International Journal of Heat and Fluid Flow 106 (2024) 109288

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow

journal homepage: www.elsevier.com/locate/ijhff

Revisiting the surface-mounted cube: An updated perspective of the near

wake and near-wall flow field
Barbara L. da Silva ∗, David Sumner, Donald J. Bergstrom
Department of Mechanical Engineering, University of Saskatchewan, 57 Campus Drive, Saskatoon, Saskatchewan, S7N 5A9, Canada

ARTICLE INFO ABSTRACT

Keywords: The flow around a surface-mounted cube has been investigated for decades, yet the fundamentals of the mean
Wake flow have not been updated to reflect the latest advances regarding the flow around surface-mounted finite-
Surface-mounted cube height square prisms in general. One of the main gaps is the flow field very near the cube walls, especially the
Near-wall flow
sides, and its relationship to the near wake. To investigate these features, large-eddy simulations of the flow
Flow structures
around a surface-mounted cube were carried out at a Reynolds number Re = 1 × 104 . Two boundary layers
Bluff body
Large-eddy simulation
were considered: a thin and laminar boundary layer and a thick and turbulent one, to provide an overview of
the flow around the cube for contrasting boundary layers. The major flow structures in the mean wake were
the horseshoe vortex, the arch vortex and the dipole structures, with other regions of vorticity also present.
The time-averaged flow fields presented similar flow features for both boundary layers, but the thicker and
turbulent one caused the horseshoe vortex to be located closer to the cube, the wake to become narrower and
shorter, the coherent structures and downwash to weaken, the pressure to decrease around the sides and top of
the cube, the drag force coefficient to decrease, the normal force coefficient to increase, and the trailing edge
vortices to weaken. Intermittent flow reattachment on the top and side faces of the cube, as well as corner
vortices, were found for both boundary layers, while the side and free end vortices were found exclusively with
the thicker boundary layer, yielding a headband vortex. The three-dimensional flow structures and features
were related to the near-wall flow field on the cube and ground plane surfaces, giving a complete and updated
description of the mean flow characteristics.

1. Introduction of the ground plane boundary layer 𝛿∕𝐷 (or 𝛿∕𝐻) and the Reynolds
number Re = 𝑈∞ 𝐷∕𝜈, where 𝑈∞ is the freestream velocity and 𝜈 is the
The flow around surface-mounted cubes is found in various en- kinematic viscosity of the fluid. A literature review of these major flow
gineering and environmental applications. Air flow around buildings, structures, shown in Fig. 1, and their relevance for the surface-mounted
with applications in fire control and pollutant dispersion (Saeedi and cube are presented as follows.
Wang, 2015), and the thermal management of electronics (Khan and A common feature to all surface-mounted finite-height square
Saha, 2022), have been of particular interest. Complex phenomena prisms is the horseshoe or necklace vortex formed upstream of the
related to flow separation, vortex shedding and its modulation near junction of the prism with the ground plane as shown in Fig. 1a and b.
the free end or junction with the ground plane have also made the The horseshoe vortex may be of laminar or turbulent nature, depending
surface-mounted cube a classic benchmark case for both experimental
on the Reynolds number Re𝛿 ∗ = 𝑈∞ 𝛿 ∗ ∕𝜈, where 𝛿 ∗ is the displacement
and numerical studies of three-dimensional wake flows (Iousef et al.,
thickness of the boundary layer. The laminar horseshoe vortex system
2017; Schröder et al., 2020; da Silva et al., 2022b).
may present multiple steady or unstable vortices (Baker, 1979; Diaz-
Most of these studies focus on the time-averaged flow field around
Daniel et al., 2017; Hwang and Yang, 2004), where the distance of the
the surface-mounted cube with one of its faces normal to the incoming
primary vortex’s center from the prism’s center correlates positively
flow (angle of incidence 𝛼 = 0◦ ). Some main components of the mean
wake of a surface-mounted cube are similar to those of surface-mounted with the Reynolds number and boundary layer thickness (Baker, 1979).
finite-height square prisms of larger aspect ratio AR = 𝐻∕𝐷, where 𝐻 Transition from a laminar to a turbulent horseshoe vortex system is
is the prism height and 𝐷 is its width. However, differences in these expected to occur in the range 400 < Re𝛿 ∗ < 1000 (Ballio et al., 1998).
structures may be observed depending mainly on AR, the thickness The mean position of the turbulent horseshoe vortex varies significantly

∗ Corresponding author.
E-mail address: [email protected] (B.L. da Silva).

https://doi.org/10.1016/j.ijheatfluidflow.2024.109288
Received 26 July 2023; Received in revised form 27 November 2023; Accepted 5 January 2024
Available online 14 January 2024
0142-727X/© 2024 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 1. Schematics of the typical time-averaged flow structures around a surface-mounted finite-height square prism and of the types of reattachment on the prism’s free end.
(a) Flow structures for a dipole-type wake, (b) flow structures for a quadrupole-type wake, (c) reattachment from the flow separated from the leading edge of the prism, based
on Nakamura et al. (2001), and (d) reattachment of the backflow, based on Cao et al. (2022) and da Silva et al. (2022a).

only with the prism’s AR, in the range of AR < 1 (Ballio et al., 1998; ‘‘base-like’’ vortices have been observed near the cube, within the
da Silva et al., 2022a). It is typically located much closer to the prism, recirculation region (Hajimirzaie, 2023), and very different structures
compared to the laminar horseshoe vortex, which induces a stronger have been found for Re ≤ 500 (Liakos and Malamataris, 2014; Goswami
flow in the junction region (Mohammadi et al., 2022). The turbulent and Hemmati, 2022). On the other hand, for all AR and Re in the order
horseshoe vortex’s fluctuations are typically aperiodic, but they display of 104 , the presence of an additional pair of vorticity regions of the
a bimodal behavior associated with a zero-flow mode and a backflow same sign as the dipole structures and opposite sign to the base vortices
mode (Devenport and Simpson, 1990; Escauriaza and Sotiropoulos, has been detected near the ground plane (Unnikrishnan et al., 2017;
2011a,b; El Hassan et al., 2015). The zero-flow mode is triggered by da Silva et al., 2022b). The origin of the ground plane vorticity has
the formation of hairpins around the horseshoe vortex’s core, causing been attributed in the past to the horseshoe vortex legs, however, the
it to break down and disrupt the backward jet-like flow beneath the horseshoe vortex legs do not hold their coherence much beyond the
vortex (Paik et al., 2007). near wake (Wang, 2019; Khan and Saha, 2022) and this proposition
The mean flow field behind a surface-mounted cube is characterized remains to be verified.
mainly by the arch or near-wake vortex (Fig. 1a and b) (Martinuzzi Minor flow structures in the mean wake of surface-mounted cubes,
and Tropea, 1993; Shah and Ferziger, 1997; da Silva et al., 2022b). Its which have not been as extensively studied, include the corner vortex,
presence is identified based on the imprints of its legs on the ground the side vortex and the free end vortex system. The conical, junction
plane and the large vortex visible in the symmetry plane, also called or corner vortex is found near both lateral faces of surface-mounted
vortex Bt (Krajnović, 2011; Sumner et al., 2017). Beyond the near- finite-height square prisms (Fig. 1a and b). It has been measured
wake region where recirculating flow takes place, dipole structures and visualized in both experiments and simulations (e.g., Okuda and
Taniike, 1993; Martinuzzi and Tropea, 1993; Sousa, 2002; Cao et al.,
(Fig. 1a), typically visualized in the wake of taller prisms, have also
2019), but since its identification in the study by Okuda and Taniike
been found for surface-mounted cubes (da Silva et al., 2022a). These
(1993) it has not been given much attention in the literature. Detailed
structures have been traditionally referred to as tip vortices, although
descriptions of the corner vortex have only recently been presented
this nomenclature is ambiguous since it has been used to refer to
in da Silva et al. (2020) and Cao et al. (2022) for a prism with AR = 3,
other structures as well (Goswami and Hemmati, 2022). The dipole
where the vortex was shown to start near the upstream corners of
structures extend into the far wake and are characterized by a high-
the prism’s junction with the ground plane. In Okuda and Taniike
magnitude streamwise vorticity, positive (counter-clockwise) on the +𝑦
(1993) and da Silva et al. (2020), which have 𝛿∕𝐻 < 1, the corner
side of the wake and negative (clockwise) on the −𝑦 side (Unnikrishnan
vortices extended into the wake, as depicted in Fig. 1a and b. However,
et al., 2017; Cao et al., 2022). From a time-average point of view, both
in Cao et al. (2022) and in other studies including surface-mounted
the streamwise dipole structures and the arch vortex originate from
cubes (Meinders et al., 1999; Krajnović and Davidson, 2001; Yakhot
the separated shear flow as it is bent inward by the downward flow
et al., 2006), the corner vortices extend vertically along the prism’s
from the free end of the prism, called the downwash (da Silva et al., height, becoming a side or lateral vortex, and often connect over the
2020; Cao et al., 2022). From a dynamic point of view, the dipole free end. Cao et al. (2022) had 𝛿∕𝐻 = 6.7 and flow attachment on
structures are derived from the averaging of loop-shaped structures that the sides of the prism, which suggests that the development of the
are shed alternately from square prisms, including surface-mounted corner vortex into a side vortex may depend on the boundary layer
cubes (Bourgeois et al., 2011; Hosseini et al., 2013; da Silva et al., thickness and the flow pattern on the side faces of the prism. However,
2022b). as mentioned previously, complete descriptions of the near-wall flow
Other structures or vorticity concentrations have been reported in and the corner vortex are not very common, and this behavior has yet
the wake of surface-mounted finite-height square prisms in general, to be systematically verified for both the surface mounted cube and
depending on AR, 𝛿∕𝐷 and Re. For AR > 1 and large 𝛿∕𝐷 and Re, a taller prisms.
pair of base vortices may appear due to an increased upwash (upward Several structures have been found over the free end of surface-
flow) in the lower part of the wake (Hosseini et al., 2013; Yauwenas mounted finite-height square prisms. Sumner et al. (2017) reported two
et al., 2019; Cao et al., 2022). This type of wake, illustrated in Fig. 1b, pairs of streamwise vorticity regions: edge vortices just over the lateral
is then classified as a quadrupole-type. For AR < 1, instead of the edges of the prism’s free end, formed due to the upward flow from the
dipole structure, a pair of vorticity regions of opposite sign, termed side faces, and another pair above them, formed due to the outward-
‘‘inner vorticity’’, was observed in the upper part of the wake (Sumner deflected flow of the shear layers. Both structures have been shown
et al., 2016; da Silva et al., 2022a,b). Note that neither the base also by da Silva et al. (2020) and Goswami and Hemmati (2022), but
vortices nor the inner vorticity regions have been reported outside these vorticity regions have not been extensively described for surface-
the recirculation region for surface-mounted cubes (AR = 1), although mounted cubes. Instead, studies have mostly observed the presence

