## Abstract

Supercritical carbon dioxide (sCO_{2}) can be utilized as a working fluid in various thermal systems including large-scale power cycles; portable power production units, centralized coolant systems, and standalone cooling devices. However, the lack of accurate prediction tools such as heat transfer coefficient correlations and insufficient research studies about the mechanisms controlling heat transfer processes are hindering its practical realization for key energy and cooling systems. The overall objective of this study is to extend fundamental knowledge about heat transfer and fluid dynamic processes in conduits pertinent to sCO_{2} with an emphasis on flow direction and inclination effects. This paper presents the study on effects of gravity, buoyancy on sCO_{2} flow at temperatures near and away from the pseudo-critical temperature. The experimental setup consists of a high temperature and pressure sCO_{2} heat transfer loop and flow testing facility. Recently, researched sCO_{2} heat exchangers can have tubes oriented at different angles such as 45 deg or 90 deg to horizontal. For optimized design of efficient and cost-effective turbomachinery components utilizing sCO_{2} as the heat transfer fluid, an understanding of convective heat transfer inside a tube/pipe is equally as important as external heat transfer. This paper presents an experimental and numerical study on sCO_{2} heat transfer at various inclinations with angles ranging from 0 deg (horizontal) to 90 deg (vertical) along with upward and downward flow direction with different inlet temperatures. Thermocouple-based temperature measurement is utilized at multiple locations within the test section axially and circumferentially to study the temperature distributions on the tube surface. Computational fluid dynamics (CFD) simulations have been performed using ANSYS Fluent to complement experimental data. The CFD and experiment have been analyzed against the well-known Gnielinski Nusselt number correlation.

## 1 Introduction

Supercritical CO_{2} heat transfer has gained a lot of recent attention due to its wide range of potential applications in areas, including, concentrated solar power plants [1], refrigeration [2,3], waste heat recovery [4,5], thermonuclear power generation [6], and air-conditioning systems [7,8]. The critical pressure and temperature for CO_{2} are 7.38 MPa and 304 K, respectively. Near the critical point, the fluid undergoes rapid changes in its thermophysical properties which makes it important to understand its heat transfer behavior if it is to be used as a working fluid in thermal engineering applications. There have been numerous experimental and numerical studies pertaining to sCO_{2} heat transfer in circular cross section tubes. Most of the literature is directed toward horizontal or vertical flow direction and there is a lack of knowledge when it comes to studying heat transfer for flow at an arbitrary inclination angle. There has been some numerical work for the latter [8], but no experimental work has been found within the public domain.

As multiple authors have shown, there can be significant influence of property variations on sCO_{2} heat transfer close to the pseudo-critical temperature [9–15]. Near the critical point of CO_{2}, Schnurr [9] reported circumferential variation in temperature at a cross section in a horizontal tube for sCO_{2}. Adebiyi and Hall [10] explained that occurrence of natural convective current from bottom surface to top surface of the horizontal tube is the reason behind enhanced heat transfer at the bottom surface. This phenomenon is represented by a sketch as shown in Fig. 1. This causes higher temperature at the top surface compared to the bottom surface of the tube. Liao and Zhao [16] reported that this effect of buoyancy on heat transfer is directly related to size of the tubes. This is mainly due to the Richardson number, which is a measure of natural convection and a ratio of Grashof number to Reynolds number squared.

In the vertical orientation, sCO_{2} heat transfer in tubes is highly dependent on flow direction. The effects of buoyancy on vertical flow heat transfer under turbulent conditions are explained by Jackson and Hall [17]. Heat transfer at the wall is directly related to variation in generation of wall shear stress. For upward direction flow, buoyancy forces are opposing wall shear stress, decreasing turbulence and hence, result in decreasing heat transfer. On the other hand, for downward direction flow, buoyancy forces increase wall shear stress, generating increased turbulence and hence, increasing heat transfer at the wall.

