1. Introduction
The phenomenon of flow separation and the onset of vortex shedding from a bluff body represent a fundamental problem in fluid mechanics (Williamson Reference Williamson1996; Zdravkovich Reference Zdravkovich1997). The near-wake structure of a bluff body is pivotal as it encapsulates the critical transformation of separated flow into shear layers that roll up into vortices. Numerous natural phenomena and engineering applications feature the complex dynamics of flow past bluff bodies such as flow around bridges, chimneys, marine risers, tall buildings, suspended cables, etc. Large-scale vortical structures are also often observed in the atmospheric wake of mountain cliffs and isolated islands (Etling Reference Etling2019). These atmospheric flows can be effectively studied using flowing soap films in a laboratory set-up, as illustrated by Seychelles et al. (Reference Seychelles, Amarouchene, Bessafi and Kellay2008) in their investigation of Jupiter’s atmospheric turbulence and the Great Red Spot.
The near-wake region of a bluff body, such as a circular cylinder, which encompasses the recirculation zone, plays a critical role in determining several important quantities, such as base pressure, wake width, drag coefficient and vortex shedding frequency. A thorough understanding of the near wake is therefore essential to develop a comprehensive understanding of the overall flow field and to further design effective flow control strategies. Extensive work has been carried out in the past, as documented in the literature, to explore and characterize the near-wake structure of a circular cylinder. Among the significant early contributions, the experimental studies of Kovasznay (Reference Kovasznay1949) and Taneda (Reference Taneda1956) provided foundational insights into the behaviour of the near-wake region during the onset of vortex shedding. Kovasznay (Reference Kovasznay1949) reported that vortices do not originate directly from the cylinder, but manifest downstream as an instability within the laminar wake. The initiation of this instability in the wake is linked to the strength and spacing of the shear layers at the end of the recirculation zone as indicated in the experimental work of Paranthoën et al. (Reference Paranthoën, Browne, Le Masson, Dumouchel and Lecordier1999). Taneda (Reference Taneda1956) used flow visualization to explore how the Reynolds number affects the mean recirculation region and downstream wake undulations. Balachandar, Mittal & Najjar (Reference Balachandar, Mittal and Najjar1997) further examined the characteristics of the mean recirculation zone across various bluff bodies, highlighting the significant contribution of Reynolds normal stress to the force equilibrium of the wake bubble. Complementary to these experimental studies, Jiang (Reference Jiang2020) employed direct numerical simulations to investigate the time-averaged flow separation location of a circular cylinder, observing that the instantaneous displacement of the separation point is governed by the dynamic interaction between the separating shear layer and another shear layer formed at the cylinder’s rear due to wake recirculation. These observations and studies highlight the profound influence of the near-wake structure on the evolution and characteristics of the overall flow field in the cylinder wake. The works mentioned here represent only a fraction of the extensive literature devoted to the near wake of circular cylinders, and the reader may refer to Roshko (Reference Roshko1954), Bloor (Reference Bloor1964), Gerrard (Reference Gerrard1966) and Williamson (Reference Williamson1996) for notable contributions.
Although the flow around bluff bodies has been extensively studied, a comprehensive understanding of its behaviour under two-dimensional experimental constraints remains limited, particularly in the case of soap films. Traditionally, most previous investigations of the onset of vortex shedding around bluff bodies have focused on water tunnels, open air jets and oil flows (Homann Reference Homann1936; Gerrard Reference Gerrard1978; Paranthoën et al. Reference Paranthoën, Browne, Le Masson, Dumouchel and Lecordier1999). In the past three decades, flowing soap films have proven to offer a valuable platform to investigate bluff-body wakes and the associated near-wake phenomena, but their unique and inherent properties, including variable kinematic viscosity, local thickness variations and surface tension effects, pose challenges that are absent in conventional water and wind tunnel experiments. In a typical bluff-body flow, a critical Reynolds number of approximately
$\textit{Re}_{c} = 47$
has been consistently reported in the literature for the transition from steady flow to periodic vortex shedding (Williamson Reference Williamson1989; Norberg Reference Norberg1994; Kumar & Mittal Reference Kumar and Mittal2006). However, this transitional behaviour has been largely under-reported in flowing soap films, with a few notable exceptions (Horváth et al. Reference Horváth, Cressman, Goldburg and Wu2000; Wen & Lin Reference Wen and Lin2001, Wu et al. Reference Wu, Wen, Yen, Weng and Wang2004; Roushan & Wu Reference Roushan and Wu2005a
); importantly, a detailed understanding of the near-wake structure at the onset of vortex shedding remains lacking, and addressing this gap is the central aim of the present study.
In the past three decades, numerous experiments have been conducted using flowing soap films to study the flow field around objects such as cylinders, filaments and foils (Schnipper, Andersen & Bohr Reference Schnipper, Andersen and Bohr2009; Chikkam & Kumar Reference Chikkam and Kumar2019; Yang, Masroor & Stremler Reference Yang, Masroor and Stremler2021). A key observation in these studies is the consistent use of fixed volume flow rate of the soap solution, as it is known to influence the average film thickness and, consequently, the kinematic viscosity of the film (Vorobieff & Ecke Reference Vorobieff and Ecke1999). With the flow rate fixed, a commonly adopted strategy to vary the flow Reynolds number is to adjust the length scale of the obstacle. The influence of the length scale of the obstacle on the surface kinematic viscosity, however, is not adequately addressed in the existing literature, with most studies using cylinder diameters ranging from a few hundred microns to a few millimetres. In addition, most studies keep the value of the flow rate fixed, a practice that, in our view, needs further investigation. In the present study, the length scale, which is the cylinder diameter, is held constant, and the flow rate is varied to change the Reynolds number. This method, as highlighted by Roushan & Wu (Reference Roushan and Wu2005b ), inherently influences the kinematic viscosity of the film and one of the objectives of the present study is to examine the near-wake behaviour of the cylinder while accounting for this dependency.
Studies focusing on the onset of vortex shedding remain limited in the context of flowing soap films, particularly regarding the evolution of the near-wake structure and its impact on
$\textit{Re}_c$
during the steady-to-unsteady transition. In the present work, we focus on the development of the near wake structure. Unlike in three-dimensional flows, an elongated recirculation bubble is observed in the present experiments, accompanied by small-scale vortices along its edges resulting from Kelvin–Helmholtz instability. These features, absent in three-dimensional flows, mark a distinct transition pathway from steady state to periodic vortex shedding. Through flow visualization and particle image velocimetry (PIV), we conduct a comparative analysis between soap film and water tunnel experiments; to the best of the authors’ knowledge, this study is the first of its kind and offers new insights to the subject of vortex shedding and bluff-body wakes.
Building on this, the present work addresses the following key questions.
-
(i) How does the near-wake structure evolve at the onset of vortex shedding in a two-dimensional soap film?
-
(ii) What is the mechanism behind the existence of an elongated recirculation bubble in soap films?
-
(iii) What is the role of small-scale vortices and the resulting intermittency in the near-wake evolution?
-
(iv) How do the transitional dynamics in soap films compare with those observed in conventional water tunnel experiments?
Following this introduction, § 2 describes the experimental set-up and methodology employed in the present study. Section 3 details the comparative results from both the soap film and water tunnel experiments. Finally, § 4 summarizes the main findings and conclusions.
Schematic of the flowing soap film experimental set-up.

