betaFeedback

WorksPhysics and Astronomy

Stability of acoustic streaming jets

Abstract

We study the stability of a steady Eckart streaming jet that is acoustically forced at one end of a closed cylindrical cavity and impinges the wall at the other end, where a recirculation forms. This configuration generically represents industrial processes where acoustic forcing offers a contactless means of stirring or controlling confined flows. Successfully doing so, however, requires sufficient insight into the topology of the acoustically forced flow. This raises the question of whether the base acoustic streaming jet is stable and, when not, of which alternative states emerge. Using Linear Stability Analysis (LSA) and three-dimensional nonlinear simulations, we identify the instability mechanisms and determine the nature of the bifurcations that ensue. We show that the ratio CRC_R between the cavity and the maximum beam radii determines the dominant unstable mode. For 4≤CR≤64 \leq C_R \leq 6, a non-oscillatory perturbation rooted in the jet impingement triggers a supercritical bifurcation. For CR=3C_R = 3, the flow destabilises through a subcritical non-oscillatory bifurcation. Further reducing CRC_R increases the shear within the flow, and gradually relocates the instability in the shear layer between impingement-induced vortices: for CR=2C_R = 2, an unstable travelling wave grows out of a subcritical bifurcation, which becomes supercritical for CR=1C_R=1. For each geometry, the nonlinear 3D simulations validate the LSA, identify the saturated nonlinear state and its stability. This study offers fundamental insight into the stability of acoustically-driven flows in general, but also opens possible pathways to either induce turbulence acoustically, or to avoid it in realistic configurations.

Banner appropriate to article type will appear here in typeset article

Key words: —

MSC Codes (Optional) Please enter your MSC Codes here

Introduction

This work is concerned with the stability of a jet flow driven by an axisymmetric beam of attenuated travelling sound waves in a closed cavity. We focus here on the effect of confinement on the stability of the jet, and shed light on the very unstable nature of such flows.

Sound-driven flows are referred to as acoustic streaming, and result from a nonlinear process producing momentum out of the attenuation of sound waves. Acoustic streaming can be divided into two categories, depending on whether the flow is forced by either standing (Rayleigh-Schlichting streaming [58] (Westervelt 1953)) or travelling (Eckart streaming [26]) sound waves. Since this work focuses exclusively on flows driven by travelling sound waves, any mention of acoustic streaming from now shall refer to Eckart streaming, unless specified otherwise.

As they allow for remote flow creation, both Rayleigh and Eckart streaming recently received a renewed interest in fields for which geometrical constraints prevent fluid actuation by mechanical means, e.g., using a propeller. This is typically the case in microfluidics, where Rayleigh and Eckart streaming are commonly used to promote flow in microcavities and microchannels [33, 31, 30, 54], in droplets [2, 59], and around oscillating bubbles [16, 28] (Doinikov et al. 2022).

At a larger scale, acoustic streaming offers an advantageous mean of actuating fluids whilst avoiding the contamination inherent to mechanical contact. This situation commonly arises in metals and semiconductors solidification processes, during which flows in the melt may affect the quality of the solid ingot. By stabilising convective flows within the melt [24], acoustic streaming is known to enhance the homogenity of the solid ingots [41]. Solidification processes may also benefit from a local control of the grain growth obtained by directing a streaming jet towards the solidification front [45]. More recently, Absar et al. [1] showed that nanocomposites manufacturing can also benefit from acoustic streaming to homogenise the distribution of fibers and particles within the metal matrix. All these applications of acoustic streaming to materials manufacturing rely on low-frequency and high-power ultrasounds (typically a few kilohertz and hundreds of watts) which, close to the source, create a cloud of cavitation bubbles acting as the main sound attenuation mechanism driving the flow. Although this setting leads to strong jets, it suffers from the loss of beam coherence at short distances from the source. This prevents any remote control of the flow. We are instead interested in using beams of low-power and megahertz sound waves to extend the range over which controlled flow actuation is possible [39, 48]. By acting deep into the bulk of the flow, such a choice of parameters may open new possibilities of using acoustic streaming in solidification processes. Examples include the local enhancement of mass transfer at the solidification front with an elongated jet [27], and the creation of three-dimensional flows using multiple reflections of a single beam to stir the melt [68]. Still, whether the purpose is to drive a complex chaotic flow [8, 44] (Qu et al. 2022) or to stabilise the melt [24], these applications require a detailed understanding of the flow patterns for a given configuration and beam intensity. Hydrodynamic instabilities may lead to the formation of flow patterns with a topology that significantly differs from that of the forcing beam and even to turbulence. Whether such states are desired or not, understanding how a streaming jet may destabilise is key to designing optimal industrial setups involving acoustic streaming.

The destabilisation of Eckart streaming jets has been observed in several experiments, such as in those of Green et al. [32]. The authors studied a streaming flow in a vertical cylindrical cavity filled with a mixture of diethyl phthalate and ethanol. The circular transducer, placed at the top endwall of the cavity, fired bursts of ultrasonic waves travelling towards the opposite end of the fluid volume, where a layer of sound-absorbing material prevented the backward reflection of the incident sound waves. This setting created a steady streaming flow, which however transitioned towards another state after some time due to a thermal instability arising from the sound-absorbing plate progressively heating up. Moudjed et al. [52] also reported strong distortion of the streaming jet when increasing the forcing magnitude. These severe and intermittent jet oscillations systematically occurred near the wall facing the transducer, suggesting that the impingement plays a major role on the flow stability.

These experimental observations of acoustic streaming destabilisation motivated theoretical stability analyses. The first acoustic streaming stability work traces back to Dridi et al. [25], who considered an analytical 1D-1C (one-dimensional, one-component) jet velocity profile in a 2D infinite layer. For various ratios of the source height to the layer thickness, the authors showed that the flow is destabilised by a wave travelling in the direction facing the acoustic forcing. The frequency of that wave decreases as the layer height is increased with respect to the source height. Later, multidimensional simulations in a closed cavity revealed that the destabilisation is indeed rooted in the jet impingement ([4]), as observed by ([52]). Since then, numerous studies focused on how acoustic streaming may stabilise flows driven by other phenomena such as lateral ([23]) and vertical (i.e., Rayleigh-Bénard) heating ([36]), gradient of chemical concentration ([47]) and thermodiffusion ([13, 14]). All these studies, however, suffer from a rather crude modelling of the acoustic field driving the flow. First, none of these works considered beam diffraction. The radial enlargement of the beam not only prescribes a radial length scale to the jet ([51, 67]), but also locally reduces the magnitude of the acoustic forcing due to the conservation of acoustic power along the beam. As such, these two effects create longitudinal gradients of the streamwise jet velocity ([39, 48, 19]) that cannot be recovered in a non-diffracting approximation. Second, none of the former stability studies accounted for the decay of the forcing along the beam caused by sound attenuation. This approximation violates two fundamental properties of streaming jets: First, attenuation causes the streamwise jet momentum flux to saturate at infinite distances from the source. Second, attenuation defines the length scale over which the jet builds up momentum ([46]). As a consequence, this non-attenuated approximation actually severely overestimates the jet velocity, and fails to capture the jet’s structure. Therefore, the causes of destabilisation of realistic jets forced by a diffracting beam of attenuated sound waves still remains an open question.

The aim of this work is precisely to determine how a laminar streaming jet driven by an attenuated and diffracting ultrasound beam may become unstable. We shall consider the simplest geometry for this purpose, i.e. a closed cylindrical cavity with impermeable and sound-absorbing boundaries. The acoustic source, placed at one cavity end, radiates an attenuated beam driving a swirl-free axisymmetric jet. The key novelty is to use a realistic model for the jet accounting for both attenuation and diffraction based on our recent work ([67]). With these models for the cavity and acoustic beam, the problem loses streamwise invariance. The base flow must be determined by numerical simulations and its bi-global stability is carried out numerically too. To assess the influence of confinement, we consider different cavity sizes based on the attenuation length of the sound waves and on the maximum beam diameter. We address the problem by means of numerical methods based on high-order spectral elements to answer the following questions:

  1. What are the mechanisms underpinning the flow’s instability and what are the conditions of their onset?

  1. What is the type (oscillatory or not) of the unstable perturbation, and how does its topology change when varying the cavity size with respect to the size of the forcing field?

  1. What is the nature of the bifurcations leading to these modes, and in particular, are they subcritical or supercritical?

  1. What are the nonlinear states arising out of the instability?

This work is organised as follows: we shall first define the problem along with the governing equations in § 2, before presenting in § 3 the computational methods together with the grid convergence analyses in § 3. We then analyse the effect of confinement on the flow structure in § 4, before turning to the LSA in § 5. Then, we shall determine the bifurcation criticalities for all cavity sizes in § 6. Concluding remarks are given in § 7.

Formulating the stability problem for jets driven by attenuated and diffracting progressive waves

Studied configuration and governing equations

We consider an axisymmetric acoustic streaming jet flowing in a closed cylindrical cavity (Figure 1). The problem is defined using a cylindrical coordinate system in which ex\mathbf{e}_x, er\mathbf{e}_r and eθ\mathbf{e}_\theta refer to the axial, radial and azimuthal directions, respectively. The spatial coordinates in the corresponding directions are xx, rr and θ\theta; the origin is centred on the upstream wall (Figure 1) so that the cavity axis is r=0r=0. Finally, we shall define the hydrodynamics and acoustics problems using the same dimensionless parameters and variables as in our previous work ([67]).

