WorksPhysics and Astronomy
Numerical Study of Viscoelastic Upstream Instability
Abstract
In this work, we report numerical results on the flow instability and bifurcation of a viscoelastic fluid in the upstream region of a confined cylinder in a narrow channel. Two-dimensional direct numerical simulations based on the FENE-P model (the finite-extensible nonlinear elastic model with the Peterlin closure) are conducted with numerical stabilization techniques. Our results show that the macroscopic viscoelastic constitutive relation can capture the viscoelastic upstream instability reported in previous experiments for low-Reynolds-number flows. The numerical simulations reveal that the non-dimensional recirculation length () is affected by the cylinder blocking rate (), the Weissenberg number (), the viscosity ratio (), and the maximum polymer extension (). Close to the onset of upstream recirculation, depending on the values of and , the bifurcation can be supercritical or subcritical. has a nonlinear influence on the upstream recirculation length. This work contributes to our theoretical understanding of this new instability mechanism in viscoelastic wake flows.
Introduction
A mixture of a Newtonian fluid and high-molecular-weight polymers (even at extremely low concentration) exhibits viscoelasticity. In the low-Reynolds-number range, when the elasticity becomes the dominant source of nonlinearity, this flow can manifest many interesting and novel flow phenomena that have not been studied extensively, especially on flow instabilities, such as symmetry breaking ([4, 55, 29, 26]), secondary flow ([83, 17]), time dependency and even elastic instability and turbulence ([63, 39, 23, 24, 62, 22, 67] (Varshney & Steinberg 2019)). These instability phenomena widely exist in processes ranging from plastic manufacturing [18, 74] to human blood plasma [11, 70] and porous media flows [78, 32]. However, detailed understandings of the underlying mechanisms in many instances remain vague. In these flows, because the extra elastic stress plays a key role, it is essential to know its value and how it influences the flow field. Experimental measurement and numerical simulation are the two main approaches to obtain the elastic stress. Compared to a large number of reported experimental studies with regards to elastic instability phenomena, numerical method is applied relatively less [54] to this problem and deserves to be employed more frequently because of its advantages such as more feasible parametric studies and direct analysis of the elastic stress field. However, a notoriously difficult problem in numerical simulation of viscoelastic flow has challenged researchers for decades, namely, the numerical instability when simulating flow at high Weissenberg number (), which is called the high Weissenberg number problem (HWNP). The reader is referred to the recent review paper by Alves et al. [2] for a more complete account on these issues. In the following, we will focus on explaining an idealized flow model, that is, viscoelastic flow around a circular cylinder and the recent studies on its upstream instability.
Viscoelastic flows past bluff bodies are frequently encountered in industrial applications (e.g., filtration processes and oil extraction) and natural phenomena (e.g., flow in the soil and blood flow in cardiovascular valves and brain tissues) [40, 34, 43]. In addition, porous media are frequently modelled by ordered and disordered arrays of microfluidic circular cylinders ([78, 32]. Placing one or more cylinders in channel or pipe inlet section is often used as a disturbance source to study the channel viscoelastic instability in experiments ([58] (Varshney & Steinberg 2018). Viscoelastic flow around a circular cylinder has also been regarded as a benchmark case for numerical and experimental studies ([72, 19]. In Newtonian wake flows, a downstream recirculation zone and vortex shedding can be observed because of the global instability in this flow ([81, 7, 66, 69]. However, in viscoelastic wake flows, interestingly, an upstream recirculation can form in cylinder wake flows confined between two plane plates.
The recirculation in front of circular cylinder was firstly reported by Kenney et al. [37]. Later, Shi et al. [65], Zhao et al. [84], Qin et al. [57], Haward et al. [27] and Hopkins [30] successively studied this phenomenon experimentally. These experiments were performed using Boger solutions, which are elastic without shear thinning, such as dilute polyethylene oxide (PEO) or Polyacrylamide (PAA) solutions, or using wormlike fluids with a strong shear thinning behaviour, such as cationic surfactant cetyltrimethylammonium bromide (CTAB) and stable hydrotropic salt 3-hydroxy naphthalene-2-carboxylate (SHNC) solutions. Regarding the geometry of the experimental setup, the blockage ratio (), i.e., the ratio of cylinder diameter to channel width, in these experiments are not less than 50% and the depth-diameter ratio () is no more than 1.2. Zhao et al. [84] summarized the flow patterns in the space with five states: Newtonian like state, bending streamlines, vortex growth upstream, unsteady downstream, and 3D time-dependent chaotic upstream, which are shown in Figure 1. Qin et al. [57] found that this flow is inherently three-dimensional and observed symmetry breaking as well as strong upstream propagation effects via elastic waves. At low (), there also exists a mild upstream instability, with only streamline bending, which was reported by Ribeiro et al. [60], Nolan et al. [47] and Haward et al. [29]. Therefore, it seems that high is one of the conditions for (strong) upstream instability. Besides, the experiments by Shi et al. [64] and Varshney & Steinberg [75] (2018) also showed that when the cylinders are arranged in tandem at low , the recirculation zone is likely to appear in the upstream region of the rear cylinders. Few studies quantitatively investigated the length of upstream recirculation. A schematic diagram of the upstream recirculation length () is shown in Figure 2, which is the distance from the foremost end of the upstream recirculation to the foremost end of cylinder. Using the cylinder diameter to normalize as , Zhao et al. [84] obtained a Landau-type behaviour of the dimensionless upstream recirculation length ( increasing with ). In the experiment of Qin et al. [57], is almost linear to the Weissenberg number based on the first normal stress difference , where is the first normal stress difference, is the characteristic shear rate defined by the mean centerline velocity, and is the shear viscosity at this . It is worth pointing out that the definitions of the Weissenberg number in these two studies are different.

Figure 1. Summary for the – state space of flow patterns, which is adapted from Zhao et al. [84]. Panel (e) is adapted from Qin et al. [57].

Figure 2. Schematic diagram of upstream recirculation length.
In the simulations of viscoelastic fluid flow, an additional elastic stress divergence term is often linearly introduced to the right hand side of Navier-Stokes equations. The elastic stress is closed by the constitutive relation with conformation tensor, such as upper-convected Maxwell model (UCM model, Olsson & Yström [49]), Oldroyd-B model (Oldroyd [48]), FENE series models (Herrchen & Öttinger [31]), Phan-Thien and Tanner model (PTT model, Thien & Tanner [71]), etc. The calculated is often limited to a low value due to the aforementioned numerical instability. Varchanis et al. [73] simulated the asymmetric viscoelastic flow around a cylinder, which agreed well with their experiment. Their numerical simulation was still limited to a low .
Flow around a cylinder in a channel is a typical mixed flow, including shear flow and tensile flow. Using full-field time-resolved flow induced birefringence (FIB) imaging, Zhao et al. [84] found that upstream instabilities are associated with high stress in the fluid that accelerates in the narrow gap between the cylinder surface and the channel wall when is large. In the narrowest area between the cylinder surface and the channel wall, the fluid parcels are strongly stretched, which results in high elastic stress there. This sharp growth in elastic stress may cause loss of convergence and trigger the well-known HWNP in numerical simulation (Hulsen et al., [33]). It is a challenging task to balance the numerical accuracy and stability. Up to now, there are no systematic numerical simulations to investigate this flow upstream phenomenon (Poole [54]). In this paper, we present numerical simulations of the viscoelastic upstream instability recently observed experimentally in front of a cylinder in a narrow channel. The finite extensible nonlinear elastic model with the Peterlin closure (FENE-P) is adopted to describe the rheological constitutive behaviour of the dilute polymer solution. The square root reconstruction method (Balci et al. [6]) is selected as the stabilization technique for the numerical simulation to handle HWNPs.
The rest of the paper is organized as follows. Section 2 introduces our specific problem, governing equations, solution method, mesh generation, grid convergence test. In Section 3, we discuss our numerical results on the upstream instability. We conclude the present work in Section 4. In the Appendices, we also provide more results on the validation, some additional results on the FENE-CR model, the effect of Peclet number and the effect of different boundary conditions.
Problem formulation and Numerical method
Problem description
Figure 3(a) shows the computational domain, which consists of a pressure-driven flow from left to right past a circular cylinder in a confined channel. It is true that Qin et al. [57] revealed the inherent three-dimensional nature of the upstream recirculation zone. However, we will study the corresponding flow instability in this two-dimensional setting as the first attempt. The width of the channel is . The diameter of the cylinder is . Three different blockage ratios (), i.e., 50%, 62.5% and 75%, are considered. The distance between the cylinder center and the inlet is (upstream region ), and the distance between the cylinder center and the exit is (downstream region ). A parabolic flow profile is adopted as the inlet velocity, where is the average velocity. The zero-pressure and zero velocity gradient conditions are adopted at the outlet. On the upper wall, the lower wall and the cylindrical wall, the no-slip boundary condition is imposed. The treatment of boundary conditions for the conformation tensor will be discussed below and also in appendix A-4.