This paper presents experimental and numerical analysis on sCO_{2} heat transfer in a heated tube at five different inclinations; 0 deg, 30 deg, 45 deg, 60 deg, and 90 deg. Both upward and downward direction flow cases are performed here. Two inlet temperatures are considered here; one near the pseudo-critical temperature, and one away, to study their effects on heat transfer.

## 2 Experimental Methodology

### 2.1 Description of Experimental Rig.

A schematic of the experimental rig can be seen in Fig. 2. The test loop is setup in a closed-loop configuration to conserve usage of CO_{2} and is designed for maximum operating pressure and temperature of 100 bar and 150 °C. As seen from the right image in Fig. 3, the test section is located within a structural frame to facilitate variation of inclination by adjusting winches, hooked to the test section frame. Before filling the loop with CO_{2}, the loop is vacuumed to remove any air, ensuring close to pure concentrations of CO_{2} for testing. The loop is also equipped with an exhaust system to safely discharge CO_{2} from the loop. A CO_{2} cylinder and booster pump are used in series to pressurize the loop to the desired pressure. After the filling process, the booster pump is disconnected from the loop using an ON/OFF valve. A micropump gear pump is used to circulate sCO_{2} through the loop, with a maximum operating pressure of 105 bar. A buffer tank is used downstream of the gear pump to absorb any possible pressure fluctuations and deliver steady mass flowrate to the test section. A Coriolis mass flowrate meter suitable for CO_{2} flows in gas, liquid, and critical phases is used to measure mass flowrate. Inlet temperature to the test section is controlled using a preheater made from wrapped-around electric rope heaters and voltage controlling variacs supplying constant voltage.

Electric heating of the test section is achieved using a DC power supply, which is connected to the test section tube using machined copper busbars. These machined copper busbars ensure low thermal contact resistance achieved from good contact with test section tube. A busbar, which is in loose contact with the tube, might create local spikes in resistance, resulting in heat generation, making it difficult to account for heat transferred to the flow. Seamless stainless steel tubing of outer diameter 12.7 mm is used in the test section and tubing connections throughout the experimental loop. The test section is made of ASTM A213 SS316 tube with outer diameter of 12.7 mm and inner diameter of 10.92 mm, enclosed in a polyvinyl chloride pipe and insulated using mineral wool insulation. A water cooler is used downstream of the test section before the recirculating pump to prevent from exceeding temperature above working limit of the pump, as well as to achieve overall steady-state during testing. The water cooler consists of a tube array immersed in a cold water drum where the water temperature is controlled and recycled using a chiller.

Inlet flow temperature is measured using two different instruments to add redundancy in measurement. These consist of a K-type thermocouple probe and a RTD probe inserted at the same axial location using a cross-union fitting. The same setup is used to measure outlet flow temperature as well. Static pressure is measured at the inlet using a pressure transducer. The test section includes an unheated section with *L*/*D*_{int} = 74, heated section with *L*/*D*_{int} = 123 and unheated section *L*/*D*_{int} = 74 before measuring outlet flow temperature and the busbar sections *L*/*D*_{int} = 1.15. These values of *L*/*D*_{int} ratios are chosen based on conventional values necessary to fully develop the hydrodynamic boundary layer and thermal boundary layer. The data reduction process takes into account outside wall temperatures in the test section, which are measured using T-type thermocouples instrumented on the tube surface. At each axial location, four surface thermocouples are attached to the surface at 90 deg tangential interval to measure temperatures at top, bottom, right, and left wall locations (as shown in right side of Figs. 2 and 4(b)). Electric heating is accounted by measuring current provided by the DC power supply and voltage measured between two busbars.

### 2.2 Heat Loss Tests.

*T*

_{w−}_{amb}), outer tube area and area-specific equivalent conductance (

*U*

_{loss}(W/m

^{2}K)) between outer wall of tube and ambient as shown in Eq. (1). The goal of heat loss tests was to calculate this equivalent conductance,

*U*

_{loss}, since the other two parameters are known. Δ

*T*

_{w−}_{amb}is measured using thermocouples instrumented on outer wall of tube and ambient. To calculate

*U*

_{loss}, the test loop is vacuumed so that the inside wall of the tube does not experience any surface heat convection. Here, it is assumed that the axial conduction from the heated section of the tube to unheated section of the tube is negligible. That is why Eq. (1) holds true when the test loop is vacuumed

Three heat loss tests are performed with three different values of electric power to cover the range of Δ*T _{w −}*

_{amb}expected from flow experiments.