2
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

of a recirculation bubble, with the possibility of flow attachment and 𝛿∕𝐻 > 0.3. In the present study, to generate two opposing cases that
separation similar to the side faces (Meinders et al., 1999; Krajnović represent extreme variations in the flow, a laminar and thin boundary
and Davidson, 2001; Nakamura et al., 2001; Depardon et al., 2005; layer with 𝛿∕𝐷 = 0.2 and a turbulent and thick boundary layer with
Yakhot et al., 2006; Kozmar, 2021). Flow attachment on the free end 𝛿∕𝐷 = 0.8 (for which experimental data are available for validation)
of a cube was found to come from two different sources: the flow are considered. As such, the present study provides a comprehensive
separated from the leading edge of the cube (Fig. 1c), or the backward and updated description of the mean flow field around a cube based on
flow that separated from the rear edge (Fig. 1d) (da Silva et al., 2022a). LES results. The flow dynamics will be investigated in future studies.
The reattachment of the leading-edge separated flow is the main type
observed in most studies. It becomes more likely to happen for larger 2. Computational methods
𝛿∕𝐷 and turbulent intensity (Castro and Robins, 1977; Hearst et al.,
2016; Kozmar, 2021; Chen et al., 2023), and it is expected to cause the The flow around a surface-mounted cube was evaluated through
appearance of a bound free end vortex (Meinders et al., 1999; Krajnović large-eddy simulation (LES). In this approach, the large scale incom-
and Davidson, 2001; Ito et al., 2006). Backflow attachment can take pressible flow of a Newtonian fluid is described by the filtered Navier–
place as well. This type of attachment is found also for surface-mounted Stokes equations
prisms of AR as large as 9 (Sumner et al., 2017), and it has a higher 𝜕𝑢𝑖
probability of occurring for smaller 𝛿∕𝐷 (da Silva et al., 2022a). For = 0, (1)
𝜕𝑥𝑖
either type of flow attachment, variations of the pressure coefficient on sgs
𝜕𝑢𝑖 𝜕 𝜕 2 𝑢𝑖 𝜕𝑝 𝜕𝜏𝑖𝑗
the free end of the surface-mounted cube are not striking (Heng and + (𝑢𝑖 𝑢𝑗 ) = 𝜈 − − , (2)
Sumner, 2022), which makes identification of reattachment difficult 𝜕𝑡 𝜕𝑥𝑗 𝜕𝑥𝑗 𝜕𝑥𝑗 𝜕𝑥𝑖 𝜕𝑥𝑗
through pressure measurements alone. where 𝑢𝑖 or (𝑢, 𝑣, 𝑤) are the filtered velocity components along the
Recent studies that address the flow around surface-mounted cubes Cartesian coordinates 𝑥𝑖 or (𝑥, 𝑦, 𝑧), 𝑝 is the filtered kinematic pressure
have examined the effects of turbulence intensity on pressure profiles (i.e., 𝑝 = 𝑃 ∕𝜌, where 𝑃 is the static pressure and 𝜌 is the density
on the cube faces (Kozmar, 2021), effects of the depth ratio on the sgs
of the fluid) and 𝜏𝑖𝑗 is the subgrid-scale stress tensor. The tensor
three-dimensional flow field for Re ≤ 500 (Goswami and Hemmati, accounts for the small scales of the flow, which have been modeled
2022), free end pressure distributions as functions of AR and 𝛿∕𝐷 (Heng using the dynamic Lagrangian subgrid-scale model of Meneveau et al.
and Sumner, 2022), wind loads as function of elevation (Melaku et al., (1996). This model has shown good performance in LES of bluff body
2022), streamwise vorticity distributions downstream of the prism (da wakes (Cianferra et al., 2018; Gupta and Wan, 2019; Puthan et al.,
Silva et al., 2022a; Hajimirzaie, 2023), and flow control strategies 2020), as long as the scale-invariance hypothesis is valid (Bou-Zeid
involving holes and jets (Khan and Saha, 2022; Li et al., 2023). As et al., 2005). Its performance was also verified in preliminary tests
shown in the present literature review, although similar, not all ele- regarding the flow around surface-mounted finite-height square prisms,
ments in the mean wake of surface-mounted finite-height square prisms not shown for brevity. Note that, although only time-averaged fea-
are the same as the ones in the wake of a surface-mounted cube. Most tures are the focus of the present study, the LES (in comparison with
studies involving AR = 1 are focused on specific features, such as wind approaches based on solving the unsteady Reynolds-averaged Navier–
loads and pressure distributions, or have been carried out for limited Stokes (URANS) equations) provides accurate results for future analyses
flow conditions, such as for a single boundary layer. Therefore, recent of the flow dynamics around a surface-mounted cube.
advances about the three-dimensional flow around surface-mounted A schematic of the computational domain is shown in Fig. 2a, based
finite-height square prisms at high Reynolds numbers remain to be on the width of the cube, 𝐷 = 0.06 m. The computational domain
verified for cubes. extends 25𝐷 in the streamwise (𝑥) direction, 15𝐷 in the transverse (𝑦)
In addition, little research has been done regarding the flow struc- direction and 9𝐷 in the vertical (𝑧) direction. The inlet was located
tures in close vicinity of the surface-mounted cube, such as the corner 7.5𝐷 upstream of the cube center, and the origin of the domain is
vortex and side vortex. A complete description of these features is made located at the center of the junction of the cube with the ground plane,
difficult by two factors: first, their apparent dependence on the ground which consists of the bottom boundary in Fig. 2a. The extent of the
plane boundary layer, which is typically held constant in the majority computational domain follows recommendations of other studies to
of studies that evaluate flow structures; and second, the requirement avoid blockage effects (Saha, 2013; Cao et al., 2022). In addition, a
to consider the three-dimensional flow field simultaneously with the sensitivity analysis (not included for brevity) was carried out to verify
near-wall flow field. Considering this gap in the literature and the effects of the domain size in the 𝑦 and 𝑧 directions, and variations in
overdue need to update the characterization of the three-dimensional the mean velocity field were found to be minimal.
flow around a surface-mounted cube, the present study aims to more A grid composed of 6 731 160 hexahedral cells was used to discretize
thoroughly describe the mean flow around a surface-mounted cube the problem. A 𝑥-𝑧 slice of the grid in the vicinity of the cube is shown
(AR = 1) and its relationship to the near-wall flow field. Large-eddy in Fig. 2b, highlighting the increased resolution near the cube and the
simulations have been conducted for Re = 1 × 104 , which is at the be- ground plane surfaces. The cube’s edges contain 60 grid cells along the
ginning of the range where Reynolds number effects can be considered 𝑥 and 𝑦 directions and 69 cells in the 𝑧 direction, concentrated near
negligible for the mean flow field (Martinuzzi and Havel, 2000; Lim the corners and the ground plane. The wall-normal distance 𝑦𝑤 of the
et al., 2007), to allow for comparisons with literature results. closest cell to the cube and ground plane surfaces gave maximum values
As for the boundary layer thickness 𝛿∕𝐷, the present literature of 𝑦+ +
𝑤 = 𝑦𝑤 𝑢𝜏 ∕𝜈 = 0.7 for the ground plane and 𝑦𝑤 = 1.1 for the cube
review has revealed that both the laminar or turbulent nature of the surfaces, where 𝑢𝜏 is the friction velocity.
boundary layer and its thickness may have significant effects on the A grid sensitivity analysis was carried out using additional coarser
flow. A variety of conditions have been adopted in the literature, with and finer grids, with 3 260 024 cells and 9 805 752 cells, respectively. The
laminar boundary layers ranging from 𝛿∕𝐷 = 0.025 to 3 (e.g., Castro 𝑦𝑤 value of the grid cells closest to the cube and ground plane was kept
and Robins, 1977; Krajnović and Davidson, 2001; Hwang and Yang, the same between different grids, and all other sizes were scaled by a
2004; Diaz-Daniel et al., 2017; Goswami and Hemmati, 2022; Khan and factor of 1.25 (for the coarse grid) or 0.8 (for the fine grid). For refer-
Saha, 2022; Hajimirzaie, 2023; Li et al., 2023) and turbulent boundary ence, the maximum grid size 2𝐷 away from the cube’s faces was 0.08𝐷
layers as thin as 𝛿∕𝐷 = 0.2 (Depardon et al., 2005) up to atmospheric for the intermediate (present) grid, which then increased at a ratio of
boundary layers, many times thicker than the cube’s width (e.g., Lim 1.01 further upstream/downstream and moving laterally away from the
et al., 2007; Wang and Lam, 2019). Sakamoto and Arie (1983) have, cube. The numerical uncertainty of the present grid with 6 731 160 cells
however, reported little influence from turbulent boundary layers for was estimated with the grid convergence index (GCI) method (Roache,

3
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 2. (a) Schematic of the surface-mounted cube of width 𝐷 and height 𝐻 = 𝐷, and computational domain dimensions and boundary conditions. At the inlet, either a uniform
velocity profile or turbulent boundary layer data were adopted to simulate a laminar and a turbulent boundary layer, respectively. (b) Slice of the grid in the 𝑥-𝑧 plane, highlighting
the higher resolution by the corners of the cube.

1994; Celik et al., 2008), along lines at (𝑥∕𝐷, 𝑧∕𝐷) locations of (2, 0.5),
(2, 1), (4, 0.5) and (4, 1), and (𝑥∕𝐷, 𝑦∕𝐷) locations of (0, 0), (1, 0),
(2, 0) and (3.5, 0), i.e., in both the near and intermediate wake. The
three grids had similar performance, with maximum GCI (equivalent
to uncertainty) of ±0.13 for the mean streamwise component 𝑢∕𝑈∞ ,
±0.08 for the mean transverse component 𝑣∕𝑈∞ and ±0.09 for the mean
vertical component 𝑤∕𝑈∞ , and an average GCI of less than ±0.01 for
all components. Note that oscillatory convergence (Celik et al., 2008)
ranged from 10% to 76% of the points, with larger ratios for cases
where the profiles mostly overlap.
The quality of the LES was further examined by comparing the
local grid size 𝛥 to the local integral length scale of the flow 𝐿 and to
the Kolmogorov length scale 𝜂 = (𝜈 3 ∕𝜀)1∕4 , where 𝜀 is the total time-
averaged dissipation rate of the turbulent kinetic energy, including
both the resolved and subgrid-scale dissipation rates. Values of 𝐿 were
checked near main flow features like the horseshoe vortex legs and the
dipole structure, and the local grid size 𝛥 was found to be at least one Fig. 3. Mean velocity and Reynolds stress profiles of the precursor channel flow
order of magnitude smaller than 𝐿. As for the 𝛥∕𝜂 ratio, its maximum simulation of a turbulent boundary layer, in inner boundary layer coordinates (𝑢+ =
+
𝑢∕𝑢𝜏 , 𝑢𝑖 𝑢𝑗 = 𝑢𝑖 𝑢𝑗 ∕𝑢2𝜏 and 𝑧+ = 𝑧𝑢𝜏 ∕𝜈) for Re𝜏 = 335, in comparison with Re𝜏 = 392
value was 30 for the laminar boundary layer simulation and 26 for the
in Moser et al. (1999). The profiles obtained in the main cube simulation at 𝑥∕𝐷 = 0
turbulent boundary layer simulation, with average ratios of 1.7 and 1.9, (without the cube) are included.
respectively.
Boundary conditions used for the velocity field comprise a no-slip
condition (𝑢𝑖 = 0) on the ground plane and cube surfaces, a free slip
𝛥𝑦+ = 14. Periodic boundary conditions were used on the boundaries
condition (𝑢normal = 0, 𝜕𝑢tangential ∕𝜕𝑥normal = 0) on the lateral and top
normal to the 𝑥 and 𝑦 directions, and the no-slip condition was used
boundaries, and a convective outflow condition (𝜕𝑢𝑖 ∕𝜕𝑡 + 𝑈𝑐 𝜕𝑢𝑖 ∕𝜕𝑥 = 0,
normal to the 𝑧 direction. An exception was during the start-up of
where 𝑈𝑐 is the convective velocity) at the outlet. For the kinematic
the channel simulation, during which a 1/7th power law profile with
pressure and the integral functions of the dynamic Lagrangian model,
random fluctuations of 20% of the velocity magnitude was prescribed
zero-gradient conditions were specified almost everywhere. Exceptions
at the inlet, and a zero-gradient condition at the outlet. This approach
are the kinematic pressure, which had 𝑝 = 0 set at the outlet, and the
was necessary to trigger turbulence in the LES, after which the inlet and
𝐿𝑀 integral function, which was set to zero on the ground plane and
cube surfaces (Meneveau et al., 1996). As for the velocity field at the outlet boundaries were switched to periodic conditions, and a pressure
inlet, different approaches were taken depending on the boundary layer gradient source term was added to the 𝑥-momentum equation.
at the location of the cube. For the thin and laminar boundary layer, The resulting Reynolds number was Re𝜏 = 𝑢𝜏 ℎ∕𝜈 = 335. Fig. 3
a uniform velocity with 𝑈∞ = 2.5 m∕s was prescribed normal to the presents the mean velocity and Reynolds stress profiles of the turbulent
inlet. This condition led to 𝛿∕𝐷 = 0.2 at 𝑥∕𝐷 = 0, and a Reynolds boundary layer of the channel flow, which follow closely the direct
number Re = 1×104 for the present value of 𝜈 = 1.5×10−5 m2 /s. For the numerical simulation (DNS) results of Moser et al. (1999) for Re𝜏 =
turbulent boundary layer, a thickness of 𝛿∕𝐷 = 0.8 was desired at the 392. The channel spanned one third of the main simulation domain’s
location of the cube, to allow for comparisons with the experimental width, so the channel flow fields were mapped to three sections on
velocity data of da Silva et al. (2022a). This condition was achieved the inlet of the main simulation at each time step. Accounting for
by simultaneously running a precursor channel simulation with the the coarser grid resolution of the main simulation, which caused the
velocity at the center of the channel equal to 𝑈∞ , to generate turbulent inflow turbulence to lose some of its quality (dashed line in Fig. 3),
boundary layer velocity fields that were mapped onto the lower part of the turbulent boundary layer at 𝑥∕𝐷 = 0 still captures all the essential
the inlet boundary at each time step. This method led to 𝛿∕𝐷 = 0.8 at characteristics of a turbulent boundary layer. In addition, it shows
the location of the cube. good agreement with the experimentally observed boundary layer in da
The precursor simulation consisted of a plane channel of half-height Silva et al. (2022a). This agreement is illustrated by the parameters in
ℎ = 42 mm, with a domain size of 6ℎ (length) × 7.1ℎ = 5𝐷 (width) Table 1, which include the boundary layer thickness 𝛿∕𝐷, momentum
× 2ℎ (height). The channel grid had 84 × 167 × 96 elements, with thickness 𝜃∕𝐷, displacement thickness 𝛿 ∗ ∕𝐷, Reynolds number based
𝑧+ = 0.14 for the closest elements to the channel wall, 𝛥𝑥+ = 24 and on the displacement thickness Re𝛿∗ and shape factor, for both boundary