Figure 1. Sketch of the investigated setup in the (x,r)(x,r) plane, where xx and rr are the axial and radial coordinates, respectively. The circular transducer of unit diameter (grey rectangle) at x=0x=0 emits a beam of linear sound waves (green shaded area) within an elongated cylindrical cavity of length LL and diameter DD. As the acoustic pressure waves travel across the cavity, their amplitude decay at a rate N/LN/L, where NN is the ratio between the cavity length and the acoustic pressure attenuation length. Attenuation of these waves yields a body force driving the streaming flow of velocity u\mathbf{u}. All the boundaries are impermeable walls and completely absorb the acoustic waves.

The cavity, of diameter DD and length LL, is entirely filled with a Newtonian fluid. The cavity is fitted with a plane circular transducer of unit diameter at x=0x=0. The transducer, whose axis is aligned with ex\mathbf{e}_x, radiates an axisymmetric beam of linear acoustic pressure waves. As these waves travel towards the opposite cavity end, their amplitude decays for two reasons: diffraction, which also causes the beam radius to increase at a rate SS (Figure 1), and sound attenuation. The latter gives rise to an axisymmetric body force driving a streaming flow of velocity u\mathbf{u}. The exponential decay rate of the acoustic pressure wave amplitudes is defined by the sound attenuation coefficient N/LN/L, where NN is the ratio between the cavity length and the attenuation length of the acoustic pressure wave amplitude. Finally, all cavity walls absorb incident sound waves thus preventing the development of standing sound waves.

Whilst the sound field is characterised by small length scales and fast time scales, the streaming flow they give rise to is significantly slower than the speed of sound and involves length scales that greatly exceed the acoustic wavelength (Lighthill [46]; Orosco & Friend [55]). These scale differences allow to model streaming flows using the continuity and Navier-Stokes equations for an incompressible Newtonian fluid together with a body force (Moudjed et al. [51]):

∇⋅u=0,(1)\nabla\cdot\mathbf{u} = 0, \tag*{(1)}
∂u∂t+(u⋅∇)u=−∇p+∇2u+Fac,(2)\frac{\partial\mathbf{u}}{\partial t} + (\mathbf{u} \cdot\nabla)\mathbf{u} = -\nabla p + \nabla^{2}\mathbf{u} + \mathbf{F}_{ac}, \tag*{(2)}

in which pp refers to pressure, and

Fac=GracI~(3)\mathbf{F}_{ac} = Gr_{ac}\widetilde{\mathbf{I}} \tag*{(3)}

is the body force coupling the acoustics and hydrodynamics problems. The spatial structure of Fac\mathbf{F}_{ac} is set by the normalised acoustic intensity I~\widetilde{\mathbf{I}}, which is bounded by unity for an unattenuated sound field. The magnitude of the forcing is defined by the Grashof number GracGr_{ac}, which compares the magnitude of the acoustic force to the magnitude of the viscous forces at the scale of the transducer (El Ghani et al. [27]). Finally, the differences between the acoustic and flow scales prevent any alteration of the forcing field due to the flow; therefore, Fac\mathbf{F}_{ac} is time-independent.

The cavity walls are impermeable boundaries and thus impose:

u=0\mathbf{u} = 0

at x∈{0,L}x \in\{0,L\} and r=D/2r = D/2. Besides, the flow is axisymmetric and swirl-free, so that for all xx and rr:

uθ=∂ux∂θ=∂ur∂θ=∂p∂θ=0,u_{\theta} = \frac{\partial u_{x}}{\partial\theta} = \frac{\partial u_{r}}{\partial\theta} = \frac{\partial p}{\partial\theta} = 0,

where uxu_x, uru_r and uθu_{\theta} refer to the axial, radial and azimuthal velocity components, respectively. Axisymmetry of both u\mathbf{u} and pp further imposes

ur=∂ux∂r=∂p∂r=0u_r = \frac{\partial u_x}{\partial r} = \frac{\partial p}{\partial r} = 0

on the axis (r=0r = 0).

The transducer is modelled as a circular and plane baffled piston vibrating at a uniform velocity. The vibrating surface of the transducer is represented by an infinite number of acoustic point sources, each contributing to the acoustic field at any point (x,r)(x,r) of the fluid domain. The classical equations for a plane baffled piston in a semi-infinite domain, such as those found in Blackstock [7], were adapted by Vincent et al. [67] to account for viscous attenuation. The authors showed that I~\widetilde{\mathbf{I}} can be obtained from a set of integrals defining the acoustic pressure p~ac\tilde{p}_{ac} and velocity u~ac\tilde{\mathbf{u}}_{ac} at any point (x,r)(x,r) in the fluid domain:

I~=R{(p~acu~ac)}(4)\tilde{I}=\mathfrak{R}\left\{\begin{pmatrix}\tilde{p}_{ac}\\ \tilde{\mathbf{u}}_{ac}\end{pmatrix}\right\} \tag*{(4)}
p~ac=i8×1.22π2∫01/2∫02πk2Se−ik∥d∥∥d∥ r′ dr′ dθ′(5)\tilde{p}_{ac}=\frac{i}{8\times1.22\pi^2}\int_{0}^{1/2}\int_{0}^{2\pi}\frac{k^2S e^{-ik\lVert\mathbf{d}\rVert}}{\lVert\mathbf{d}\rVert}\,r'\,\mathrm{d}r'\,\mathrm{d}\theta' \tag*{(5)}
u~ac=14π∫01/2∫02π(1+ik∥d∥)e−ik∥d∥∥d∥2d∥d∥ r′ dr′ dθ′(6)\tilde{\mathbf{u}}_{ac}=\frac{1}{4\pi}\int_{0}^{1/2}\int_{0}^{2\pi}\left(1+ik\lVert\mathbf{d}\rVert\right)\frac{e^{-ik\lVert\mathbf{d}\rVert}}{\lVert\mathbf{d}\rVert^2}\frac{\mathbf{d}}{\lVert\mathbf{d}\rVert}\,r'\,\mathrm{d}r'\,\mathrm{d}\theta' \tag*{(6)}
k=2.44πS−iNL(7)k=\frac{2.44\pi}{S}-i\frac{N}{L} \tag*{(7)}

where ∗* refers to the complex conjugate and R\mathfrak{R} to the real part. In equations (2.4), d=x ex+[r−r′cos⁡(θ′)]er−r′sin⁡(θ′)eθ\mathbf{d}=x\,\mathbf{e}_x+[r-r'\cos(\theta')]\mathbf{e}_r-r'\sin(\theta')\mathbf{e}_{\theta} is the distance between any point source on the transducer surface (located by the set of radial and azimuthal coordinates r′r' and θ′\theta'), and any point at (x,r)(x,r). Finally, I~\tilde{I} is mainly shaped by the complex wave number kk: the real part of kk sets the beam diffraction, whilst the imaginary part of kk involves the sound attenuation coefficient N/LN/L and thus adds attenuation to I~\tilde{I}.

Equations (2.4) define a beam-shaped axisymmetric acoustic field. As (i) the beam expands radially at a rate SS along xx, and (ii) the beam radius matches the transducer radius (0.5 in dimensionless coordinates) at x=0x=0, an approximate description of the beam radius RbeamR_{beam} is:

Rbeam=0.5+Sx(8)R_{beam}=0.5+Sx \tag*{(8)}

from which we define the confinement ratio CRC_R:

CR=D/2Rbeam∣x=L(9)C_R=\frac{D/2}{R_{beam}|_{x=L}} \tag*{(9)}

From now, we shall use CRC_R as the main indicator of the expected radial confinement of the flow; small (respectively large) CRC_R values correspond to cases of high (respective weak) confinement.

To investigate the effect of confinement on the flow stability, we shall consider six different cavity sizes by varying either CRC_R or $N (table {xref}1). We take $C_R\in[1,6]$ to be consistent with the experiments of {citet}Kamakura et al. [39], {citet}Mitome [48] and {citet}Moudjed et al. [51] for which $C_R=2.9$, 8.5 and 4.1, respectively. As these experiments were made for $N<1$, we set $N=0.25$ for five cases. Nevertheless, from a process standpoint, it is preferable to have $N\ge1$ to ensure that most of the acoustic kinetic energy radiated by the transducer is used to create flow kinetic energy ({citet}Vincent et al. 2024a`). We shall thus investigate a last case for which (N,CR)=(1,6)(N, C_R) = (1, 6) (Table 1). For all six cases, the parameters S=0.03S = 0.03 and N/L=0.003N/L = 0.003 are kept constant; these values are based on the 2 MHz water experiments of Moudjed et al. [51] and Moudjed et al. [53]. Therefore, the structure of Fac\mathbf{F}_{ac} in our work is consistent with the forcing fields of former experiments. These experiments covered a wide range of GracGr_{ac}; it typically ranged from 5.6×1025.6 \times10^{2} Mitome [48] to 1.2×1051.2 \times10^{5} Moudjed et al. [51]. In the present work, we shall adjust GracGr_{ac} on a case-by-case basis to identify the instability onset.

CRC_RNNSSLLDD
10.250.0383.36.02
20.250.0383.312.04
30.250.0383.318.06
40.250.0383.324.08
60.250.0383.336.12
610.03333.3126.51

Table 1. Values of the parameters defining each setup. The parameter CRC_R is defined as the ratio between the cavity radius D/2D/2 and the approximate beam radius evaluated at x=Lx=L using equation (8). The values of SS and of the acoustic pressure attenuation coefficient N/LN/L are chosen to match the 2 MHz water experiments of Moudjed et al. [51].