Figure 3. (a) Schematic diagram of computational domain and (b) the mesh topology near the cylinder.
Governing equations
The viscoelastic fluids in the experiments of Qin et al. (2019) and Pan et al. [50] show both the shear thickening of elongational viscosity and the shear thinning of the shear viscosity. Thus, the FENE-P model is adopted in the present study, which takes into account the finite elongation of polymer molecules and the bounded stress. The governing equations of the flow combined with the FENE-P constitutive model read as follows [10] (Bird, & Hassager 1987):
with
where , , , , , , and are fluid density, flow velocity, pressure, time, the statistical relaxation time of the polymers, conformation tensor, stress tensor of polymer and identity tensor, respectively, and are the polymeric contribution to zero-shear rate viscosity and the solvent viscosity, respectively, denotes trace operator of tensor, the extensibility parameter measures the maximum stretching of the polymer chains [10, 56] (Bird & Hassager 1987). When is set as and tends to 1, the FENE-P model returns to the Oldroyd-B model [48].
Note that another common FENE type model is the FENE-CR model, which has been widely applied to predict the flow behaviour of viscoelastic fluid with a constant shear viscosity and a bounded elongational viscosity. We have also performed numerical simulations based on the FENE-CR model. More information about the FENE-CR model and the comparison of the numerical results obtained by the FENE-P model and the FENE-CR model are provided in Appendix A-3.
The main dimensionless parameters for this simulation are the Reynolds number and the Weissenberg number , which are defined as and , respectively. It is emphasized that the present study uses the half channel width as the reference length. Thus, a factor of 2 should be applied to the corresponding defined by using the whole channel width as the reference length when comparing with the published data. is the total viscosity of the solution at zero-shear rate. is the solvent viscosity ratio, a measurement of polymer concentration and molecular characteristics of polymers. In this study, is set to 0.59 in most of the cases. Besides, , 0.75, 0.45, 0.3 and 0.15 are also considered in some cases. Four values of , 2500, 10000 and 40000 are considered. It is suggested from the definitions of and that the fluid viscoelasticity is more prominent than the fluid inertia for micro-scale flows. Fluid viscoelasticity can be characterized by . In this work, we will mostly only change to probe how the elasticity influences the upstream instability.
The drag coefficient acted on the cylinder is computed as
where is a vector normal to each face element of the cylinder boundary, whose magnitude is equal to the area of face element, and is a unit vector aligned with the streamwise direction.
The fluctuation frequency () of flow field and elastic stress wave in the direction can be obtained by fast Fourier transform (FFT) of the time series of . The Strouhal number is defined as,
In the experiment of Qin et al. [57] and Pan et al. [50], the Weissenberg number is defined using the first normal stress difference. In order to compare our numerical results with their experimental results, we calculate the first normal stress difference of the FENE-P model under simple shear flow as follows (Purnode & Crochet [56]),
where,
The shear viscosity for the special strain rate () reads
In experiment, the material properties of a viscoelastic fluid may be complicated and not easy to fit the parameters in the existing viscoelastic constitutive models, including the existence of branched chains of polymer molecules, the spectrum distribution of relaxation time and the certain distribution of chain length of polymer molecules. Herein, we use a normal stress similarity to re-calibrate the rheological parameters. The Weissenberg number based on the first normal stress difference can then be defined as the ratio of the first normal stress difference to twice the Newtonian shear stress ([57, 58]), i.e.,
which is different from for the FENE-P model. The relationship for is plotted in Figure 4(a). For the Oldroyd-B model, tends to infinite and we have,

Figure 4. (a) The relationship and (b) the relationship for the FENE-P model with .
It is worth mentioning that the above definitions for and do not include the effect of . Thus, we adopt another definition proposed in Pan et al. [50] here,
Specially, for the Oldroyd-B model with , one has,
This definition reflects the influence of . becomes lower for a larger . The corresponding – relationship for is shown in Figure 4(b). Note that for and , is equal to 9.391 when , which is close to the maximum in the experiments of Qin et al. [57, 58] and Pan et al. [50].
Numerical method
The governing equations are solved by the open-source CFD platform OpenFOAM ([80] and the rheotool toolbox ([51]). Because no numerical studies have reported the upstream instability in viscoelastic wake flows, we detail our numerical method below and provide extensive validation of the numerical code. In order to ensure the boundedness of in the FENE-P model, an implicit algorithm is used for pre-calculation before each time step ([59]). Tracing the transport equation of the conformation tensor yields,
By defining
Eq. (2.11) can be rewritten as,
The scalar is solved at each time step, and then saved for the calculation of the next time step. Once is obtained at a given time step, the conformation tensor can be calculated by Eq. (2.12).
Because of the high singularity of the constitutive governing equations, a small global artificial dissipation term is added to the right hand side of the transport equation of conformation tensor in order to avoid divergence. In our simulations, we set . Molecular dissipation of polymer should be considered at this . The presence of this additional diffusive term can be justified by the diffusivity of polymer in solvent, which was estimated to be over the range of to ([25]). The Schmidt number , which is defined as the ratio between the zero-shear rate viscosity and the polymer molecular diffusivity, is introduced to quantify the artificial dissipation. For dilute solution of polymer dissolved in water, the density is around . The viscosity may vary over a wide range. For example, is equal to in the experiment of Qin et al. [57]. Thus, is estimated over the range of to . The normalized transport equation of conformation tensor is featured by two dimensionless numbers, and the Péclet number (). The value of , instead of , will be specified in our discussions. In the present study, is set to be 0.0001. The corresponding range of is from 10 to 1000. The effect of on the flow behaviour is discussed in Appendix A-2. By comparing the numerical results and experimental results of Qin et al. [57], is fixed at 40 for the final simulation.
In order to improve the stability of numerical calculation, stabilization technique must be adopted. More information on common stabilization techniques can be found in Appendix A-§ 1. We perform a preliminary study to test two stabilization techniques, i.e., the logarithmic reconstruction method and the square-root reconstruction method. In the numerical results based on the logarithmic reconstruction method, we do not observe the upstream recirculation found in the experiments of Qin et al. [57]. Moreover, the logarithmic reconstruction method was also adopted by Mokhtari et al. [45] and Kumar & Ardekani [38] and the upstream recirculation was not reported in their simulations. However, the square root reconstruction method can successfully predict the upstream recirculation, which is, therefore, adopted in the present simulation. A new symmetric tensor is introduced satisfying . The transport equation of conformation tensor can then be rewritten as,
where,
As advised by Balci et al. [6], the term in Eq. (14) is ignored in our numerical study. Note that in Eq. (14) is an antisymmetric tensor, which can be written in the form of components as,
The components of could be calculated by solving the following equations,
where,
where is the components of . For a detailed description of this method, the reader can refer to Balci et al. [6].
The second-order backward scheme is used for time discretization. The time step in our simulation is set small enough to ensure numerical stability. The MINMOD scheme is adopted for discretization of . The second-order upwind scheme is adopted for discretization of . For the tensor , the linear extrapolation boundary with second-order accuracy is imposed on the upper and lower walls, and no-flux boundary condition with first-order accuracy on the cylindrical wall, to ensure the stability of numerical calculation. More discussion on the boundary condition for the tensor can be found in Appendix A-4. The pUcoupled algorithm is adopted for the pressure-velocity coupling [35, 52]. Our simulation is performed using the rheoFoam solver module of rheoTool in OpenFOAM extend 4.0 [51]. The validation of the present numerical method is provided in Appendix A-1.
The subcritical bifurcation can be distinguished by checking whether the simulation result is path dependence or not, as shown in Figure 5. In the present study, we systematically vary the controlling parameter and the initial condition to examine their effects on the final simulation result. In the increasing process, we gradually and slowly increase and use the numerical result of the current state to initialize the simulation for the next simulation. In the decreasing process, we gradually and slowly decrease and use the numerical result of the current state to initialize the simulation for the next state. We will explicitly point out which process is applied to perform the simulation only if the results at the same obtained in the increasing and decreasing processes are not identical. For an instability indicator around the linear critical condition (, as shown in Figure 5a), finite-amplitude solutions can exist even if is less than , which results in a different increasing path from a decreasing path (the hysteresis phenomenon). This method on identifying subcritical bifurcation is widely used in the previous studies [8, 50, 13]. In this paper, , the root mean square upstream recirculation length , the time-averaged drag coefficient , the root mean square drag coefficient , and an asymmetry parameter (will be defined in Eq. (26)) are selected as the instability indicators.