4
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Table 1 2 are presented along vertical lines in the symmetry plane

𝑢′ 𝑤′ ∕𝑈∞
Flat-plate boundary layer parameters at 𝑥∕𝐷 = 0. (Fig. 4a–d), and the three velocity components are presented in one
Boundary layer 𝛿 [mm] 𝛿∕𝐷 𝜃∕𝐷 𝛿 ∗ ∕𝐷 Re𝛿∗ Shape factor horizontal line (Fig. 4e). The subgrid-scale stress 𝜏𝑢𝑤SGS ∕𝑈 2 is plotted
∞
Laminar 10 0.2 0.01 0.04 389 2.7 ′ ′ 2
together with the resolved stress 𝑢 𝑤 ∕𝑈∞ , showing that its contribution
Turbulent 45 0.8 0.07 0.10 1.0 × 103 1.4
is in general negligible. As for the resolved flow field, excellent agree-
da Silva et al. (2022a) 49 0.8 0.07 0.09 6.4 × 103 1.2
ment is observed at most locations. The most notable disagreement is
for the normal and shear Reynolds stresses at 𝑥∕𝐷 = 0 and 𝑦∕𝐷 =
0 (Fig. 4a). The stresses show the same trends as the experimental
layers considered in the present study and the turbulent boundary layer data in the profiles, but the peak where the separated shear layer
in da Silva et al. (2022a). The Reynolds number Re𝛿 ∗ for the laminar is found has been underpredicted by the LES, even accounting for
boundary layer is in the upper range of the laminar horseshoe vortex the subgrid-scale contribution. In addition, differences are observed
regime of Ballio et al. (1998), but, as will be mentioned later, the between the LES and PIV results for 𝑢∕𝑈∞ in Fig. 4a very near the
horseshoe vortex has started to transition to turbulence in this set of cube. This localized disagreement may be due to the lower Reynolds
results. number in the LES, different shape factor of the boundary layer and/or
All simulations were carried out with the OpenFOAM toolbox ver- absence of fluctuations in the incoming freestream flow. Nevertheless,
sion 7, which uses the finite-volume method to approximate a solution the present LES offered a good prediction of both the aerodynamic force
to the Navier–Stokes equations. The PISO algorithm was used to solve coefficients and the mean velocity components, and it is found to be
the system of equations, and the central differencing scheme was used acceptable to study the mean flow features around the surface-mounted
to discretize almost all spatial terms. An exception was the transport cube.
equations of the integral functions of the subgrid-scale model, for
which advective terms were discretized with a first order upwind 3. Results and discussion
scheme. A second order implicit scheme was used for time discretiza-
tion, and a fixed time step of 4 × 10−6 s was chosen to give a maximum Results will be presented for the main features of the time-averaged
Courant–Friedrichs–Lewy number smaller than 1. The equations for the flow around a surface-mounted cube in Section 3.1, including tradi-
momentum and the subgrid-scale model functions were solved with tional planar projections and the three-dimensional flow field. Specific
the preconditioned bi-conjugate gradient method and the diagonal- near-wall characteristics will then be addressed in Section 3.2, followed
incomplete lower–upper triangular (LU) preconditioner, and the pres- by final remarks in Section 3.3.
sure equation (derived from the discretized continuity equation) was
solved with the OpenFOAM generalized geometric-algebraic multi- 3.1. Main flow features
grid linear solver with a Gauss–Seidel smoother. The adopted scaled
2 ) and
residual convergence criteria were 1 × 10−6 for the pressure equation Fig. 5 presents the mean pressure coefficient 𝐶 𝑝 = 𝑝∕(0.5𝑈∞
and 1 × 10−8 for the other equations. The simulations were run until velocity field in the symmetry plane, for the boundary layers with
a regular quasi-periodic state was reached, indicated by the history 𝛿∕𝐷 = 0.2 (Fig. 5a) and 𝛿∕𝐷 = 0.8 (Fig. 5b). The same features are
of the force coefficients of the cube and velocity probes in the wake. found in both wakes, though with differences depending on the inflow
After that, the flow was averaged in time over 10 s or approximately conditions. A remarkable difference is that the horseshoe vortex is
40 shedding periods, based on the dominant frequency identified in located farther from the cube and closer to the ground with the laminar
the wake. Strouhal numbers of St = 0.11 and 0.09 were found for boundary layer. The position of its center, indicated by a focus point,
the laminar and turbulent boundary layers, respectively, in agreement is given in Table 3 together with the locations of other notable critical
with Sakamoto and Arie (1983). In addition, since the time-averaged points. For the topological definition of critical points which include
flow field is expected to be symmetric, it was also averaged over the saddle, node, focus, half-saddle and half-node points, the reader is
𝑥-𝑧 symmetry plane of the domain to smooth out minor differences referred to Hunt et al. (1978). Note that not all critical points have been
between its +𝑦 and −𝑦 sides. highlighted in Fig. 5 for clarity, but they meet the condition proposed
by Hunt et al. (1978) for two-dimensional plane sections of the flow:
2.1. Validation ( ) ( )
1 1
𝛴𝑁 + 𝛴𝑁 ′ − 𝛴𝑆 + 𝛴𝑆 ′ = 1 − 𝑛, (3)
2 2
The present results have been compared with experimental data where 𝛴𝑁 is the sum of nodal (and focus) points, 𝛴𝑁 ′ is the sum of
in the literature to evaluate the performance of the LES in terms of half-nodes, 𝛴𝑆 is the sum of saddle points, 𝛴𝑆 ′ is the sum of half-
aerodynamic force coefficients and mean velocity profiles, especially saddles, and 𝑛 is the connectivity of the section of the flow. In this
for the turbulent boundary layer with 𝛿∕𝐷 = 0.8, for which more case, 𝑛 = 1 for a single-connected region (Hunt et al., 1978; Cao et al.,
data are available. Table 2 presents the mean drag force coefficient 2022).
𝐶 𝐷 , root-mean-squared lift force coefficient 𝐶𝐿′ and mean normal force Upon inspection of the instantaneous flow field (not shown), tran-
coefficient 𝐶 𝑁 for the cube in the present LES and values reported in sition to turbulence was found to take place in the horseshoe vortex
the literature. The mean drag force coefficient of the cube with 𝛿∕𝐷 = with the laminar boundary layer, triggered by the adverse pressure
0.8 agrees well with the measurements by Sakamoto and Oiwake (1984) gradient upstream of the cube as in the DNS of Vinuesa et al. (2015).
and Heng and Sumner (2020), and is close to da Silva et al. (2022a) The mean position of the horseshoe vortex’s center is, however, close
where a higher value was measured. Higher drag force coefficients are to that of laminar horseshoe vortices, following Ballio et al. (1998).
expected for thinner boundary layers (Sakamoto and Oiwake, 1984), Its horizontal position is farther upstream than suggested by the cited
which is the case in the present LES with 𝛿∕𝐷 = 0.2 and in the ESDU study’s correlation, but it is likely an effect of the transitioning state
(1978) dataset. The fluctuating lift force coefficient is similar between of the horseshoe vortex with the laminar boundary layer. The position
the two present cases and to the value reported in Sakamoto and of the horseshoe vortex core with the turbulent boundary layer, on the
Oiwake (1984), and the mean normal force coefficient also shows good other hand, agrees with the fully turbulent horseshoe vortices in Ballio
agreement with the literature. et al. (1998) and da Silva et al. (2022a).
Fig. 4 compares mean velocity profiles obtained with the turbulent Regarding the bimodal behavior of the horseshoe vortex, Fig. 6a
boundary layer condition with the PIV and seven-hole pressure probe and b present the resolved turbulent kinetic energy 𝑘∕𝑈∞ 2 computed
measurements of da Silva et al. (2022a). The mean velocity components in the 𝑥-𝑧 symmetry plane for both boundary layers. This behavior is
𝑢∕𝑈∞ and 𝑤∕𝑈∞ and resolved correlations 𝑢′ 𝑢′ ∕𝑈∞ 2 , 𝑤′ 𝑤′ ∕𝑈 2 and characterized by an approximate C shape of high-magnitude 𝑘 contours
∞

5
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 4. Mean velocity and fluctuation profiles of the present simulation with a turbulent boundary layer with 𝛿∕𝐷 = 0.8 compared with experimental results of da Silva et al.
(2022a) along vertical lines with 𝑦∕𝐷 = 0 and (a) 𝑥∕𝐷 = 0, (b) 𝑥∕𝐷 = 1, (c) 𝑥∕𝐷 = 2 and (d) 𝑥∕𝐷 = 3.5, and (e) along a horizontal line with 𝑥∕𝐷 = 4 and 𝑧∕𝐷 = 1.

Table 2
Mean drag force coefficient 𝐶 𝐷 , root-mean-squared lift force coefficient 𝐶𝐿′ and mean normal force coefficient 𝐶 𝑁 for the
surface-mounted cube in the present simulations and select experimental studies.
Case 𝛿∕𝐷 Re 𝐶𝐷 𝐶𝐿′ 𝐶𝑁
Laminar 0.2 1 × 104 1.25 0.14 0.56
Turbulent 0.8 1 × 104 0.97 0.13 0.62
ESDU (1978) <0.2 1 × 104 –1 × 106 1.16 – 0.6
Sakamoto and Oiwake (1984) 0.67 2.82 × 104 –1.69 × 105 0.96 0.13 –
Heng and Sumner (2020) 0.8 1.1 × 105 0.97 – –
Heng and Sumner (2022) 0.8 6.5 × 104 – – 0.65
da Silva et al. (2022a) 0.7 7.5 × 104 1.13 – 0.47

Table 3 near the horseshoe vortex core, originally observed by Devenport and
Location of the notable critical points identified in Fig. 5.
Simpson (1990). The present results reveal that this feature is found
𝛿∕𝐷 = 0.2 𝛿∕𝐷 = 0.8 only for the turbulent boundary layer case, since the intermittent
Center of the horseshoe vortex, HSV 𝑥∕𝐷 = −1.58 𝑥∕𝐷 = −0.82 merging and breakdown of vortices, associated with the bimodal behav-
𝑧∕𝐷 = 0.05 𝑧∕𝐷 = 0.11
ior and the high-𝑘 signature (Escauriaza and Sotiropoulos, 2011b), is
Front stagnation point, N1 𝑧∕𝐷 = 0.30 𝑧∕𝐷 = 0.66 typically observed for turbulent horseshoe vortex systems (Mohammadi
Center of the free end vortex, Ft 𝑥∕𝐷 = 0.30 𝑥∕𝐷 = 0.23 et al., 2022). The signature double-peaked probability density function
𝑧∕𝐷 = 1.32 𝑧∕𝐷 = 1.22
of the bimodal behavior in the vicinity of the horseshoe vortex core in
Free end saddle point, St 𝑥∕𝐷 = 0.31 𝑥∕𝐷 = 0.45 the 𝑥-𝑧 symmetry plane is also found only for the turbulent boundary
𝑧∕𝐷 = 1.32 𝑧∕𝐷 = 1.22
layer case (Fig. 6e). However, a weak bimodal behavior is detected
Free end reattachment point, Rt 𝑥∕𝐷 = 0.19 𝑥∕𝐷 = 0.41
for the transitioning horseshoe vortex closer to the leading edges of
Center of vortex Bt , FBt 𝑥∕𝐷 = 1.30 𝑥∕𝐷 = 0.99 the cube, as shown in Fig. 6d. This finding suggests that the mecha-
𝑧∕𝐷 = 1.14 𝑧∕𝐷 = 1.03
nism of development of this instability is triggered downstream in the
Rear stagnation point, N2 𝑧∕𝐷 = 0.14 𝑧∕𝐷 = 0.12
transitioning horseshoe vortex.
Center of vortex Nw , FNw 𝑥∕𝐷 = 0.59 𝑥∕𝐷 = 0.58 The thicker boundary layer and the size and proximity of the
𝑧∕𝐷 = 0.04 𝑧∕𝐷 = 0.04
turbulent horseshoe vortex to the cube led to a higher location of
Wake reattachment point, Rw 𝑥∕𝐷 = 2.84 𝑥∕𝐷 = 2.39
the stagnation point on the cube’s frontal face (N1 ), in Fig. 5b, which
reduces the portion of the impinging flow that is directed upward
and separates from the top edge. Over the free end of the cube,
flow reattachment from the backward flow onto the top face is found