Linear stability analysis

Although the base flow is 2D-2C (two-dimensional, two-component), it may nevertheless be destabilised by non-axisymmetric 3D-3C perturbations under certain forcing and geometrical conditions. We shall therefore decompose the flow fields into a steady axisymmetric part corresponding to the base flow, and an infinitesimal time-dependent perturbation:

u=U(x,r)+u′(x,r,θ,t),(10)\mathbf{u} = \mathbf{U}(x,r) + \mathbf{u}'(x,r,\theta,t), \tag*{(10)}
p=P(x,r)+p′(x,r,θ,t).(11)p = P(x,r) + p'(x,r,\theta,t). \tag*{(11)}

The perturbations (u′,p′)(\mathbf{u}', p') are purely hydrodynamic i.e., they dot not correspond to any acoustic signals as feedback effects of the flow on the acoustic fields are discarded. The linear stability equations are then obtained by inserting the decompositions (10) and (11) into the Navier-Stokes (2) and continuity (1) equations, and by linearising the resulting equations around the base flow. Since (U,P)(\mathbf{U}, P) satisfy equations (2) and (1), the linear equations for (u′,p′)(\mathbf{u}', p') are thus:

∂u′∂t+(U⋅∇)u′+(u′⋅∇)U=−∇p′+∇2u′,(12)\frac{\partial\mathbf{u}'}{\partial t} + (\mathbf{U} \cdot\nabla)\mathbf{u}' + (\mathbf{u}' \cdot\nabla)\mathbf{U} = -\nabla p' + \nabla^{2}\mathbf{u}', \tag*{(12)}
∇⋅u′=0.(13)\nabla\cdot\mathbf{u}' = 0. \tag*{(13)}

As the base flow is invariant along eθ\mathbf{e}_{\theta}, a generic three-dimensional perturbation q′\mathbf{q}' is expressed using Fourier series:

q′(x,r,θ,t)=∑m=−∞+∞q^m′(x,r,t)eimθ,(14)\mathbf{q}'(x,r,\theta,t) = \sum_{m=-\infty}^{+\infty} \widehat{\mathbf{q}}'_m(x,r,t)e^{im\theta}, \tag*{(14)}

where the integer mm is the azimuthal wavenumber. Since equations (12) and (13) are linear, the LSA is thus decoupled for each mm. The linearity of equations (12) and (13) decouples the Fourier modes, so that the LSA (Linear Stability Analysis) equations can be solved individually for each mm. Finally, the perturbations satisfy the same boundaries conditions as the base flow, except at r=0r = 0 where mm-dependent kinematic conditions are imposed to allow for three-dimensional perturbations to develop whilst avoiding the geometrical singularity [6].

The linear stability problem is addressed using a timestepper approach. From equations (12) and (13), the perturbation at a time t+τt + \tau results from the action of an evolution operator A(τ)\mathcal{A}(\tau) applied to the same perturbation at a previous time tt [3]:

q′(t+τ)=A(τ)q′(t).\mathbf{q}'(t+\tau) = \mathcal{A}(\tau)\mathbf{q}'(t).

The complex eigenvalue ζk\zeta_k of A(τ)\mathcal{A}(\tau), associated with the eigenvector qk′\mathbf{q}'_k, is linked to the eigenvalue (σk+iωk)(\sigma_k + i\omega_k) of the linear problem (12)–(13) through an exponential mapping:

ζk=e(σk+iωk)τ,\zeta_k = e^{(\sigma_k+i\omega_k)\tau},

from which the growth rate σk\sigma_k and frequency ωk\omega_k can be retrieved:

σk=ln⁡(∣ζk∣)τ,ωk=φkτ,(15)\sigma_k=\frac{\ln\left(\lvert\zeta_k\rvert\right)}{\tau},\qquad\omega_k=\frac{\varphi_k}{\tau}, \tag*{(15)}

where φk\varphi_k is the complex argument of ζk\zeta_k. A given base flow (U,P)(\mathbf{U},P) is said to be unstable if, for any mm, there is at least one perturbation (u′,p′)(\mathbf{u}',p') for which σ>0\sigma>0 (or, equivalently, ∣ζ∣>1\lvert\zeta\rvert>1). Conversely, the base flow is stable if the growth rates associated with all the eigenvalues of the linear problem are strictly negative. The minimum value of GracGr_{ac} for which σ=0\sigma=0 is reached is called the critical acoustic Grashof number GraccGr_{ac}^{c} and indicates the onset of an instability.

Computational methodology

Numerical methods

Our work involves four different numerical calculations. First, we calculate Fac\mathbf{F}_{ac} at the grid points in the fluid domain. Second, we compute the steady axisymmetric 2D-2C streaming flow driven by Fac\mathbf{F}_{ac} by time-stepping equations (2)–(1) until a steady-state is reached. Third, we perform the LSA of the axisymmetric 2D-2C flow for each case listed in Table 1. The leading eigenmode and its eigenvalue are computed by solving the eigenvalue problem for 0≤m≤100\leq m\leq10. This allows to determine GraccGr_{ac}^{c} and identify the shape of the unstable perturbation for each case. Fourth, nonlinear 3D-3C unsteady simulations are run at slightly supercritical GracGr_{ac} for two purposes: (i) to validate the LSA, and (ii) to determine the bifurcation nature (sub- or supercritical). The latter is inferred by modelling the growth and saturation of a perturbation near GraccGr_{ac}^{c} with a Stuart-Landau equation (Landau & Lifschitz [43]), the parameters of which being evaluated from time series of the flow variables (Sheard et al. [63]).

The forcing field is obtained by numerically evaluating the integrals in equation (2.4) at the grid points. The transducer is discretised by a homogeneous distribution of NsN_s point sources in both the radial and the azimuthal directions. The total number of point sources used to estimate I~\tilde{\mathbf{I}} (hence Fac\mathbf{F}_{ac}) at any grid point in the fluid domain is thus Ns2N_s^2.

Both the steady axisymmetric 2D-2C and the unsteady 3D-3C flows are computed using the open-source spectral-element code Semtex (Blackburn et al. [5]). The (x,r)(x,r) plane is split into non-overlapping elements, inside which the flow variables are expanded using a tensor product of Lagrange polynomials defined on the Gauss–Lobatto–Legendre points (Blackburn & Sherwin [6]). Convergence of the solution is then achieved by increasing the polynomial degree NpN_p of the expansion basis. For the 3D-3C unsteady simulations, the azimuthal direction is discretised using a Fourier spectral method (Blackburn & Sherwin [6]).

The (x,r)(x,r) plane is discretised using a grid of quadrilateral elements (Figure 2). For instance, the mesh used for the (N,CR)=(0.25,4)(N,C_R)=(0.25,4) comprises 24 and 9 elements in the axial and radial directions, respectively. For each case, the grid is refined near the upstream and downstream walls, so that the ratio of the lengths of two neighbouring elements does not exceed 1.35. In the radial direction, the element height is increased by a factor 1.3 as the distance from the axis is increased. The grid density in the radial direction is set such that there is at least one element in 0≤r≤0.50 \leq r \leq0.5. We also adapted the total number of elements to keep a similar spatial resolution of FacF_{ac} in the very-near acoustic field (x≤1)(x \leq1) for each case (Table 2).

Figure 2. Typical mesh used to discretise the fluid domain. The circular transducer is placed at x=0x=0 and its axis is aligned with the r=0r=0 line. The beam, of approximate radius RbeamR_{\mathrm{beam}} (equation (8)), is shown in green. The displayed mesh corresponds to the (N,CR)=(0.25,4)(N,C_R)=(0.25,4) case.

NNCRC_RElements in axial directionElements in radial directionTotal number of elements
0.251245120
0.252247168
0.253249216
0.254249216
0.2562410240
163414476

Table 2. Characteristics of the meshes used for each case. See Table 1 for a complete list of the computational parameters defining each case.

Equations ((2.1)–(2)) are discretised in time using the operator splitting scheme of Karniadakis et al. [40]. The scheme integrates explicitly the inertia terms of equation (2), and implicitly its viscous terms. For the unsteady 3D-3C simulations, the temporal integration scheme is used in its third-order formulation. For the 2D-2C base flow calculations, as only steady solutions of equations (1)–((2)) are sought, the scheme is used in its first-order version. A steady state is then considered to be reached when the time series of the flow variables, recorded at different locations within the fluid domain, do not display variations with time up to a precision of seven significant figures.

We carry out the LSA using the open-source code DOG (Direct Optimal Growth, Barkley et al. [3]), which is based on a time-stepper method and a spectral element spatial discretisation. The code marches forward in time the linear stability equations (12)–((13)) using the second-order formulation of the operator splitting scheme of Karniadakis et al. [40]. The eigenvalues and associated eigenvectors of the stability problem are then recovered by applying the iterative algorithm of Barkley et al. [3].

Finally, for a given set of values of NN, CRC_R and GracGr_{ac}, the initial condition for the 2D-2C flow calculations is either set to u=0\boldsymbol{u}=0, or to a previous solution obtained for the same (N,CR)(N,C_R) but for a different GracGr_{ac} to accelerate the convergence towards a steady state. The unsteady 3D-3C simulations at slightly supercritical GracGr_{ac} are initialised with the steady axisymmetric base flow duplicated along eθe_\theta, plus a Gaussian white noise. The noise is evenly distributed over all the azimuthal wavenumbers, thus ensuring that the initial disturbance is not biased towards a preferred shape. The standard deviation of the white noise was set to 1 for the 3D-3C unsteady simulations run for CR=6C_R=6. As we reduced CRC_R, such standard deviation yielded very large initial disturbances, masking the early exponential growth of the perturbation impossible to observe. We thus reduced the noise standard deviation to 0.1 for CR∈{3,4}C_R \in\{3,4\}, and to 0.01 for CR∈{1,2}C_R \in\{1,2\}.