*U*

_{loss}is calculated for each case and plotted against Δ

*T*

_{w −}_{amb}as show in Fig. 5 to obtain an expression for

*U*

_{loss}. Heat loss is negligible, with a maximum heat loss percentage of 0.056% calculated between the electrical power and the heat added to the test section, for all experimental cases. Heat loss tests to calculate the equivalence conductance were performed for completeness. This plot is shown including the expression for

*U*

_{loss}as a function of Δ

*T*

_{w −}_{amb}in Fig. 5.

### 2.3 Data Reduction.

The test section is divided into seven zones as shown in Fig. 4(a) to account for an accurate energy balance. These zones include energy balance based on difference in bulk enthalpy between outlet and inlet. Figure 4(b) also shows all possible pathways of heat transfer in a domain, where Nusselt number is calculated. The approach to calculate local bulk enthalpy includes axial energy balance marching considering heat transferred to CO_{2}. This energy marching starts at the inlet where inlet enthalpy is calculated using REFPROP [18] from measured bulk temperature and pressure. Figure 4(b) shows the domain for calculation of heat transfer coefficient along with pathways of heat transfer to CO_{2}. Each heat transfer domain consists of 8.7 *× D*_{int} length around a thermocouple station and 90 deg sector (quadrant) of annular cross section of tube, each for four circumferential locations considered—top, bottom, left, and right. Data reduction process to calculate local bulk temperature and local Nusselt numbers follows the same equations as reported in the previous study [19].

*q*

_{vl}) is calculated using total electrical heat generation (

*Q*

_{electric,total}) and solid metal volume of the heated portion of the tube, as shown in Eq. (3). Thermal conductivity of stainless steel being a function of temperature is calculated using Eq. (4) [20]. Internal wall temperature is then obtained using Eq. (5).

*U*

_{loss}and temperature difference between external wall and ambient (Δ

*T*

_{w −}_{amb}), as shown in Eq. (6). Area of heat loss considered is the local external surface area of the tube (

*L*

_{local}

*× π × D*

_{ext})

Figure 4(b) shows the domain for calculation of heat transfer coefficient along with pathways of heat transfer to CO_{2}. Each heat transfer domain consists of a length 8.7 *×* D_{int} around a thermocouple station and a 90 deg sector (quadrant) of annular cross section of tube, each for four circumferential locations considered (top, bottom, left, and right.)

_{2}; hence, electrical heat added minus heat lost to surrounding, as shown in the following equation:

*h*) at a specific axial station (i) is considered common for all quadrants.

_{b,i}*h*is calculated by adding Δ

_{b,i}*h*to the enthalpy calculated at the previous axial station (

_{b,i}*h*

_{b,i −}_{1}), as shown by Eq. (8). Δ

*h*is calculated using Eq. (9)

_{b,i}*T*

_{bulk}

*at a specific axial station (i) is calculated using REFPROP [18] with local enthalpy and pressure as inputs*

_{,i}_{2}bulk flow thermal conductivity,

*k*, is calculated using REFPROP with bulk flow temperature and pressure as inputs. This value of CO

_{b}_{2}thermal conductivity is used in calculation of local Nusselt number as shown in the following equation:

Since mass flowrate for all the cases presented here is constant, Reynolds number equation with mass flowrate is adopted here.

### 2.4 Uncertainty Analysis.

The uncertainty analysis is based on methods described by Kline [22] and the Test Uncertainty Standard PTC 19.1 by the American Society of Mechanical Engineers (ASME) [23]. Uncertainties of different measured parameters are listed in Table 1. Table 2 shows a summary of uncertainty calculated for Nusselt numbers with a maximum uncertainty of 6%.