6
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 5. Mean pressure coefficient 𝐶 𝑝 and velocity field in the 𝑥-𝑧 symmetry plane for (a) the laminar boundary layer with 𝛿∕𝐷 = 0.2 and (b) the turbulent boundary layer with
𝛿∕𝐷 = 0.8. The yellow line corresponds to 𝑢∕𝑈∞ = 0, and the following critical points are identified: the center of the horseshoe vortex, HSV; the front stagnation point, N1 ;
the center of the free end vortex, Ft ; the free end reattachment point, Rt ; the free end saddle point, St ; the center of the near-wake vortex (vortex Bt ), FBt ; the rear stagnation
point, N2 ; the center of vortex Nw , FNw ; and the wake reattachment point, Rw . The insets to the right present a closer view of the rear free end region. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)

2
Fig. 6. Turbulent kinetic energy 𝑘∕𝑈∞ in the 𝑥-𝑧 symmetry plane for (a) the laminar boundary layer with 𝛿∕𝐷 = 0.2 and (b) the turbulent boundary layer with 𝛿∕𝐷 = 0.8.
Probability density functions (PDF) are shown for the (c,e) streamwise velocity component 𝑢∕𝑈∞ at the point indicated by a cross in the turbulent kinetic energy plots, and for
the (d,f) radial velocity component 𝑢𝑟 ∕𝑈∞ at the crosses indicated in the inset to the right, for (c,d) the laminar boundary layer with 𝛿∕𝐷 = 0.2 and (e,f) the turbulent boundary
layer with 𝛿∕𝐷 = 0.8.

for both boundary layers, yet it is more notable for the laminar one the near wake for the laminar boundary layer due to the strengthened
(Fig. 5a). The bubble formed ahead of the trailing edge is the dominant vortex shedding under this condition (Mohammadi et al., 2022), in
feature above the free end — such that the free end vortex that appears contrast to the maximum value of 𝑘 being found above the free end
for the turbulent boundary layer in Fig. 5b has almost merged with the for the turbulent boundary layer, as in the measurements of Hussein
free end saddle point St , and the markers are mostly overlapped. For and Martinuzzi (1996). As will be demonstrated in this study, the
the turbulent boundary layer in Fig. 5b, there is a very small bubble, latter behavior is related to the formation of a headband vortex and
which explains why the reattachment point could not be detected by da intermittent attachment and separation on the free end of the cube.
Silva et al. (2022a) with PIV. Still, the present results agree with da The overall length of the recirculation region is also reduced in Fig. 5b,
Silva et al. (2022a) in the sense that the backflow reattachment is more indicated by the wake reattachment point in Table 3. The location
evident for a thinner boundary layer, and confirm the role of this type of vortex Bt , consequently, moved closer to the cube and the ground
of reattachment in the formation of trailing-edge vortices for a surface- plane for the thicker boundary layer, in agreement with da Silva et al.
mounted cube as well, previously observed by Cao et al. (2022) for (2022a). The same does not happen for the smaller vortex Nw , present
AR = 3. In contrast, higher fluctuations are found with the turbulent in the near wake near the junction of the cube with the ground plane:
boundary layer above the free end of the cube, as shown in Fig. 6b, the location of FNw and its associated stagnation point N2 does not
which may disrupt this process. change significantly for the two boundary layers.
The recirculation region behind the cube is characterized by the An overview of the three-dimensional flow is presented in Fig. 7,
large near-wake or arch vortex in Fig. 5, also called vortex Bt , for where the isosurfaces of 𝜆2 = −0.01 (Jeong and Hussain, 1995) in
both boundary layers. However, while the minimum pressure is found Fig. 7a and c, and 𝜔𝑥 𝐷∕𝑈∞ = ±0.45 in Fig. 7b and d, have been
around the focus point of this vortex for the laminar boundary layer spatially filtered to clarify and highlight the most important structures.
(Fig. 5a), a significant difference observed in the present results is The major flow structures in Fig. 7a and c are the horseshoe vortex, the
that the minimum pressure occurs over the free end for the turbulent arch vortex and the dipole structures (or tip vortices) in the far wake. As
boundary layer (Fig. 5b), the same behavior as in Yakhot et al. (2006). identified previously, the horseshoe vortex is located farther from the
The opposite occurs regarding the turbulent kinetic energy distribution cube with the laminar boundary layer. Its legs are spread wider than for
in the symmetry plane: Fig. 6 shows that it is significantly higher in the turbulent boundary layer, in agreement with the laminar boundary

7
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 7. Isosurfaces of 𝜆2 = −0.01 colored by the mean streamwise vorticity 𝜔𝑥 𝐷∕𝑈∞ for (a) the laminar boundary layer with 𝛿∕𝐷 = 0.2 and (c) the turbulent boundary layer with
𝛿∕𝐷 = 0.8. Isosurfaces for the mean streamwise vorticity 𝜔𝑥 𝐷∕𝑈∞ = ±0.45 close to the cube are shown in (b) and (d) for 𝛿∕𝐷 = 0.2 and 𝛿∕𝐷 = 0.8, respectively, together with
streamlines of the mean velocity near regions of large streamwise vorticity. The green isosurface indicates the mean vertical velocity 𝑤∕𝑈∞ = −0.2, and the magenta isosurface
indicates 𝑤∕𝑈∞ = 0.4 in the insets. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

layer in Mohammadi et al. (2022). The near-wake vortex, whose center 𝛿∕𝐷 = 0.2 than for 𝛿∕𝐷 = 0.8, as indicated by the isosurface of the
was identified as FBt in Fig. 5, is completely merged with the separated vertical velocity component 𝑤∕𝑈∞ = −0.2 in green (no isosurface is
shear layers in Fig. 7a. Some disconnection between the structures is present for the turbulent boundary layer case because the minimum
apparent in Fig. 7c with the turbulent boundary layer, especially near vertical velocity in the wake is only 𝑤∕𝑈∞ = −0.17). The weakening
the downstream corners of the cube. of the downwash due to a thicker boundary layer has already been
The dipole structures in Fig. 7a and c protrude from the near-wake reported for prisms of AR = 5–10 (Wang et al., 2006; Behera and
vortex and extend outward into the far wake as for surface-mounted Saha, 2019; Chen et al., 2022), where it was attributed to the increased
finite-height square prisms of larger AR (Yauwenas et al., 2019; Chen upwash in the bottom part of the wake. However, Crane et al. (2022)
et al., 2022). The dipole structures are characterized mainly by a higher recently have shown that downwash and upwash do not share a causal
magnitude of the streamwise vorticity, which is the most intense right behavior. In the present results, the weaker downwash with the thicker
outside of the recirculation region (da Silva et al., 2020), i.e., 𝑥∕𝐷 ≈ 2.5 boundary layer is due to the lower momentum of the separated flow,
for the laminar boundary layer and 𝑥∕𝐷 ≈ 2 for the turbulent boundary compounded by the weaker adverse pressure gradient induced behind
layer. For the laminar boundary layer and the chosen threshold of the cube. This effect can be seen in the higher vertical position of
𝜆2 = −0.01, these structures have their largest cross-section at about the front stagnation point N1 in Fig. 5, which indicates that a smaller
𝑥∕𝐷 ≈ 6, and by 𝑥∕𝐷 = 10 it decreases in size as the intensity of the fraction of the flow is diverted upward, and in the smaller upward
streamwise vorticity decreases and the structures diffuse downstream velocity of the flow in front of the cube, shown by the insets of Fig. 7b
in the wake. For the turbulent boundary layer, there is a large gap and d. In summary, the weaker downwash with the thicker boundary
between the small, starting part of the dipole structures and the large layer is a consequence of mass conservation and a weaker pressure
characteristic strands downstream, even for values of 𝜆2 closer to zero. gradient behind the surface-mounted cube.
The streamwise vorticity shows, however, a high magnitude in this re- The second main difference is related to the mean streamwise
gion (shown in Fig. 8e), satisfying the solenoidal condition (Crane et al., vorticity regions that are found in the vicinity of the cube, hidden by
2022). The dipoles reappear after about 𝑥∕𝐷 ≈ 3, where they show an the near-wake vortex and shear layers of Fig. 7a and c. Above the
irregular shape, composed of a main circular strand and an adjunct part free-end, the top separated shear layer shows an outward trajectory
in the outer side of the dipoles, closer to the ground plane. The dipole that corresponds to the negative (blue) vorticity on the +𝑦 side of
structures are compatible with the streamwise wake vorticity pattern the cube and positive (orange) vorticity on the −𝑦 side, in Fig. 7b
in da Silva et al. (2022a), in which case the adjunct part corresponds and d. These vorticity regions have been identified for taller prisms
to the ground plane vorticity, to be further investigated later in this in Sumner et al. (2017) and da Silva et al. (2020), and called secondary
section. tip vortices in Goswami and Hemmati (2022). While these vorticity
Fig. 7b and d show isosurfaces of the mean streamwise vorticity regions are inclined upward and away from the surface of the cube
closer to the surface-mounted cube and streamlines of the mean veloc- for the laminar boundary layer, for the turbulent boundary layer they
ity field seeded near the location of strongest vorticity for the dipole are larger and stretch to the downstream corners of the cube. The sign
structures. The vorticity close to the ground plane (𝑧∕𝐷 < 0.1) has been of the streamwise vorticity of the outward-deflected flow conflicts with
clipped, for clarity. The streamlines show that the dipole structures, that of the dipole structures, which causes their weakening in the near
from a time-average perspective, are related to the inward bending of wake. This phenomenon is also shown in the disconnection of the shear
the outer separated shear layers by the downwash, as described by da layers and arch vortex near the corners, in Fig. 7c.
Silva et al. (2020) and Cao et al. (2022) for surface-mounted prisms of Besides the near-wall vorticity distributions, Fig. 7b and d also
larger AR. However, two main differences are observed in this process reveal the corner vortices by the cube’s junction with the ground plane,
for the mean flow with the laminar (Fig. 7b) and turbulent (Fig. 7d) and the ground plane vorticity. To examine the ground plane vorticity
boundary layers. First, the downwash is substantially more intense for and other streamwise vorticity regions in the near wake, Fig. 8 presents

8
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 8. Isosurfaces of the mean streamwise vorticity 𝜔𝑥 𝐷∕𝑈∞ = ±0.45 for 𝑧∕𝐷 ≤ 0.3 and streamlines of the mean velocity field with (a) the laminar boundary layer with 𝛿∕𝐷 = 0.2
and (d) the turbulent boundary layer with 𝛿∕𝐷 = 0.8, clipped to the −𝑦 half of the domain. The streamlines are colored dark blue for 𝑤∕𝑈∞ < 0 and dark red for 𝑤∕𝑈∞ > 0.
Contours of 𝜔𝑥 𝐷∕𝑈∞ and vectors of the mean velocity field in 𝑦-𝑧 planes with 𝑥∕𝐷 = 2 and 4 are shown in (b) and (c) for 𝛿∕𝐷 = 0.2 and in (e) and (f) for 𝛿∕𝐷 = 0.8. The dashed
magenta lines show the outline of the cube, for reference, and the yellow lines correspond to 𝑢∕𝑈∞ = 0. Vectors are interpolated into a uniform grid with spacing of 0.12𝐷. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