Grid sensitivity analysis

We assessed the suitability of the grids discretising the transducer and the fluid domain. For the transducer discretisation, we previously derived an analytical expression for the on-axis intensity I~x(x,r=0)\widetilde{I}_x(x,r=0), which we compared to the numerical evaluation of equation (2.4) using the methodology outlined in § 3.1 (Vincent et al. [67]). We showed that discretising the transducer with Ns=300N_s=300 grid points in both the radial and azimuthal directions ensures a local error of less than 1 % between the analytical and numerical intensity profiles on the axis (represented by the black curve and red stars, respectively, in Figure 3). All the results presented in this work were thus obtained with Ns=300N_s=300.

Figure 3. On-axis normalised acoustic intensity profile I~x(x,r=0)\widetilde{I}_x(x,r=0) for (N,CR)=(0.25,4)(N,C_R)=(0.25,4). The analytical profile, proposed by Vincent et al. [67], is shown in black. The transition between the near and far acoustic fields occurs at the last intensity peak located at approximately the Fresnel distance LF≈1.22/(4S)L_F \approx1.22/(4S) (purple dashed vertical line). The values obtained by numerically evaluating equations (2.4) at the collocation points of the mesh (with an expansion basis of degree Np=8N_p=8 in each element) are represented by red stars. The numerical points are obtained by discretising the transducer with Ns=300N_s=300 point sources in both the radial and the azimuthal directions.

As shown in Figure 3, the forcing field features strong gradients in the near acoustic field. These sharp intensity variations occur over very short length scales and would require an extremely fine grid to be accurately captured. As these length scales are significantly smaller than those of the studied streaming flow (Vincent et al. [67]), the very near acoustic field (i.e., x≤1x \le1) is undersampled to keep the flow calculations tractable. We nevertheless performed pp-refinement to determine the minimum NpN_p above which the variations of the leading eigenvalue remained below 0.5 %. An example of grid sensitivity analysis for the (N,CR)=(0.25,4)(N,C_R)=(0.25,4) case at Grac=7000Gr_{ac}=7000 (the largest GracGr_{ac} for that setup) is shown in Table 3. The leading perturbation is an unstable non-oscillatory m=2m=2 mode. We chose Np=8N_p=8, since variations of σ\sigma with respect to the most refined grid were at most ≈0.1\approx0.1 % for larger NpN_p. We then repeated this pp-refinement study for all (N,CR)(N,C_R) to ensure that our stability results are mesh-independent.

NpN_pσmax⁡\sigma_{\max}Relative difference (%)
50.1024933.11
60.073794.17
70.077921.20
80.076950.05
90.077100.13
100.077090.12
110.076960.05
120.077020.03
130.077010.02
140.07700—

Table 3. Evolution of the leading mode growth rate σmax⁡\sigma_{\max} with the polynomial degree NpN_p of the expansion basis. The growth rates are obtained for (N,CR)=(0.25,4)(N,C_R)=(0.25,4) with Grac=7000Gr_{ac}=7000 and m=2m=2. The error is relative to the value of σmax⁡\sigma_{\max} obtained for Np=14N_p=14.

For the 3D-3C unsteady simulations, u\boldsymbol{u} and pp were expanded along eθ\boldsymbol{e}_\theta using NFN_F Fourier modes. We assessed the effect of NFN_F by computing the flow kinetic energy over the entire cavity after the perturbation had saturated. For (N,CR)=(0.25,1)(N,C_R)=(0.25,1) with Grac=15500Gr_{ac}=15500 for instance, the relative difference between the kinetic energies computed with either NF=32N_F=32 or 64 was ≈0.03\approx0.03 %. Even smaller differences were obtained for the CR∈{4,6}C_R \in\{4,6\} cases at supercritical forcings. The 3D-3C unsteady simulations were thus performed with NF=32N_F=32 for all cases.

Finally, we adapted the time step Δt\Delta t to each case so that the Courant–Friedrichs–Lewy (CFL) number remained below unity to ensure the numerical stability. For instance, the (N,CR)=(0.25,4)(N,C_R)=(0.25,4) case at Grac=7000Gr_{ac}=7000 and with Np=8N_p=8 was time-stepped with Δt=10−4\Delta t=10^{-4}, yielding a maximum CFL number of 0.476.

Two-dimensional axisymmetric base flow

Impinging jet structure of the streaming flow in a weakly confined setting

Before studying its stability, we shall first discuss the steady two-dimensional and axisymmetric base flow. Figure 4 shows a typical base flow considered for LSA, with the (N,CR)=(0.25,6)(N,C_R)=(0.25,6) case taken as illustrative example. The main flow features are identified by the patterns of streamlines together with critical point theory ([15]). The critical point analysis is commonly used to identify structures within complex 3D flows ([38]). It was first introduced to analyse the wall-stress fields of flows around bluff bodies in wind-tunnels but we showed that its application to the complex flow field produced by 3D DNS (Direct Numerical Simulations) of separated flows around bluff bodies ([21, 22]) or complex conduits ([56]) offered a powerful way to reliably extract the main structures governing the flow’s dynamics. The main idea is to extract 2D streamlines or stresslines along symmetry planes or solid walls, and to classify the critical points, where both components of velocity or stress cancel, into nodes and saddles. First, the theoretical constraint on the number of saddles and nodes offers a convenient way to validate the numerical simulations ([38]). Second, these points offers a simple parametrisation of the entire flow field. When a flow transitions between regimes, topological changes are straightforwardly captured by changes in the number of critical points, and the nature of these points often points to the physical mechanism underpinning the transition. We use it here for the first time in the context of stability analysis, and apply it separately to the base flow and to the perturbations.

Map of the steady velocity magnitude

Figure 4. Map of the steady velocity magnitude ∥U∥\lVert U\rVert computed for (N,CR)=(0.25,6)(N,C_R)=(0.25,6) at Grac=6.4×103Gr_{ac}=6.4\times10^{3}. The streamlines are displayed together with the base flow critical points: half saddles (zero wall skin friction, red squares) are labelled as Si′S_i^{\prime} and NiN_i are nodes (U=0U=0, purple discs). The green line depicts the approximate beam radius given by equation (8). The base flow is axisymmetric and is reflected about the r=0r=0 axis for the sake of visualisation.

The critical points where U=0U=0 are shown in purple (labelled SiS_i for saddles and NiN_i for nodes, respectively) and points of zero skin friction on the solid boundaries, displayed in red, are half saddles labelled as Si′S'_i. The base flow field typically comprises a jet aligned with the cavity axis. Being driven by the acoustic forcing, the jet originates at S1′S'_1 and impinges the wall facing the transducer. The jet impingement at S8′S'_8 then gives rise to a large roll centred on N4N_4. This roll features weak velocity amplitudes, and spreads over the entire space available between the jet and the lateral cavity wall. The presence of this roll causes the return flow to separate from and reattach to the wall at S5′S_5^{\prime} and S4′S_4^{\prime}, respectively. A recirculation bubble centred at N3N_3 lies between these two points.

The distribution and type of critical points within a flow field obey a strict topological rule. More precisely, the number of nodes ΣN\Sigma_N, of half-nodes ΣN′\Sigma_{N^{\prime}}, of saddles ΣS\Sigma_S and half-saddles ΣS′\Sigma_{S^{\prime}} in a slice of UU are related to each other through (Hunt et al. [38]; Foss [29]):

2ΣN+ΣN′−2ΣS−ΣS′=χ,(16)2\Sigma_N+\Sigma_{N^{\prime}}-2\Sigma_S-\Sigma_{S^{\prime}}=\chi, \tag*{(16)}

where χ\chi is a parameter defining the topology of the surface in which the streamline patterns are studied (χ=2\chi=2 for the plane shown in Figure 4 (Foss [29])). Inspection of the streamlines in Figure 4 gives:

ΣN=8,ΣN′=0,ΣS=0,ΣS′=14,\Sigma_N=8,\qquad\Sigma_{N^{\prime}}=0,\qquad\Sigma_S=0,\qquad\Sigma_{S^{\prime}}=14,

so that the left-hand side of equation (16) is indeed equal to χ=2\chi=2. The topological rule (16) is thus satisfied, meaning either that the entire flow topology is captured, or at least that unresolved parts of the flow are consistently cut off. Indeed, only one out of the infinite series of corner vortices theorised by Moffatt [49] is captured in each corner. Increasing the resolution would reveal a greater number of them. By “consistently”, we mean that with each additional vortex, a node and two half-saddles would add to the tally of critical points, without invalidating equation (16).

Effect of confinement on the base flow topology and on the jet velocity

Reducing the cavity size by either varying NN or CRC_R may affect the 2D-2C base flow structure. However, any topological modification of the base flow shall comply with the topological rule 4.1. Figure 5 shows the steady axisymmetric field obtained at the same GracGr_{ac} as in Figure 4, but for the smaller value of CR=4C_R=4. The increased radial confinement yields two additional critical points: a node N5N_5 at x≈66x\approx66 and a saddle S1S_1 at x≈72x\approx72. This new critical point combination keeps the left-hand side of the topological rule (16) identical to the CR=6C_R=6 case (Figure 4); the base flow for CR=4C_R=4 shown in Figure 5 is thus again valid from a topological perspective. Asides of these examples, a similar change of the base velocity field topology has only been observed when reducing CRC_R from 2 to 1 and for forcing magnitudes close to their respective Gracc\mathrm{Gr}_{ac}^{c}.

Figure 5. Map of the steady velocity magnitude ∥U∥\lVert\boldsymbol{U}\rVert computed for (N=0.25N=0.25, CR=4C_R=4) for Grac=6.4×103\mathrm{Gr}_{ac}=6.4\times10^3. The streamlines are displayed together with the base flow critical points: half saddles (zero wall skin friction, red squares) are labelled as Si′S_i', NiN_i are nodes (U=0\boldsymbol{U}=0, purple discs) and SiS_i are saddles (U=0\boldsymbol{U}=0, purple squares). The green line depicts the approximate beam radius given by equation (8). The base flow is axisymmetric and is reflected about the r=0r=0 axis for the sake of visualisation.