### 2.5 Air Validation.

Due to the unconventional nature of the experimental setup described here, it is important to validate the correctness of experimental results and the data reduction process obtained from this setup. For this paper, validation experiments with high pressure air were conducted to establish the baseline confidence interval for the sCO_{2} tests and the data reduction procedure.

Figure 6(a) shows comparison of experimentally obtained Nusselt numbers and Gnielinski Nusselt number for a case with the mentioned testing conditions. The match for this case was within *±*3%. Overall, for all air tests, the Nusselt number calculated from experiments fit well with predicted Nusselt numbers from Gnielinski correlations. This is plotted in Fig. 6. The largest error between the predicted Nusselt numbers and calculated values was less than *±*6% for the range of Reynolds number tested. This is within maximum uncertainty observed in the experimentally calculated Nusselt numbers.

## 3 Numerical Methodology

### 3.1 Geometry and Numerical Modeling.

The companion numerical simulations performed in this study aim to model the experimental configuration as closely as possible; this includes geometry and boundary conditions. To allow for direct comparison to experimental results, surface temperature probes and bulk temperatures are monitored at identical streamwise, circumferential locations as those of the physical experiment and the same volumetric heat generation for the heated section (zone 4) of the tube as shown in Fig. 4. Solid domain pipe dimensions are identical to that of the experiment and are given in Sec. 2. Numerical boundary conditions for the companion simulations used a pressure inlet (based on case experimental values of pressure and bulk temperature), mass flow exit, volumetric heat generation for the solid domain, and an external wall temperature-dependent convective boundary condition for the solid domain, as shown in Table 3 This temperature-dependent convective boundary condition uses the experimental heat loss relation from Fig. 5, to model experimental heat loss.

Inlet pressure (MPa) | 8.0 |

Mass flow rate (Kg/s) | 8.33 × 10^{−}^{3} |

Inlet temperature (pseudo-critical) (°C) | 35 |

Inlet temperature (°C) | 55 |

Volumetric heat generation (W/m^{3}) | 1.45 × 10^{7} |

Angles of inclination (°) | 0 (horizontal), 30, 40, 60, 90 (vertical) |

Flow directions | Upward and downward |

Inlet pressure (MPa) | 8.0 |

Mass flow rate (Kg/s) | 8.33 × 10^{−}^{3} |

Inlet temperature (pseudo-critical) (°C) | 35 |

Inlet temperature (°C) | 55 |

Volumetric heat generation (W/m^{3}) | 1.45 × 10^{7} |

Angles of inclination (°) | 0 (horizontal), 30, 40, 60, 90 (vertical) |

Flow directions | Upward and downward |

The simulations for this study were run with the finite volume commercial solver ANSYS Fluent. Steady-state conjugate heat transfer simulations for sCO_{2} were run with double precision and 2nd order upwind discretization schemes. Fluent's pressure-based solver was used with the fully coupled scheme for pressure–velocity coupling, with the PRESTO! scheme used for pressure interpolation. With strong expected buoyancy effects, the PRESTO! scheme was chosen for its ability to model such flows. The current problem is treated as a single-phase flow and thus can be solved with the general fluid and heat transfer equations. The general equations are given as:

For this study, the *k* − *ω* Shear Stress Transport turbulence model [24] was used based on its documented success in challenging, buoyancy-impacted sCO_{2} flows [25–27]. Since this study focuses on inlet bulk temperatures close to the critical point where thermophysical property variations are considerably high, CO_{2} is modeled as a real gas with variable properties. Property lookup tables are generated within Fluent's UDF library using the NIST Database Version 9.1 (REFPROP v9.1). The temperature resolution of the generated lookup table is 0.065 °C. Solid domain material properties for 316 Stainless Steel used a constant density and temperature dependent functions for specific heat and thermal conductivity.

Numerical meshes for the fluid and solid domains were generated within ANSYS workbench and consisted of unstructured tetrahedral and hexahedral cells, respectively. The first cell height in the fluid domain prism mesh was set to maintain wall *y*^{+ }*< *1 throughout the entire domain. Twenty-two prism layers were used in the inflation layer. Due to the presence of buoyancy-driven secondary flows within the heated section of the domain, additional volumetric refinement was performed (zones 2 through 6) to improve resolution in the core flow area. The solid domain mesh maintained similar cell base size to that of the fluid domain for consistency and robustness.

The pipe domain mesh maintained 10 cells in the radial direction, with smaller cell heights biased toward the inner and outer wall boundaries.

Mesh independence was verified by the evaluation of external wall temperature distributions, across four numerical meshes of varying count (7, 9, 13.7, 16, and 20 million). Figure 7(a) shows the minimum, maximum, and average temperature difference of each mesh compared to the finest mesh of 16 million. “Minimum” and “Maximum” refer to the highest and lowest absolute temperature deviations between the respective meshes. Mesh independence is reached with a cell count of *∼* 13.7 million, where the deviation between it and the 20 million mesh is less than 1 °C. Note that all meshes evaluated retained a *y*^{+}*<* 1. Coarsening/refinement was accomplished via base size adjustment and volumetric refinement within the heated portion of the domains. All simulation results presented herein use the 20 million count mesh.

The CFD data are validated using temperature, by comparing the zone 4 temperatures for the 35 °C and 55 °C to the experimental external wall temperatures. In order to obtain the wall temperature, the simulation is ran until the wall temperatures reached steady-state and convergence is achieved. The results of the simulation study are presented in Figs. 8 and 9 along with the experimental data, the CFD effectively predicts the wall temperature trend although for the case of 35 °C in Fig. 8, the trend matches for both CFD and experiment in the axial direction, but there is mismatch in the radial direction, particularly at the upper wall of the pipe. With the 35 °C case being near the pseudo-critical temperature, the heat transfer prediction becomes challenging, as is noted in other numerical works in the literature. The 55 °C inlet temperature case shows excellent match with the experimental results.

## 4 Results and Discussion

This study analyzed the effect of inclination on the heat transfer properties of sCO_{2} flowing in a tube at an operating pressure of 8.0 MPa, for a mass flux of 90 kg/m^{2}s and heat flux of 18 kW/m^{2} and 12 kW/m^{2}, with inlet bulk temperatures of 35 °C and 55 °C, respectively. The effect of parameters, such as varying pressure and mass flux, has been studied by Pidaparti et al., which showed that when the bulk temperature is slightly lower than the pseudo-critical temperature, the heat transfer coefficient is highest for lower operating pressures and when the pseudo-critical temperature is higher than the bulk temperature, the heat transfer coefficient is higher for higher operating pressures [13]. The results will show the effect of inclination on the heat transfer properties of the flow. The plots and images shown below are for parameters in zone 4, from Fig. 4, zone 4 consists of 8 thermocouple stations numbered 5–12. The plots are for stations 6–11 while the contour profiles are for stations 5–11.

### 4.1 Effect of Inlet Bulk and Wall Temperature.

The results of the 35 °C and 55 °C cases for horizontal and 90 deg flow direction show the effect of inlet bulk temperature on the flow heat transfer correlation. At 35 °C, which is approximately the pseudocritical temperature at 8 MPa, for the horizontal case, the wall temperature at the top is significantly higher than the wall temperature at the side and the bottom, with the bottom showing the lowest temperature in the radial direction of the tube. This trend is consistent across all stations as shown in Fig. 10.

This indicates that there is temperature variation of wall temperature in the tube for the horizontal case, which becomes less prevalent as the tube is inclined. This circumferential variation of wall temperature for the horizontal case is due to the density difference in the tube. The density variation in the tube creates a region where the high density fluid settles at the bottom of the fluid domain, as indicated in Fig. 1. The phenomenon causes buoyancy and flow acceleration due to the difference in density of the bulk flow and the near wall flow, hence creating turbulent shear stress at the higher density region (bottom of the tube), which locally enhances the fluid's thermal conductivity and heat transfer; this is verified in Figs. 11 and 12, which shows that the bottom of the tube, especially for the horizontal case, has the highest Nusselt number. The phenomenal effect of buoyancy and gravity defining the thermophysical properties of the flow is further discussed in the next subsection, effect of flow direction and buoyancy.