the mean streamwise vorticity distribution close to the ground plane wake. The vorticity regions of the top shear layer are larger in size for
and in 𝑦-𝑧 planes with 𝑥∕𝐷 = 2 and 4, for the laminar boundary layer the turbulent boundary layer (Fig. 8e), while the dipole structures are
with 𝛿∕𝐷 = 0.2 in Fig. 8a–c and the turbulent boundary layer with significantly smaller.
𝛿∕𝐷 = 0.8 in Fig. 8d–f. The mean streamwise vorticity isosurfaces have Below the dipole structures, weaker vorticity regions of opposite
been clipped for 𝑧∕𝐷 ≤ 0.3 and the −𝑦 side of the domain in Fig. 8a sign of rotation are found especially for the laminar boundary layer,
and d. The ground plane vorticity is highlighted in both figures as a within the recirculation region. These regions are related to the down-
clockwise (blue) vorticity region that begins within the recirculation stream inclination of the arch vortex. As shown in Fig. 9a, the vorticity
region. Select streamlines show that this vorticity is derived from the line near 𝑥∕𝐷 = 2 is inclined so that a positive vorticity region is pro-
entrainment of fluid into the low-pressure region in the near wake, sim- jected on the −𝑦 side of the 𝑦-𝑧 plane in Fig. 8b, and a negative region
ilar to the dipole structures; this flow ends up being slightly deflected is found on the +𝑦 side. The arch vortex for the turbulent boundary
downward by the downwash, which dominates the entire wake for both layer, on the other hand, appears mostly vertical in Fig. 9b, with only
boundary layers (see Fig. 5). This vorticity is, therefore, different from a slight inclination upstream at its top. Some vorticity of the same
the horseshoe vortex and accessory to the dipole structures in the upper sign as the ground plane vorticity is also present between it and the
part of the wake, yet enhanced by its proximity to the ground plane. In dipole structures in Fig. 8b and e (next to the arch vortex projection).
addition, the ground plane vorticity spreads above the vorticity of the While this vorticity resembles the descending vortices of Kindree et al.
reformed and developing ground plane boundary layer, shown as the (2018) and Mohammadi et al. (2022), the flow topology is overall
orange sheet downstream in Fig. 8a and d. different from the one in the cited studies. In addition, these vorticity
The ground plane vorticity and flow field can also be visualized in a regions do not constitute a flow structure based on vortex identification
two-dimensional perspective in Fig. 8b, c, e, and f. While it first exists parameters like 𝜆2 and 𝑄 (Hunt et al., 1988; Jeong and Hussain, 1995).
independently from the horseshoe vortex, downstream in the wake the The corner vortex, which is hidden behind the arch vortex in Fig. 7,
horseshoe vortex legs diffuse and their vorticity merges with the ground is more clearly observed in Fig. 10 for both boundary layers. Isosurfaces
plane vorticity, a process especially visible for the turbulent boundary of the mean pressure coefficient 𝐶 𝑝 = −0.65 are used to help identify
layer, given its proximity to the cube. The present results reveal that the relevant structures. For the laminar boundary layer with 𝛿∕𝐷 = 0.2
ground plane vorticity is involved in the formation of flow structures: (Fig. 10a), the arch vortex appears as a dominant structure, where
after the horseshoe vortex legs join with the ground plane vorticity, its lowest-pressure region is in agreement with Fig. 5a. The corner
vortex cores reappear for the turbulent boundary layer as the adjunct vortex is located close to the ground plane and is vertically inclined
parts for 𝑥∕𝐷 > 5, based on the 𝜆2 threshold specified in Fig. 7c. Fig. 8b as characteristic of its positive streamwise vorticity component, shown
and e also show the vorticity of the top shear layer in the upper part in Figs. 7 and 8. Like for the AR = 3 prism of da Silva et al. (2020),
of the plane, as it is in these planes deflected inward into the near the corner vortex is initially formed from the streamlines that separate

9
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

3.2. Near-wall flow

In this section, limiting streamlines of the velocity field (from the

closest grid node to the surfaces, at a distance of 0.0006𝐷) are discussed
for the ground plane, followed by an analysis of their distribution
on the cube faces, and their relationship to the three-dimensional
structures. Critical points have been identified based on this limiting
velocity field, under the assumption that it is close enough to the cube
and ground plane surfaces to represent skin-friction lines.
Fig. √
11 presents the magnitude of the mean skin-friction coefficient
2 2
|𝐶 𝑓 | = 𝐶 𝑓 𝑥 + 𝐶 𝑓 𝑦 and limiting streamlines of the mean velocity field
on the ground plane for the laminar boundary layer with 𝛿∕𝐷 = 0.2
and the turbulent boundary layer with 𝛿∕𝐷 = 0.8. The streamline
patterns agree qualitatively with the flow visualizations of Mohammadi
et al. (2022) and Martinuzzi and Tropea (1993) (for the turbulent
boundary layer), as well as the DNS of Yakhot et al. (2006). An
interesting outcome of the transitioning state of the horseshoe vortex
in the laminar boundary layer case is that the flow pattern is similar
to that of a turbulent horseshoe vortex but, as identified previously,
the transitioning horseshoe vortex is located farther from the cube
(Fig. 11a), as is the case for a laminar horseshoe vortex. Its primary
separation line and the corresponding separation saddle point are also
Fig. 9. Streamlines of the mean velocity field (gray) and vorticity lines (purple) in the
farther, located at 𝑥∕𝐷 = −2.5 for the laminar boundary layer versus
vicinity of the arch vortex for (a) the laminar boundary layer with 𝛿∕𝐷 = 0.2 and (b)
the turbulent boundary layer with 𝛿∕𝐷 = 0.8. (For interpretation of the references to 𝑥∕𝐷 = −1.8 for the turbulent one. The horseshoe vortex core (approx-
color in this figure legend, the reader is referred to the web version of this article.) imated from a plane located at 𝑧∕𝐷 = 0.1, i.e., above the limiting
streamlines plane, and indicated by the dashed green lines) is located
far from the main separation line for both boundary layers, close to
near the upstream corner of the cube, as shown in Fig. 10b. Following a line of low streamwise shear also found by Devenport and Simpson
Fig. 10c and d, it stays contained by the lateral face of the cube only (1990), Paik et al. (2007) and Kindree et al. (2018). In Fig. 11, these
up to 𝑧∕𝐷 ≈ 0.2, at which point the vortex is not contained in the side lines delimit the regions of low shear stress upstream and high shear
region anymore and extends downstream. The flow recirculates into stress closer to the cube, with higher magnitude of the stress found for
the side area for higher planes, showing a flow field similar to that of the laminar boundary layer case. The legs of the transitioning horseshoe
Fig. 10d. Also notable for the laminar boundary layer is the occurrence vortex (Fig. 11a) follow the shape of the near-ground streamlines in
of backward flow reattachment over the entire lateral surface of the their narrowest point close to where wake reattachment takes place,
cube, similar to its top surface (see Fig. 5a). The resulting recirculation while the turbulent horseshoe vortex legs (Fig. 11b) have a straighter
region has also been identified by Cao et al. (2022), who referred to it shape.
as the trailing-edge vortex. Upstream of the cube, a node for the laminar boundary layer or a
As identified previously, the pressure distribution in the wake is node-saddle combination for the turbulent boundary layer is found. The
different for the turbulent boundary layer with 𝛿∕𝐷 = 0.8 (Fig. 10e). former pattern has been reported in Cao et al. (2022), while the latter
The arch vortex is weakened, and the lowest pressures are located was found in Paik et al. (2007) and related to a secondary separation.
around the sides and top of the cube. The corner vortex is still present The near-wall flow field on the ground plane in the vicinity of the
near the cube’s junction with the ground plane, but for the turbulent cube’s side face is shown in more detail in the insets of Fig. 11. Both
boundary layer it is only partially released into the wake. Fig. 10f–h boundary layer cases show a focus point related to flow separation
show that part of the vortex is formed by the lateral face of the cube from the ground plane into the corner vortex, a smaller focus point
and mostly stays contained, with higher planes showing a flow field close to the upstream corner and a saddle point that delimits the side
like that of Fig. 10h. This behavior characterizes the side vortex, which flow region and the near wake region. The mean flow with the laminar
extends and connects over the top of the cube in Fig. 10e, forming
boundary layer (Fig. 11a) shows, however, a more complex mean flow
a ‘‘headband’’ vortex. The same pattern with the presence of a side
distribution with additional focus and saddle points on each side of
vortex has appeared in Meinders et al. (1999), Krajnović and Davidson
the cube, which are related to the reverse-flow reattachment region
(2001) and Yakhot et al. (2006) for surface-mounted cubes, but in the
(or trailing-edge vortex, following Cao et al. (2022)). These points are
present study this is the first time that this structure is scrutinized
absent for the case of the turbulent boundary layer (Fig. 11b) because
in conjunction with the corner vortex for a cube. The earlier studies
the flow mostly follows the contours of the cube, and only a half-saddle
adopted either laminar and turbulent boundary layers, but all of them
and a half-node are present instead, near the trailing edge.
were carried out with boundary layer thicknesses equal to or larger
than the prism’s height (i.e., 𝛿∕𝐻 ≥ 1). The literature suggests, there- Behind the cube, the flow distribution agrees with Depardon et al.
fore, that the development of the side vortex is strongly conditioned (2005), Yakhot et al. (2006) and Cao et al. (2022) and is similar
to the viscous shear imposed by a larger boundary layer thickness, a for both boundary layers, aside from a narrower mean wake for the
trend which agrees with the appearance of the side vortex for AR = 3 turbulent one due to the proximity of the horseshoe vortex’s legs. The
and 𝛿∕𝐻 = 6.7 (Cao et al., 2022), but not for AR = 4 and 𝛿∕𝐻 = shear stress distribution is weaker in this region, but low pressure
0.52 (Okuda and Taniike, 1993). The corner vortices are, however, coefficients (not shown) are found especially for the laminar boundary
still present closer to the ground plane. They are characterized by a layer close to the pair of foci, which corresponds to the location of the
significant streamwise vorticity component, making them a more three- arch vortex’s footprints. Near the attachment region of the near wake,
dimensional branch of the corner-side vortex system, and they extend different patterns with either a single node or a node-saddle–node are
into the wake of the cube, in contrast to the bounded side vortex. In found, both of which have appeared in the literature (Depardon et al.,
addition, the present results indicate that corner vortices are expected 2005; Yakhot et al., 2006; da Silva et al., 2020). Furthermore, the
to be formed regardless of the boundary layer characteristics. critical points have also been verified to satisfy Eq. (3), with 𝛴𝑁 = 11,

10
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 10. Isosurfaces of the mean pressure coefficient 𝐶 𝑝 = −0.65 and streamlines of the mean velocity field for (a) the laminar boundary layer with 𝛿∕𝐷 = 0.2 and (e) the turbulent
boundary layer with 𝛿∕𝐷 = 0.8. Contours of the mean vertical vorticity 𝜔𝑧 𝐷∕𝑈∞ and velocity field in 𝑥-𝑦 planes near the lateral face of the cube at 𝑧∕𝐷 = 0.01, 0.1 and 0.2 are
shown in (b–d) for 𝛿∕𝐷 = 0.2 and in (f–h) for 𝛿∕𝐷 = 0.8.

Fig. 11. Contours of the magnitude of the mean skin-friction coefficient 𝐶 𝑓 and limiting streamlines of the mean velocity field on the ground plane for (a) the laminar boundary
layer with 𝛿∕𝐷 = 0.2 and (b) the turbulent boundary layer with 𝛿∕𝐷 = 0.8. The dashed green lines correspond to the approximate location of the horseshoe vortex based on
an elevated plane at 𝑧∕𝐷 = 0.1. Node (□), saddle (◊) and focus (◦) points are indicated as white squares, diamonds and circles, respectively. The insets to the right show a
close-up of the flow near the side, upstream and downstream of the cube, where triangles (△) indicate half-node points and inverted triangles (▽) indicate half-saddle points.
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

11
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 12. Contours of the mean pressure coefficient 𝐶 𝑝 and limiting streamlines of the mean velocity field on the surface-mounted cube’s faces for (a) the laminar boundary layer
with 𝛿∕𝐷 = 0.2 and (b) the turbulent boundary layer with 𝛿∕𝐷 = 0.8. The white lines on the side and top faces correspond to attachment lines (AL) or separation lines (SL)
with 𝑢∕𝑈∞ = 0. Node (□), saddle (◊) and focus (◦) points are indicated as white squares, diamonds and circles, respectively. Notable critical points are highlighted: the front
stagnation point, N1 ; the rear stagnation point, N2 ; the main lateral saddle point, SL ; the center of the corner vortex, Fc ; the main (reattachment) top saddle point Rt ; the secondary
(separation) top saddle point St2 ; and the top attachment node, Nt .