2. Experimental set-up
The experiments were conducted in the Low Speed Aerodynamics Laboratory and the Fluid Dynamics Laboratory of the Department of Aerospace Engineering, Indian Institute of Technology Kanpur. The first set of experiments were conducted in a gravity-driven flowing soap film set-up and a further set of experiments were carried out in a recirculating water tunnel. The objective was to clearly distinguish between the two with respect to the transition from the steady to vortex shedding state.
2.1. Flowing soap film set-up
A gravity-driven flowing soap film facility as illustrated in figure 1 is used to conduct the flowing soap film experiments. The soap solution comprised 2 % by volume of commercially available dish washing liquid (Dettol) in water. A glass rotameter connected to the upper reservoir regulated the volume flow rate of the soap solution. The soap solution, from the outlet of the glass rotameter, traversed over a pair of nylon guide wires (diameter 1 mm) into the lower reservoir. The inclination of the soap film formed was 6.5°. The guide wires were supported by four pulling wires (diameter 0.8 mm) at the ends forming a 1.2 m-long test section. The width of the film was maintained at 12 cm and the cylinder model was placed 20D (where D is the cylinder diameter) downstream of the test section entrance. Throughout the experiments, readings on the rotameter were cross-validated by collecting the soap solution in a measuring beaker at the bottom reservoir for a period of 60 s in each case. The soap solution temperature was monitored at 23
$\pm$
1 °C with the aid of a thermometer. The experiments were conducted at night to minimize external disturbances. Illumination of the soap film was achieved using a monochromatic light source. The camera was fixed to a traverse mechanism, capturing images at right angles to the direction of the flowing soap film. The circular cylinder model used in the experiments was a 10 cm-long stainless-steel rod, 8 mm in diameter, which was painted black. The blockage ratio in the test section was 6.66 %. The blockage ratio in soap film experiments is defined as the ratio between the lateral dimension of the body, which is the diameter of the cylinder, and the width of the test section, which is the width of the soap film. The cylinder was positioned at the same angle as the soap film inclination, which is 6.5° to maintain a normal flow condition.
2.2. Water tunnel
A recirculating water tunnel (Model: 0710, Rolling Hills Inc.) powered by a 1.5 HP centrifugal pump is used in the present experiments. The set-up is illustrated in figure 2. The tunnel test section measures 0.25 m in depth, 0.46 m in length and 0.18 m in width. The sidewalls are made of 6 mm thick tempered glass that permits the light source to illuminate the designated visualization region. The upper part of the test section consists of a transparent Plexiglas (acrylic material) plate. The water level in the test section was maintained such that the top plate was in continuous contact with the free surface of the water. The required flow velocity in the test section was controlled using a digital interface. The water temperature within the tunnel was monitored at 23
$\pm$
1 °C with the aid of a thermometer. At the bottom of the test section, an anodized aluminium plate was positioned. The circular cylinder model used was a 280 mm-long stainless-steel rod, 8 mm in diameter, which was painted black. The blockage ratio in the test section was 4.44 %. In the water tunnel experiments, the blockage ratio is defined as the ratio between the lateral dimension of the body, which is the diameter of the cylinder, and the width of the test section.
Schematic of the water tunnel experimental set-up.