The inlet bulk temperature affects the heat transfer properties for the 35 °C horizontal case; the experimental Nusselt number increases as the bulk temperature increases, which is opposite of the Gnielinski correlation Nusselt number as shown in Fig. 11 for the horizontal flow case. The Nusselt number trend for inlet bulk temperature 55 °C for the horizontal case showed better matching and prediction for the sidewall experimental Nusselt number and trend in the same direction. This further shows that the Gnielinski correlation can be used in cautiously predicting the Nusselt number for inlet temperatures which are significantly higher than the pseudo-critical temperatures. Figure 10 shows the difference between the top and bottom wall temperatures and can be seen to decrease as the inlet bulk temperature increases. Due to the difference in the bulk density and the near wall density, decreasing buoyancy effects are further observed compared to the 35 °C inlet bulk temperature case.

### 4.2 Effect of Flow Direction and Buoyancy.

Figure 10 shows the temperature profile for the horizontal cases of 35 °C and 55 °C, there is notable difference between the wall and bulk temperatures, which can be explained as the effect of radial convection as heat is transferred from the tube to the fluid, as the inlet temperature is increased, the effect of buoyancy is diminished. The plots show the external wall and bulk temperatures at locations inside the heated section. There is a temperature gradient at each *z*/*D*_{int} location, which is due to radial density variation. There is also an increase in the bulk and wall temperatures with the top wall temperature being the highest. As the inlet bulk temperature is increased from 35 °C to 55 °C, there is significant increase in the wall and bulk temperatures in the axial direction, although the temperature gradient at each *z*/*D*_{int} location (radial direction) does not reflect the increase; from Fig. 10, the temperatures at the wall and bulk maintain about the same trendline in the axial direction; hence, the difference between the wall temperatures and the difference in any wall temperature and at each *z*/*D*_{int} location is approximately the same. The temperature difference between the top and bottom wall is decreasing as from Δ*T* = 22 °C for the 35 °C horizontal case, to Δ*T* = 16 °C for the 55 °C horizontal case; this shows that with an increase the inlet bulk temperature away from the pseudo-critical temperature, the radial temperature gradient of the wall decreases while the temperature difference between the sidewall and the bulk temperature at the *z*/*D*_{int} location increases.

The Nusselt number plots in Figs. 11 and 12 show the trend of Nusselt number calculated from the internal wall temperature using heat transfer coefficient and Gnielinski correlation, the Gnielinski Nusselt number follows a downward trend for both cases, which indicates that as the bulk temperature is increasing in the axial direction, the Nusselt number decreases, which is relatively similar to the experimental Nusselt number trend for the 55 °C case. The Nusselt number plot in Fig. 11 for the 35 °C case shows the experimental Nusselt number increasing in the axial direction while the Gnielinski Nusselt number decreases, however.

This singularity of trend difference in the plot is as a result of the different inlet bulk temperature and its thermophysical properties. In Fig. 10, it is shown that as inlet bulk temperature increases to be sufficiently higher than the pseudo-critical temperature, there is an increase in ΔT between the wall and bulk temperatures. Analyzing Fig. 13, we see the effect of inlet bulk temperature on thermal conductivity, which is inversely proportional to the experimental Nusselt number. The rate of decrease of the thermal conductivity for inlet bulk temperature of 35 °C closer to the pseudo-critical temperature is very rapid when compared to that of the 55 °C case; it is interesting to note that this, coupled with the high thermal capacitance of the fluid at these near pseudo-critical bulk temperatures (less than 40 °C), greatly affects the upward and downward trend of the experimental Nusselt number in Figs. 11 and 12.