𝛴𝑁 ′ = 6, 𝛴𝑆 = 12, 𝛴𝑆 ′ = 6 for Fig. 11a, 𝛴𝑁 = 10, 𝛴𝑁 ′ = 5, 𝛴𝑆 = 10, to −0.5% and −0.2% for the turbulent and laminar cases, respectively,
𝛴𝑆 ′ = 7 for Fig. 11b, and 𝑛 = 2 for the ground plane. which is a lower contribution than observed for a prism with AR = 3 (da
The mean pressure coefficient distribution and limiting (near- Silva et al., 2020).
surface) streamlines of the velocity field are presented in Fig. 12 for the The characteristics of the side and top faces of the cube can be
front, back, side and top faces of the cube, and for the two boundary remarkably different depending on the boundary layer, as already
layers. Regarding the critical points distribution on a surface-mounted hinted in Figs. 5 and 10. For the laminar boundary layer, the lateral
body, the established rule is that the total number of nodes on both face contains a main saddle point SL at approximately half-height,
the body and the ground plane must be equal to the total number of contrary to Cao et al. (2019, 2022) who identified two saddle points
saddles, i.e., 𝛴𝑁 − 𝛴𝑆 = 0 (Hunt et al., 1978). The present results that separated the free-end, 2D-like and junction regions for a prism
have 𝛴𝑁 − 𝛴𝑆 = 2 for the laminar and turbulent boundary layer cases. with AR = 3. For the cube, such distinct regions of the near-wall flow
The rule is not met, but a similar outcome has been obtained in other field are not as well defined. The main lateral saddle point for the cube
studies (Liakos and Malamataris, 2014; Cao et al., 2022), possibly due is located on an attachment line near the trailing edge, formed from the
to structural bifurcations (Chapman and Yates, 1991; Depardon et al., reattachment of the reverse flow that separated from the back edges
2005) which were not fully captured by the LES. of the cube. Flow separation takes place again at the corner vortex’s
The frontal pressure and velocity distribution are shown in the focus point, Fc . This distribution is very similar to that visualized and
leftmost panels of Fig. 12a and b. As observed previously in Table 3, the measured by Depardon et al. (2005) for a cube with 𝛿∕𝐷 = 0.2 and Re =
stagnation point N1 is located higher for the turbulent boundary layer 4 × 104 and 1.6 × 105 , yet this is the first time that the near-wall features
with 𝛿∕𝐷 = 0.8 (Fig. 12b). The lower pressure coefficient magnitude have been connected to structures such as the corner vortex and the
for the thicker boundary layer, especially away from the stagnation trailing-edge vortex. As for the pressure field, it is nearly uniform and
point, occurs due to the lower momentum of the freestream flow that slightly lower close to the rear edge, in agreement with Castro and
impinges on the cube. The flow spreads out and separates from the Robins (1977) (𝛿∕𝐷 = 0.025) and the ESDU 71016 report (ESDU, 1978)
leading edges, except near the ground plane, where a saddle point (𝛿∕𝐷 < 0.2).
indicates flow separation and the formation of a small junction vortex With the turbulent boundary layer, the flow on the lateral face
between the cube and the horseshoe vortex for both boundary layers, becomes more complex. The main lateral saddle point SL is located
also delimited by a node upstream of the cube on the ground plane higher and closer to the trailing edge, indicating that the trailing-edge
(Fig. 11). An additional, even smaller junction vortex is present in front vortex is significantly smaller than for the laminar boundary layer.
of the cube with the turbulent boundary layer, as indicated by the node Aside from the main attachment line where SL is located, there is a
on the front of the cube and the saddle on the ground plane (Fig. 11b), nearly vertical separation line and a secondary attachment line, which
close to their junction. are consistent with the flow near the side of the cube in Fig. 10f–
On the back of the cube, the velocity field is the same for both h. This pattern is related to the formation of the side vortex for the
boundary layers, but the pressure coefficient distribution shows higher turbulent boundary layer, so far only having been implied for surface-
(closer to zero) pressure coefficients for the turbulent one, following mounted cubes based on flow visualizations (Martinuzzi and Tropea,
the pressure distributions in Figs. 5 and 10. The increased pressure 1993; Nakamura et al., 2001). The corner vortex’s focus point, Fc ,
recovery for a thicker boundary layer has been verified also by Heng is found on the separation line, and the pressure coefficient is much
and Sumner (2022), but based on pressure measurements on the top lower near this point. Higher pressures occur closer to the rear edge
face. The weaker pressure distributions on the front and back faces due to pressure recovery, in agreement with Castro and Robins (1977)
of the cube with the turbulent boundary layer translate into a lower and Sakamoto et al. (1982).
drag force coefficient of 𝐶 𝐷 = 0.97, shown in Table 2, versus 𝐶 𝐷 = Some of the characteristics identified on the side faces carry over
1.25 for the laminar boundary layer. The pressure drag is by far the to the top face of the cube. With the laminar boundary layer, the
most important component, with the skin-friction drag corresponding large trailing-edge vortex is delimited by the attachment line where the

12
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

top reattachment saddle point, Rt , is located. Reverse flow then takes

place over most of the cube’s free end, with a distribution that agrees
with Khan and Saha (2022). The top face of the cube with the turbulent
boundary layer has a smaller trailing-edge vortex, in agreement with
Fig. 5, and the additional separation and secondary attachment lines
from the side faces carry on over the top of the cube. This behavior
corresponds to the presence of the headband vortex composed by a
free end vortex connecting the side vortices, as shown by the low-
pressure region in Fig. 10e. The saddle St2 and node Nt are found on
the separation and secondary attachment lines, respectively, and are
expected to merge shortly above the free end of the cube where reverse
flow is dominant. Similar flow fields on the top face of cubes and
prisms were obtained by Depardon et al. (2005) and Cao et al. (2019),
although a pair of foci on the separation line and a concave main (rear)
attachment line have been observed in other studies (Nakamura et al.,
2001; Yakhot et al., 2006).
As for the pressure distribution on the top of the cube, there is
little variation in agreement with Castro and Robins (1977), ESDU
(1978) and Heng and Sumner (2022), with slightly higher pressure
coefficients found where flow attachment takes place. The overall pres-
sure coefficients are, however, lower (more negative) for the turbulent
boundary layer. It leads to a larger mean normal force coefficient 𝐶 𝑁 =
0.62 (more suction), whereas the laminar boundary layer condition Fig. 13. Contours of the probability that the instantaneous streamwise velocity is equal
has 𝐶 𝑁 = 0.56. The unique distribution of the pressure on the free- to zero (𝑢=0 ) and limiting streamlines of the mean velocity field on the surface-
mounted cube’s side and top faces for (a) the laminar boundary layer with 𝛿∕𝐷 = 0.2
end of the cube with the turbulent boundary layer, associated with
and (b) the turbulent boundary layer with 𝛿∕𝐷 = 0.8. The white lines on the side and
the formation of the headband vortex, justifies the counter-intuitively top faces correspond to attachment or separation lines with 𝑢∕𝑈∞ = 0. The node (□),
higher value of 𝐶 𝑁 that has been obtained for a thicker boundary layer, saddle (◊) and focus (◦) points of Fig. 12 are included.
also in experimental data (da Silva et al., 2022a). As was the case for
the drag force coefficient, the pressure contribution is also the most
important for the total 𝐶 𝑁 : skin-friction contributions amount to 1.2% as much near Fc , suggesting that the corner vortex’s separation focus
and 2.1% for the turbulent and laminar boundary layer conditions, point may not always be found. For the turbulent boundary layer (left
respectively. panel of Fig. 13b), high values of 𝑢=0 are found along the attachment
The overall modification of the flow on the sides and top of the cube and separation lines, but also on the lower half of the cube between
with the development of the headband vortex for larger 𝛿∕𝐷 is hypoth- the main attachment line and the separation line containing Fc . Above
esized to be related to the higher proximity and lower momentum of this region, the likelihood of attachment or separation is low, which
the shear layers separated from the cube’s leading edges, in comparison indicates the persistence of the side vortex.
with the thin boundary layer. The lower momentum carries over to the
recirculating flow found near the side and top faces of the cube, which 3.3. Closure
in turn does not have enough energy to stay attached (upstream of the
trailing-edge vortices). The reverse flow then separates toward joining The near-wall flow distribution on the cube faces, and associated
the shear layers, as shown also in Cao et al. (2022), and in agreement structures like the trailing-edge and side vortices, are connected to the
with the low-pressure region that develops in the headband vortex. flow features around and downstream of the cube. One of these features
To further investigate the occurrence of flow attachment and sep- is the dipole structures, which were shown in the previous section to
aration on the side and top faces of the cube, the near-wall locations possibly be weakened by the enhanced presence of the opposite-sign
with instantaneous 𝑢 = 0 were tracked along 1000 snapshots equally vorticity from the outward-deflected top shear layer for the surface-
spaced during the time-averaging of the LES. The probability that the mounted cube with the turbulent boundary layer. To further look into
instantaneous streamwise velocity is equal to zero, 𝑢=0 , is presented this phenomenon, Fig. 14 presents streamlines of the three-dimensional
for the side and top faces of the cube in Fig. 13 for the two boundary flow field near the side of the cube, colored by the mean streamwise
layers. The critical lines with 𝑢∕𝑈∞ = 0, the limiting mean velocity vorticity 𝜔𝑥 𝐷∕𝑈∞ . With the laminar boundary layer (Fig. 14a), the
streamlines and the critical points of Fig. 12 are included for refer- near-wake flow that recirculates toward the back face of the cube is
ence. The considered probability 𝑢=0 indicates the likelihood of flow deflected to its top and sides, following the back face’s near-wall flow
attachment or separation, without distinguishing between these events. distribution. The flow closest to the cube separates to form the trailing-
Still, it brings some insight into the attachment and separation lines edge vortices, while the outer recirculating flow bypasses these regions,
identified in the mean flow field. as shown in the figure. The streamlines are then deflected upward away
The maximum probability does not go above 55% (aside from ex- from the region occupied by the corner vortex in the bottom half of
actly at the leading edges, where it is 100%). Note that an intermittent the cube. In contrast to the counter-clockwise streamwise vorticity of
reattachment is expected for large 𝛿∕𝐷 and turbulence intensity (Hearst the corner vortex, these recirculating streamlines are characterized by
et al., 2016), but there were locations with 𝑢 = 0 for all snapshots in a clockwise (blue) streamwise vorticity component (in agreement with
the present results, both with laminar and turbulent boundary layers. the isosurfaces in Fig. 7b), which matches the vorticity of the dipole
Higher probabilities of 𝑢 = 0 are generally concentrated near the mean structure.
attachment and separation lines, with considerable spread. These high- With the turbulent boundary layer (Fig. 14b), the streamlines that
magnitude regions take place especially close to mid-height (𝑧∕𝐷 ≈ 0.5) enter the side region are deflected upward close to the cube, but the
for the side face and in the middle (𝑦∕𝐷 ≈ 0) for the top face, as the formation of the side vortex between the separation line and the main
shear layers are disrupted near the edges or the junction of the cube. attachment line on the side faces of the cube disrupts the streamline
On the side face of the cube with the laminar boundary layer (left panel trajectory observed for the laminar boundary layer case. Like the corner
of Fig. 13a), 𝑢=0 is high by the reverse-flow reattachment line, but not vortex, the side vortex is characterized by a counter-clockwise (orange)

13
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

laminar boundary layer with 𝛿∕𝐷 = 1, while Yakhot et al. (2006)

showed a mostly vertical arch vortex for a turbulent boundary layer
with 𝛿∕𝐷 = 1.5, which suggests that the inclination of the arch vortex
is most likely a consequence of the laminar or turbulent state of the
boundary layer. Other features include the higher turbulent kinetic en-
ergy for laminar boundary layers (Mohammadi et al., 2022), narrower
horseshoe vortices for turbulent boundary layers, and the development
of the side vortex, which appear to be mostly dependent on 𝛿∕𝐷. On
the other hand, most features are due to both 𝛿∕𝐷 and the state of
the boundary layer, which tend to change together (i.e., a turbulent
boundary layer typically has a larger thickness). Notable examples are
the length of the near-wake reverse flow region, which decreases for
larger 𝛿∕𝐷 and higher turbulence (Wang et al., 2006; Chen et al., 2023),
and the strength of the dipole structures, which has been shown to
decrease for both a thicker and a turbulent boundary layer due to
the disruption of the shedding process of coherent structures in the
wake (da Silva et al., 2022b; Mohammadi et al., 2022).