2.3. Flow visualization
2.3.1. Soap film experiments
Visualization in a soap film occurs due to the interference patterns formed upon incidence of a monochromatic light based on the local changes in the film thickness (Vorobieff & Ecke Reference Vorobieff and Ecke1999). The thickness variations are in turn due to local changes in flow velocity. In the present study, a low-pressure sodium vapour lamp (55 W General Electric SOX) is used as the light source. A large white plate served as a reflector, aiding in achieving a consistent light intensity across the film. A high-speed camera (Integrated Design Tools, Inc) was used to capture the images. The capture rate was 500 frames per second and the duration was 1 s. A 100 mm focal length lens (ZEISS 100 mm Makro–Planar) was used to capture the near-wake region which covered an area of 10D
$\times$
5D and a 50 mm focal length lens (Nikon NIKKOR AF 50 mm f/1.8D) was used to capture the midwake region which is 21D
$\times$
11D.
2.3.2. Water tunnel experiments
Flow visualization in the water tunnel was accomplished using the planar laser-induced fluorescence technique. The plane designated for visualization was exposed to a 3 W continuous laser, with the beam transformed into a sheet by a cylindrical lens housed in a collimator. The collimator facilitated adjustments to both the thickness and alignment of the sheet. A 42 MP digital camera (SONY – Alpha 7R III) recorded the wake structure at a rate of 60 frames per second, employing an 100 mm lens (ZEISS 100 mm Makro–Planar) to capture an area corresponding to a distance of 11D
$\times$
7D. Rhodamine-B solution served as the dye substance, and a pressurized apparatus was utilized in the dye injection system to continuously pump the dye liquid into the tunnel test section. Within the test section, an L-shaped stainless-steel hollow tube with a diameter of 0.9 mm was placed 15D upstream of the cylinder. The dye visualization apparatus is similar to the one used in earlier studies by Khan et al. (Reference Khan, Sunil, Bhattacharyya, Yadav, Poddar and Kumar2023) and Bhattacharyya et al. (Reference Bhattacharyya, Khan, Verma, Kumar and Poddar2022) in the same experimental set-up which may be referred to for more details.
2.4. Particle image velocimetry
Velocity field data of the wake are obtained through time averaged PIV measurements. The method employed in the present work for carrying out PIV experiments was similar for both soap film and water tunnel cases. Unlike the conventional approach of using a pulsed laser and a synchronized imaging system, a continuous laser was used in the present study along with a high-speed camera. Given the considerable difference in vortex shedding frequency in soap film and water tunnel test cases, the frame rate of the camera was set accordingly. A 5 W continuous laser (532 nm) and a high-speed camera (Integrated Design Tools, Inc.) were used to capture the images. The laser beam was transformed into a sheet using a cylindrical lens (TSI - 610026, focal length = −15 mm). It was housed in a collimator, serving to regulate both the thickness and alignment of the sheet.
The water tunnel was seeded with 10 μm diameter hollow glass spheres (Dantec Dynamics) and illuminated horizontally at the midsection of the cylinder with the laser sheet, which was focused to a thickness of less than 1 mm. The dimensions of the images captured were 2336
$\times$
1080 pixels corresponding to 8.5D
$\times$
4D in the physical space. The capture rate was set at 30 frames per second and 1800 images were captured for each case giving a time period of 60 s (observed to capture approximately 20 cycles of vortex shedding at Re = 150). In the case of soap film experiments, the soap solution was seeded with 5 μm polyamide seeding particles (Dantec Dynamics). The soap film was illuminated by the laser sheet which was aligned at the same angle which is 6.5°. The dimensions of the images captured were 2336
$\times$
848 pixels corresponding to 9.5D
$\times$
3.5D in the physical space. Since the vortex shedding frequency is relatively higher in the case of soap film, the frame rate was set at 1500 frames per second and 3000 images were captured giving a time period of 2 s (observed to capture approximately 28 cycles of vortex shedding at 31 ml min−1).
2.5. Flow field characterization
2.5.1. Velocity measurement
Time averaged PIV measurements were made in the tunnel without the model for both water tunnel and soap film cases. For the soap film experiments, an area corresponding to 7D cross-stream and 3D streamwise was used at the location of the cylinder. Images were captured at a rate of 733 frames per second for a duration of 3 s. For the water tunnel experiments, the captured area was 8.5D
$\times$
4D with a capture rate of 30 frames per second and a period of 10 s. These measurements were made for all the test cases considered in the present work and for each case, the value of velocity was obtained upon averaging.
2.5.2. Estimation of Reynolds number and kinematic viscosity
In the water tunnel experiments, determining the Reynolds number was straightforward: the kinematic viscosity was obtained from the measured water temperature, the flow velocity was determined using PIV and the cylinder diameter was known beforehand. By varying the flow velocity in the tunnel test section, a range of Reynolds numbers could be systematically explored.
In contrast, in soap film experiments, a direct determination of the Reynolds number is not feasible owing to the complex dependence of film properties on flow rate, thickness and local flow conditions. Consequently, an indirect approach originally proposed by Gharib & Derango (Reference Gharib and Derango1989) is typically employed. In this method, the vortex shedding frequency is first measured to obtain the Strouhal number, and the Reynolds number is then inferred from a known empirical Re–St relation. In the present work, the relation reported by Williamson & Brown (Reference Williamson and Brown1998) was adopted, expressed as
This expression has been widely used in previous soap film studies (e.g. Wu et al. Reference Wu, Wen, Yen, Weng and Wang2004; Chikkam & Kumar Reference Chikkam and Kumar2019; Yang & Stremler Reference Yang and Stremler2019; Yang et al. Reference Yang, Masroor and Stremler2021), and notably, the investigation by Chikkam & Kumar (Reference Chikkam and Kumar2019) employed the same experimental set-up as that used in the present work.
At a fixed soap solution flow rate of 31 ml min−1, using the corresponding free stream velocity and the known cylinder diameter, a Reynolds number of 167 was obtained. Using this value, the kinematic viscosity was calculated as
$28.8 \times 10^{-6}\,\mathrm{m}^2\,\mathrm{s}^{-1}$
. The vortex shedding frequency was determined from high-speed imaging by measuring the time taken for 12 pairs of vortices to traverse a location approximately 20D downstream of the cylinder.
A distinct feature of the present methodology, however, is the variation of the soap solution flow rate while keeping the cylinder diameter fixed; and implementing the above indirect approach under these conditions becomes challenging because the viscosity at each flow rate is not known a priori. From the current measurements, it was observed that the St values at different flow rates do not follow the expected trend commonly observed in water tunnel experiments, varying only slightly even as the inferred Reynolds numbers change substantially. This behaviour arises because, for a fixed cylinder diameter, it was observed that the decrease in free stream velocity with decreasing flow rate is far more pronounced than the corresponding decrease in vortex shedding frequency. This implies a reduction in the film’s effective viscosity, weakening the stabilizing effect of viscous damping and rendering the applicability of any Re–St relation strongly viscosity dependent.
In this context, it is also worth noting the study by Roushan & Wu (Reference Roushan and Wu2005a ), who demonstrated that a single Re–St relation,
can represent a broad range of Reynolds numbers across both soap film and bulk fluid experiments. However, when the above expression was applied to the present data with the constants A and B from their experimental data fits, significant deviation was observed in the magnitude of Re in our present case. This may likely be due to the constant viscosity condition and a different surfactant concentration in their study. In contrast, the present experiments exhibit considerable variation in viscosity with flow rate and film thickness, naturally leading to discrepancies from fixed viscosity correlations. While viscosity in bulk fluids is known a priori, its determination in soap films remains a persistent challenge because it varies dynamically with experimental conditions.
Only a limited number of studies have examined flowing soap films near the onset of vortex shedding, most notably those by Roushan & Wu (Reference Roushan and Wu2005a
) and Wu et al. (Reference Wu, Wen, Yen, Weng and Wang2004), both conducted at fixed flow rates. To the authors’ knowledge, Horváth et al. (Reference Horváth, Cressman, Goldburg and Wu2000) is the only study that varied the flow rate across the onset of vortex shedding, but at the same time, explicitly noting that ‘the mean velocity
$\overline {V}$
is used here rather than the Reynolds number because of an uncertainty concerning the value of the viscosity of the flowing soap film’. The present results reinforce this challenge, highlighting that the treatment of kinematic viscosity in flowing soap films remains poorly understood and needs further investigation.
Based on the measured St and Re values at a flow rate of 31 ml min−1, the kinematic viscosity is evaluated as
$28.8 \times 10^{-6}\,\mathrm{m}^2\,\mathrm{s}^{-1}$
. This viscosity value is then used to calculate the corresponding Reynolds numbers at lower flow rates using the measured free stream velocities, without employing the Re–St relation for each case. Moreover, at flow rates below 26 ml min−1, the wake exhibited intermittent vortex shedding, leading to unreliable and overestimated Reynolds numbers. When a state of steady flow was reached and vortex shedding ceased entirely, a critical Reynolds number,
$\textit{Re}_c$
, of 104 is obtained. To distinguish the indirectly determined Reynolds numbers from those obtained directly in the water tunnel experiments, they are denoted as
$\textit{Re}^*$
throughout this study. Table 1 summarizes the flow rates considered and the corresponding
$\textit{Re}^*$
values.
Volume flow rate of the soap solution and the corresponding effective Reynolds number (
$\textit{Re}^*$
).