The effect of buoyancy, which is largely due to density effects, can be reduced by inclination, which affects the turbulent shear stress at the wall of the tube. The effect of buoyancy can be characterized by a nondimensional parameter Grashof and/or Richardson number, for flow in a pipe or tube. In order to investigate the effect of buoyancy on the flow, we analyzed the Nusselt number plot and CFD contours for horizontal, 45 deg and 90 deg case for inlet bulk temperatures of 35 °C and 55 °C. Buoyancy forces act upwards and opposite to gravitational forces. Figure 14 shows the Richardson number obtained from experiment, plotted against the axial distance. For the case of 45 deg downward flow, the effect of buoyancy, which is characterized by Richardson number, is decreasing rapidly in the axial direction for the 35 deg case. The 55 deg case shows a decrease in the Richardson number in the axial direction, but it is evident that for inlet bulk temperatures closer to the pseudo-critical temperature, radial density variation at each *z*/*D*_{int} is predominant in the flow.

In order to visualize the effect of buoyancy in the flow, CFD was performed. It is observed in Figs. 15 and 16 that there is density variation at each *z*/*D*_{int} location with the heavier fluid at the bottom of the tube. The temperature at the tube inlet is uniform, but as the flow enters the heated section and heat is added, the fluid develops a radial temperature gradient with peak temperature at the top wall location. This creates the radial density variation shown in the contour plot, and hence the buoyancy forces deteriorate as the fluid local bulk temperature is increased in the axial direction and as the flow direction changes. Buoyant forces act on the fluid creating an “m”-shaped density field with the secondary flow fluid at the top of the tube, and the heavier fluid at the bottom. Figures 11 and 12 also validate the density contour plots; the effect of buoyancy can be seen on the separation between the Nusselt number at the top and bottom wall. As the flow direction changes, the separation between the top and bottom reduces until the flow experiences negligible buoyant forces.

In Fig. 16(d), contour fields of in-plane velocity magnitude show slightly increased boundary layer thickness when compared to velocity contours of horizontal flow in Fig. 15(e). One can also note lower near-wall velocity magnitudes when compared to horizontal, or 45 deg upward flow. As the angle of inclination is increased for the case of downward flow, the turbulent shear stress is enhanced by the buoyant forces acting on the fluid which then creates localized drops in the wall temperature, thereby increasing the heat transfer and the Nusselt number. In the case of upward flow, as the angle of inclination is increased, buoyant forces reduce to become relatively negligible at the limiting case of 90 deg. Both upward and downward 90 deg cases are highlighted by the lack of large scale secondary buoyancy driven structures. Velocity contours are not shown for these cases, as the in-plane velocity magnitudes are an order of magnitude smaller, than that of the horizontal and 45 deg inclined cases. The density fields seen for Figs. 15(a) and 16(a) are axisymmetric, which support the observed experimental station wall temperature uniformities.

## 5 Conclusions

This work was performed to understand the fluid dynamic and heat transfer properties associated with high pressure and varying inlet temperature flow of sCO_{2} in a tube. Experimental and CFD studies were performed at a pressure of 80 MPa, mass flux of 90 Kg/m^{2}s with inlet bulk temperatures of 35 °C and 55 °C. The study showed that the flow direction and inlet temperature are major parameters affecting the heat transfer properties of sCO_{2}. Nondimensional parameters such as Nusselt and Richardson number are plotted from experimental data to show the effect of the varying parameters on heat transfer and fluid dynamics properties of the flow. CFD simulation is further utilized to visualize the flow regime in the radial direction.

The study showed that the Gnielinski and experimental Nusselt number trend in the same axial direction for inlet bulk temperatures further from the pseudo-critical temperature, however, cannot be used to effectively predict the experimental Nusselt number close to the pseudo-critical temperature. As the tube is inclined, the buoyant forces become negligible and there is no radial variation in temperature as the inclination angle approaches 90 deg. For downward flow, buoyant forces enhance the turbulent shear stress and heat transfer coefficient increases. It can be concluded that for inlet bulk temperatures near the pseudo-critical temperature, buoyant forces are stronger, but reduce as the inlet bulk temperature is increased. Maximum buoyant forces and heat transfer occur in the horizontal flow, and for inlet bulk temperatures near the pseudo-critical temperature. The results of this study provide insight that can be applied to design and development of pressure vessels, turbomachinery, and heat exchanger components in power plants or cycles that may use sCO_{2} as its working fluid.

## Data Availability Statement

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

## Nomenclature

*A*_{cs}=cross-sectional area of tube

*A*_{local}=local convection heat transfer area

*D*_{ext}=external diameter of tube = 12.57 mm

*D*_{int}=internal diameter of tube = 10.92 mm

*f*=Darcy friction factor

*g*=acceleration due to gravity

- Gr =
Grashof number

*h*=_{b,i}bulk enthalpy at station

*i**h*=_{b}bulk enthalpy

- HTC
_{local}= local heat transfer coefficient

- I =
current flowing through test section

*k*=thermal conductivity

*k*=_{b}thermal conductivity calculated using bulk CO

_{2}properties*k*_{SS}=thermal conductivity of Stainless steel

*L*_{local}=local length of heat transfer calculation domain

*L*_{total}=total heated length of tube

*m*˙ =mass flow rate

- Nu
=_{B} local experimental Nusselt number at bottom surface

- Nu
_{Gn}= Gnielinski Nusselt number

- Nu
_{local}= local experimental Nusselt number

- Nu
=_{T} local experimental Nusselt number at top surface

- Pr
=_{b} Prandtl number calculated using bulk CO

_{2}properties*q*_{vl}=volumetric heat generation in tube

- $q\u02d9\u2033$ =
heat flux

*Q*_{gen,electric}=electric power generated in test section

*Q*_{loss,radial}=heat loss to surrounding in radial direction

- Re
=_{b} Reynolds number calculated using bulk CO

_{2}properties*Ri*=Richardson number

- sCO
_{2}= supercritical Carbon Dioxide

*T*_{amb}=ambient temperature

*T*=_{b}bulk temperature

*T*=_{b,i}bulk temperature at station i

*T*_{ext}=temperature at external wall surface

*T*_{int}=temperature at internal wall surface

*T*=_{w}wall temperature

*u*=_{i}component of velocity vector in

*i*direction*U*_{loss}=coefficient of radial heat loss

*V*=voltage provided to test section

*V*_{xy}=velocity magnitude in xy plane

*x*=_{i}cartesian coordinate in

*i*direction- Δ
*h*=_{b,i} bulk enthalpy change at station

*i**ΔQ*_{CO2,local}=heat transferred to CO

_{2}in heat transfer calculation domain*ΔT*_{w −}_{amb}=temperature difference between tube external wall and ambient

*δ*=_{i j}Kronecker delta function

*μ*=dynamic viscosity

*μ*=_{t}turbulent viscosity

*ρ*=_{b}bulk flow CO

_{2}density*ρ*_{wall}=CO

_{2}density at wall*ρ*=density

## References

_{2}Cycles for Power Generation in CSP Solar Tower Plants

_{2}Brayton Cycle for Simple Cycle Gas Turbines Exhaust Heat Recovery Using Genetic Algorithm

_{2}Waste Heat Recovery Power Systems

_{2}Symposium

_{2}Power System

_{2}in Circular Tubes at Various Inclination Angles

_{2}at Supercritical Pressures in a Horizontal Tube

_{2}at a Supercritical Pressure Flowing Vertically Upward in Tubes and an Annular Channel

_{2}Flow in a Horizontal Tube

_{2}Flowing in a Mini Tube Under Heating Conditions

*Turbulent Forced Convection in Channels and Bundles: Theory and Applications to Heat Exchangers and Nuclear Reactors*,

_{2}Heat Transfer in Circular Pipes

*Fire Design of SteelStructures: Eurocode 1: Actions on Structures; Part 1-2: General Actions–Actions on Structures Exposed to Fire; Eurocode 3: Design of Steel Structures; Part 1-2: General Rules–Structural Fire Design*,

*Introduction to Heat Transfer*,