4. Conclusion

Large-eddy simulations (LES) of the flow around a surface-mounted

cube with Reynolds number Re = 1 × 104 have been carried out to
carefully investigate the time-averaged three-dimensional and near-
wall flow fields. In addition, effects of a laminar boundary layer with
thickness 𝛿∕𝐷 = 0.2 and a turbulent boundary layer with 𝛿∕𝐷 = 0.8
have been considered.
Fig. 14. Selected three-dimensional streamlines of the mean velocity field colored
One of the main contributions of this study pertains to the charac-
by the mean streamwise vorticity 𝜔𝑥 𝐷∕𝑈∞ for (a) the laminar boundary layer with
𝛿∕𝐷 = 0.2 and (b) the turbulent boundary layer with 𝛿∕𝐷 = 0.8. The limiting terization of the flow on and near the sides of the cube. Corner vortices
streamlines of the mean velocity field on the surface-mounted cube’s faces are included were found near the junction with the ground plane for both boundary
for reference. (For interpretation of the references to color in this figure legend, the layers, but the thicker boundary layer caused side vortices to develop in
reader is referred to the web version of this article.) addition. These side vortices were related to the presence of additional
separation and attachment lines on the side face of the cube, in addition
to the separation focus points characteristic of the corner vortices. The
streamwise vorticity. The present results show that the presence of the side vortices connected with the free end vortex over the cube’s free end
side vortex prevents the separated shear layers from getting directly en- forming the headband vortex, collocated with a low-pressure region in
trained into the near wake in a clockwise trajectory, which contributes the flow located around the cube, in contrast with the lower pressure
to weakening the formation of dipole structures for this case. being located in the near wake for the thinner boundary layer. The
The main features of the mean three-dimensional flow field around presence of the side and free end vortices for the thicker boundary layer
the surface-mounted cube with a laminar boundary layer with 𝛿∕𝐷 = affected the mean flow field characteristics in different ways. The lower
0.2 and a turbulent boundary layer with 𝛿∕𝐷 = 0.8 are presented in pressure related to the free end vortex caused the overall pressure on
Fig. 15, with the new features and characteristics identified in this the top face of the cube to be lower, which resulted in a higher normal
study included. The horseshoe vortex is found in both cases, yet it is force coefficient. In addition, the side vortex contains a streamwise
wider and farther from the cube for the laminar boundary layer. Corner vorticity of opposite sign to the dipole structures’ vorticity, which may
vortices are located near the junction of the cube with the ground interfere with the organization of these structures in the wake.
plane for both cases, but these vortices branch into the side vortices The mean three-dimensional flow field in the near wake of the cube
for the thick boundary layer and connect over the free end, forming has also been updated considering recent advances in the literature of
a headband vortex. Trailing-edge vortices appear due to reverse flow taller surface-mounted finite-height square prisms. Besides the corner
reattachment on the side and top faces of the cube, especially for the and side vortices, the two boundary layer cases presented a horseshoe
thin boundary layer. In both cases, the wake contains the arch or vortex upstream of the cube, dipole structures and ground-plane vor-
near-wake vortex, the dipole structures and a pair of ground plane ticity that extended into the far wake, and an arch vortex behind the
vorticity, but the flow structures are smaller or weaker in strength cube. The effects of the boundary layer were observed, however, for
for the turbulent boundary layer. A pair of vorticity regions near the all structures and on the near-wall flow field. The turbulent nature of
ground plane accompanies the dipole structures, and may combine with the boundary layer caused the horseshoe vortex to be located closer
the horseshoe vortex’s legs farther in the wake. to the cube, which, together with the larger boundary layer thickness,
The present flow model is representative of the flow around a increased the height of the front stagnation point. The ground plane
surface-mounted cube under varying boundary layers, as confirmed vorticity formed independently of the horseshoe vortex. It merged with
by comparisons of the identified features with the literature, and at the horseshoe vortex legs in the far wake, causing the appearance of
large Reynolds numbers. For Re < 1 × 104 , different structures may vortex cores adjacent to the dipoles, especially for the turbulent bound-
appear, as shown in Goswami and Hemmati (2022) for very small ary layer. The turbulent nature of the boundary layer also weakened the
Re. It is also expected that different combinations of the features of flow structures and gave the arch vortex a more upright form. With the
Fig. 15a and b may take place, depending on the flow conditions. While turbulent and thicker boundary layer, the downwash was weakened,
further investigation is necessary to separate the effects of the boundary the pressure showed greater variation on the top and side faces of
layer thickness and its laminar or turbulent nature, observations can the cube, which resulted in a larger normal force coefficient, and the
be made based on the existing literature. For example, Krajnović and overall pressure on the back face increased, which resulted in a smaller
Davidson (2001) obtained a downstream-inclined arch vortex for a drag force coefficient. The turbulent and thicker boundary layer also

14
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Fig. 15. Schematic of the main flow structures and features for the mean flow around a surface-mounted cube with (a) a laminar boundary layer with 𝛿∕𝐷 = 0.2 and (b) a
turbulent boundary layer with 𝛿∕𝐷 = 0.8.

caused the trailing-edge vortices, related to attachment of the reverse References

flow on the cube’s top and side faces, to decrease in size.
The present analysis has provided a comprehensive description of Baker, C.J., 1979. The laminar horseshoe vortex. J. Fluid Mech. 95 (2), 347–367.
http://dx.doi.org/10.1017/S0022112079001506.
the characteristics of the mean flow field around a surface-mounted
Ballio, F., Bettoni, C., Franzetti, S., 1998. A survey of time-averaged characteristics
cube, which, despite being widely studied, has had different inter- of laminar and turbulent horseshoe vortices. J. Fluids Eng. 120 (2), 233–242.
pretations and conjectures that required clarification. It contributes http://dx.doi.org/10.1115/1.2820639.
to knowledge about the flow around surface-mounted finite-height Behera, S., Saha, A.K., 2019. Characteristics of the flow past a wall-mounted finite-
length square cylinder at low Reynolds number with varying boundary layer
square prisms by reconciling information scattered in the literature, thickness. J. Fluids Eng. 141 (6), 061204. http://dx.doi.org/10.1115/1.4042751.
and also providing new insight into the relationship between the three- Bou-Zeid, E., Meneveau, C., Parlange, M., 2005. A scale-dependent Lagrangian dynamic
dimensional flow structures and the near-wall fields including varia- model for large eddy simulation of complex turbulent flows. Phys. Fluids 17 (2),
tions expected due to different boundary layer types. Having laid the 025105. http://dx.doi.org/10.1063/1.1839152.
Bourgeois, J.A., Sattari, P., Martinuzzi, R.J., 2011. Alternating half-loop shedding in the
foundation of the mean flow around a surface-mounted cube, the next
turbulent wake of a finite surface-mounted square cylinder with a thin boundary
step is to consider the characteristics of the dynamic flow and how it is layer. Phys. Fluids 23 (9), 095101. http://dx.doi.org/10.1063/1.3623463.
affected by different boundary layers. That is the goal of a future study Cao, Y., Tamura, T., Kawai, H., 2019. Investigation of wall pressures and surface flow
using the present validated dataset. patterns on a wall-mounted square cylinder using very high-resolution Cartesian
mesh. J. Wind Eng. Ind. Aerodyn. 188, 1–18. http://dx.doi.org/10.1016/j.jweia.
2019.02.013.
Cao, Y., Tamura, T., Zhou, D., Bao, Y., Han, Z., 2022. Topological description of near-
CRediT authorship contribution statement
wall flows around a surface-mounted square cylinder at high Reynolds numbers.
J. Fluid Mech. 933, A39. http://dx.doi.org/10.1017/jfm.2021.1043.
Castro, I.P., Robins, A.G., 1977. The flow around a surface-mounted cube in uniform
Barbara L. da Silva: Conceptualization, Formal analysis, Inves-
and turbulent streams. J. Fluid Mech. 79, 307–335. http://dx.doi.org/10.1017/
tigation, Methodology, Software, Validation, Visualization, Writing – S0022112077000172.
original draft, Writing – review & editing. David Sumner: Concep- Celik, I.B., Ghia, U., Roache, P.J., Freitas, C.J., Coleman, H., Raad, P.E., 2008.
tualization, Funding acquisition, Supervision, Validation, Writing – Procedure for estimation and reporting of uncertainty due to discretization in CFD
applications. J. Fluids Eng. 130 (7), 078001. http://dx.doi.org/10.1115/1.2960953.
review & editing. Donald J. Bergstrom: Conceptualization, Funding
Chapman, G.T., Yates, L.A., 1991. Topology of flow separation on three-dimensional
acquisition, Resources, Supervision, Writing – review & editing. bodies. Appl. Mech. Rev. 44 (7), 329–345. http://dx.doi.org/10.1115/1.3119507.
Chen, G., Li, X.-B., Sun, B., Liang, X.-F., 2022. Effect of incoming boundary layer
thickness on the flow dynamics of a square finite wall-mounted cylinder. Phys.
Declaration of competing interest Fluids 34 (1), 015105. http://dx.doi.org/10.1063/5.0076541.
Chen, G., Li, X.-B., Xue, R.-D., He, K., Wang, H., Liang, X.-F., 2023. Effects of incoming
free-stream turbulence on the flow dynamics of a square finite wall-mounted
The authors declare that they have no known competing finan- cylinder. Phys. Fluids 5.0137846. http://dx.doi.org/10.1063/5.0137846.
cial interests or personal relationships that could have appeared to Cianferra, M., Armenio, V., Ianniello, S., 2018. Hydroacoustic noise from different
influence the work reported in this paper. geometries. Int. J. Heat Fluid Flow 70, 348–362. http://dx.doi.org/10.1016/j.
ijheatfluidflow.2017.12.005.
Crane, R., Popinhak, A., Martinuzzi, R., Morton, C., 2022. Tomographic PIV investiga-
tion of vortex shedding topology for a cantilevered circular cylinder. J. Fluid Mech.
Data availability
931, R1. http://dx.doi.org/10.1017/jfm.2021.904.
da Silva, B.L., Chakravarty, R., Sumner, D., Bergstrom, D.J., 2020. Aerodynamic forces
Data will be made available on request. and three-dimensional flow structures in the mean wake of a surface-mounted
finite-height square prism. Int. J. Heat Fluid Flow 83, 108569. http://dx.doi.org/
10.1016/j.ijheatfluidflow.2020.108569.
da Silva, B.L., Hahn, D.G.H., Sumner, D., Bergstrom, D.J., 2022a. Mean wake and
Acknowledgments aerodynamic forces for surface-mounted finite-height square prisms of very small
aspect ratio. Phys. Fluids 34 (11), 115118. http://dx.doi.org/10.1063/5.0123259.
da Silva, B.L., Sumner, D., Bergstrom, D.J., 2022b. Mean and dynamic aspects of
This work was supported by the National Sciences and Engineering
the wakes of a surface-mounted cube and block. J. Fluids Eng. 144 (1), 011302.
Research Council of Canada (Discovery Grants Program numbers RG- http://dx.doi.org/10.1115/1.4051428.
PIN 2022-04981 and 2018-03760) and the University of Saskatchewan’s Depardon, S., Lasserre, J.J., Boueilh, J.C., Brizzi, L.E., Borée, J., 2005. Skin friction
Dean’s Scholarship program, Canada. Computational resources on the pattern analysis using near-wall PIV. Exp. Fluids 39 (5), 805–818. http://dx.doi.
org/10.1007/s00348-005-0014-8.
Plato cluster from Research Computing at the University of Saskatchewan
Devenport, W.J., Simpson, R.L., 1990. Time-dependent and time-averaged turbulence
and from the Digital Research Alliance of Canada are gratefully ac- structure near the nose of a wing-body junction. J. Fluid Mech. 210, 23–55.
knowledged. http://dx.doi.org/10.1017/S0022112090001215.

15
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Diaz-Daniel, C., Laizet, S., Vassilicos, J.C., 2017. Direct numerical simulations of a Martinuzzi, R.J., Havel, B., 2000. Turbulent flow around two interfering surface-
wall-attached cube immersed in laminar and turbulent boundary layers. Int. J. Heat mounted cubic obstacles in tandem arrangement. J. Fluids Eng. 122 (1), 24.
Fluid Flow 68, 269–280. http://dx.doi.org/10.1016/j.ijheatfluidflow.2017.09.015. http://dx.doi.org/10.1115/1.483222.
El Hassan, M., Bourgeois, J., Martinuzzi, R., 2015. Boundary layer effect on the Martinuzzi, R., Tropea, C., 1993. The flow around surface-mounted, prismatic obstacles
vortex shedding of wall-mounted rectangular cylinder. Exp. Fluids 56 (2), 33. placed in a fully developed channel flow. J. Fluids Eng. 115 (1), 85. http://dx.doi.
http://dx.doi.org/10.1007/s00348-014-1882-6. org/10.1115/1.2910118.
Escauriaza, C., Sotiropoulos, F., 2011a. Lagrangian model of bed-load transport in Meinders, E.R., Hanjalić, K., Martinuzzi, R.J., 1999. Experimental study of the local
turbulent junction flows. J. Fluid Mech. 666, 36–76. http://dx.doi.org/10.1017/ convection heat transfer from a wall-mounted cube in turbulent channel flow. J.
S0022112010004192. Heat Transfer 121 (3), 564–573. http://dx.doi.org/10.1115/1.2826017.
Escauriaza, C., Sotiropoulos, F., 2011b. Reynolds number effects on the coherent Melaku, A.F., Doddipatla, L.S., Bitsuamlak, G.T., 2022. Large-eddy simulation of
dynamics of the turbulent horseshoe vortex system. Flow Turbul. Combust. 86 (2), wind loads on a roof-mounted cube: application for interpolation of experimental
231–262. http://dx.doi.org/10.1007/s10494-010-9315-y. aerodynamic data. J. Wind Eng. Ind. Aerodyn. 231, 105230. http://dx.doi.org/10.
ESDU, 1978. Fluid Forces, Pressures and Moments on Rectangular Blocks, ESDU 71016, 1016/j.jweia.2022.105230.
Amendment (C), 01 Nov 1978. ESDU International PLC, London, UK. Meneveau, C., Lund, T.S., Cabot, W.H., 1996. A Lagrangian dynamic subgrid-scale
Goswami, S., Hemmati, A., 2022. Mechanisms of wake asymmetry and secondary model of turbulence. J. Fluid Mech. 319, 353–385. http://dx.doi.org/10.1017/
structures behind low aspect-ratio wall-mounted prisms. J. Fluid Mech. 950, A31. S0022112096007379.
http://dx.doi.org/10.1017/jfm.2022.824. Mohammadi, A., Morton, C., Martinuzzi, R.J., 2022. Effect of boundary layer state on
Gupta, V., Wan, M., 2019. Low-order modelling of wake meandering behind turbines. the wake of a cantilevered square cylinder of aspect ratio 4. Phys. Rev. Fluids 7
J. Fluid Mech. 877, 534–560. http://dx.doi.org/10.1017/jfm.2019.619. (8), 084702. http://dx.doi.org/10.1103/PhysRevFluids.7.084702.
Hajimirzaie, S.M., 2023. Experimental observations on flow characteristics around Moser, R.D., Kim, J., Mansour, N.N., 1999. Direct numerical simulation of turbulent
a low-aspect-ratio wall-mounted circular and square cylinder. Fluids 8 (1), 32. channel flow up to Re𝜏 = 590. Phys. Fluids 11 (4), 943–945. http://dx.doi.org/10.
http://dx.doi.org/10.3390/fluids8010032. 1063/1.869966.
Hearst, R.J., Gomit, G., Ganapathisubramani, B., 2016. Effect of turbulence on the wake Nakamura, H., Igarashi, T., Tsutsui, T., 2001. Local heat transfer around a wall-
of a wall-mounted cube. J. Fluid Mech. 804, 513–530. http://dx.doi.org/10.1017/ mounted cube in the turbulent boundary layer. Int. J. Heat Mass Transfer 44 (18),
jfm.2016.565. 3385–3395. http://dx.doi.org/10.1016/S0017-9310(01)00009-6.
Heng, H., Sumner, D., 2020. Wind loading of a finite prism: aspect ratio, incidence Okuda, Y., Taniike, Y., 1993. Conical vortices over side face of a three-dimensional
and boundary layer thickness effects. Wind Struct. Int. J. 31 (3), 255–267. http: square prism. J. Wind Eng. Ind. Aerodyn. 50, 163–172. http://dx.doi.org/10.1016/
//dx.doi.org/10.12989/was.2020.31.3.255. 0167-6105(93)90071-U.
Heng, H., Sumner, D., 2022. The mean pressure distribution on the free end of a
Paik, J., Escauriaza, C., Sotiropoulos, F., 2007. On the bimodal dynamics of the
square prism. Int. J. Heat Fluid Flow 96, 109005. http://dx.doi.org/10.1016/j.
turbulent horseshoe vortex system in a wing-body junction. Phys. Fluids 19 (4),
ijheatfluidflow.2022.109005.
045107. http://dx.doi.org/10.1063/1.2716813.
Hosseini, Z., Bourgeois, J.A., Martinuzzi, R.J., 2013. Large-scale structures in dipole
Puthan, P., Jalali, M., Ortiz-Tarin, J.L., Chongsiripinyo, K., Pawlak, G., Sarkar, S., 2020.
and quadrupole wakes of a wall-mounted finite rectangular cylinder. Exp. Fluids
The wake of a three-dimensional underwater obstacle: Effect of bottom boundary
54 (9), 1595. http://dx.doi.org/10.1007/s00348-013-1595-2.
conditions. Ocean Model. 149, 101611. http://dx.doi.org/10.1016/j.ocemod.2020.
Hunt, J.C.R., Abell, C.J., Peterka, J.A., Woo, H., 1978. Kinematical studies of the flows
101611.
around free or surface-mounted obstacles; applying topology to flow visualization.
Roache, P.J., 1994. Perspective: a method for uniform reporting of grid refinement
J. Fluid Mech. 86 (1), 179–200. http://dx.doi.org/10.1017/S0022112078001068.
studies. J. Fluids Eng. 116 (3), 405–413. http://dx.doi.org/10.1115/1.2910291.
Hunt, J.C.R., Wray, A.A., Moin, P., 1988. Eddies, Streams, and Convergence Zones
Saeedi, M., Wang, B.-C., 2015. Large-eddy simulation of turbulent flow and dispersion
in Turbulent Flows. Technical Report Technical Report No. CTR-S88, Center for
over a matrix of wall-mounted cubes. Phys. Fluids 27 (11), 115104. http://dx.doi.
Turbulence Research. Stanford University, CA, USA, pp. 193–208.
org/10.1063/1.4935112.
Hussein, H.J., Martinuzzi, R.J., 1996. Energy balance for turbulent flow around a
Saha, A.K., 2013. Unsteady flow past a finite square cylinder mounted on a wall at
surface mounted cube placed in a channel. Phys. Fluids 8 (3), 764–780. http:
low Reynolds number. Comput. & Fluids 88, 599–615. http://dx.doi.org/10.1016/
//dx.doi.org/10.1063/1.868860.
j.compfluid.2013.10.010.
Hwang, J.-Y., Yang, K.-S., 2004. Numerical study of vortical structures around a
Sakamoto, H., Arie, M., 1983. Vortex shedding from a rectangular prism and a circular
wall-mounted cubic obstacle in channel flow. Phys. Fluids 16 (7), 2382–2394.
cylinder placed vertically in a turbulent boundary layer. J. Fluid Mech. 126,
http://dx.doi.org/10.1063/1.1736675.
147–165. http://dx.doi.org/10.1017/S0022112083000087.
Iousef, S., Montazeri, H., Blocken, B., van Wesemael, P.J.V., 2017. On the use of
Sakamoto, H., Moriya, M., Taniguchi, S., Arie, M., 1982. The form drag of three-
non-conformal grids for economic LES of wind flow and convective heat transfer
dimensional bluff bodies immersed in turbulent boundary layers. J. Fluids Eng.
for a wall-mounted cube. Build Environ. 119, 44–61. http://dx.doi.org/10.1016/j.
104 (3), 326. http://dx.doi.org/10.1115/1.3241841.
buildenv.2017.04.004.
Ito, S., Okuda, Y., Kikitsu, H., Ohashi, M., Taniguchi, T., Taniike, Y., 2006. Experimental Sakamoto, H., Oiwake, S., 1984. Fluctuating forces on a rectangular prism and a circular
study on flow and pressure fields over the roof of a cube by PIV measurements. cylinder placed vertically in a turbulent boundary-layer. J. Fluids Eng. 106 (2),
In: Proceedings of the CWE2006. Yokohama, Japan. July 16-19, pp. 435–438. 160–166. http://dx.doi.org/10.1115/1.3243093.
Jeong, J., Hussain, F., 1995. On the identification of a vortex. J. Fluid Mech. 285, Schröder, A., Willert, C., Schanz, D., Geisler, R., Jahn, T., Gallas, Q., Leclaire, B.,
69–94. http://dx.doi.org/10.1017/S0022112095000462. 2020. The flow around a surface mounted cube: a characterization by time-
Khan, B.A., Saha, A.K., 2022. Turbulent flow and heat transfer characteristics of an resolved PIV, 3D shake-the-box and LBM simulation. Exp. Fluids 61, 189. http:
impinging jet over a heated wall-mounted cube placed in a cross-flow. Phys. Fluids //dx.doi.org/10.1007/s00348-020-03014-5.
34 (2), 025120. http://dx.doi.org/10.1063/5.0079956. Shah, K.B., Ferziger, J.H., 1997. A fluid mechanicians view of wind engineering: Large
Kindree, M.G., Shahroodi, M., Martinuzzi, R.J., 2018. Low-frequency dynamics in eddy simulation of flow past a cubic obstacle. J. Wind Eng. Ind. Aerodyn. 67–68,
the turbulent wake of cantilevered square and circular cylinders protruding a 211–224. http://dx.doi.org/10.1016/S0167-6105(97)00074-3.
thin laminar boundary layer. Exp. Fluids 59 (12), 186. http://dx.doi.org/10.1007/ Sousa, J., 2002. Turbulent flow around a surface-mounted obstacle using 2D-3C DPIV.
s00348-018-2641-x. Exp. Fluids 33 (6), 854–862. http://dx.doi.org/10.1007/s00348-002-0497-5.
Kozmar, H., 2021. Flow, turbulence and surface pressure on a wall-mounted cube Sumner, D., Rostamy, N., Bergstrom, D.J., Bugg, J.D., 2017. Influence of aspect ratio
in turbulent boundary layers. J. Wind Eng. Ind. Aerodyn. 210, 104503. http: on the mean flow field of a surface-mounted finite-height square prism. Int. J. Heat
//dx.doi.org/10.1016/j.jweia.2020.104503. Fluid Flow 65, 1–20. http://dx.doi.org/10.1016/j.ijheatfluidflow.2017.02.004.
Krajnović, S., 2011. Flow around a tall finite cylinder explored by large eddy simulation. Sumner, D., Unnikrishnan, S., Teng, M., Beitel, A., Das, A., Fulton, M., 2016. The mean
J. Fluid Mech. 676, 294–317. http://dx.doi.org/10.1017/S0022112011000450. wake of low-aspect-ratio surface-mounted finite-height square prisms and the effects
Krajnović, S., Davidson, L., 2001. Large eddy simulation of the flow around a three- of incidence angle. In: Proceedings of the 8th International Colloquium on Bluff
dimensional bluff body. In: Proceedings of the 39th Aerospace Sciences Meeting and Body Aerodynamics and Applications. Boston, Massachusetts, USA. June 7-11, p.
Exhibit. Reno, NV, USA. January 8-11, p. 0432. http://dx.doi.org/10.2514/6.2001- 65.
432. Unnikrishnan, S., Ogunremi, A., Sumner, D., 2017. The effect of incidence angle on
Li, J., Rinoshika, H., Han, X., Dong, L., Zheng, Y., Rinoshika, A., 2023. Evolution and the mean wake of surface-mounted finite-height square prisms. Int. J. Heat Fluid
control of multiscale vortical structures in a wall-mounted cube wake. Phys. Fluids Flow 66, 137–156. http://dx.doi.org/10.1016/j.ijheatfluidflow.2017.05.012.
35 (1), 015128. http://dx.doi.org/10.1063/5.0132761. Vinuesa, R., Schlatter, P., Malm, J., Mavriplis, C., Henningson, D.S., 2015. Direct
Liakos, A., Malamataris, N.A., 2014. Direct numerical simulation of steady state, three numerical simulation of the flow around a wall-mounted square cylinder under
dimensional, laminar flow around a wall mounted cube. Phys. Fluids 26 (5), various inflow conditions. J. Turbul. 16 (6), 555–587. http://dx.doi.org/10.1080/
053603. http://dx.doi.org/10.1063/1.4876176. 14685248.2014.989232.
Lim, H.C., Castro, I.P., Hoxey, R.P., 2007. Bluff bodies in deep turbulent boundary Wang, Y.Q., 2019. Effects of Reynolds number on vortex structure behind a surface-
layers: Reynolds-number issues. J. Fluid Mech. 571, 97–118. http://dx.doi.org/10. mounted finite square cylinder with AR = 7. Phys. Fluids 31 (11), 115103.
1017/S0022112006003223. http://dx.doi.org/10.1063/1.5123994.

16
B.L. da Silva et al. International Journal of Heat and Fluid Flow 106 (2024) 109288

Wang, F., Lam, K.M., 2019. Geometry effects on mean wake topology and large- Yakhot, A., Liu, H., Nikitin, N., 2006. Turbulent flow around a wall-mounted cube:
scale coherent structures of wall-mounted prisms. Phys. Fluids 31 (12), 125109. A direct numerical simulation. Int. J. Heat Fluid Flow 27 (6), 994–1009. http:
http://dx.doi.org/10.1063/1.5126045. //dx.doi.org/10.1016/j.ijheatfluidflow.2006.02.026.
Wang, H.F., Zhou, Y., Chan, C.K., Lam, K.S., 2006. Effect of initial conditions on Yauwenas, Y., Porteous, R., Moreau, D.J., Doolan, C.J., 2019. The effect of aspect ratio
interaction between a boundary layer and a wall-mounted finite-length-cylinder on the wake structure of finite wall-mounted square cylinders. J. Fluid Mech. 875,
wake. Phys. Fluids 18 (6), 065106. http://dx.doi.org/10.1063/1.2212329. 929–960. http://dx.doi.org/10.1017/jfm.2019.522.

17