., WEI, M., WESSLING, S., ZHANG, X., ZWIRNER, L. & SHISHKINA, O. 2022 Aspect ratio dependence of heat transfer in a cylindrical Rayleigh-Bénard cell. Phys. Rev. Lett. 128, 084501.DOI
  2. [2]AHLERS, G., GROSSMANN, S. & LOHSE, D. 2009 Heat transfer and large-scale dynamics in turbulent Rayleigh-Bénard convection. Rev. Mod. Phys. 81, 503.DOI
  3. [3]CASTAING, B., GUNARATNE, G., HESLOT, F., KADANOFF, L., LIBCHABER, A., THOMAE, S., WU, X. Z., ZALESKI, S. & ZANETTI, G. 1989 Scaling of hard thermal turbulence in Rayleigh-Bénard convection. J. Fluid Mech. 204, 1–30.DOI
  4. [5]CIONI, S., CILIBERTO, S. & SOMMERIA, J. 1997 Strongly turbulent Rayleigh-Bénard convection in mercury: comparison with results at moderate Prandtl number. J. Fluid Mech. 335, 111–140.DOI
  5. [6]CREYSSELS, M. & MARTINAND, D. 2024 Stability analysis of sheared thermal boundary layers and its implication for modelling turbulent rayleigh–bénard convection. European Journal of Mechanics - B/Fluids 105, 97–103.DOI
  6. [7]GROSSMANN, S. & LOHSE, D. 2000 Scaling in thermal convection: a unifying theory. J. Fluid Mech. 407, 27.
  7. [8]GROSSMANN, S. & LOHSE, D. 2011 Multiple scaling in the ultimate regime of thermal convection. Phys. Fluids 23 (4), 045108.DOI
  8. [9]GUNASEGARANE, G. S. & PUTHENVEETTIL, B. A. 2014 Dynamics of line plumes on horizontal surfaces in turbulent convection. J. Fluid Mech. 749, 37–78.DOI
  9. [10]HE, J. C., BAO, Y. & CHEN, X. 2022 Scaling transition of thermal dissipation in turbulent convection. Physics of Fluids.DOI
  10. [11]HE, X., FUNFSCHILLING, D., NOBACH, H., BODENSCHATZ, E. & AHLERS, G. 2012 Transition to the ultimate state of turbulent Rayleigh-Bénard convection. Phys. Rev. Lett. 108, 024502.pubmed.ncbi.nlm.nih.gov/22324688
  11. [12]HOWARD, L.N. 1966 Convection at high rayleigh number. In Proceedings of the 11th Congress of Applied Mechanics, Munich (ed. H. Görtler), Applied Mechanics p. pp. 1109–1115.DOI
  12. [13]IYER, K. P., SCHEEL, J. D., SCHUMACHER, J. & SREENIVASAN, K. R. 2020 Classical 1/3 scaling of convection holds up to Ra = 10¹⁵. Proceedings of the National Academy of Sciences 117 (14), 7594–7598.DOI
  13. [14]KRAICHNAN, R. H. 1962 Turbulent Thermal Convection at Arbitrary Prandtl Number. The Physics of Fluids 5 (11), 1374–1389, arXiv: https://pubs.aip.org/aip/pof/article-pdf/5/11/1374/12772254/1374_1_online.pdf.DOI
  14. [15]LAM, S., SHANG, X., ZHOU, S. Q. & XIA, K. Q. 2002 Prandtl number dependence of the viscous boundary layer and the Reynolds numbers in Rayleigh-Bénard convection. Phys. Rev. E 65, 066306.pubmed.ncbi.nlm.nih.gov/12188827
  15. [16]LINDBORG, E. 2023 Scaling in rayleigh–bénard convection. Journal of Fluid Mechanics 956, A34.DOI
  16. [18]Pera, L. & Gebhart, B. 1973a Natural convection boundary layer flow over horizontal and slightly inclined surfaces. Int. J. Heat Mass Trans. 16 (6), 1131–1146.DOI