Figure 5. Method to identify subcritical bifurcation. The solid and dashed lines denote the stable and unstable solutions, respectively. denotes the critical condition. IW and DW denote the increasing and decreasing processes, respectively.
Mesh generation and grid convergence test
The block-structured mesh is generated for the computational domain by using the commercial software ANSYS ICEM. The surrounding region of the cylinder is discretized by the O-type mesh, as shown in Figure 3(b). The remainder of the computational domain is discretized using several blocks of quadrilateral meshes. In the direction, (for , 201 for and 75%) grid points are unevenly distributed. In the direction, . The dense mesh is applied near the cylinder. The O-type mesh comprises ( and ) or 440 () grid points uniformly distributed along the cylinder perimeter and ( and ) or 120 () grid points stretched over an exponential progression along the radial direction to ensure a fine mesh near the cylinder surface. We set the first cell side-by-side to the cylinder surface in the radial direction to . In the direction, grid points are set in the upstream region, and grid points are unevenly arranged in the downstream region. In the direction, (201 for and ) grid points are unevenly distributed. The total number of cells in the computational domain is approximately 280,000 ( and ) or 150,800 (). The detailed mesh generation description can be found in our previous publication [53].
The parameter set is selected to check the mesh dependency and the corresponding simulation is performed on three sets of meshes, which are listed in table 1. The effect of the mesh size on is investigated. In our numerical simulation, is measured as the horizontal distance from the most remote upstream stagnation point ( and ) to the front end of the cylinder. The simulation results show that for Mesh1 is time dependent and its variation range is recorded in table 1. is steady for Mesh2 and Mesh3 and the difference between the results on Mesh2 and Mesh3 is 3%. We thus consider that our numerical results on Mesh2 are accurate enough and the final simulations are performed on this mesh.
| Mesh | Total | |||||||
| Mesh1 | 301 | 201 | 51 | 50 | 71 | 39000 | ||
| Mesh2 | 701 | 301 | 121 | 121 | 71 | 150800 | 1.2329 | |
| Mesh3 | 1201 | 501 | 201 | 201 | 121 | 436000 | 1.2726 |
Table 1. . and denote the grid numbers of the upstream and downstream regions along the direction, respectively. and denote the grid numbers in the and directions, respectively. denotes the grid numbers radiating from the cylinder surface. is the grid size of the innermost grid near the cylinder wall.
Result and Discussions
In this section, we report the numerical results of the viscoelastic wake flow in a confined channel. The parameters considered are and , 2500, 10000 or 40000 in Section 3.1; , , 2500, 10000 or 40000 and to , to study the effect of in Section 3.2; and or , or 10000 and - 0.9 to study the effect of in Section 3.3. Table 2 summarizes the parameter space of the present study. In a certain parameter range, the flow is unsteady. We use the data in 5 statistical cycles to calculate the time-averaged PSD and quantities after the flow reaches the statistically steady-state.
| Cases | ||||
| Section 3.1 | 50% | 0.59 | 400, 2500, 10000 | 0~100 |
| 50% | 0.59 | 40000 | 0~60 | |
| Section 3.2 | 10%~50% | 0.59 | 2500 | 0~100 |
| 62.5% | 0.59 | 2500, 10000, 40000 | 0~20 | |
| 75% | 0.59 | 400, 2500, 10000 | 0~10 | |
| 75% | 0.59 | 40000 | 0~9 | |
| Section 3.3 | 50% | 0.1-0.9 | 2500 | 0~30 |
| 75% | 0.3-0.9 | 10000 | 0~20 | |
| 75% | 0.15 | 10000 | 0~15 |
Table 2. Parameter space of the present study.
Before we perform the flow simulations, the effect of the length of downstream region on the upstream recirculation is examined. Here we consider the parameter set . Three different of , and are investigated. The upstream recirculation lengths at different are summarized in table 3, which indicates that has a very small effect on the upstream recirculation. To save computational time, is selected for the final simulation.
| () | 2.3012 | 2.3001 | 2.2992 |
Table 3. Effect of on the upstream recirculation length.
High-Weissenberg simulations for the FENE-P model at
Simulation is first performed for the case with the parameter set . is often used as a benchmark case for numerical simulation. Due to the numerical instability and heavy calculation burden, was often set to less than 0.5 in previous studies (e.g., Fan et al. [20]; Alves et al. [3]; Hulsen et al. [33]). In our numerical simulation, relatively high- flows (up to 100) are calculated by considering polymer molecular diffusion and using a small time step. Lee et al. [42] indicated that the effect of molecular dissipation of polymer can be negligible only when is large () and is small (). Since the maximum considered is 100 and is equal to 40 in the present study, molecular dissipation of polymer should be taken into account.
The streamline distributions and the -velocity profiles are shown in Figure 6. In a Newtonian fluid, the flow is almost symmetry with respect to the cylinder as shown in Figure 6(a). As increases, the velocity decreases gradually along the directions approaching the cylinder both upstream and downstream, and the downstream velocity decreases faster. Figure 6(f) indicates that the downstream -velocity almost linearly increases with except for the region near the cylinder wall for different , which is consistent with the experimental results of Haward et al. [29]. Note that the strain rate at the rear stagnation point of the cylinder is equal to zero (Haward et al. [28]). At , a recirculation zone starts to form upstream of the cylinder. Correspondingly, a region of negative velocity can be observed in the upstream recirculation region as shown in Figure 6(f). With a further increase in , the upstream recirculation bubble becomes larger.

Figure 6. Streamlines for Newtonian (a) and viscoelastic (b-e) fluids at , and for different and the corresponding -velocity profiles along (f). The contours in (a-e) are colored by .
To better understand the upstream recirculation, we follow Lee et al. [41] and discuss the flow type parameter defined as
where and are the magnitudes of the deformation rate tensor
and the vorticity tensor , respectively, which can be locally evaluated using the components of the velocity vectors. is all coordinate invariant [41]. Here, indicates solid body rotation, simple shear, and pure extension. The flow strength in the extensional regions is quantified by the principle strain rate or the eigenvector of [5],
In two dimensional space, is coordinate invariant. In order to better describe the tensile strength in the extension area, a dimensionless parameter is defined by multiplying and ,
Since and are coordinate invariant, is also coordinate invariant. Define a local stretch Weissenberg number as,
which describes the local effect on the Weissenberg number caused by stretching (positive ) or compression (negative ) of fluid parcels along the flow streamline direction.
The and distributions are shown in Figure 7(a) and (b), respectively. In a Newtonian fluid, both and distribute symmetrically against the and axes with respect to the cylinder. Fluid parcel around the cylinder exhibits high extension. Just upstream of the cylinder, the flow velocity is reduced, resulting in compression. In the narrow gap between the cylinder surface and the channel wall with , flow velocity increases, resulting in stretching. Figure 7 shows that the upstream compression or stretching regions further expand upstream along the middle line () and the gap middle line () as increases. The distribution demonstrates that the strongest tensile strength region appears in the gap between the cylinder surface and the channel wall.

Figure 7. (a) Flow type parameter distributions, and (b) dimensionless distributions.

Figure 8. Variation of the local elastic stretch Weissenberg number along the direction at (a) and (b) .

Figure 9. Contours of for different . is positively correlated with the elastic stress.
The local elastic stretch Weissenberg numbers () at and are extracted and shown in figure 8, respectively. As defined above, the positive and negative denote fluid stretching and compression, respectively. At , the initial decrease in the flow velocity results in a negative for a long distance. initially decreases and then increases along the direction. Generally, the location of the valley point moves upstream and the corresponding minimum decreases with increasing . For , the minimum is about -19.8. At , the increase in the flow velocity results in a positive for a long distance. Before approaching , rapidly increases to the peak value and then experiences a sudden drop. The maximum is about 68.1 for , which occurs in the gap between the cylinder surface and the channel wall. The absolute in the gap (= 68.1) is much larger than that in front of the cylinder (= 19.8). The large in the gap reflects that the high elastic stress is concentrated there, which extends upstream near the wall region as shown in figure 9. A similar observation has been reported by Zhao et al. [84] in wormlike fluids.
A well-known dimensionless parameter which rationalizes these streamline instabilities is the parameter introduced by McKinley et al. [44].
where is the characteristic length over which perturbations to the base stress and velocity fields relax, is the streamline radius of curvature, is the streamwise tensile stress. We will convert this definition of the criterion to a form amenable to local evaluation in flows by the substitution of characteristic values with local values. To consider local effective relaxation time () and ignore solid-like rotational flows, Cruz et al. [16] proposed a modified Pakdel–McKinley criterion,
where . Here we note that is the material derivative of the velocity vector, which is equivalent to for steady-state flow. The Frobenius norm () is used, so that the resulting will vary between zero, when the normal stress is weak, and one, when the tensile normal stress dominates, in highly elastic shear or extensional flows. Therefore in strongly extensional flows, is approximately given by .
Contours of for different are shown in figure 10. The largest region occurs on the gap between the cylinder and the outer wall. In the upstream recirculation regions, we can also find non-negligible, though smaller, values of . From the definition of , high happens in highly elastic shear or extensional flows. Although the flow curvature in the recirculation zone is large, we only see relatively small value of in the recirculation region because the extension in this region is not larger than that in the gap regions (figure 8).
The above results indicate that and the stress dominate in the gap, compared with those in front of the cylinder. In this sense, the flow across the gap can be regarded as a main flow, while the upstream recirculation is a secondary flow. A primary-secondary flow model shown in figure 11 could be applied to characterize this upstream recirculation, i.e., a high-speed stretching gap flow and a relatively low-speed upstream recirculation.

Figure 10. Contours of in the viscoelastic cylindrical wake flow with upstream instability for different .
The time evolutions of the drag coefficient for , 80, 90 and 100 are plotted in Figure 12. To save computing time, the numerical result of (at time of ) is set as the initial value for these higher- flows. The drag acted on the cylinder increases with when (refer to data in table A.1 for low range and data here in high range), no matter whether the flow is steady ([45]) or unsteady. The time-averaged drag coefficient () is at but at . A sharp increment in is observed when increases from 80 to 82.5 as shown in Figure 14(a), which results from the enhancement in the additional extensional viscosity due to flow fluctuations ([12]) as shown in Figure 14b. A similar sudden increase of drag was also reported in the numerical simulation of Grili et al. (2013). Therefore, the increment in with increasing is sharper in the unsteady flow than the steady flow.

Figure 12. The solid black lines denote the temporal histories of the drag coefficient at (, , ) = (50%, 0.59, 2500) for (a) = 70, (b) = 80, (c) = 90, and (d) =100. The dash brown line in (d) denotes the temporal history of .

Figure 14. Streamlines at time instants (a) , (b) , (c) , and (d) marked in Figure 12(d) for .
When , our calculation indicates that the flow is steady after developing for a long time. However, the flow becomes unsteady when . At , the fluctuation of drag coefficient exhibits a fully-developed periodic state with a single frequency demonstrated by the FFT analysis shown in Figure 13(a). For in Figure 12(d), the variation of drag coefficient roughly maintains a periodic state that features more discrete frequencies demonstrated by the FFT analysis shown in Figure 13(b). Interestingly, the quantitative analysis demonstrates , , and , indicating the nonlinear effect of the period-doubling. Figure 12(d) also shows the corresponding time evolution of for one period, which exhibits similar trend to the curve. The maximum drag of the cylinder corresponds to the longest upstream recirculation. The corresponding streamlines at four time instants within one period are shown in Figure 14.

Figure 13. Fast Fourier transform of the drag coefficients for at (a) and (b) .
A sharp increase of drag between from and under the increasing process (i.e., the discontinuity on the - curve near ) shown in Figure 15(a) implies a subcritical transition. We use the result of as the initial field to study this transition behavior and gradually reduce (i.e., the decreasing process). It is found that the - curve maintains continuity until and then experiences a sudden drop between to as shown in Figure 15(a). Both Figure 15 and Figure 16 show that the unsteady flow becomes steady when is decreased from to . The hysteresis phenomenon shown in Figure 15(b) implies that the transient transition is a subcritical Hopf bifurcation.

Figure 15. Variations of (a) the time averaged drag coefficient () and (b) the root mean square drag coefficient () with for . IW and DW denote the increasing and decreasing processes, respectively.

Figure 16. The temporal histories of the drag coefficient at for (a) and (b) under the decreasing process. The numerical result of is selected as the initial field. DW denotes the decreasing process.
For and , we also observe the upstream recirculation. The non-dimensional recirculation length () as a function of for , and is summarized in Figure 17(a). At high , the flow becomes unsteady for , as discussed above. For , the flow also becomes unsteady when . However, for , the unsteady behaviour is only observed when . When the flow is unsteady and varies with time, the maximum and minimum values of at the corresponding are provided, as shown in Figure 17(a). Note that we only present the result up to for here. A further increase in causes the simulation divergence due to numerical instability, even we use a very small time step. We observe a sudden shooting up of before the simulation divergence.

Figure 17. (a) The ~ relationship and (b) the ~ relationship at . When the flow becomes unsteady and varies with time at high , the time averaged is provided while the error bar indicates the corresponding maximum and minimum values. IW and DW denote the increasing and decreasing processes, respectively.
for is shorter than those for or . However, Qin et al. [57] implies the upstream recirculation length is positively correlated with extensional viscosity. Variation of with is plotted in Figure 17(b). As mentioned above, the definition of is similar to that in Pan et al. [50]. The time averaged measured by Qin et al. [57] is also plotted in Figure 17(b). Behaviour of at is mostly consistent with the experimental results of Qin et al. [57]. It is worth pointing out that in their experiment, the upstream recirculation length varies strongly with time when is beyond 4.35. However, the fluctuation of is not very obvious in our simulation when for . Note that the present simulation is two-dimensional and the present rheological model and parameters may be different from those for the material used in the experiment. Besides, may be different between our simulation and their experiment. Thus, the temporal behaviour of deviates from that of Qin et al. [57].
| 14.97 | 16.252 | 16.83 | 17.94 | |
| 2.8440 | 0.5164 | 0.8423 | 2.2303 | |
| -0.00003 | -0.00002 | -0.00011 | -0.0001 |
Table 4. The fitting parameters for in Eq. (25)
Figure 17 clearly shows that there is a critical Weissenberg number for the onset of the upstream recirculation at each . Close to the transition, increases smoothly from zero to a non-zero value with increasing , and the relationship between and can be well described by a simple Landau-type quartic potential minimized as:
where denotes the growth rate coefficient and quantifies system imperfection that biases a transition to a favored branch. Figure 18 shows the relationship between and for both the numerical results and the fitted data around the onset of the upstream recirculation region. The corresponding values of and are listed in table 4. The fitted and increase with while is always close to zero. When , the fitted curve only collapses well with the numerical results when is less than 40. This implies that at , variation in may cause inherent change in flow state when . It may result from the velocity increase in the gap between the cylinder surface and the channel wall. The time-averaged streamlines and contours of the dimensionless velocity for and 50 are shown in Figure 19(a) and (b). The streamlines indicate that the upstream recirculation elongates when increases from 40 to 50. Meanwhile, the contours of the dimensionless velocity show that the maximum velocity in the gap between the cylinder surface and the channel wall increases from 3.31 at to 3.65 at . The maximum time averaged dimensionless -direction velocity in the flow field is extracted and plotted in Figure 19(c). Obviously, a significant increase in velocity exits between to 50. At high , the elastic stress is concentrated near the channel and cylinder walls, which acts to narrow the gap and accelerate the flow velocity there. For example, the maximum velocity in the gap is 3.65 for , however, is about 3 for a Newtonian fluid. The effect of on the upstream recirculation will be discussed in the next section.

Figure 18. Variation of with at for (a) , (b) , (c) , and (d) . The solid line denotes the data fitted with Landau-type quartic potential while the symbols ●, ♦, ★, and ⁕ denote the numerical results. The fitted for each is summarized in table 4. IW and DW denote the increasing and decreasing processes, respectively.

Figure 19. Time-averaged flow streamlines and contours of the dimensionless velocity at for and . Variation of the maximum time-averaged dimensionless velocity with .
Effect of on viscoelastic upstream instability
The existing experiments indicated that the upstream recirculation occurs when is not less than 50%. When is low, such as 10%, only upstream streamline bending exists [29]. Thus, there is a critical value of which is around 50% and determined by and , below which the upstream recirculation will not occur. When is 50% and is low in the simulations discussed in § 3.0.1, the velocity in the midline changes from 1.5 at the inlet to 0 at the cylinder surface, which results in a compression . However, the velocity along changes from about 1.5 at the inlet to about 3.0 in the gap, which leads to a stretching . The maximum compression is almost equal to the maximum stretching at this situation. When exceeds 50%, the gradient of flow velocity in the gap is steeper than that upstream of the cylinder. That is, the stretching upstream is larger than the compression in the gap. The compression in the gap plays a leading role, which results in a primary-secondary flow as shown in Figure 11. Moreover, with the increase of , the flow velocity inside the gap may slightly increase (discussed in last paragraph in § 3.0.1). This assembles that the blockage effect is more severe and the apparent is increased (The increase of the flow rate in the center line of the gap can be equivalent to the increase of ). Therefore, the upstream recirculation also occurs when .

Figure 11. The primary-secondary flow model for the present flow configuration.
It is reasonable to speculate that a certain boundary exists in the space of to distinguish the regimes with and without the upstream recirculation. Thus, we studied the effect of on for , as shown in Figure 20. Note that does not exceed 100 in this test. Here is not obtained by fitting Eq. (25). Instead, at a given , multiple simulations are performed near . The bisection method is adopted to determine by checking whether the upstream recirculation occurs. Our results imply that the upstream recirculation only occurs when is more than 16.7%, i.e., the curve asymptotically approaches .
Figure 20. Variation of with for .
In the rest of this subsection we mainly consider high of and and discuss its effect on the upstream recirculation. Figure 21(a) and (Figure 21b) plots as a function of for and , respectively. When , , and (75%, 2500), the growth of upstream recirculation with close to the transition also satisfies Eq. (25), as shown in Figure 22(a)-(Figure 22b) and 23(a)-(b). However, when , , and , the relationship for the increasing process does not satisfy Eq. (25), as shown in Figure 21(a)-(Figure 21b) and 23(d). In fact, there is a sudden shooting up of when is larger than a certain value, indicating that the relationship is discontinuous there. For example, when and , equals zero at . However, suddenly changes to when . Thus, the bifurcation for these parameter sets is subcritical. For decreasing , the results for and still satisfy Eq. (25), as shown in Figure 23(b) and (c). The flow instability affected by external disturbance. In previous experiments [37, 65, 84, 57, 27], no sudden shooting-up in the relationship was observed, which may cause by strong disturbance in their experiment. The appearance of upstream recirculation is a more stable form. These scenarios are consistent with our results obtained in the decreasing process, where strong disturbance is also introduces by the initial fields.
Figure 21. The relationship for (a) and (b) (for the decreasing process, the numerical results of are selected as the initial fields for , while the numerical results of are selected as the initial fields for ). Variation in the root mean square with at for (c) and (d) . IW and DW denote the increasing and decreasing processes, respectively.

Figure 22. The growth of upstream recirculation with at for (a) , and (b) . The solid line denotes the data fitted with Landau-type quartic potential while the symbols ♦ and ★ denote the numerical results.

Figure 23. The growth of upstream recirculation with at , for (a) , (b) , (c) , and (d) . The solid line denotes the data fitted with Landau-type quartic potential while the symbols ●, ♦, ■, and ⁕ denote the numerical results. DW denotes the decreasing process.
Figure 21(c) and (Figure 21d) plots the root mean square as a function of at for and , respectively. Comparison between Figure 21(a) and 21(c) indicates that, the critical for the onset of the upstream recirculation and the non-zero fluctuation in the upstream recirculation region almost coincide at . The variations of the root mean square with for both the increasing and decreasing processes in Figure 21(d) also indicate that bifurcation is subcritical.
for different could be obtained from Figure 18 and Figure 21. With an increase in , decreases. With an increase in , becomes larger (the increasing process). Yamani & McKinley [82] defined an important parameter for flow instability. This parameter implies that is positively related to , which is consistent with the present result. Therefore, the onset of the upstream recirculation is promoted and is smaller when is smaller. Moreover, is influenced by how the flow field is initialized, i.e., the increasing or decreasing process mentioned in Section 2.3. for the decreasing process may be much less than that for the increasing process. For example, at is about 2.75 for the decreasing process and about 6.75 for the increasing process.
At high , such as or , only slightly changes with the variation in . may reach the saturation value in this situation. For or , the saturation value is large when is large, which is different from the trend for . A large , i.e., a longer molecular chain length, means higher potential of extensional viscosity. Qin et al. [57] implies the upstream recirculation length is positively correlated with extensional viscosity, which is qualitatively consistent with our results at or .
When or , the upstream recirculation exhibits strong time dependence behaviour at high when . The -velocity contours and streamlines at time instant for is shown in Figure 24(a). Under this parameter set, the flow is unsteady and the length and shape of the upstream recirculation vary with time. However, the fluctuation amplitude in the upstream recirculation region is not the maximum in the flow field. Instead, the strongest temporal fluctuation occurs in the gap region, as shown by the distribution in Figure 24(b). Downstream the cylinder, the high velocity region is concentrated on both sides of the centerline, as shown in Figure 24(a). Correspondingly, the flow fluctuation also is high there.

Figure 24. (a) The -velocity contours and streamlines at time instant , and (b) the corresponding distribution. The parameter set used in this simulation is .
We carefully examine the symmetry of the flow field and find that the symmetry against the horizontal centerline is obviously broken for and when is large, such as . Figure 25 plots the variations of the averaged velocity (space average) in the upper and lower gaps with time at . Both averaged velocity show strongly oscillating behaviour. The corresponding oscillating amplitudes for upper and lower averaged velocities are opposite, in order to satisfy the continuity equation. The upper averaged velocity fluctuates around a mean value of 4.168, which is larger 3.898 for the lower averaged velocity.

Figure 25. Time histories of the averaged velocity (space average) in the upper and lower gaps for .
A parameter, which follows that of Hopkins et al. (2019), is defined to evaluate the flow asymmetry,
where and are the volumetric flow rate through the upper and lower gaps, respectively. Variation of with for is presented in Figure 26. The flow symmetry only maintains at , i.e., for the Newtonian flow. When , the viscoelastic flow exhibits complex asymmetric behavior for high and high . For the increasing process, initially increases with until . Then decreases with further increase in until . Note that for , the onset of the upstream recirculation occurs at . When , increases again with . For the decreasing process, decreases with decreasing from to . However, increases with a further decrease in until , which is corresponding to below which the upstream recirculation disappears. When , decreases again with decreasing . The increasing and decreasing paths do not coincide between , which indicates a typical hysteresis phenomenon.

Figure 26. Variation of flow asymmetry parameter with for . IW and DW denote the increasing and decreasing processes, respectively.
Effect of and on viscoelastic upstream instability
Figure 18(b) shows that the relationship for approximately satisfies Landau-type quartic potential near . However, Figure 21(b) the variation of with for under the increasing process does not approximately fit Landau-type quartic potential near , but shows the hysteresis phenomenon. Thus, these two parameter sets are specially selected to investigate the effect of . For , ranges from 0.1 to 0.9 and is no more than 30. For , ranges from 0.15 to 0.9 and is no more than 20. Within these parameter spaces, the flow fluctuations are small and only varies weakly with time. Thus, the effect of on time-dependent stability is not discussed here.

Figure 27. The – relationship at for different .
In experiments, decreases with increasing polymer concentration. The FENE-P model can describe the shear thinning behaviour of viscoelastic fluid, usually for the case when is small. The smaller , the more obvious the shear-thinning effect in the flow. In our study, is large and the influence of on shear-thinning could be ignored ([68]). However, a variation in may affect the distribution of elastic stress as discussed below. We first consider . The variations of with for different are shown in figure 27. Near the upstream recirculation transition point, the - curves all approximately satisfies the Landau-type quartic potential. A smaller corresponds to a smaller . For example, is between 23 to 24 at , while is between 8 to 9 at . For the same , a lower corresponds to a larger . For example, is 0.3242 at while is 2.1010 at , for a fixed .

Figure 28. (a) The - relationship and (b) the - relationship at . In panel (b), the larger is selected if obtained in the increasing and decreasing processes (denoted by IW and DW respectively) at the same parameters is not identical.
For , for different are counted and shown in figure 28. has a non-linear effect on . For both the increasing and decreasing processes, before the onset of upstream recirculation, the - relationships are identical and satisfies Eq. (25) at or 0.75. However, the - relationships for the increasing and decreasing processes at , 0.45, 0.3, or 0.15 are different. For the increasing process, experiences a sudden jump with increasing , and the - relationship shows a discontinuity. Interestingly, the discontinuity point () locates between 6.5 to 7.0 for all . These features indicate that the transition is subcritical bifurcation at , 0.45, 0.3$, or 0.15. Thus, we can conclude that affects the bifurcation type, which due to the amplifying effect on elastic stress with decreasing . Note that also shows the similar effect in other viscoelastic flows. For example, in the elasto-inertial pipe flow, recent experimental and theoretical studies indicated supercritical bifurcation for large and subcritical bifurcation for small [15, 14, 61, 79]. For the decreasing process, all the - relationships are still uninterrupted to satisfy Eq. (25).
The effect of on the upstream flow behaviour has been discussed scatteredly in the above subsections. We thus provide a brief summary in this subsection. has a noticeable effect of for the onset of the upstream recirculation. Generally, a larger corresponds to a higher as shown in Figure 17(a) and Figure 21(a-Figure 21b). also affects subcritical behaviour. For large with high (such as for , and or for ), the upstream instability is subcritical. The effect of on subcritical behaviour is similar to that of discussed in § 3.2. For larger (e.g. and ), the unsteady flow is more likely to appear at a larger .
Conclusion Remarks
In this paper, we confirm that the existing macroscopic viscoelastic constitutive relationship (FENE-P model) is still qualitatively applicable to predict the viscoelastic upstream instability through numerical simulation. By applying the square root reconstruction method and considering the molecular dissipation effect to stabilize the numerical simulations, we studied the viscoelastic flow over a circular cylinder in a narrow channel at a very low and over a wide range of .
The upstream recirculation is observed when is beyond a certain critical value and is larger than around . The occurrence of upstream recirculation is related to high local stretch Weissenberg number and the high elastic stress in the narrow gap between the cylinder surface and the channel wall, which is consistent with that reported by Zhao et al. [84]. For , our simulation results on the upstream recirculation are basically consistent with the experimental results of Qin et al. [57] for the time-averaged behaviour.
Higher could precipitate the onset of the upstream recirculation, that is, becomes lower. has less effect on (the discontinuous point, the increasing process), but has a nonlinearity influence on at high () and relative high (). For and , a lower corresponds to a lower . Close to the onset of upstream recirculation, with satisfy Landau-type quartic potential when , , and are not very large, e.g., . Close to the unsteady transition, the occurrence of the hysteresis phenomenon (signaling a subcritical bifurcation) depends on and [13]. For the cases with large , small and large , the flow often exhibits subcritical behavior. In this situation, strong disturbance should be added in the initial flow field so that the corresponding simulation results can be consistent with the experimental results.
When exceeds about a critical at (the critical depends on ), the drag on the cylinder increases with , although the flow remains steady. When continues to increase beyond a certain threshold, the upstream recirculation becomes unsteady. The drag on the cylinder increases more sharply in this range.
Our results advance the understanding the underlying mechanism of this novel flow instability, which may also be relevant to a number of interesting phenomena observed in viscoelastic flows, e.g., stationary dead zones associated with complex geometries such as multiple tandem circular cylinders [64] and porous media [36].
The cylinder in a channel is often used as a prototype configuration to experimentally study the elastic instability. The upstream instability may extend downstream when is larger. It is interesting to study the elastic wave in the channel through numerical simulations. However, only the elastic instability in the channel without cylinder was studied [46]. Our simulation could provide a platform for flows in the channel with cylinder(s).
The upstream recirculation phenomena are governed by many parameters and the present study does not explore all the parameter space of (, , , , ). These parameters need to be investigated in more detail. In experiment of Shi et al. [65], the upstream instability eventually occurs when the elastic Mach number is beyond ∼10 at , which is affected by both channel geometry and fluid properties. We believe that is an important dimensionless number for such instability. However, the effect of is not discussed in this paper and will be studied in our future work.
Declaration of Interests
The authors report no conflict of interest.
Acknowledgement
The author P Yu would like to thank the financial support from Shenzhen Science and Technology Innovation Commission (Grant No. JCYJ20180504165704491), Guangdong Provincial Key Laboratory of Turbulence Research and Applications (Grant No. 2019B21203001), the National Natural Science Foundation of China (NSFC, Grant No. 12172163, 12002148, 12071367). This work is supported by Center for Computational Science and Engineering of Southern University of Science and Technology. The authors would like to thank Prof. Victor Steinberg from Weizmann Institute of Science and Jia-yu Li from Southern University of Science and Technology for the useful discussions.
Appendix
Validation
To validate the present numerical method, a series of simulations based on the Oldroyd-B model are performed for the flow past cylinder in a channel with (Mesh2) and the numerical results are compared with published data. In these simulations, the Reynolds number is set as zero, i.e., the convection terms in the momentum equations are ignored. Note that the polymer molecular dissipation is also not considered in this section. The viscosity ratio is fixed at 0.59 while is varied from 0 to 0.55. At low , the viscoelastic equations could be numerically solved directly. However, previous numerical tests ([20, 33, 3] (Alves et al. 2009)) indicates that the stabilization technique must be adopted to avoid simulation divergence at high . Commonly used stabilization techniques include the elastic viscous split stress (EVSS) method ([20]), the log-conformation representation method (Labelled as Log, Fattal and Kupferman 2004; [33, 1]) and the square root reconstruction method (Labelled as Sqrt, Alves et al. 2009). The present study adopts the square root reconstruction method.
The values of for different are compared with the published data and listed in Table A1. Our results are in good agreement with those reported in the literature. For example, the drag coefficient for obtained in our simulation is , which is consistent with in Hulsen et al. [33] and in Fan et al. [20]. Besides, the stress profiles along the upper cylinder wall and the downstream center line for and 0.45 are shown in figure A1. We can see a good agreement between our result and that in Aleves et al. [3] for . However, a difference among our results and those of Alves et al. [3] and Afonso et al. [1] can be observed in the region immediately behind the cylinder for , even the local refined mesh is utilized there in our simulation as shown in Figure A1(b). Note that a higher peak of the local elastic stress occurs behind the cylinder when is larger, which is very difficult to be accurately resolved by numerical simulation. Anyway, the comparison indicates that our simulation could capture the main flow characteristics downstream since the behaviour of the local elastic stress downstream is similar for the three simulations.

Figure A1. Stress profile along the upper cylinder wall and the downstream center line for the Oldroyd-B fluid at (a) and (b) .
| Hulsen et al. [33] Log | Fan et al. [20] EVSS | Alves et al. [3] | Afonso et al. [1] Log | Afonso et al. [1] Sqrt | Present Sqrt | |
| 0 | 132.358 | 132.36 | 132.378 | - | - | 132.36 |
| 0.05 | 130.363 | 130.36 | 130.355 | - | - | 130.30 |
| 0.1 | 126.626 | 126.62 | 126.632 | - | - | 126.565 |
| 0.15 | 123.193 | 123.19 | 123.210 | - | - | 123.133 |
| 0.2 | 120.596 | 120.59 | 120.607 | - | - | 120.555 |
| 0.25 | 118.836 | 118.83 | 118.838 | 118.818 | 118.821 | 118.808 |
| 0.3 | 117.775 | 117.78 | 117.787 | 117.774 | 117.776 | 117.776 |
| 0.35 | 117.315 | 117.32 | 117.323 | 117.323 | 117.324 | 117.33 |
| 0.4 | 117.373 | 117.36 | 117.357 | 117.364 | 117.370 | 117.36 |
| 0.45 | 117.787 | 117.80 | 117.851 | 117.817 | - | 117.776 |
| 0.5 | 118.501 | 118.49 | 118.518 | 118.680 | - | 118.508 |
| 0.55 | 119.466 | - | - | 119.780 | - | 119.51 |
Table A1. Variation of with at for the Oldroyd-B fluid.
Effect of Pe
In this section, the effect of on the upstream flow behavior is examined. The parameter set investigated is . Five values of 10, 40, 64, 80 and 400 are considered. Figure A2 shows that, when is equal to 10, 40 or 64, the flow field is similar to the experimental results of Qin et al. [57]. However, when is set to be 80 or 400, the upstream recirculation becomes short and one or two recirculation regions attached to the channel wall appear. Thus, we select a relative lower in the final simulation.

Figure A2. Instantaneous streamlines for at (a) , (b) , (c) , (d) and (e) .
The FENE-CR model
The governing equations based on the FENE-CR model are similar to those for the FENE-P model (Eqs. (1) to (3)) except for the polymer stress definition, which is expressed as,
The FENE-CR model is degenerated to the Oldroyd-B model if is set to 1 in Eq. (A.1).
Tracing the transport equation of the conformation tensor for the FENE-CR model yields,
By using the scalar defined in Eq. (12), Eq. (A.2) can be rewritten as,
When the square root reconstruction method is applied, the transport equation of conformation tensor for the FENE-CR model can then be rewritten as,
All other related equations are the same as those for the FENE-P model and thus not repeated here. The simulation based on the FENE-CR model is also performed using the rheoFoam solver module of rheoTool in OpenFOAM extend 4.0 [51].
The parameter set considered here is . Haward et al. [28] suggested that the purely elastic instability is only observed when both shear-thinning and elasticity exist. The simulation based on the FENE-CR model would help us to identify whether shear-thinning is necessary for the upstream instability at high . Variation of with for the results based on both the FENE-CR and FENE-P models is shown in Figure A3. The results indicate that the simulations based on both the models can predict the occurrence of the upstream recirculation. Although deviations can be observed for the - curves based on the two models, the general trends are similar. Therefore, shear-thing is not a necessary condition for this upstream instability.

Figure A3. Variation of with for . The results based on both the FENE-CR and FENE-P models are provided. IW and DW denote the increasing and decreasing processes, respectively.
Boundary condition for the tensor b
When simulating the viscoelastic fluid flow in OpenFOAM, the boundary condition for the elastic stress on the wall needs to be explicitly implemented, in order to obtain the stress gradient of the center point of a grid element close to the wall ([2]). In our numerical simulation, the b tensor adopts the linear extrapolation boundary condition on the channel walls and the no flux boundary on the cylinder wall. These uneven implementations are based on the following considerations. First, if the linear extrapolation boundary condition is imposed on the cylinder wall, a very small time step should be used in the simulation. For example, for the parameter set , the dimensionless time step must be set as . It takes more than one month for the simulation to reach statistically steady state with the parallel computing on 40 CPU-cores. Second, if the linear extrapolation boundary condition is imposed on the cylinder wall, the simulation results show unstable trend. For example, the upstream recirculation behaves symmetrically in the experiment of Haward et al. [27]. However, the corresponding flow field at exhibits obvious asymmetric behaviour, which implies that too strong disturbance is introduced into the simulation. Finally, if the no-flux boundary condition is imposed on both the channel walls and the cylinder wall, the upstream recirculation only occurs at very high in the simulation. The deviation between the numerical and experimental results is unacceptable.
Figure A4 shows the variations of and the root mean square with at by using different boundary conditions for the tensor b on the cylinder wall. The boundary condition has a noticeable effect on and the overall upstream recirculation behaviour. The time averaged -velocity contours and streamlines and the corresponding distribution at are presented in figure A5(a) and A5(b), respectively. The simulation is performed based on the linear extrapolation boundary condition for the tensor on the cylinder wall. The flow asymmetry could be clearly seen in the distribution in Figure A5. The flow asymmetry parameter is also computed and compared. is 0.0423 for the liner extrapolation boundary condition and 0.0335 for the no flux boundary condition. Compared to that for the no-flux boundary condition, flow fluctuation is more severe for the liner extrapolation boundary, which results in more additional elongational viscosity and thus larger drag force. The comparison indicates that the results obtained by the no flux boundary condition on the cylinder wall are closer to the experimental observations, which is therefore adopted by the present study.

Figure A4. Variation of (a) and (b) the root mean square with at for different boundary conditions of the tensor on the cylinder wall. IW and DW denote the increasing and decreasing processes, respectively.

Figure A5. (a) The time averaged -velocity contours and streamlines, and (b) the corresponding distribution at . The linear extrapolation boundary condition for the tensor is imposed on the cylinder wall.
References
- [1]AFONSO, A., OLIVEIRA, P. J., PINHO, F. T., & ALVES, M. A. 2009 The log-conformation tensor approach in the finite-volume method framework. J. Non-Newton. Fluid Mech., 157(1-2), 55-65.DOI
- [2]ALVES, M. A., OLIVEIRA, P. J., & PINHO, F. T. 2021 Numerical Methods for viscoelastic fluid flows. Ann. Rev. Fluid Mech., 53, 509-541.DOI
- [3]ALVES, M. A., PINHO, F. T., & OLIVEIRA, P. J. 2001 The flow of viscoelastic fluids past a cylinder: finite-volume high-resolution methods. J. Non-Newton. Fluid Mech., 97, 207-232.DOI
- [4]ARRATIA, P. E., THOMAS, C. C., DIORIO, J., & GOLLUB, J. P. 2006 Elastic instabilities of polymer solutions in cross-channel flow. Phys. Rev. Lett., 96 (14), 144502.pubmed.ncbi.nlm.nih.gov/16712081
- [5]ASTARITA, G. 1979 Objective and generally applicable criteria for flow classification. J. Non-Newton. Fluid Mech., 6 (1), 69-76.DOI
- [6]BALCI, N., THOMASES, B., RENARDY, M., & DOERING, C. R. 2011 Symmetric factorization of the conformation tensor in viscoelastic fluid models. J. Non-Newton. Fluid Mech., 166 (11), 546-553.
- [7]BARKLEY, D. 2006 Linear analysis of the cylinder wake mean flow. Europhys. Lett. 75, 750–756.DOI
- [8]BECHERER, P., MOROZOV, A. N., & VAN SAARLOO, W. 2009 Probing a subcritical instability with an amplitude expansion: An exploration of how far one can get. Physica D., 238(18), 1827-1840.DOI
- [10]BIRD, R. B., DOTSO, P. J., & JOHNSON, N. L. 1980 Polymer solution rheology based on a finitely extensible bead—spring chain model. J. Non-Newton. Fluid Mech., 7 (2-3), 213-235.DOI
- [11]BRUST, M., SCHAEFER, C., DOERR, R., PAN, L., GARCIA, M., ARRATIA, P. E. & WAGNER, C. 2013 Rheology of human blood plasma: viscoelastic versus Newtonian behavior. Phys. Rev. Lett., 110, 078305.DOI
- [12]BROWNE, C. A., & DATTA, S. S. 2021 Elastic turbulence generates anomalous flow resistance in porous media. Sci. Adv., 7(45), eabj2619.DOI
- [13]BURSHTEIN, N., ZOGRAFOS, K., SHEN, A. Q., POOLE, R. J., & HAWAR, S. J. 2017 Inertioelastic flow instability at a stagnation point. Phys. Rev. X, 7(4), 041039.
- [14]CHANDRA, B., SHANKAR, V. & DAS, D. 2020 Early transition, relaminarization and drag reduction in the flow of polymer solutions through microtubes. J. Fluid Mech., 885, A47.DOI
- [15]CHOUEIRI, G. H., LOPEZ, J. M., VARSHNEY, A., SANKAR, S. & HOF, B. 2021 Experimental observation of the origin and structure of elasto-inertial turbulence. Proc. Natl Acad. Sci. USA, 118(45).DOI
- [16]CRUZ, F. A., POOLE, R. J., AFONSO, A. M., PINHO, F. T., OLIVEIRA, P. J., & ALVES, M. A. 2016 Influence of channel aspect ratio on the onset of purely-elastic flow instabilities in three-dimensional planar cross-slots. J. Non-Newton. Fluid Mech., 227, 65-79.
- [17]DAVOODI, M., LEROUGE, S., NOROUZI, M., & POOLE, R. J. 2018 Secondary flows due to finite aspect ratio in inertialess viscoelastic Taylor–Couette flow. J. Fluid Mech., 857, 823-850.DOI
- [18]DENN, M. M. 2001 Extrusion instabilities and wall slip. Annu. Rev. Fluid Mech., 33, 265–287.
- [19]DHAHIR, S. A., & WALTERS, K. 1989 On non-Newtonian flow past a cylinder in a confined flow. J. Rheol., 33 (6), 781-804.DOI
- [20]FAN, Y., TANNER, R. I., & PHAN-THIEN, N. 1999 Galerkin/least-square finite-element methods for steady viscoelastic flows. J. Non-Newton. Fluid Mech., 84, 233-256.DOI
- [22]GRILLI, M., VÁZQUEZ-QUESADA, A. & ELLERO, M. 2013 Transition to turbulence and mixing in a viscoelastic fluid flowing inside a channel with a periodic array of cylindrical obstacles. Phys. Rev. Lett., 110, 174501.pubmed.ncbi.nlm.nih.gov/23679735
- [23]GROISMAN, A. & STEINBERG, V. 2000 Elastic turbulence in a polymer solution flow. Nature, 405, 53–55.pubmed.ncbi.nlm.nih.gov/10811214
- [24]GROISMAN, A. & STEINBERG, V. 2001 Efficient mixing at low Reynolds numbers using polymer additives. Nature, 410, 905–908.DOI
- [25]HAGGERTY, L., SUGARMAN, J. H., & PRUDHOMME, R. K. 1988 Diffusion of polymers through polyacrylamide gels. Polymer, 29(6), 1058-1063.
- [26]HAWARD, S. J., HOPKINS, C. C., & SHEN, A. Q. 2020 Asymmetric flow of polymer solutions around microfluidic cylinders: Interaction between shear-thinning and viscoelasticity. J. Non-Newton. Fluid Mech., 278, 104250.DOI
- [27]HAWARD, S. J., HOPKINS, C. C., VARCHANIS, S., & SHEN, A. Q. 2021 Bifurcations in flows of complex fluids around microfluidic cylinders. Lab Chip, 21, 4041-4059.DOI
- [28]HAWARD, S. J., KITAJIMA, N., TODA-PETERS, K., TAKAHASHI, T., & SHEN, A. Q. 2019 Flow of wormlike micellar solutions around microfluidic cylinders with high aspect ratio and low blockage ratio. Soft Matter, 15(9), 1927-1941.DOI
- [29]HAWARD, S. J., TODA-PETERS, K., & SHEN, A. Q. 2018 Steady viscoelastic flow around high-aspect-ratio, low-blockage-ratio microfluidic cylinders. J. Non-Newton. Fluid Mech., 254, 23-35.DOI
- [30]HOPKINS, C. C., HAWARD, S. J., & SHEN, A. Q. 2022 Upstream wall vortices in viscoelastic flow past a cylinder. Soft Matter, 18(26), 4868-4880.DOI
- [31]HERRCHEN, M., & ÖTTINGER, H. C. 1997 A detailed comparison of various FENE dumbbell models. J. Non-Newton. Fluid Mech., 68 (1), 17-42.DOI
- [32]HOPKINS, C. C., HAWARD, S. J., & SHEN, A. Q. 2021 Tristability in viscoelastic flow past side-by-side microcylinders. Phys. Rev. Lett., 126 (5), 054501.DOI
- [33]HULSEN, M. A., FATTAL, R., & KUPFERMAN, R. 2005 Flow of viscoelastic fluids past a cylinder at high Weissenberg number: stabilized simulations using matrix logarithms. J. Non-Newton. Fluid Mech., 127 (1), 27-39.DOI
- [34]ILIFF, J. J., WANG, M., LIAO, Y., PLOGG, B. A., PENG, W., GUNDERSEN, G. A., BENVENISTE, H., VATES, G. E., DEANE, R., GOLDMAN, S. A., NAGELHUS, E. A. & NEDERGAARD, M. 2012 A paravascular pathway facilitates CSF flow through the brain parenchyma and the clearance of interstitial solutes, including amyloid β. Sci. Transl. Med., 4 (147), 147ra111.DOI
- [35]JARETEG, K. 2012 Block coupled calculations in OpenFOAM. Project within course: CFD with OpenSource software. Chalmers University of Technology.
- [36]KAWALE, D., MARQUES, E., ZITHA, P. L., KREUTZER, M. T., ROSSEN, W. R., & BIUKANY, P. E. 2017 Elastic instabilities during the flow of hydrolyzed polyacrylamide solution in porous media: effect of pore-shape and salt. Soft Matter, 13(4), 765-775.DOI
- [37]KENNEY, S., POPER, K., CHAPAGAIN, G., & CHRISTOPHER, G. F. 2013 Large Deborah number flows around confined microfluidic cylinders. Rheol. Acta, 52 (5), 485-497.DOI
- [38]KUMAR, M., & ARDEKANI, A. M. 2022 Hysteresis in viscoelastic flow instability of confined cylinders. Phys. Rev. Fluids, 7(9), 093302.DOI
- [39]LARSON, R. G. 2000 Turbulence without inertia. Nature, 405, 27–28.pubmed.ncbi.nlm.nih.gov/10811202
- [40]LARSON, R. G. 1999 The Structure and Rheology of Complex Fluids. Oxford University Press.
- [41]LEE, J. S., DYLLA-SPEARS, R., TECLEMARIAM, N.-P., & MULLER, S. J. 2007 Microfluidic four-roll mill for all flow types. Appl. Phys. Lett., 90, 074103.DOI
- [42]LEE, J., HWANG, W. R., & CHO, K. S. 2021 Effect of stress diffusion on the Oldroyd-B fluid flow past a confined cylinder. J. Non-Newton. Fluid Mech., 297, 104650.DOI
- [43]MARSDEN, A. L. 2014 Optimization in cardiovascular modeling. Annu. Rev. Fluid Mech., 46, 519-546.DOI
- [44]McKinley, G. H., Pakdel, P., & Öztekin, A. 1996 Rheological and geometric scaling of purely elastic flow instabilities. J. Non-Newton. Fluid Mech., 67, 19-47.DOI
- [45]MOKHTARI, O., LATCHÉ, J. C., QUINTARD, M., & DAVIT, Y. 2022 Birefringent strands drive the flow of viscoelastic fluids past obstacles. J. Fluid Mech., 948, A2.DOI
- [46]MOROZOV, A. 2022 Coherent structures in plane channel flow of dilute polymer solutions with vanishing inertia. Phys. Rev. Lett., 129(1), 017801.DOI
- [47]NOLAN, K. P., AGARWAL, A., LEI, S., & SHIELDS, R. 2016 Viscoelastic flow in an obstructed microchannel at high Weissenberg number. Microfluid. Nanofluid., 20 (7), 1-12.DOI
- [48]OLDROYD, J.G. 1950 On the formulation of rheological equations of state. Proc. Roy. Soc. London Ser A 200, 523–541.DOI
- [49]OLSSON, F., & YSTRÖM, J. 1993 Some properties of the upper convected Maxwell model for viscoelastic fluid flow. J. Non-Newton. Fluid Mech., 48, 125-145.
- [50]PAN L., MOROZOV A., WAGNER C., ARRATIA P. E. 2013 Nonlinear elastic instability in channel flows at low Reynolds numbers, Phys. Rev. Lett., 110 (17), 174502.pubmed.ncbi.nlm.nih.gov/23679736
- [51]PIMENTA, F., & ALVES, M. A., Rheotool 2018 https://github.com/fppimenta/rheoTool.
- [52]PIMENTA, F., & ALVES, M. A. 2019 A coupled finite-volume solver for numerical simulation of electrically-driven flows. Comput. Fluids, 193, 104279.DOI
- [53]PENG, S., LI, J. Y., XIONG, Y. L., XU, X. Y., & YU, P. 2021 Numerical simulation of two dimensional unsteady Giesekus flow over a circular cylinder. J. Non-Newton. Fluid Mech., 294, 104571.
- [54]POOLE, R. J. 2019 Three-dimensional viscoelastic instabilities in microchannels. J. Fluid Mech., 870, 1-4.DOI
- [55]POOLE, R. J., ALVES, M. A., & OLIVEIR, P. J. 2007 Purely elastic flow asymmetries. Phys. Rev. Lett., 99 (16), 164503.pubmed.ncbi.nlm.nih.gov/17995258
- [56]PURNODE, B., & CROCHET, M. J. 1998 Polymer solution characterization with the FENE-P model. J. Non-Newton. Fluid Mech., 77, 1-20.DOI
- [57]QIN, B., SALIPANTE, P. F., HUDSON, S. D. & ARRATIA, P. E. 2019a Upstream vortex and elastic wave in the viscoelastic flow around a confined cylinder. J. Fluid Mech., 864, R2.
- [58]QIN, B., SALIPANTE, P. F., HUDSON, S. D., & ARRATIA, P. E. 2019b Flow resistance and structures in viscoelastic channel flows at low Re. Phy. Rev. Lett., 123 (19), 194501.DOI
- [59]RICHTER, D., IACCARINO, G., & SHAOFEH, E. S. 2010 Simulations of three-dimensional viscoelastic flows past a circular cylinder at moderate Reynolds numbers. J. Fluid Mech., 651, 415-442.DOI
- [60]RIBEIRO, V. M., COELHO, P. M., PINHO, F. T., & ALVES, M. A. 2014 Viscoelastic fluid flow past a confined cylinder: Three-dimensional effects and stability. Chem. Eng. Sci., 111, 364-380.
- [61]SAMANTA, D., DUBIEF, Y., HOLZNER, M., SCHÄFER, C., MOROZOV, A.N., WAGNER, C. & HOF, B. 2013 Elasto-inertial turbulence. Proc. Natl Acad. Sci. USA, 110 (26), 10557–10562.DOI
- [62]SCHIAMBERG, B. A., SHEREDA, L. T., HU, H. & LARSON, R. G. 2006 Transitional pathway to elastic turbulence in torsional, parallel-plate flow of a polymer solution. J. Fluid Mech., 554, 191–216.DOI
- [63]SHAQFEH, E. S. 1996 Purely elastic instabilities in viscometric flows. Annu. Rev. Fluid Mech., 28(1), 129-185.
- [64]SHI, X., & CHRISTOPHER, G. F. 2016 Growth of viscoelastic instabilities around linear cylinder arrays. Phys. Fluids, 28 (12), 124102.DOI
- [65]SHI, X., KENNEY, S., CHAPAGAIN, G., & CHRISTOPHER, G. F. 2015 Mechanisms of onset for moderate Mach number instabilities of viscoelastic flows around confined cylinders. Rheol. Acta, 54 (9), 805-815.DOI
- [66]SIPP, D., & LEBEDEV, A. 2007 Global stability of base and mean flows: a general approach and its applications to cylinder and open cavity flows. J. Fluid Mech., 593, 333-358.DOI
- [67]STEINBERG, V. 2021 Elastic turbulence: an experimental view on inertialess random flow. Annu. Rev. Fluid Mech., 53, 27–58.DOI
- [68]TAMANO, S., HAMANAKA, S., NAKANO, Y., MORINISHI, Y., & YAMADA, T. 2020 Rheological modeling of both shear-thickening and thinning behaviors through constitutive equations. J. Non-Newton. Fluid Mech., 283, 104339.DOI
- [69]TANG, T., YU, P., SHAN, X., LI, J., & YU, S. 2020 On the transition behavior of laminar flow through and around a multi-cylinder array. Phys. Fluids, 32(1), 013601.DOI
- [70]THIÉBAUD, M., SHEN, Z., HARTING, J., & MISBAH, C. 2014 Prediction of anomalous blood viscosity in confined shear flow. Phys. Rev. Lett., 112 (23), 238304.pubmed.ncbi.nlm.nih.gov/24972235
- [71]THIEN, N. P., & TANNER, R. I. 1977 A new constitutive equation derived from network theory. J. Non-Newton. Fluid Mech., 2 (4), 353-365.DOI
- [72]ULTMAN, J. S., & DENN, M. M. 1971 Slow viscoelastic flow past submerged objects. Chem. Eng. J., 2 (2), 81-89.DOI
- [73]VARCHANIS, S., HOPKINS, C. C., SHEN, A. Q., TSAMOPOULOS, J., & HAWARD, S. J. 2020 Asymmetric flows of complex fluids past confined cylinders: A comprehensive numerical study with experimental validation. Phys. Fluids, 32 (5), 053103.DOI
- [74]VARCHANIS, S., PETTAS, D., DIMAKOPOULOS, Y., & TSMOPOULOS, J. 2021 Origin of the Sharkskin Instability: Nonlinear Dynamics. Phys. Rev. Lett., 127 (8), 088001.DOI
- [75]VARSHNEY, A., & STEINBERG, V. 2017 Elastic wake instabilities in a creeping flow between two obstacles. Phys. Rev. Fluids, 2 (5), 051301.DOI
- [78]WALKAMA, D. M., WAISBORD, N., & GUASTO, J. S. 2020 Disorder suppresses chaos in viscoelastic flows. Phys. Rev. Lett., 124 (16), 164501.DOI
- [79]WAN, D., SUN, G., & ZHANG, M. 2021 Subcritical and supercritical bifurcations in axisymmetric viscoelastic pipe flows. J. Fluid Mech., 929, A16.DOI
- [80]WELLER, H. G., TABOR, G., JASAK, H., & FUREBY, C. 1998 A tensorial approach to computational continuum mechanics using object-oriented techniques. Comput. Phys., 12(6), 620.DOI
- [81]WILLIAMSON, C. H. K. & ROSIRKO, A. 1988 Vortex formation in the wake of an oscillating cylinder. J. Fluids Struct., 2, 355-381.DOI
- [82]YAMANI, S., & MCKINLEY, G. H. 2022 Master curves for FENE-P fluids in steady shear flow. arXiv preprint arXiv:2208.12393.arxiv.org/abs/2208.12393
- [83]YUE, P., DOOLEY, J., & FENG, J. J. 2008 A general criterion for viscoelastic secondary flow in pipes of noncircular cross section. J. Rheol., 52 (1), 315-332.DOI
- [84]ZHAO, Y., SHEN, A. Q., & HAWARD, S. J. 2016 Flow of wormlike micellar solutions around confined microfluidic cylinders. Soft Matter, 12 (42), 8666-8681.pubmed.ncbi.nlm.nih.gov/27722471