An additional viewpoint on the assessment of viscosity in soap films is worth noting. Two approaches have been proposed in the literature to describe the effective viscosity of soap films: a three-dimensional formulation, (Vorobieff & Ecke Reference Vorobieff and Ecke1999) and a two-dimensional formulation (Prasad & Weeks Reference Prasad and Weeks2009). The applicability of these approaches depends directly on the film thickness. A soap film may be regarded as a thin liquid layer bounded by two surfactant layers, and the relative contribution of these surface layers strongly influences the effective kinematic viscosity. Vorobieff & Ecke (Reference Vorobieff and Ecke1999) demonstrated that this contribution is inversely proportional to the film thickness: for relatively thin films, the surface viscosity can reach values six to seven times higher than the bulk fluid viscosity. Regardless of the thickness, in experimental studies the viscosity is typically regarded as a three-dimensional quantity to ensure consistency with the use of the Reynolds number and the Re–St relationship. The two-dimensional formulation, on the other hand, is particularly useful in numerical simulations, where the viscosity must be assigned a priori. In such cases, an appropriate value based on the specified film thickness can be prescribed and normalized by the thickness to maintain dimensional consistency. This contrasts with the experimental approach, where the Re is first obtained from the measured St and the effective viscosity is then inferred from the Re–St relationship.
In the present experiments, the flow rate is used as the primary control parameter. Varying the flow rate simultaneously alters the film thickness and the free stream velocity, and for each value of flow rate a unique thickness-velocity pair is obtained. Consistent with the reasoning of Vorobieff & Ecke (Reference Vorobieff and Ecke1999), these pairs are treated as fixed, from which the vortex shedding frequency and the corresponding St could be determined. The effective viscosity is then inferred from the Re–St relationship and regarded as a three-dimensional quantity to ensure dimensional consistency with the definition of Re.
3. Results and discussion
A detailed examination of the experimental results obtained from both the soap film and water tunnel set-ups is presented in the following subsections. The analysis begins with a comparative study of flow visualization across the two set-ups, highlighting the differences in flow structures, near-wake development and vortex shedding characteristics. This is followed by the discussion of quantitative results including shear layer behaviour, which are analysed in detail using both PIV and high-speed imaging. Finally, a synthesis of these observations is presented to delineate the distinct near-wake behaviour observed in the soap film and water tunnel experiments, providing insights into the underlying mechanisms governing the onset of vortex shedding.
3.1. Flow visualization
Qualitative aspects of the wake structure are captured using flow visualization experiments. The primary objective of the present work is to explore the near-wake behaviour at the onset of vortex shedding in the flowing soap film, which is achieved by selecting a relatively broad range of flow rates of the soap solution, spanning from 17.5 to 32 ml min−1, which corresponds to
$\textit{Re}^*$
values from 89 to 172, respectively. For the experiments conducted in water tunnel, the velocity within the test section is adjusted to attain both steady flow state and vortex shedding state with Re values ranging from 30 to 150.
Figure 3 shows the wake structure variation with
$\textit{Re}^*$
in soap film experiments showing both near-wake and midwake regions. Further discussion focusing only on the near-wake structure is provided separately in figures 4 and 5. One of the primary observations in the present investigation is the presence of an elongated recirculation zone in the near-wake region at low flow rates (see figure 3). The recirculation bubble is observed to extend approximately 5–6D downstream, an observation that appears to be novel and potentially unique to soap film flows. A more detailed discussion regarding the possible mechanisms underlying this elongation is provided in the description of figure 4.
Variation of the wake structure with
$\textit{Re}^*$
in the soap film experiments, x / D = 0–20. Here
$\textit{Re}^*$
= 104–167. Refer to table 1 for corresponding flow rate values.

Variation of the near-wake region with
$\textit{Re}^*$
in the soap film experiments, x / D = 0–8. Here
$\textit{Re}^*$
= 104–167. Refer table 1 for corresponding flow rate values.

Zoomed-in view of the recirculation bubble during its mode shape transition. The bubble transitions from an elongated structure to a relatively shorter form.

Figure 3 shows the evolution of wake morphology in terms of the development of the vortex street. At a low flow rate of 20 ml min−1 (
$\textit{Re}^*$
= 104), a steady flow state is observed to exist in the wake, devoid of any signs of vortex shedding. Increasing the flow rate to 21 ml min−1 (
$\textit{Re}^*$
= 108) introduces slight undulations in the wake indicating the onset of unsteadiness. This instability progresses into intermittent bursts of vortices as the flow rate rises through 22–25 ml min−1 (
$\textit{Re}^*$
= 113–134). The recirculation bubble becomes more susceptible to instabilities giving rise to emergence of small-scale vortices along its edges due to Kelvin–Helmholtz instability (more clearly visible in figure 4). Intermittent vortex shedding is observed throughout the 22–25 ml min−1 range as a result of lateral deflection of the recirculation bubble. An interesting phenomenon is observed when the flow rate is increased slightly beyond 25 ml min−1 (
$\textit{Re}^*$
= 134): a mode shape change occurs in the near-wake structure, characterized by a significant reduction in the size of the bubble and a shift in the mechanism of vortex shedding. Vortices are observed to shed distinctly from the separating shear layers, in contrast to the earlier regime where their formation appeared to be influenced by the deflection of the recirculation bubble. The instability in the shear layers manifests into differential growth of the two vortices leading to alternate shedding into the wake as described by Gerrard (Reference Gerrard1966). The intermittency in vortex shedding also ceases at this flow rate. Further increase in flow rate decreases the recirculation bubble size and narrows the spacing between the wake vortices. With a shorter recirculation bubble, the Kelvin–Helmholtz instability loses effectiveness along the edges, and the vortices evolve in a more deterministic manner. Both the near and far wake regions resemble the von Kármán vortex street. This observation was made consistently across multiple repetitions, with the flow rate meticulously increased from 25 to 26 ml min−1. Upon further increase in flow rate beyond 26 ml min−1 (
$\textit{Re}^*$
= 141), no significant change is observed in near-wake structure, with only a slight reduction in the size of the recirculation region.
To gain more insight into the structure of the recirculation zone, the near-wake region is visualized in a higher resolution setting, as shown in figure 4. Expanding on earlier observations, an attempt is made to discuss the near-wake structure in detail, highlighting specific mechanisms responsible for both the pronounced elongation of the recirculation bubble and the emergence of small-scale vortices.
The elongated structure of the recirculation bubble as clearly observed in figure 4, at low flow rates, is likely a consequence of two characteristic properties of soap films: surface tension gradients and spatially varying surface kinematic viscosity. Prior studies have reported that the kinematic viscosity exhibits a strong dependence on the local film thickness (Vorobieff & Ecke Reference Vorobieff and Ecke1999; Prasad & Weeks Reference Prasad and Weeks2009). Experimental work by Vorobieff & Ecke (Reference Vorobieff and Ecke1999) demonstrated that regions of lower film thickness are associated with higher surface viscosity. In the present context, the shear layers bounding the recirculation bubble exhibit high velocity gradients and are correspondingly thinner and more viscous, which acts to dampen instabilities. This damping promotes more retention of fluid immediately downstream of the cylinder, which contributes to the persistence of the steady recirculation bubble. The fluid within the bubble is relatively thicker, less viscous, and has a lower average velocity due to reverse flow. Such behaviour is notably absent in traditional three-dimensional flows, where the viscosity is constant throughout the recirculation zone. The recirculation bubble is further characterized by the presence of small-scale vortical structures, resembling features of two-dimensional turbulence (Martin et al. Reference Martin, Wu, Goldburg and Rutgers1998).
Another pertinent characteristic of soap films is the presence of surface tension gradients. In the present investigation, local variations in film thickness give rise to corresponding gradients in the surfactant concentration. These concentration differences, in turn, induce surface tension gradients through the Marangoni effect (Kim & Mandre Reference Kim and Mandre2017; Sane, Mandre & Kim Reference Sane, Mandre and Kim2018). Specifically, the faster-moving, thinner shear layers that bound the recirculation zone, at relatively low flow rates, are characterized by a reduced surfactant concentration. Conversely, the comparatively thicker fluid within the recirculation bubble, being a region of slower or stagnant flow, possesses higher surfactant concentration. This difference in surface tension gives rise to Marangoni stresses which induce a surface-parallel flow across the shear layer, directed from the lower surface tension region (the thicker bubble interior) towards the higher surface tension region (the thinner, stretched shear layer). This Marangoni-driven flow serves as a self-restoring force by actively replenishing fluid towards the thinner region of the shear layer, thus helping to prevent film rupture. We propose that this continuous stabilization, working in conjunction with the elevated surface viscosity effect at the edges of the shear-layer, plays a key role in the formation of an elongated recirculation bubble and its remarkable persistence to higher Reynolds numbers than typically observed in traditional water tunnel experiments.
The structural changes occurring in the near-wake region with incremental increase in
$\textit{Re}^*$
are shown in figure 4. At 20 ml min−1 (
$\textit{Re}^*$
= 104), the recirculation region maintains a relatively stable configuration. Here, disturbances along the edges of the bubble are effectively dampened, resulting in a complete absence of vortex shedding in the wake. As the flow rate increases to 21 ml min−1 (
$\textit{Re}^*$
= 108) and beyond, Kelvin–Helmholtz instability sets in, driving the propagation of the small-scale vortices along the edges of the bubble. The unsteady and non-uniform nature of these vortices causes significant lateral deflection at the tip of the recirculation region, subsequently leading to the intermittent shedding of vortices. This phenomenon becomes increasingly pronounced as the flow rate approaches 25 ml min−1 (
$\textit{Re}^*$
= 134). At 25 ml min−1, an interesting transformation occurs in the near wake where even a slight increase in the flow rate induces significant changes in the recirculation bubble, as shown separately in a zoomed-in view of figure 5. As the inertial forces in the flow increase, a point is reached beyond which the elongated recirculation bubble can no longer be sustained. The enhanced damping provided by the thin shear layers becomes insufficient to counteract the growing instability, leading to the collapse of the bubble into a shorter region. Consequently, the recirculation bubble shortens and the small-scale vortices along its edges become weak or disappear altogether. This evolution follows the well-established regular von Kármán vortex street, in which the two shear layers on either side of the recirculation region interact and induce counter-rotation that effectively cuts off the vorticity of the forming vortices leading to their alternate shedding. The intermittency that characterized the earlier flow regime is eliminated entirely indicating a transition to more organized and periodic vortex shedding process. The evolution of small-scale vortices is discussed separately in more detail in § 3.4.
It is worth noting that the onset of vortex shedding in a soap film has previously been reported to exhibit hysteresis (Horváth et al. Reference Horváth, Cressman, Goldburg and Wu2000). To verify this behaviour in the present study, experiments were repeated by both increasing and decreasing the flow rate. It was observed that the sudden and pronounced change in the length of the recirculation bubble consistently occurred at a flow rate of 25.5 ml min−1. It is observed that the wake becomes completely steady only after a sequence of changes in the recirculation bubble, including the intermittency and the emergence of shear layer vortices. When focusing on the point where the recirculation bubble abruptly contracts, no noticeable hysteresis was observed. However, when the entire transitional regime is considered, a range of velocities exists in which the flow is neither fully steady nor exhibits a well-defined shedding frequency. Our assessment is that in the experiments of Horváth et al. (Reference Horváth, Cressman, Goldburg and Wu2000), the velocity was measured at a point which was 50D downstream of the model, and it is likely that the near-wake structure had an influence on their observation of hysteresis. It is plausible that the regime identified as ‘intermittent’ in the present study corresponds to the ‘hysteretic’ regime described by Horváth et al. (Reference Horváth, Cressman, Goldburg and Wu2000).
Following the observation of the distinctive near-wake behaviour in the soap film experiments, a corresponding investigation was carried out in the water tunnel to observe the key differences between the two scenarios. Figure 6 presents the instantaneous wake structures obtained at various Reynolds numbers. Figure 6(a) shows the wake structure at Re = 49. For Re
$\lt$
49, a clearly observable steady flow state was observed in the wake. With an increase in Re, the near-wake structure exhibits indications of shear layer instability, characterized by undulations of the dye streak. The recirculation region exhibits asymmetry, as noted by Taneda (Reference Taneda1963), who referred to it as the ‘Z’ structure appearance. Gerrard (Reference Gerrard1978) reported that such undulations exist in the wake in the range 55
$\lt$
Re
$\lt$
70. In this range, the wake with a waving oscillation at low Re transitions to a fully developed vortex street at higher Re. This is illustrated at Re = 65 in figure 6(b). An important observation made at this point is the absence of intermittency in the wake structure concerning the periodicity of the shed vortices. While waviness is evident in both soap film experiments and water tunnel tests, a notable difference arises: in water tunnel experiments, once vortex shedding is established, a consistent stream of vortices emanates from the separating shear layers. Conversely, in soap film experiments, the elongated near-wake structure contributes to both wake waviness and distinct vortices hence giving rise to intermittency. With further increase in Re to 80, a regular von Kármán vortex street is clearly observed in the wake. This behaviour prevailed for all the Re values further. The primary difference in wake structures observed at Reynolds numbers exceeding Re = 80 relates to variations in vortex shedding frequency and intervortex spacing. Another notable observation is the relatively smaller sized near-wake region in water tunnel experiments across all Reynolds numbers compared with the cases observed in soap film experiments. The difference in the size of the recirculation bubble in water tunnel and soap film experiments appears to be a noteworthy finding.
Variation of the wake structure with Reynolds number in the water tunnel experiments, x / D = 0–10. The number in the corner of the frame is the value of the Reynolds number based on the cylinder diameter.

3.2. Velocity distribution in the wake – PIV results
To complement the qualitative flow visualization results discussed above and to gain insight into the near-wake dynamics, PIV measurements in both soap film and water tunnel experiments were performed. The present objective was to evaluate the variation in the velocity field within the elongated recirculation bubble and their relation to intermittent vortex shedding. Figure 7 presents a comparison of the
${U_{\textit{mean}}}$
profiles between water tunnel and soap film cases at various downstream locations. The Reynolds numbers considered for this comparative study cover both the steady flow and the vortex shedding regimes, providing insights related to the transition. The profiles are plotted at three downstream locations (x / D = 2, 4 and 6). At x / D = 2, in the soap film case (refer to figure 7
a) the profiles are narrower compared with the water tunnel data and velocity overshoots are observed along the y-axis for all
$\textit{Re}^*$
values. In the water tunnel case (refer to figure 7
d), a consistent velocity deficit is evident across all Reynolds numbers, indicating the presence of low base pressure in the near wake regardless of Re. The peak deficit in the soap film case is notably higher for higher values of
$\textit{Re}^*$
whereas W-shaped profiles are observed at low
$\textit{Re}^*$
values. These W-shaped profiles are a manifestation of the standing eddies and suggest that the x / D = 2 is upstream of the vortex centres within the recirculation bubble. Moving downstream to x / D = 4 (see figure 7
b,e), the velocity profiles in the water tunnel case for Re = 150 and 100 show a reduction in the deficit, due to the widening of the wake resulting from fluid entrainment into its core. A similar trend is observed at
$\textit{Re}^*$
= 167 and 145 in the soap film cases. However, at lower values of
$\textit{Re}^*$
, the profiles remain within the recirculation bubble and a transition from W-shape to a parabolic shape can be noted indicating that the location x / D = 4 surpasses the position of the vortex centres within the recirculation bubble. At x / D = 6 in the downstream region (refer to figure 7
c, f), the wake broadens further and the eddies weaken due to viscous dissipation and diffusion, leading to a reduction in the deficit. In the soap film case it can be noted that the velocity profiles no longer have negative component and the width of the profiles begin to broaden.
Comparison of
${U_{\textit{mean}}}$
profiles at different x / D locations: soap film (a–c) versus water tunnel (d–f) experiments with parameters as
$\textit{Re}^*$
(a–c) and Re (d–f). The vertical lines represent the U = 0 condition. Also shown are the successive
${U_{\textit{mean}}}$
profiles in the near wake for one of the soap film cases at 21 ml min−1 (
$\textit{Re}^*$
= 108).

To further investigate the mean velocity distribution within the recirculation bubble, the variation of
$U_{\textit{mean}}$
recovery along the wake centreline is examined. Wake recovery is closely linked to base pressure and overall drag; an increase in the recirculation bubble length leads to a corresponding increase in base pressure and a resultant drag reduction Bearman (Reference Bearman1967). This phenomenon is generally observed as a delay in the recovery of the mean velocity along the centreline. Figure 8 presents the variation of
$U_{\textit{mean}}$
recovery along the wake centreline with
$Re$
and
$\textit{Re}^*$
for several representative cases from both the water tunnel and soap film experiments. The wake stagnation point, where the transition from negative to positive
$U_{\textit{mean}}$
velocity occurs along the wake centreline can be clearly observed. In the regular vortex shedding regime (Re = 150 for water tunnel and
$\textit{Re}^*$
= 167 for soap film cases, respectively), the wake stagnation point is located at x / D = 2 and 2.4 for the water tunnel and soap film cases, respectively. While the difference is not very noticeable, the dip in velocity within this region is significantly different for these two cases (
$U/U_{\infty }$
= −1.5 and −3.8). The substantial deficit in
$U_{\textit{mean}}$
in the soap film case is due to the longer recirculation bubble, resulting in higher acceleration of the fluid towards the rear stagnation point of the cylinder. At lower Re values close to
$\textit{Re}_{c}$
, distinct wake stagnation points are observed for the soap film and water tunnel cases. At Re = 53, the wake stagnation point occurs at x / D = 3 in the water tunnel case, which is consistent with data from Paranthoën et al. (Reference Paranthoën, Browne, Le Masson, Dumouchel and Lecordier1999) and Qu et al. (Reference Qu, Norberg, Davidson, Peng and Wang2013). In the soap film case, two curves are shown in the figure representing each, the beginning and end of the transition regime, namely,
$\textit{Re}^*$
= 104 and 134. The wake stagnation points differ considerably, ranging from x / D = 4.6 up to 6.15. The velocity recovery rate is lower due to the presence of an elongated recirculation bubble.
Recovery of
${U_{\textit{mean}}}$
and a closer look at the back flow velocity variation across the wake stagnation point. Blue and red curves represent water tunnel and soap film data, respectively. Data from Paranthoën et al. (Reference Paranthoën, Browne, Le Masson, Dumouchel and Lecordier1999) and Qu et al. (Reference Qu, Norberg, Davidson, Peng and Wang2013) are included for reference.

Following the analysis of mean velocity variations, a more detailed investigation was undertaken to examine the instantaneous velocity fluctuations in the soap film experiments. The occurrence of intermittency necessitated a closer examination of the fluctuating velocity field, as both intermittency and the instantaneous position of the recirculation bubble influence the onset of vortex shedding. The instantaneous variations of streamwise (
${U}$
) and cross-stream (
${V}$
) velocity components, along with the power spectrum of the temporal fluctuations, are extracted and plotted as shown in figures 9 and 10. A point downstream (
$x/D = 8.0$
,
$y/D = 0.5$
) was considered, and the time series of instantaneous velocity was extracted. It can be observed that both the extent and the nature of fluctuations in
${U}$
and
${V}$
differ significantly at lower and higher values of
$\textit{Re}^*$
. The time series data clearly demonstrates the evolution of unsteadiness with increasing flow rate. At lower
$\textit{Re}^*$
, the velocity signals remain largely steady, indicating the absence of vortex shedding. As
$\textit{Re}^*$
increases to moderate values, intermittent oscillations begin to appear, particularly in the transverse velocity component,
${V}$
, reflecting irregular and weak vortex shedding associated with early-stage shear layer instabilities. These signals result from intermittent lateral deflections of the tip of the recirculation bubble. At higher
$\textit{Re}^*$
, the signals exhibit strong periodicity and increased amplitude, corresponding to well-established, regular vortex shedding. These observations support the interpretation that the flow in the soap film experiments transitions from steady to intermittent and eventually to fully periodic behaviour.
Instantaneous velocity variation with
$\textit{Re}^*$
. Both streamwise and cross-stream velocity components are shown. The streamwise location is x / D = 8; y / D = 0.5.

Power spectrum density (PSD) of the fluctuating velocity components at different
$\textit{Re}^*$
. The streamwise location is x / D = 8; y / D = 0.5.

The frequency content of the wake, represented by the power spectrum of the velocity fluctuations, is shown in figure 10. At lower values of the flow rate (
$\textit{Re}^*$
= 104), the spectrum displays multiple peaks across a broad range of frequencies, which is indicative of the absence of well-defined periodic vortex shedding. In this steady flow regime, there is nearly a constant spectral energy density across frequencies, reflecting the lack of coherent structures. As the
$\textit{Re}^*$
increases, the flow transitions towards a more organized shedding regime, and this transition is reflected in the frequency spectrum: a single, dominant peak emerges, signifying the presence of regular, periodic vortex shedding at a characteristic frequency. The sharpening and increasing amplitude of this peak with increasing
$\textit{Re}^*$
further confirms the growing coherence and strength of the vortex shedding as the flow evolves.
3.3. The recirculation bubble length
The recirculation bubble and its variation with Reynolds number are examined with the objective of distinguishing between the water tunnel and soap film datasets particularly in the context of onset of vortex shedding. The goal was to determine and validate the
$\textit{Re}_c$
value by first ensuring the consistency and reliability of the water tunnel measurements through comparison of the values obtained with the existing literature. Figure 11 shows the variation of recirculation bubble length (
$L_r$
) with Reynolds number for the water tunnel data. The x-axis in figure 11 is expressed as the difference between
$\textit{Re}_c$
and the actual Reynolds number. Here
$L_r$
is defined as the distance along the wake centreline from the cylinder axis to the point where the mean streamwise velocity (
$U_{\textit{mean}}$
) changes from negative to positive. This definition is consistent with those employed by Paranthoën et al. (Reference Paranthoën, Browne, Le Masson, Dumouchel and Lecordier1999) and Qu et al. (Reference Qu, Norberg, Davidson, Peng and Wang2013). Experiments are carried out across Reynolds numbers ranging from 30 to 150, with particular emphasis on the interval between 30 and 60. The length of the recirculation bubble is maximum when
$Re$
equals
$\textit{Re}_c$
(Paranthoën et al. Reference Paranthoën, Browne, Le Masson, Dumouchel and Lecordier1999; Taneda Reference Taneda1956, Qu et al. Reference Qu, Norberg, Davidson, Peng and Wang2013). In the present experiments, the
$Re$
value corresponding to the peak in
$L_r$
is obtained as 53. It is observed that the size of the recirculation bubble increases with
$Re$
in the steady flow state of the wake and reaches a peak value of 2.42 at the onset of vortex shedding. Subsequently, as Reynolds number increases,
$L_r$
decreases progressively. The present experimental data closely align with the findings of Paranthoën et al. (Reference Paranthoën, Browne, Le Masson, Dumouchel and Lecordier1999) and validate the value of
$L_r$
reported by Qu et al. (Reference Qu, Norberg, Davidson, Peng and Wang2013). In the vortex shedding regime, the data agree well with the experimental work of Nishioka & Sato (Reference Nishioka and Sato1978). Also shown is the experimental work of Taneda (Reference Taneda1956) for reference.
Variation of the length of the recirculation bubble with Reynolds number in water tunnel experiments.

Figure 12 shows the comparison of recirculation bubble length variation between soap film and water tunnel test cases from the time averaged PIV data. The variables used in figure 12 are the calculated values of
$\textit{Re}^*$
and Re for soap film and water tunnel cases, respectively. The data points for the soap film cases are presented along with their standard deviation values to assess the impact of intermittency on the recirculation bubble. A notable contrast is evident in the magnitude of
$L_{r}$
and its abrupt change in the soap film experiments. From figure 12, three distinct zones or regions can be identified in the soap film curve, a steady flow state (flow rate
$\lt$
21 ml min−1;
$\textit{Re}^*$
$\lt$
108), a transition state (flow rates between 21 and 26 ml min−1; 108
$\lt$
$\textit{Re}^*$
$\lt$
141) and the regular vortex shedding state (flow rates
$\gt$
26 ml min−1;
$\textit{Re}^*$
$\gt$
141). At relatively higher flow rates, the variation in
$L_{r}$
is minimal, corresponding to a short recirculation region. A sudden increase in
$L_{r}$
occurs when the flow rate is reduced from 26 to 25 ml min−1 (
$\textit{Re}^*$
reduced from 141 to 134), signifying a mode shape change in the near-wake structure. Between flow rates of 24 and 20 ml min−1 (
$\textit{Re}^*$
between 127 and 104),
$L_{r}$
gradually increases with decreasing flow rate, reaching a peak value of 6.15 at 20 ml min−1 (
$\textit{Re}^*$
= 104). This regime corresponds to the transition from regular vortex shedding mode to the steady flow mode. A higher standard deviation is evident in this regime, indicating increased variability that is attributed to intermittency and its effect on the average length of the recirculation bubble. For flow rates
$\lt$
20 ml min−1 (
$\textit{Re}^*$
$\lt$
104), intermittency becomes negligible resulting in a steady flow state in the wake. Further reduction in flow rate beyond this point leads to a decrease in
$L_{r}$
, as evident in the figure at 17.5 ml min−1 (
$\textit{Re}^*$
= 89).
Comparison of the recirculation bubble length,
$L_{r}$
in water tunnel and soap film experiments. The inset shows the
$U_{\textit{mean}}$
contour and the recirculation bubble length for one of the soap film cases.

Instantaneous variation of the recirculation bubble length at
$\textit{Re}^*$
= 87, 120 and 167.

The instantaneous variation of the recirculation bubble length is presented in figure 13. Three representative cases are examined: one within the steady flow regime (
$\textit{Re}^*$
= 87), one in the intermittent regime (
$\textit{Re}^*$
= 120) and one in the fully developed vortex shedding regime (
$\textit{Re}^*$
= 167). At
$\textit{Re}^*$
= 87, the recirculation bubble tip exhibits minimal fluctuations, and the length remains nearly constant. Conversely, at
$\textit{Re}^* = 120$
, substantial scatter is observed in the instantaneous length due to the intermittency associated with the recirculation bubble tip. At
$\textit{Re}^*$
= 167, the periodic vortex shedding induces a more deterministic oscillation of the bubble length, corresponding to the periodic extension and retraction of the bubble tip. The recirculation bubble length is also notably reduced in this regime, consistent with the mode shape shift discussed earlier.
3.4. Evolution of the near wake – a closer look
In this section, a detailed examination of the near-wake evolution is presented to describe the complex transitional behaviour observed in the soap film experiments. In addition, the small-scale vortices and their evolution is also discussed subsequently in detail.
Figure 14 shows the various stages of development within the recirculation region during the onset of vortex shedding. A schematic representation of these stages is provided on the right-hand side of the figure for illustration. Stage 1 corresponds to a state of complete flow steadiness. It is characterized by two ‘standing’ vortices in the recirculation bubble and absence of any unsteadiness in the wake. The flow rate maintained in the current experiments to achieve this state is below 20 ml min−1 (specifically 17.5 ml min−1 as depicted in figure 14) which corresponds to
$\textit{Re}^* \lt 104$
. Upon marginally increasing the flow rate beyond 20 ml min−1, a transition occurs within the recirculation bubble, shifting it from the ‘standing vortices’ state to a state characterized by two-dimensional turbulence. This transition is depicted as Stage 2 in figure 14. The length of the bubble slightly increases and the disturbances emerging along the periphery of the recirculation bubble are effectively suppressed, impeding the formation of vortices. The tip of the recirculation bubble, however, undergoes mild transverse oscillations and consequently the wake, although devoid of vortex structures, exhibits waviness. With further increase in flow rate to approximately 22 ml min−1 (
$\textit{Re}^*$
= 113), small-scale vortices emerge along the edges of the recirculation bubble. This corresponds to Stage 3 in figure 14. These vortices roll up as a result of Kelvin–Helmholtz instability in the shear layers. The occurrence frequency of these vortices is low, and for the most part, the recirculation bubble resembles that of Stage 2. Occasionally, these small-scale vortices merge to form a significantly larger vortex, which persists downstream into the wake, contributing to the intermittent nature of the vortex street in the far wake. Additionally, the average length of the recirculation bubble slightly decreases. Upon further increasing the flow rate from 22 ml min−1 to 25 ml min−1 (increasing
$\textit{Re}^*$
from 113 to 134), the frequency of the small-scale vortices increases and coalescence occurs more predominantly. The intermittency in the wake vortices increases as a result and the average length of the bubble reduces further. This is referred to as Stage 4 in figure 14. When the flow rate reaches a critical threshold (25.5 ml min−1 in the present case), a notable transformation in the shape and dimensions of the recirculation bubble takes place. The near-wake region no longer hosts the small-scale vortices and appears shortened. The intermittency of wake vortices diminishes to zero, resembling a typical von Kármán vortex street. Essentially, the shear layers are observed to separate from the recirculating region and evolve into vortices. This evolution is illustrated as Stage 5 in figure 14.
Schematic of the developmental stages of the recirculation zone at the onset of vortex shedding.

Following the detailed account of the near-wake evolution, an important aspect that is evident is the development of small-scale vortical structures along the recirculation bubble boundary. The presence of these vortices happens to be one of the most salient and clearly identifiable features observed in this investigation. The occurrence of such shear layer vortices in the recirculation region of a bluff body were first documented by Bloor (Reference Bloor1964) and have since been extensively studied in the past (Unal & Rockwell Reference Unal and Rockwell1988; Wu et al. Reference Wu, Sheridan, Hourigan and Soria1996; Prasad & Williamson Reference Prasad and Williamson1997). Often referred to as Bloor–Gerrard (BG) vortices by some researchers (Unal & Rockwell Reference Unal and Rockwell1988; Prasad & Williamson Reference Prasad and Williamson1997; Khor, Sheridan & Hourigan Reference Khor, Sheridan and Hourigan2011), these structures typically emerge at the critical Reynolds number (
$\textit{Re}_{\textit{crit}}$
) ranging from approximately 1000–3000. In all the earlier studies, the near-wake structure was observed to comprise of a relatively short (approximately 2.5D) recirculation zone with the BG vortices observed along the edges of the shear layers. In the present study, these vortices (named as small-scale vortices to distinguish them from the BG vortices from the literature) emerge in a completely different range of Reynolds number (close to the
$\textit{Re}_{c}$
). As described earlier, the recirculation zone undergoes transition from steady state to the vortex shedding state via the presence of these small-scale vortices unlike previous observations where vortex shedding begins at
$\textit{Re}_{c}$
and BG vortices appear at
$\textit{Re}_{\textit{crit}}$
. We believe this novel finding can offer new insights into understanding two-dimensional transition in bluff-body wakes.
In previous studies (Bloor Reference Bloor1964; Unal & Rockwell Reference Unal and Rockwell1988; Wu et al. Reference Wu, Sheridan, Hourigan and Soria1996; Prasad & Williamson Reference Prasad and Williamson1997), the frequency of shear layer vortices,
$f_{\textit{BG}}$
or
$f_{sl}$
, is typically normalized by the Kármán vortex shedding frequency,
$f_v$
, and expressed as a power law function of Reynolds number (Prasad & Williamson Reference Prasad and Williamson1997) (see 3.1):
However, in the present investigation, the intermittency at the onset of vortex shedding complicates normalization using
$f_v$
. Instead, the small-scale vortex frequency,
$f_{ss}$
, is non-dimensionalized by cylinder diameter,
$D$
, and free stream velocity,
$U_{\infty }$
. Figure 15 shows the variation of the non-dimensional frequency,
$ (f_{ss} D / U_{\infty } )$
and wavelength,
$ (\lambda _{ss} / D )$
of the small-scale vortices with
$\textit{Re}^*$
. The results show a quadratic relationship between
$f_{ss}$
and
$\textit{Re}^*$
up to a threshold of 25.5 ml min−1 (
$\textit{Re}^*$
= 138), after which
$f_{ss}$
drops to zero. For flow rates below 20 ml min−1 (
$\textit{Re}^*$
= 104), the edges of the recirculation bubble remain steady, damping the small-scale vortices. The experimental data is approximated by a second-order polynomial represented by
The small-scale vortex wavelength,
$\lambda _{ss}$
, normalized by
$D$
, is also shown in figure 15. Assuming the convective velocity to be
$U_{sl}$
= (
$U_1 + U_2$
)/2, where
$U_1$
and
$U_2$
are velocities on the high- and low-speed sides of the shear layer, we found
$U_{sl} \approx 0.62 \times U_{\infty }$
and
$U_{2}$
$\approx$
0, following Wu et al. (Reference Wu, Sheridan, Hourigan and Soria1996). As
$\textit{Re}^*$
increases,
$f_{ss}$
rises faster than
$U_{sl}$
, which leads to a decline in
$\lambda _{ss}$
as approximated by the polynomial fit
Variation of the average time interval of small-scale vortices with
$\textit{Re}^*$
including the standard deviation bars.

In order to further characterize the temporal behaviour of the small-scale vortices, the time intervals between their successive formation and advection were examined. Figure 16 presents the variation of these intervals for vortices that develop along the shear layers at the recirculation bubble edge. The data were extracted through visual inspection of high-speed image sequences by recording the time difference between the formation of consecutive small-scale vortices. In the intermittent regime, a wide distribution of time intervals is observed, indicating a high degree of randomness in the vortex formation process. As
$\textit{Re}^*$
increases, the standard deviation of these intervals decreases along with reduction in the time interval between successive vortices, reflecting a more continuous stream of small-scale vortices and an increased frequency.
Before making the concluding remarks, it is worth mentioning that the complex, richer transitional behaviour observed in the soap film experiments, deviates from the canonical understanding of instability in bluff-body wakes. The onset of vortex shedding behind a circular cylinder has been extensively studied in the framework of bifurcation theory (Sreenivasan, Strykowski & Olinger Reference Sreenivasan, Strykowski and Olinger1987; Huerre & Monkewitz Reference Huerre and Monkewitz1990), where the transition from steady to unsteady flow is commonly described by the Stuart–Landau equation (Provansal, Mathis & Boyer Reference Provansal, Mathis and Boyer1987). However, the present soap film experiments reveal a more complex transition, extending beyond the conventional Hopf bifurcation. This highlights the need to revisit the vortex shedding phenomenon in two-dimensional soap films from the perspective of bifurcation theory to explore the driving factors of this richer transitional behaviour.
4. Conclusion
This study primarily investigates the mechanisms underlying the evolution of the near-wake structure of a circular cylinder during the onset of vortex shedding in a flowing soap film. Our findings indicate that the transition from steady flow to periodic vortex shedding is not abrupt. Rather, it proceeds through a gradual pathway involving the emergence of shear layer instabilities, formation of small-scale vortices and intermittent behaviour of the recirculation bubble. It emphasizes the intricate transition from steady flow to regular vortex shedding and the significant impact of the recirculation bubble. Two key findings emerge from this investigation: first, the identification of an elongated recirculation bubble; and second, the emergence of small-scale vortices during the onset of vortex shedding. It is proposed that a continuous stabilization mechanism, wherein, the surface tension gradients, working in conjunction with the elevated surface viscosity effects at the edges of the shear layer, is responsible for the elongated shape of the recirculation bubble and its remarkable persistence to higher Reynolds numbers than typically observed in traditional water tunnel experiments. Furthermore, in contrast to three-dimensional flows, where shear layer vortices generally develop in the range
$1000 \lt Re \lt 3000$
, our findings show that such vortices emerge in the
$\textit{Re}_{c}$
range, suggesting a novel flow behaviour characteristic of soap films. Moreover, the presence of turbulent-like structures within the elongated recirculation bubble is observed to be a distinctive feature of soap film flows.
Further investigation is needed to fully elucidate the precise causes and interdependencies underlying these phenomena, particularly the role of kinematic viscosity and local thickness variations in the soap film in governing the dynamics of the near wake.
Acknowledgements
The authors are grateful to Professor K. Poddar and the support staff of the Low Speed Aerodynamics Lab of the Department of Aerospace Engineering, IIT Kanpur for their help.
Declaration of interests
The authors report no conflict of interest.



