  17. [19]Pera, L. & Gebhart, B. 1973b Natural convection boundary layer flow over horizontal and slightly inclined surfaces. Int. J. Heat Mass Transfer 16 (6), 1131–1146.DOI
  18. [21]Puthenveettil, B. A. & Arakeri, J. H. 2005 Plume structure in high Rayleigh number convection. J. Fluid Mech. 542, 217–249.DOI
  19. [22]Puthenveettil, B. A., Gunasegarane, G. S., Agrawal, Y. K., Schmeling, D., Bosbach, J. & Arakeri, J. H. 2011 Length of near-wall plumes in turbulent convection. J. Fluid Mech. 685, 335–364.DOI
  20. [25]Roche, P-E, Gauthier, F, Kaiser, R & Salort, J 2010 On the triggering of the ultimate regime of convection. New Journal of Physics 12 (8), 085014.DOI
  21. [26]Rotem, Z. & Claassen, L. 1969 Natural convection above unconfined horizontal surfaces. J. Fluid Mech. 39, 173–192.DOI
  22. [27]Rotem, Z. & Classen, L. 1969 Natural convection above unconfined horizontal surfaces. J. Fluid Mech. 39(part1), 173–192.DOI
  23. [28]Samuel, R. J., Bode, M., Scheel, J. D., Sreenivasan, K. R. & Schumacher, J. 2024 Boundary layers in thermal convection are fluctuation-dominated, arXiv: 2403.12877.
  24. [29]Scheel, J. D., Kim, E. & White, K. R. 2012 Thermal and viscous boundary layers in turbulent Rayleigh–Bénard convection. J. Fluid Mech. 711, 281–305.DOI
  25. [31]Shevkar, P. P., Gunasegarane, G. S., Mohanan, S. K. & Puthenveettil, B. A. 2019 Effect of shear on coherent structures in turbulent convection. Phys. Rev. Fluids 4, 043502.
  26. [32]Shevkar, P. P., Mohanan, S. K. & Puthenveettil, B. A. 2023a Effect of shear on local boundary layers in turbulent convection. Journal of Fluid Mechanics 962, A41.DOI
  27. [33]Shevkar, P. P. & Puthenveettil, B. A. 2024 Boundary layer velocity field in turbulent convection. under preparation .
  28. [34]Shevkar, P. P., Sadu, R. R. & Puthenveettil, B. A. 2023b Tomographic PIV investigation of thermal plumes in turbulent convection. IET Conference Proceedings pp. 131–136(5).DOI
  29. [35]Shevkar, P. P., Vishnu, R., Mohanan, S. K., Vipin, K., Mathur, M. & Puthenveettil, B. A. 2022 On separating plumes from boundary layers in turbulent convection. J. Fluid Mech. 941, A5.DOI
  30. [36]Shraiman, B. I. & Siggia, E. D. 1990 Heat transport in high-rayleigh-number convection. Phys. Rev. A 42, 3650–3653.pubmed.ncbi.nlm.nih.gov/9904456
  31. [37]Skrbek, L. & Urban, P. 2015 Has the ultimate state of turbulent thermal convection been observed? Journal of Fluid Mechanics 785, 270–282.DOI
  32. [38]Stevens, R., Poel, E. P., Grossmann, S. & Lohse, D. 2013 The unifying theory of scaling in thermal convection: the updated predfactors. J. Fluid Mech. 730, 295–308.DOI
  33. [39]Theerthan, S. A. & Arakeri, J. H. 1998 A model for near wall dynamics in turbulent Rayleigh-Bénard convection. J. Fluid Mech. 373, 221–254.DOI
  34. [40]Theerthan, S. A. & Arakeri, J. H. 2000 Planform structure and heat transfer in turbulent free convection over horizontal surfaces. Phys. Fluids 12, 884–894.DOI

Paper details

vlnc:W_mf688c5h
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents