Fact-checked by Grok 2 weeks ago

Nusselt number

The Nusselt number (Nu) is a dimensionless in that represents the ratio of convective to pure conductive across a boundary under identical conditions. It is mathematically defined as Nu = (h L) / k, where h is the convective , L is the of the surface, and k is the thermal conductivity of the . A Nusselt number greater than 1 indicates that enhances beyond conduction alone, while Nu = 1 corresponds to purely conductive transfer. Named after the German mechanical engineer Ernst Kraft Wilhelm Nusselt (1882–1957), the number emerged from his pioneering work on during the early . Nusselt, a professor at the Technical Universities of and , introduced foundational dimensionless analyses in his 1915 paper on convective , which laid the groundwork for modern correlations involving the Nusselt number. His contributions extended to for and , influencing the number's application in multiphase flows. In engineering practice, the Nusselt number is essential for predicting and optimizing convective processes in both forced and natural convection regimes. For forced convection, such as in pipe flows or over flat plates, empirical correlations express Nu as a function of the (Re) and (Pr), typically in the form Nu = C Re^m Pr^n, where C, m, and n are constants depending on and flow conditions. In natural convection, driven by , Nu is correlated with the (Gr) and Pr, for example, Nu = f(Gr Pr), enabling assessments of from vertical plates or enclosures without external forcing. These relations are widely applied in designing heat exchangers, cooling systems, and safety analyses to ensure efficient thermal management.

Fundamentals

Definition

The Nusselt number, denoted as Nu, is a dimensionless parameter fundamental to the analysis of convective heat transfer processes in fluids. It quantifies the enhancement of heat transfer due to convection relative to pure conduction across a fluid layer. Mathematically, the Nusselt number is expressed as Nu = \frac{h L}{k}, where h is the convective heat transfer coefficient (in W/m²·K), L is a characteristic length scale of the geometry (in m), and k is the thermal conductivity of the fluid (in W/m·K). The convective heat transfer coefficient h itself is defined through Newton's law of cooling as h = \frac{q}{T_s - T_\infty}, with q representing the surface heat flux (in W/m²), T_s the temperature at the solid surface (in K), and T_\infty the far-field or bulk fluid temperature (in K). This formulation arises from the local energy balance at the fluid-solid interface, where h encapsulates the combined effects of conduction within the fluid and advection due to fluid motion. Distinctions exist between local and average forms of the Nusselt number. The local Nusselt number, Nu_x, is evaluated at a specific position x along a surface using the corresponding local h_x, reflecting variations in the thermal boundary layer development. In contrast, the average Nusselt number, Nu, integrates the local values over the entire surface area to yield an overall measure, typically computed as Nu = \frac{1}{A} \int_A Nu_x \, dA, where A is the surface area. As a dimensionless group, Nu inherently lacks units, emerging from the normalization of convective resistance by conductive resistance alone. Specifically, it represents the of total (convective) to the heat transfer that would occur by conduction alone across a stagnant layer of thickness L; a value of Nu = 1 indicates pure conduction, while Nu > 1 signifies convective enhancement.

Physical Significance

The Nusselt number serves as a dimensionless measure of the enhancement in due to relative to pure conduction across a layer adjacent to a surface. Specifically, it represents the of the total heat transfer (convective plus conductive) to the conductive heat transfer alone under the same conditions, where a value of Nu = 1 corresponds to heat transfer dominated entirely by conduction, as if the were . Values greater than 1 indicate the increasing influence of motion in augmenting heat transfer, with typical ranges spanning 1 to 100 for natural processes driven by and 10 to 1000 for scenarios involving external flow, such as in turbulent regimes. In the context of boundary layers, the Nusselt number quantifies how fluid motion within the thermal near the surface steepens the , thereby enhancing the convective compared to a stagnant scenario. This enhancement arises because thins the effective thermal resistance layer, making the Nusselt number inversely proportional to the ; as velocities increase, the compresses, leading to higher Nusselt values that reflect more efficient dissipation or absorption at the . Practically, the Nusselt number is essential for optimizing the design of thermal systems, such as and electronic cooling apparatus, where higher values signify improved efficiency in transferring without excessive size or input. For instance, engineers use it to predict and enhance in compact configurations, ensuring adequate cooling in applications like power plants or HVAC systems by correlating flow conditions to rates. While the Nusselt number can theoretically approach for infinitesimally thin layers under idealized high-velocity conditions, in real-world applications it remains finite due to factors like , instabilities, and the inherent limits of , preventing unbounded enhancement.

Theoretical Foundations

Derivation from Equations

The derivation of the Nusselt number begins with the fundamental energy equation governing in a , which accounts for conduction, , and other effects. The general form of the energy equation for a is \rho c_p \left( \frac{\partial T}{\partial t} + \mathbf{u} \cdot \nabla T \right) = \nabla \cdot (k \nabla T) + \Phi, where \rho is , c_p is specific heat at constant pressure, T is , \mathbf{u} is , k is thermal conductivity, and \Phi represents viscous dissipation. For many applications involving steady-state, with negligible viscous dissipation and constant properties, the equation simplifies to \rho c_p \mathbf{u} \cdot \nabla T = k \nabla^2 T. This balance between convective transport on the left and diffusive (conductive) transport on the right forms the basis for non-dimensional analysis in convective heat transfer. To non-dimensionalize the equation, characteristic scales are introduced: a reference length L, a reference velocity U, and a reference temperature difference \Delta T = T_s - T_\infty, where T_s is the surface temperature and T_\infty is the free-stream temperature. Dimensionless variables are defined as \mathbf{x}^* = \mathbf{x}/L, \mathbf{u}^* = \mathbf{u}/U, \theta = (T - T_\infty)/\Delta T, and t^* = t U / L. Substituting these into the simplified energy equation yields \mathbf{u}^* \cdot \nabla^* \theta = \frac{1}{\mathrm{Pe}} \nabla^{*2} \theta, where \mathrm{Pe} = U L / \alpha is the Péclet number, with thermal diffusivity \alpha = k / (\rho c_p). Since \mathrm{Pe} = \mathrm{Re} \cdot \mathrm{Pr}, where \mathrm{Re} = U L / \nu is the Reynolds number and \mathrm{Pr} = \nu / \alpha is the Prandtl number, this non-dimensional form reveals the relative importance of convective to diffusive heat transport. The Péclet number governs the scaling of temperature gradients in the flow field. The Nusselt number emerges naturally from the boundary condition at a , where the convective q = h (T_s - T_\infty) equals the conductive flux at the wall, q = -k (\partial T / \partial y)|_{y=0}. Thus, the h is h = -k (\partial T / \partial y)|_{y=0} / \Delta T. In non-dimensional terms, \partial T / \partial y = (\Delta T / L) (\partial \theta / \partial y^*), so h = -\frac{k}{L} \left( \frac{\partial \theta}{\partial y^*} \right)_{y^*=0}. The local Nusselt number is then defined as \mathrm{Nu} = h L / k = - (\partial \theta / \partial y^*)|_{y^*=0}, representing the dimensionless at the wall. This links the Nusselt number directly to the non-dimensional solution of the energy equation under the specified boundary conditions. From a perspective, the conduction across a characteristic length L scales as k \Delta T / L, while the convective scales as \rho c_p U \Delta T. The ratio of these, adjusted for the , yields \mathrm{Nu} \sim h L / k, where \mathrm{Nu} quantifies the enhancement of due to over pure conduction. In the non-dimensional energy equation, the $1/\mathrm{Pe} indicates that when \mathrm{Pe} \gg 1, thin thermal boundary layers form, leading to large temperature gradients and thus \mathrm{Nu} > 1. This underscores the Nusselt number as the inverse of the non-dimensional convective thermal resistance.

Relation to Other Dimensionless Numbers

The Nusselt number (Nu) is interconnected with other dimensionless parameters in convective heat transfer, arising from Buckingham π theorem applications and governing equation , which reveal functional dependencies for scaling heat transfer coefficients. In forced convection, Nu depends primarily on the (Re = \rho u L / \mu, the ratio of inertial to viscous forces) and the (Pr = \nu / \alpha, the ratio of momentum to ), expressed as Nu = f(Re, Pr); this reflects how flow inertia and fluid property ratios influence convective enhancement over conduction. In free convection, buoyancy-driven flows lead to Nu = f(Gr, Pr), where the (Gr = g \beta \Delta T L^3 / \nu^2, the ratio of buoyancy to viscous forces squared) captures effects driving the flow. These relations enable predictive correlations without dimensional specifics, prioritizing conceptual scaling in engineering design. A key linkage is through the Peclet number (Pe = Re Pr), which quantifies the relative importance of convective to diffusive heat transport; at high Pe (typically Pe > 100), where dominates, Nu scales with Pe, often as Nu \sim Pe^{1/3} in thin approximations for external flows over objects like spheres or plates. This scaling emerges because the decreases as \delta_t \sim L Pe^{-1/3}, leading to enhanced local rates proportional to the inverse of this thickness. For instance, in creeping flows past flat plates or streaming around spheres, asymptotic analyses confirm Nu \approx c Pe^{1/3} (with c a geometry-dependent constant), highlighting Pe's role in unifying behaviors across varying Re and Pr. Similarity solutions in rely on matching these groups to achieve self-similar and profiles, allowing experimental results from one scale to predict performance at another; for example, equating , in forced flows or , in free flows ensures dimensionless boundary layer equations are identical, facilitating direct scaling of without resolving full PDEs. This approach, rooted in theory, is particularly effective for laminar regimes where exact solutions exist, such as vertical plates in free . The Colburn analogy further ties Nu to momentum transfer by introducing the Colburn j-factor for heat, defined as j_H = \frac{\mathrm{Nu}}{\mathrm{Re} \, \mathrm{Pr}^{1/3}} = \frac{f}{8}, where f is the Darcy friction factor; this relation extends the Reynolds analogy to account for Pr \neq 1, enabling predictions of heat transfer from skin friction data in turbulent flows and drawing parallels to mass transfer (via Sherwood number). Valid for 0.6 < Pr < 60 and Re > 10^4, it underscores the coupled nature of transport phenomena in engineering correlations.

Free Convection Applications

Vertical Surfaces

In free convection along vertical surfaces, such as an isothermal or uniformly heated plate immersed in a quiescent , buoyancy forces drive the flow parallel to the surface due to density variations induced by differences. The characteristic length for these flows is typically the of the plate, L, and the governing dimensionless is the , defined as Ra_L = Gr_L Pr, where Gr_L = g β (T_s - T_∞) L^3 / ν^2 is the and Pr = ν / α is the , with g as , β the thermal expansion coefficient, T_s and T_∞ the surface and ambient s, ν the kinematic , and α the . For laminar flow regimes, applicable over 10^4 < Ra_L < 10^9, the average Nusselt number is correlated empirically as Nu_L = 0.59 Ra_L^{1/4}, providing a reliable estimate of the convective heat transfer enhancement relative to conduction across the boundary layer. This correlation, developed by Churchill and Chu, encompasses a wide range of Prandtl numbers (0.5 < Pr < 10,000) and stems from integrating experimental data with theoretical boundary layer predictions. At higher Rayleigh numbers, Ra_L > 10^9, the flow transitions to turbulence, where the average Nusselt number follows Nu_L = 0.10 Ra_L^{1/3}, again per Churchill and Chu, reflecting the increased mixing and thinner effective boundary layer in turbulent conditions. Correlations for surfaces with uniform wall temperature (T_s = constant) require adjustment when the boundary condition is instead uniform (q = constant), as the surface varies along the plate. In the laminar regime, the Nusselt number based on , Nu_q, is approximately Nu_q = (4/3) Nu_T, where Nu_T is the isothermal value, accounting for the linear increase in local difference with distance and the resulting thicker boundary layer. This factor arises from similarity solutions to the equations, ensuring consistent predictions across boundary conditions. These empirical relations originate from boundary layer theory, where the momentum and energy equations are solved for self-similar velocity and temperature profiles in the thin layer adjacent to the surface. Ostrach's numerical analysis established the foundational profiles, showing that velocity peaks near the wall and decays to zero in the quiescent ambient, while temperature transitions from T_s at the wall to T_∞ asymptotically, validating the power-law dependencies in the correlations through comparison with experiments.

Horizontal Surfaces and Enclosures

In free convection heat transfer from horizontal plates, the orientation significantly affects the Nusselt number due to buoyancy-driven flow stability. For the upper surface of a hot plate facing upward or the lower surface of a cold plate facing downward, where the density gradient promotes instability, the average Nusselt number is correlated as \overline{\mathrm{Nu}}_L = 0.54 \mathrm{Ra}_L^{1/4} for $10^4 < \mathrm{Ra}_L < 10^7. This empirical relation, developed by McAdams based on experimental data for air and other fluids, captures the laminar boundary layer development over the surface. Conversely, for the lower surface of a hot plate facing downward or the upper surface of a cold plate facing upward, the configuration is stable against buoyancy, leading to weaker convection and a reduced : \overline{\mathrm{Nu}}_L = 0.27 \mathrm{Ra}_L^{1/4} in the same Rayleigh number range. These correlations assume isothermal surfaces and provide estimates for engineering applications like solar collectors or electronic cooling plates. For enclosed spaces heated from below, such as horizontal fluid layers between parallel plates, the heat transfer transitions from conduction to convection at a critical Rayleigh number \mathrm{Ra}_c = 1708, marking the onset of instability in the fluid layer with rigid boundaries. Below \mathrm{Ra}_c, the Nusselt number remains unity, indicating purely conductive transfer across the enclosure height. Above this threshold, buoyancy initiates cellular convection patterns, known as , which enhance overall heat transfer; however, practical correlations emphasize average rather than local flow structures. In the turbulent regime for $3 \times 10^5 < \mathrm{Ra}_L < 7 \times 10^9, experimental measurements yield \overline{\mathrm{Nu}}_L = 0.069 \mathrm{Ra}_L^{1/3} \mathrm{Pr}^{0.074}, a relation derived by from mercury and silicone oil experiments, showing weak Prandtl number dependence for liquids. This scaling reflects the dominance of inertial and viscous effects in vigorous convection. In rectangular enclosures, the aspect ratio H/L (height to horizontal length) influences the Nusselt number by altering flow confinement and boundary effects; for shallow cavities (H/L < 1), correlations like those of Hollands et al. incorporate H/L to adjust for reduced endwall influences, yielding higher Nu values at fixed Ra compared to tall enclosures where three-dimensional effects prevail. Such dependencies are critical for applications like building insulation or nuclear reactor pools, where geometry modulates convective efficiency.

Forced Convection in External Flows

Laminar Flow over Flat Plates

In forced convection over a flat plate under laminar conditions, the Nusselt number quantifies the enhancement of heat transfer due to convection relative to conduction across the fluid film. The foundational correlations stem from the similarity solution to the boundary layer momentum and energy equations, utilizing the Blasius velocity profile for incompressible flow with constant properties. This approach assumes a zero-pressure gradient, negligible viscous dissipation, and a Prandtl number Pr > 0.6 to ensure the thermal boundary layer is thinner than the hydrodynamic one. For a flat plate maintained at uniform surface , the local Nusselt number at x from the is expressed as \Nu_x = 0.332 \Re_x^{1/2} \Pr^{1/3}, where \Re_x = U_\infty x / \nu is the local based on free-stream velocity U_\infty and kinematic viscosity \nu. This relation arises from numerically solving the self-similar for the profile, yielding a constant of 0.332 that reflects the convective enhancement. The corresponding average Nusselt number over plate length L is \Nu_L = 0.664 \Re_L^{1/2} \Pr^{1/3}, obtained by integrating the local value along the plate, effectively doubling the local \Nu_x at x = L due to the square-root dependence on length. These expressions highlight how heat transfer rates increase with flow speed and thermal capacity while decreasing with fluid viscosity and thermal diffusivity. When an unheated starting length \xi precedes the heated section, the thermal boundary layer develops within an already established hydrodynamic layer, altering the temperature gradient at the wall. The corrected local Nusselt number becomes \Nu_x = \frac{0.332 \Re_x^{1/2} \Pr^{1/3}}{\left[1 - (\xi / x)^{3/4}\right]^{1/3}} for x > \xi, accounting for the delayed onset of heating and resulting in higher heat transfer coefficients downstream compared to fully heated plates. This correction factor approaches unity as \xi / x \to 0 and grows as the ratio decreases, emphasizing the influence of upstream flow history on local convection. Under constant heat flux conditions, where the wall heat flux q_w is uniform, the surface temperature varies along the plate, leading to a modified similarity solution for the energy equation. The local Nusselt number, defined using the local temperature difference \Delta T(x), is \Nu_x = 0.453 \Re_x^{1/2} \Pr^{1/3}, approximately 36% higher than the isothermal case due to the sustained driving potential for heat transfer. The average value follows similarly by integration, maintaining the same functional form. These correlations apply within the laminar regime, typically for \Re_x < 5 \times 10^5, beyond which transition to turbulence occurs and alters the boundary layer structure. They presume low free-stream turbulence intensities (typically below 0.5%), as elevated levels can destabilize the laminar layer, advancing transition and invalidating the similarity assumptions without directly modifying the Nusselt number in fully laminar regions.

Flow over Spheres

In forced convection heat transfer over a sphere, the incoming uniform flow develops a thin boundary layer on the forward stagnation region, where the thermal boundary layer thickness influences the local heat transfer rates. As the flow progresses around the sphere, adverse pressure gradients cause the boundary layer to separate at an angle typically between 80° and 120° from the stagnation point, depending on the Reynolds number, leading to a recirculating wake region at the rear. This separation reduces local heat transfer coefficients in the wake compared to the front, but the average Nusselt number, based on the sphere's diameter, integrates contributions from both the attached boundary layer and the separated wake to characterize overall convective enhancement. At very low Reynolds numbers (Re ≪ 1), inertial effects are negligible, and the flow approximates ; in this limit, the approaches Nu = 2, corresponding to pure conduction heat transfer from an isothermal sphere to an infinite surrounding medium, as derived from solving the for the temperature field. This baseline value represents the minimum convective enhancement, with deviations increasing as Re rises due to convective transport within the developing boundary layers. A widely adopted empirical correlation for the average in forced convection over spheres, applicable across a broad range of conditions, is the Whitaker correlation: \overline{\mathrm{Nu}}_D = 2 + \left(0.4 \mathrm{Re}_D^{1/2} + 0.06 \mathrm{Re}_D^{2/3}\right) \mathrm{Pr}^{0.4} \left( \frac{\mu_\infty}{\mu_s} \right)^{1/4} valid for 3.5 ≤ Re_D ≤ 7.6 × 10^4 and 0.7 ≤ Pr ≤ 380, where all fluid properties are evaluated at the free-stream temperature T_∞ except for the surface viscosity μ_s at the sphere temperature T_s. The constant term 2 recovers the conduction limit at low Re, while the Re-dependent terms capture boundary layer thinning at the front and wake influences at higher Re; the exponent reflects the relative thermal and momentum boundary layer development. This correlation was developed by curve-fitting experimental data from multiple studies, ensuring accuracy within ±5% for gases and liquids. The viscosity ratio correction \left( \frac{\mu_\infty}{\mu_s} \right)^{1/4} accounts for variable fluid properties arising from temperature differences between the sphere surface and free stream, which can significantly alter the boundary layer velocity profile and thus heat transfer rates, particularly for large temperature gradients in gases where viscosity increases with temperature. For isothermal conditions (μ_∞ = μ_s), this factor equals unity, simplifying the expression.

Forced Convection in Internal Flows

Fully Developed Laminar Flow in Pipes

In fully developed laminar flow through pipes, the velocity profile follows the Hagen-Poiseuille distribution, given by u(r) = 2 U_m \left(1 - \left(\frac{r}{R}\right)^2 \right), where U_m is the mean velocity and R is the pipe radius. This parabolic profile arises from solving the Navier-Stokes equations under steady, incompressible, and no-slip boundary conditions. When the flow is also thermally fully developed, the temperature profile maintains a constant shape relative to the bulk mean temperature, leading to a constant Nusselt number independent of the Reynolds and Prandtl numbers. The energy equation in dimensionless form simplifies to \frac{u}{U_m} = \frac{1}{\eta} \frac{d}{d\eta} \left( \eta \frac{d\theta}{d\eta} \right), where \eta = r/R and \theta is the dimensionless temperature, allowing analytical solutions for the Nusselt number under specified boundary conditions. For the boundary condition of constant wall temperature in a circular tube, the Nusselt number approaches \mathrm{Nu} = 3.66 in the fully developed region, derived from the asymptotic behavior of the Graetz series solution as the axial distance becomes large. This value represents the eigenvalue-based infinite series summation for the temperature field in the thermal entrance region transitioning to fully developed conditions. For the case of constant axial wall heat flux (uniform q), an exact analytical solution yields \mathrm{Nu} = \frac{48}{11} \approx 4.36, obtained by integrating the energy equation with the parabolic velocity profile to find the quadratic temperature distribution across the radius. These constants highlight the enhanced heat transfer under uniform heat flux compared to isothermal walls due to the linear bulk temperature rise. The thermal entrance length, beyond which the fully developed Nusselt number applies, is estimated as L_t / D \approx 0.05 \, \mathrm{Re} \, \mathrm{Pr}, where D = 2R is the pipe diameter. This length scales with the product of Reynolds and Prandtl numbers, reflecting the diffusion of momentum and thermal boundary layers; for typical liquids (\mathrm{Pr} \gg 1), it exceeds the hydrodynamic entrance length. In practice, heat transfer analyses assume fully developed conditions when the pipe length significantly exceeds L_t. For non-circular ducts, the Nusselt number values are generalized through similar analytical or numerical solutions of the energy equation with the appropriate velocity profile, often tabulated based on the hydraulic diameter. For parallel plates under constant wall temperature, \mathrm{Nu} = 7.54 (based on hydraulic diameter D_h = 2b, where b is the plate spacing), equivalent to approximately 3.77 when referenced to the spacing b itself; an example value of 3.61 arises in contexts like square ducts under constant heat flux. These variations underscore the influence of cross-sectional geometry on the fully developed heat transfer coefficient.

Turbulent Flow in Pipes

In fully developed turbulent forced convection within pipes, the Nusselt number quantifies the enhancement of heat transfer over pure conduction, driven by chaotic fluid motion that promotes mixing across the cross-section. Empirical correlations, derived from extensive experimental data, provide practical estimates for the average Nusselt number under conditions of constant wall heat flux or temperature, assuming smooth pipes unless otherwise specified. These relations typically express Nu as a function of the Reynolds number (Re) and Prandtl number (Pr), with properties evaluated at the bulk fluid temperature. The Dittus-Boelter correlation represents a foundational empirical model for predicting the in turbulent pipe flows, originally developed from measurements in tubular radiators using air, water, and oils. It takes the form \Nu = 0.023 \Re^{0.8} \Pr^{n} where n = 0.4 for heating the fluid (wall hotter than bulk) and n = 0.3 for cooling (wall cooler than bulk). This relation applies to fully developed flow in smooth, circular pipes with \Re > 10^{4}, $0.7 < \Pr < 160, and length-to-diameter ratios L/D > 10 to minimize entrance effects. The correlation offers simplicity and reasonable accuracy (±25% typically) for many engineering applications but assumes negligible property variations and smooth surfaces. For broader applicability, including transitional regimes and wider property ranges, the Gnielinski correlation provides a more refined estimate by incorporating the Darcy friction factor f, which links heat transfer to momentum transport. The expression is \Nu = \frac{(f/8)(\Re - 1000)\Pr}{1 + 12.7 (f/8)^{0.5} (\Pr^{2/3} - 1)} with the smooth-pipe friction factor given by f = (0.79 \ln \Re - 1.64)^{-2}. Valid for $2300 < \Re < 5 \times 10^{6} and $0.5 < \Pr < 2000, this model achieves higher precision (±10-15%) than the Dittus-Boelter equation, particularly near the laminar-turbulent transition, by bridging analogies between heat and momentum transfer. The inclusion of f allows extension to rough pipes by substituting values from the Moody chart, thereby accounting for surface roughness effects that increase friction and enhance turbulence-induced heat transfer. When significant temperature-dependent property variations occur, such as in viscous oils, the Sieder-Tate correlation modifies the Dittus-Boelter form to correct for viscosity differences between the bulk fluid and wall surface. It is expressed as \Nu = 0.027 \Re^{0.8} \Pr^{1/3} \left( \frac{\mu}{\mu_s} \right)^{0.14} where \mu is the bulk and \mu_s is the at the wall temperature; all other properties are at the bulk temperature. Applicable for \Re > 10^{4}, $0.7 < \Pr < 16,700, and L/D > 10, this adjustment improves accuracy by up to 20% in cases with large viscosity gradients, though it remains limited to fully developed turbulent flow in smooth pipes. These correlations assume fully developed conditions, where entrance effects—such as higher initial rates due to developing and layers—are negligible beyond approximately 10-20 pipe diameters in turbulent flows. For shorter , supplementary factors or more detailed models may be needed to capture the elevated Nusselt numbers in the entrance region.