Besides potentially modifying the base flow topology, reducing CRC_R also affects the jet velocity. Figure 6 shows several profiles of the centreline base velocity Ux(x,r=0)U_x(x,r=0) computed for 1≤CR≤61\le C_R\le6 and Grac=6400\mathrm{Gr}_{ac}=6400. In a weakly-confined setting, the jet strongly accelerates over 0≤x≲200\le x\lesssim20 for all CRC_R as a result of a balance between Fac\boldsymbol{F}_{ac} and the inertia forces ([51, 67]). Decreasing CRC_R enhances the jet deceleration at larger xx. For 2≤CR≤62\le C_R\le6, this confinement-induced jet deceleration is marginal: Ux(x,r=0)U_x(x,r=0) differs by at most 5.5 %, and its decrease along the jet results from a balance between Fac\boldsymbol{F}_{ac} and viscous forces ([67]). However, the jet deceleration is significantly stronger for CR=1C_R=1: near the jet impingement, the centreline velocity is 50 % less than for the CR≥2C_R\ge2 cases. This enhanced jet deceleration is caused by the build-up of a streamwise pressure gradient as CRC_R is decreased (inset of Figure 6). This pressure force acts against the acoustic forcing and drives the return flow so that the net mass flux through a cross section is zero ([60]).

Figure 6. Profiles of the on-axis velocity Ux(x,r=0)U_x(x,r=0) along the jet illustrating the effect of flow confinement on the jet velocity for N=0.25N=0.25 and Grac=6400Gr_{ac}=6400. The profiles are obtained for N=0.25N=0.25 and 1≤CR≤61 \le C_R \le6. The dashed vertical line corresponds to the Fresnel distance LFL_F marking the transition between the near (x<LFx<L_F) and the far (x>LFx>L_F) acoustic fields. Inset: variations with CRC_R of the streamwise pressure gradient ⟨dP/dx⟩\langle\mathrm{d}P/\mathrm{d}x\rangle of the base flow at the mid cavity length L/2L/2 and averaged over the cross-sectional area.

Stability analysis

Non-oscillatory instabilities in weakly-confined settings

We now turn to the linear stability analysis of the steady 2D-2C flows similar to those shown in Figure 4 and Figure 5. The variations of the growth rate of the leading eigenmode σ\sigma with GracGr_{ac} and mm for N∈{0.25,1}N \in\{0.25,1\} and 4≤CR≤64 \le C_R \le6 are shown in Figure 7. For 4≤CR≤64 \le C_R \le6, the same non-oscillatory m=2m=2 mode is found to destabilise the flow; we shall denote the associated mode branch as “Branch I”. The unstable mode branch remains the same for 4≤CR≤64 \le C_R \le6, but GraccGr_{ac}^{c} increases from 2191 for (N,CR)=(1,6)(N,C_R)=(1,6) to 6237 for (N,CR)=(0.25,4)(N,C_R)=(0.25,4). Confining the streaming jet flow thus has a stabilising effect, similarly to confined flows past bluff bodies ([50]).

Figure 7. Growth rate σ\sigma of the leading eigenmode as a function of the azimuthal wavenumber mm for 3≤CR≤63 \le C_R \le6. Oscillatory and non-oscillatory modes are represented by filled circles and triangles, respectively. The insets show the evolution of σ\sigma with GracGr_{ac} for the indicated value of mm. The critical Grashof number GraccGr_{ac}^{c}, obtained through spline interpolation of σ\sigma, is reported in red. The filled black symbols in the inset of (b) are additional eigenvalue computations made to improve the interpolation accuracy near σ=0\sigma= 0.

The topology of leading eigenmodes from Branch I is illustrated for (N,CR)=(0.25,4)(N,C_R)=(0.25,4) in Figure 8, with the critical points of the base flow overlaid. From the topology in the (x,r)(x,r) plane (Figure 8 (a) and (Figure 8b)), the perturbation intensity as measured by ux′u_x' is strongest in a region stretching from the impingement region of the jet to the shear layer between recirculations centred on N4N_4 and N5N_5. Radially, the perturbation is located within the acoustic beam (the approximate acoustic beam radius RbeamR_{\mathrm{beam}} is marked by a green line) but remains outside the jet (the points where UxU_x is 50 % of the on-axis velocity are marked by the blue line). The perturbation then crosses the shear layer at the boundary of the jet to reach the region between the two recirculations around N4N_4 and N5N_5. Hence, the mechanism is not associated to a shear layer instability but rather coincides precisely with the strong velocity gradients in the impingement region centred on S8′S_8', where the flow turns from its incoming streamwise direction to a radial one along the end wall. The azimuthal structure is best illustrated by contours of streamwise vorticity in constant xx planes containing either N5N_5, S1S_1 or N4N_4 (Figure 8 (f)-(Figure 8h)) showing a breakdown into an m=2m=2 modulation. These figures also clearly show that the streamwise vortices associated to these structures do not follow the shear layer (marked by the purple line corresponding to Ux=0U_x=0) and so exclude a shear layer instability mechanism.

Leading mode for $(N,C_R)=(0.25,4)$ at $Gr_{ac}=6400$

Figure 8. Leading mode for (N,CR)=(0.25,4)(N,C_R)=(0.25,4) at Grac=6400Gr_{ac}=6400. The mode is non-oscillatory (ω=0\omega=0) and unstable (σ=1.59×10−2>0\sigma=1.59\times10^{-2}>0). (a) Axial velocity perturbation ux′\boldsymbol{u}'_x in the (x,r)(x,r) plane, along with the streamlines and critical points of the base velocity U\boldsymbol{U}. (b) Velocity perturbation u′\boldsymbol{u}', with the background color corresponding to its azimuthal component uθ′u'_\theta. (c) to (h): Slices of u′\boldsymbol{u}' in constant-xx planes located by the blue vertical lines in (a) and (b), with either ux′u'_x or the streamwise vorticity perturbation ωx′\omega'_x as background colour (negative in blue, positive in red). The purple circles locate the points where Ux=0U_x=0. In all figures, the green solid line represents the approximate beam radius (2.52.5), and the blue solid line locates the radius where UxU_x is 5050 % of its on-axis value. All the figures on a given row share the same colour levels.

To best capture the differences between the modes associated to different branches of instabilities, we use the lines of skin friction perturbation on the downstream wall, and classify their critical points. For case (N,CR)=(0.25,4)(N,C_R)=(0.25,4) (Figure 9 (b)), the friction lines include the following numbers of critical points:

Figure 9. Lines of the skin friction perturbation stresses on the downstream wall, with the background color corresponding to the azimuthal component of the stress vector (positive in red, negative in blue). These lines are shown along with the critical points (blue symbols) at for N=0.25N=0.25 and for different CRC_R: (a) CR=6C_R=6, (b) 4 and (c) 3. The approximate beam radius (equation (8)) is represented by the green circle. The base flow half saddle S7′S'_7 on the downstream wall is located by the solid red circle. The solid blue circle approximates the separation line going through the saddles found at a same distance from the origin. For each case, the mode shapes are computed at forcing magnitudes GracGr_{ac} slightly above the instability onset.

ΣN=8,ΣN′=4,ΣS=9,ΣS′=0.\Sigma_N=8,\qquad\Sigma_{N'}=4,\qquad\Sigma_S=9,\qquad\Sigma_{S'}=0.

Since χ=2\chi= 2 for the downstream wall ([29]), the topological rule (16) is thus again satisfied, validating the mode structure shown in Figure 9 (b). The topology features attachment and detachment nodes where regions of zero azimuthal stress cause the flow to respectively attach to and separate from the wall ([18]). The half-saddle of the base flow S′S' in the meridional plane translates into a circular line separating a ring of Moffat corners in the outmost part of the cavity and an adjacent inner ring of four structures with closed friction lines (red line in the plots of Figure 9). The friction lines patterns thus feature four centres in this region (Figure 9 (a)). These points are located between this outer separator and an inner separator, represented by the solid blue circle. In this case the outer separator is impermeable for both the base flow and the perturbation, i.e. the perturbation field involves no mass flux between outer region of the Moffatt vortices of the base flow and the main 4-vortices structures of the perturbation. Decreasing CRC_R narrows down the gap between the two separators, as seen by comparing cases with CR=6C_R = 6 and CR=4C_R = 4, respectively shown in Figure 9 (a) and (Figure 9b). Decreasing CRC_R reduces the space available for the friction lines to form closed loops and forms streamwise vortices that are ever more stretched in the azimutal direction.

For CR=3C_R = 3, another local maximum of σ(m)\sigma(m) is first to become unstable at Gracc=7752Gr_{ac}^{c} = 7752 (Figure 7 (d)). The leading modes of this new branch, which we denote as “Branch II”, correspond to non-oscillatory m=4m=4 perturbations. This change of unstable branch is the first significant effect of confinement on the flow stability, and highlights a minimum DD below which the leading perturbation is altered.

To understand the change of leading mode branch, we shall turn again to the patterns of the friction perturbation on the downstream wall. Here the confinement is increased to the point where the streamwise vortices between the separators of branch I are strecthed to breaking point. This happens as the centres collapse on the inner separator (blue circle), making impossible to form closed friction loops without crossing the separator. In other words, further reducing CRC_R necessarily requires the skin friction lines patterns to be rearranged for basic topological rules to not be violated. This topological rearrangement is made possible by a different value of mm, corresponding to a new unstable branch. The resulting Branch II

mode features a new set of saddles and nodes (Figure 9 (c)). The critical point at r=0r=0 is a “high-order” critical point: both the base flow axisymmetry and the m=4m=4 symmetry of the mode cause the gradient of the skin friction vector to vanish, meaning that higher-order derivatives are required to classify this point. We refer to the latter as a “star saddle”. Being different in nature to classical four-branch saddles of topological weight w4s=−2w_{4s}=-2 (i.e., the constant by which ΣS\Sigma_S is multiplied in the topological rule (16)), this type of critical point must have its own topological weight. Assuming that the topological rule (16) is satisfied, the topological weight of this 8-branch star saddle must be w8s=−6w_{8s}=-6.

Rise of oscillatory instabilities in confined settings

As CRC_R is further decreased, the modes destabilising the flow become oscillatory. Indeed, the CR=2C_R=2 flow is destabilised at Gracc=10513Gr_{ac}^{c}=10513 by the growth of an oscillatory m=2m=2 perturbation (Figure 10 (a)). This mode is associated with a new branch that we label as “Branch III”. Interestingly, Branch III and Branch II coexist for this CRC_R, and the leading Branch II mode becomes unstable at Grac≈10630≈1.01GraccGr_{ac}\approx10630\approx1.01Gr_{ac}^{c}. The σ\sigma of the Branch II mode increases more rapidly with GracGr_{ac} than the σ\sigma of the Branch III mode, so that Branch II becomes dominant for Grac≥11000≈1.04GraccGr_{ac}\ge11000\approx1.04Gr_{ac}^{c}.

Figure 10. Growth rate σ\sigma of the leading eigenmode as a function of the azimuthal wavenumber mm for (a) (N,CR)=(0.25,2)(N,C_R)=(0.25,2) and (b) (0.25,1)(0.25,1). Oscillatory and non-oscillatory modes are represented by filled circles and triangles, respectively. The insets show the evolution of σ\sigma with GracGr_{ac} for the indicated value of mm. The critical Grashof number GraccGr_{ac}^{c}, obtained through spline interpolation of σ\sigma, is reported in red. The filled black symbols in the insets of (a) are additional eigenvalue computations made to improve the interpolation accuracy in the vicinity of σ=0\sigma=0.

For CR=2C_R=2, the increased flow confinement and the greater GraccGr_{ac}^{c} cause the flow structures resulting from the jet impingement to shrink down and to be further concentrated near the downstream wall (Figure 11). In particular, N5N_5 and S1S_1 now lie inside the main forcing region; that was not the case for larger CRC_R. Therefore, part of the return flow near N5N_5 and S1S_1 now faces the acoustic forcing.

Leading mode for $(N,C_R)=(0.25,2)$ at $Gr_{ac}=10600$, $m=2$, $\sigma=2.54\times10^{-2}$, $\omega=3.9462$ (Branch III)

Figure 11. Leading mode for (N,CR)=(0.25,2)(N,C_R)=(0.25,2) at Grac=10600Gr_{ac}=10600. The mode is oscillatory (ω≠0\omega\ne0) and unstable (σ=2.54×10−2>0\sigma=2.54\times10^{-2}>0). (a) Azimuthal velocity perturbation uθ′u_\theta^\prime in the (x,r)(x,r) plane, along with the critical points of the base velocity U\mathbf{U}. (b) Detailed view near the impingement (region framed in red in (a)) of the axial velocity perturbation ux′u_x^\prime together with the base flow streamlines. (c) Details u′\mathbf{u}^\prime near the impingement, with uθ′u_\theta^\prime taken as background colour. In all plots, the vertical dashed blues lines locate the slices shown in Figure 12. The green solid line represents the approximate beam radius (2.5), and the blue solid line locates the radius where UxU_x is 50 % of its on-axis value.

The velocity perturbation field displays distinct features compared to the previous CR≥3C_R\ge3 cases. First, ∣uθ′∣|u'_\theta| is no longer maximum at the impingement, but near S1S_1 instead (Figure 11 (c)). The flow is thus no longer destabilised by the jet impingement, but by the shear layer N5N_5 and S1S_1 instead. This is confirmed by the shape of u′u' and the contours of ωx\omega_x at N5N_5 and S1S_1 (Figure 12): as for CR≥3C_R\ge3, the perturbation is still clustered near the shear layer around the jet where Ux≈0U_x\approx0. Unlike at lower confinement, the instability extends along a surface closely following this layer. Second, uθ′u'_\theta displays a wave-like structure in the xx-direction, the amplitude of which being significantly reduced as the distance from the downstream wall increases. This is consistent with the findings of Henry et al. (2022), who reported similar unstable waves for a plane 2D streaming flow forced by a non-diffracting and non-attenuated beam of square cross section. These features, reminiscent of a Kelvin-Helmholtz instability, suggest that a high confinement the instability originates in the shear layer, rather than at the impingement. Finally, for CR=2C_R=2, the saddles of the skin friction perturbation on the downstream wall no longer coincide with the base flow half saddle S7′S_7^\prime (Figure 13 (a)). In

Figure 12. Leading velocity perturbation u′\mathbf{u}' for (N,CR)=(0.25,2)(N,C_R)=(0.25,2) at Grac=10600Gr_{ac}=10600. The mode is displayed in constant-xx planes located by the vertical dashed blue lines shown in Figure 11 and going through the base flow node N5N_5 (slice 1, left) and saddle S1S_1 (slice 2, right). Either the streamwise velocity perturbation ux′u'_x ((a)–(b)) or the streamwise vorticity perturbation ωx′\omega'_x ((c)–(d)) is used as background colour (negative in blue, positive in red). In all figures, the purple circles locate the points where Ux=0U_x=0, the green circle represents the approximate beam radius (2.5), and the blue circle locates the radius where UxU_x is 50 % of its on-axis value. Figures on a given row share the same colour levels.

Lines of skin friction stresses exerted by the leading perturbation on the downstream wall

Figure 13. Lines of skin friction stresses exerted by the leading perturbation on the downstream wall for (a) (N,CR)=(0.25,2)(N,C_R)=(0.25,2) and (b) (N,CR)=(0.25,1)(N,C_R)=(0.25,1). In each plot, the background colour represents the azimuthal component of the stress vector (positive in red, negative in blue). The friction lines are shown along with the zero mode friction critical points (blue symbols) and the approximate beam size (green circle, as defined by equation (8)). The base flow half saddle S7′S'_7 is also reported in red.

other words, reducing the cavity radius creates a perturbation flux between the Moffatt roll delimited by S7′S'_7 and the remaining of the flow near x=Lx=L.

For the most confined case (CR=1C_R=1), a new mode branch arises (Figure 10 (b)). This branch, labelled as “Branch IV”, is defined by a leading m=3m=3 oscillatory mode. The onset occurs at Grac=15310Gr_{ac}=15310. At Grac≈GraccGr_{ac}\approx Gr_{ac}^{c}, both the enhanced confinement and the greater strength of the return flow resulting from the larger GraccGr_{ac}^{c} significantly affect the base flow topology near the downstream wall (Figure 14). First, the size of the recirculating flow structures is greatly reduced e.g., the secondary recirculation bubble on the lateral wall nearly vanishes. Second, the reduced CRC_R yields two new critical points N6N_6 and S2S_2 in the spatial structure of U\mathbf{U}. These new critical points appear at approximately the same radial location as N5N_5 and S1S_1.

Leading mode for $(N,C_R)=(0.25,1)$ at $Gr_{ac}=15500$

Figure 14. Leading mode for (N,CR)=(0.25,1)(N,C_R)=(0.25,1) at Grac=15500Gr_{ac}=15500. The mode is oscillatory (ω≠0\omega\ne0) and unstable (σ=1.701×10−1>0\sigma=1.701 \times10^{-1}>0). (a) Azimuthal velocity perturbation uθ′u'_\theta in the (x,r)(x,r) plane, along with the critical points of the base velocity U\mathbf{U}. (b) Detailed view near the impingement (region framed in red in (a)) of the axial velocity perturbation ux′u'_x together with the base flow streamlines. (c) Details of u′\mathbf{u}' near the impingement, with uθ′u'_\theta taken as background colour.

Maps of u′\mathbf{u}' in the (x,r)(x,r) plane show that ∣ux′∣|u'_x| and ∣uθ′∣|u'_\theta| are maximum near N5N_5 and S1S_1, i.e., in regions of high shear in the bulk of the base flow (Figure 14 (b) and (Figure 14c)). This confirms the trend observed for CR=2C_R=2: confining the flow shifts the locus of the instability from the impingement to the shear layer near N5N_5 and S1S_1. This shift is also favoured by the build-up of an adverse pressure gradient slowing down the jet (inset of Figure 6). Besides, u′\mathbf{u}' is again marginal in the jet (blue line and circles in Figure 14) and creates strong ωx′\omega_x' at the shear layer (Figure 15). Finally, the leading Branch IV mode features a “star saddle” critical point at r=0r=0 (Figure 13 (b)), as does the leading mode for CR=3C_R=3 (Branch II).

Figure 15. Leading velocity perturbation u′\mathbf{u}' for (N,CR)=(0.25,1)(N,C_R)=(0.25,1) at Grac=15500Gr_{ac}=15500. The mode is shown in constant-xx planes located by the blue vertical lines in Figure 14, with either ux′u'_x ((a)–(b)) or the streamwise vorticity perturbation ωx′\omega'_x ((c)–(d)) as background colour (negative in blue, positive in red). The purple circles locate the points where Ux=0U_x=0. In all figures, the green solid line represents the approximate beam radius (2.5), and the blue solid line locates the radius where UxU_x is 50 % of its on-axis value. Figures on a given row share the same colour levels.

To conclude, for N∈{0.25,1}N\in\{0.25,1\} and 3≤CR≤63\le C_R\le6, the primary instabilities are caused by non-oscillatory perturbations originating from the jet impingement. Further confining the flow (1≤CR≤2)(1 \le C_R \le2) not only relocates the origin of the instability to the shear layer between the jet and the impingement-induced recirculation structures, but also causes the leading perturbation to become oscillatory.

Bifurcation characterisation

Model for the nonlinear growth and saturation of the perturbations

To complete the stability study, we shall now determine the criticality of the bifurcations identified with LSA. Our approach relies on the model of Landau & Lifschitz [43] describing the growth and saturation of a perturbation near the onset of instability. This model has been extensively used to classify the bifurcations arising in numerous situations, such as the flow past rings (Sheard et al. [63]), the wake behind a sphere (Thompson et al. [65]) and behind one or multiple cylinders (Henderson & Barkley [35]; Henderson [34]; Carmo et al. [12]). The Stuart-Landau model reads:

dAdt=(σ+iω)A−l(1+ic)∣A∣2A+O(A5).(17)\frac{\mathrm{d}A}{\mathrm{d}t}=(\sigma+\mathrm{i}\omega)A-l(1+\mathrm{i}c)\lvert A\rvert^2A+O(A^5). \tag*{(17)}

where AA is a complex perturbation, and ll and cc are two real constants. The mechanism responsible for the saturation of AA can be easily understood from the sign of ll in equation (17). For l>0l>0, the cubic term in the right-hand side of equation (17) acts against the growth of the perturbation: this situation corresponds to a supercritical bifurcation. On the contrary, for l<0l<0, higher-order terms are required for AA to saturate: the bifurcation is thus subcritical. Therefore, only the sign of ll is needed to determine the nature (or criticality) of the bifurcation.

As explained in Sheard et al. [63] and Kumar & Potherat [42], ll may be inferred from a more convenient form of equation (17). Plugging A(t)=∣A(t)∣eiϕ(t)A(t)=\lvert A(t)\rvert e^{\mathrm{i}\phi(t)} into the Stuart-Landau model (17) gives, after separating the real and imaginary part:

dlog⁡∣A∣dt=σ−l∣A∣2+O(∣A∣4),(18)\frac{\mathrm{d}\log|A|}{\mathrm{d}t}=\sigma-l|A|^2+O(|A|^4), \tag*{(18)}
dϕdt=ω−lc∣A∣2+O(∣A∣4).(19)\frac{\mathrm{d}\phi}{\mathrm{d}t}=\omega-l_c|A|^2+O(|A|^4). \tag*{(19)}

The nature of the bifurcation is determined by plotting (dlog⁡∣A∣/dt)(\mathrm{d}\log|A|/\mathrm{d}t) against ∣A∣2|A|^2; the slope of the resulting curve is −l-l for sufficiently small ∣A∣2|A|^2, and σ\sigma is obtained from the yy-intercept. From this we compare σ\sigma to its value obtained from the LSA results and so assess the accuracy of the estimate for ll. Note that for an non-oscillatory perturbation, ϕ(t)=0\phi(t)=0, so only equation (18) is needed.

The saturated state is characterised using the steady-state forms of equations (18) and (19). For a supercritical (l>0l>0) bifurcation for instance, the first two terms in the right-hand side of the amplitude equation (18) are dominant, so that the saturated perturbation amplitude is:

∣Asat∣=σl.(20)\lvert A_{sat}\rvert= \sqrt{\frac{\sigma}{l}} . \tag*{(20)}

Furthermore, if the saturated state is a time-periodic signal of constant amplitude ∣Asat∣\lvert A_{sat}\rvert, then dϕ/dt\mathrm{d}\phi/\mathrm{d}t reduces to the saturated frequency ωsat\omega_{sat}. Using equation (20), the steady-state version of equation (19) gives, for the Landau constant cc:

c=ω−ωsatσ,(21)c = \frac{\omega-\omega_{sat}}{\sigma}, \tag*{(21)}

thus allowing for all the quantities in the truncated Stuart–Landau model (17) to be fully characterised.

We shall now define AA in terms of flow quantities. We followed Kumar & Pothé rat [42] and based ∣A∣\lvert A\rvert on a local value of a flow variable, and more particularly on uθ′(t)u'_\theta(t). Other choices are possible, including global measures of the flow unsteadiness [65, 63, 64, 61]. For oscillatory bifurcations, we considered the envelope of uθ′(t)u'_\theta(t). Choosing where to record uθ′(t)u'_\theta(t) is based on

NNCRC_R(x,r,θ)(x,r,\theta)
0.251(82, 1.8, 0)
0.252(81, 1.3, 0)
0.253(80, 2, 0)
0.254(78, 6, 0)
0.256(75, 2, 0)
16(308, 30, 0)

caption: Coordinates (x,r,θ)(x,r,\theta) of the points where the time series of the azimuthal velocity perturbation uθ′u'_{\theta} are recorded in the nonlinear 3D-3C simulations.

on the need to obtain clean signals. To maximise the signal-to-noise ratio, we thus chose points where LSA predicted large ∣uθ′∣\lvert u'_{\theta}\rvert. Depending on the values of CRC_R, these points are located near the critical points N5N_5 and S1S_1, and near the impingement. The precise locations of these points are given in table 4.

From a numerical point of view, the nonlinear 3D-3C unsteady simulations were run for slightly supercritical regimes. The relative gap to GraccGr_{ac}^{c} is defined by the criticality parameter rcr_c:

rc=GracGracc−1,(22)r_c = \frac{Gr_{ac}}{Gr_{ac}^{c}} - 1, \tag*{(22)}

with rc>0r_c > 0 (respectively rc<0r_c < 0) corresponding to a supercritical (respectively subcritical) regimes. Finally, (dlog⁡∣A∣/dt)(\mathrm{d}\log\lvert A\rvert/\mathrm{d}t) was computed from the recorded time series using centred finite differences.

Nature of the bifurcations

The Stuart-Landau analysis has been carried for all the setups listed in Table 1; the cases defined by N=0.25N = 0.25 and CR∈{1,2,4}C_R \in\{1, 2, 4\} are shown in Figure 16 as examples. For each case, the initial instants of ∣A∣\lvert A\rvert are dominated by noise (Figure 16 (a), (c) and (e)). The Stuart-Landau equation (18) for ∣A∣\lvert A\rvert is then fitted to the time series over a time interval free of the initial noise; that range is highlighted in red in all plots of Figure 16.

Figure 16. Stuart-Landau analysis for (a)-(b) (N=0.25N = 0.25, CR=4C_R = 4) at Grac=6400Gr_{ac} = 6400 (rc=0.0262r_c = 0.0262), (c)-(d) (N=0.25N = 0.25, CR=2C_R = 2) at Grac=10600Gr_{ac} = 10600 (rc=0.0083r_c = 0.0083) and (e)-(f) (N=0.25N = 0.25, CR=1C_R = 1) at Grac=15500Gr_{ac} = 15500 (rc=0.0124r_c = 0.0124). Left: growth and saturation of a white-noise-triggered perturbation injected at t=0t = 0, monitored by the time series of the azimuthal velocity perturbation uθ′u'_\theta. Right: evolution of d(log⁡∣A∣)/dt\mathrm{d}(\log|A|)/\mathrm{d}t with ∣A∣2|A|^2, where ∣A∣|A| is the amplitude of the perturbation plotted on the left figures (for the CR=2C_R = 2 and 1 cases, ∣A∣|A| is based on the envelope of uθ′u'_\theta). The growth rate σ\sigma and the Landau coefficient ll are obtained by fitting equation (18) to the red curve. For each case, the fitting range is highlighted in red.

For 4≤CR≤64 \leq C_R \leq6, l>0l > 0 near the instability onset. The leading Branch I mode thus destabilises the flow through a supercritical circular pitchfork bifurcation (Touihri et al. 1999). Further confining the flow yields l<0l < 0 for CR=3C_R = 3, indicating that an unstable Branch II mode triggers a subcritical circular pitchfork bifurcation. Although the leading mode becomes oscillatory for CR=2C_R = 2, ll is negative (Figure 16 (d)): the associated Hopf bifurcation is thus subcritical. Finally, l>0l > 0 for CR=1C_R = 1 (Figure 16 (f)), hence the unstable Branch IV mode is associated with a supercritical Hopf bifurcation. For all cases, the difference between the σ\sigma obtained with the nonlinear 3D-3C simulations and LSA is at most 6% (Table 5). This excellent agreement (i) validates the LSA, and (ii) adds further confidence on the validity of the Stuart-Landau analyses.

NNCRC_RGracGr_{ac}rcr_cσ\sigma (LSA)σ\sigma (nonlinear 3D-3C)εσ\varepsilon_\sigma (%)ω\omega (LSA)ω\omega (nonlinear 3D-3C)εω\varepsilon_\omega (%)ω−ωsat\omega-\omega_{sat}
0.251155000.01241.701×10−11.701 \times10^{-1}1.703×10−11.703 \times10^{-1}0.1221.47221.4090.290
0.252106000.00832.540×10−22.540 \times10^{-2}2.601×10−22.601 \times10^{-2}2.353.9463.9580.301.166
0.25380000.03207.245×10−27.245 \times10^{-2}7.690×10−27.690 \times10^{-2}5.7900––
0.25464000.02621.591×10−21.591 \times10^{-2}1.594×10−21.594 \times10^{-2}0.1900––
0.25650000.06351.384×10−21.384 \times10^{-2}1.383×10−21.383 \times10^{-2}0.0700––
1625000.14112.755×10−32.755 \times10^{-3}2.896×10−32.896 \times10^{-3}4.8700––

Table 5. Comparison between the growth rates σ\sigma and frequencies ω\omega returned by linear stability analysis (LSA) with those obtained from nonlinear 3D-3C unsteady simulations of supercritical regimes (rc>0r_c > 0, see equation (22)). The relative errors between the values of σ\sigma and ω\omega obtained by LSA and the nonlinear 3D-3C simulations are given by εσ\varepsilon_\sigma and εω\varepsilon_\omega, respectively.

Finding subcritical bifurcations for CR∈{2,3}C_R \in\{2, 3\} raises the question of whether the corresponding instabilities can be triggered for rc<0r_c < 0. For both cases, introducing white noise for −0.05≤rc≤−0.01-0.05 \leq r_c \leq-0.01 did not cause the flow to deviate from its axisymmetric base state. Triggering these bifurcations at Grac<GraccGr_{ac} < Gr_{ac}^{c} would require different tools, e.g. the analysis of the transient growth of non-modal perturbations [62]. Finding the minimum seeds to trigger these subcritical bifurcations is nevertheless out of the scope of the present work and shall be the subject of a dedicated study.

Saturated states and secondary instabilities

We shall finally characterise the saturated states following the perturbation growth at rc>0r_c > 0.

For Hopf bifurcations, the frequency ωsat\omega_{\mathrm{sat}} of the saturated perturbation may differ from ω\omega in the linear regime (regime of exponential perturbation growth), yielding a non-zero cc (equation (21)). A change of frequency is captured by taking the Fast Fourier Transform (FFT) of uθ′(t)u'_\theta(t) in both the linear and the saturated regimes (Figure 17).

Frequency spectrum for $N=0.25$, $C_R=1$

(a) (a) (N=0.25,CR=1), Grac=15500 (rc=0.0124)(a)\ (N = 0.25, C_R = 1),\ Gr_{ac} = 15500\ (r_c = 0.0124)

Frequency spectrum for $N=0.25$, $C_R=2$

(b) (b) (N=0.25,CR=2), Grac=10600 (rc=0.0083)(b)\ (N = 0.25, C_R = 2),\ Gr_{ac} = 10600\ (r_c = 0.0083)

Figure 17. Frequencies obtained from the time-series of the azimuthal velocity perturbation uθ′(t)u'_\theta(t) for (a) (N,CR)=(0.25,1)(N, C_R) = (0.25, 1) and (b) (0.25,2)(0.25, 2). The corresponding time-series are those displayed in Figure 16 (e) and (c), respectively. The frequencies are computed from FFTs made in the linear regime (ω\omega, blue) and in the saturated regime (ωsat\omega_{sat}, red).

For (N,CR)=(0.25,2)(N,C_{\mathrm{R}})=(0.25,2), ωsat\omega_{sat} differs by 41.8 % from ω\omega (Figure 17b). In that case, c=44.83±14.63c=44.83 \pm14.63, and the discrepancy between the values of ω\omega predicted by LSA and the nonlinear 3D-3C simulations is 0.30 % (Table 5). For (N,CR)=(0.25,1)(N,C_{\mathrm{R}})=(0.25,1), the ω\omega and ωsat\omega_{sat} obtained with the nonlinear 3D-3C simulations in the linear and saturated regimes, respectively, are almost identical (Figure 17a). They also barely differ from the frequency predicted by LSA: the observed discrepancies are of 0.29 % (Table 5) and 0.87 % in the linear and saturated regimes, respectively. Given these small differences, ω\omega and ωsat\omega_{sat} are considered to be equal, hence c=0c=0.

For (N,CR)=(0.25,2)(N,C_{\mathrm{R}})=(0.25,2), the large uncertainty on cc is due to the rather low resolution of the saturated regime spectrum (Figure 17a). That low resolution results from the short duration of the saturated regime: the latter indeed lasts approximately 10 oscillation periods after which uθ′u'_{\theta} changes abruptly (Figure 18). The saturated state is thus itself unstable and causes the flow to transition towards a highly unsteady and chaotic flow regime. Unfortunately, that secondary instability develops too abruptly to be characterised by means of a Stuart-Landau analysis.

Figure 18. Time-series of the azimuthal velocity perturbation uθ′u'_{\theta} for (N,CR)=(0.25,2)(N,C_{\mathrm{R}})=(0.25,2) and Grac=10600Gr_{ac}=10600 (rc=0.0083r_c=0.0083). The signal is recorded at (x,r,θ)=(81,1.3,0)(x,r,\theta)=(81,1.3,0).

A secondary instability was also observed for (N,CR)=(0.25,3)(N,C_R)=(0.25,3) at rc=0.0320r_c=0.0320 (Figure 19 (a)). For that case, the saturated state is unstable to a non-oscillatory perturbation. Contrarily to the (N,CR)=(0.25,2)(N,C_R)=(0.25,2) case, the slow deviation from the saturated state allows for a Stuart-Landau analysis to be made (Figure 19 (b)). As the primary bifurcation, the secondary instability is found to be subcritical. However, it remains unclear whether the temporal variations of uθ′u'_\theta observed for t≥110t\geq110 correspond to a transient regime before reaching a stable steady state, or instead are the beginning of a chaotic regime. Significantly longer simulations would be necessary to answer this question.

Figure 19. Stuart-Landau analysis for (N,CR)=(0.25,3)(N,C_R)=(0.25,3) at Grac=8000Gr_{\mathrm{ac}}=8000 (rc=0.0320r_c=0.0320). (a) Temporal evolution of the azimuthal velocity perturbation uθ′u'_\theta obtained from nonlinear 3D-3C unsteady simulations and recorded at (x,r,θ)=(80,2,0)(x,r,\theta)=(80,2,0). The perturbation originates from white noise injected at t=0t=0. (b) Evolution of d(log⁡∣A∣)/dt\mathrm{d}(\log|A|)/\mathrm{d}t with ∣A∣2|A|^2 (red curve), where ∣A∣=∣uθ′∣|A|=|u'_\theta| and is extracted from (a). The black dashed line is obtained by fitting the Stuart-Landau equation (18) to the red curve over the range highlighted in red. The positive slope indicates a subcritical transition.

Finally, the leading oscillatory modes for (N,CR)=(0.25,1)(N,C_R)=(0.25,1) and (0.25,2)(0.25,2) are defined by a pair of complex conjugate eigenvalues corresponding to counter-propagating waves in the eθ\boldsymbol{e}_\theta-direction. Depending on the complex amplitudes of these unstable modes, their combination results in either a standing wave or a travelling wave [17]. Although these complex amplitudes cannot be obtained from LSA, they do appear in the nonlinear 3D-3C unsteady simulations. The leading LSA mode shapes are compared to snapshots of the 3D-3C unsteady simulations in Figure 20 for (N,CR)=(0.25,1)(N,C_R)=(0.25,1) and (0.25,2)(0.25,2). The snapshots are taken from the early stage of the exponential growth of the perturbation to avoid nonlinear effects on the perturbation shape. For both cases, a strong agreement in found between the Branches III and IV mode shapes obtained by LSA and by the nonlinear 3D-3C simulations, and both perturbations correspond to waves travelling along eθ\boldsymbol{e}_\theta (see the supplementary movies 1 and 2).

Figure 20. Comparison between the leading velocity perturbation u′\mathbf{u}' obtained with LSA (left) and the perturbation obtained with nonlinear 3D-3C unsteady simulations (right) for (a) (N,CR)=(0.25,1)(N,C_R)=(0.25,1) and (b) (N,CR)=(0.25,2)(N,C_R)=(0.25,2). The results from LSA and nonlinear 3D-3C unsteady simulations are compared at the location of the axisymmetric base flow saddle S1S_1 (x=80.15x=80.15 in (a) and x=77.52x=77.52 in (b)). The indicated σ\sigma and ω\omega are those of LSA, and the nonlinear results are extracted from the early exponential growth of the perturbation. Features of the steady base velocity U\mathbf{U} are reported in both plots: the points where the sign of the streamwise velocity UxU_x changes (purple circles), and the points where UxU_x is equal to 50 % of its on-axis value (blue circle). The green circle represents the approximate edge of the beam defined by equation (8). See the supplementary movies (movies 1 and 2) for the animations of the velocity perturbations obtained from the nonlinear 3D-3C simulations.

To conclude, the thorough characterisation of the Stuart-Landau model (17) for all cases (Table 6) showed that the sub- or supercritical nature of the primary bifurcation is closely related to the flow confinement. For the subcritical primary bifurcations, the saturated states were found to be unstable, causing the flow to transition to an unsteady state.

NNCRC_{\mathrm{R}}GracGr_{ac}rcr_cllccBranchBifurcation type
0.251155000.0124(2.856±0.012)×10−2(2.856 \pm0.012) \times10^{-2}0IVSupercritical Hopf
0.252106000.0083−1.547±0.001-1.547 \pm0.00144.83±14.6344.83 \pm14.63IIISubcritical Hopf
0.25380000.0320−0.794±0.003-0.794 \pm0.0030IISubcritical circular pitchfork
0.25464000.02620.391±0.0040.391 \pm0.0040ISupercritical circular pitchfork
0.25650000.06350.142±0.0010.142 \pm0.0010ISupercritical circular pitchfork
1625000.1411(1.971±0.006)×10−2(1.971 \pm0.006) \times10^{-2}0ISupercritical circular pitchfork

Table 6. Summary of the Stuart-Landau analysis based on local measurements of uθ′u'_{\theta}; these measurements are obtained from nonlinear 3D-3C unsteady simulations of slightly supercritical regimes (rc>0r_c > 0, see equation (22)). The ll estimates are bounded by their 95 % confidence interval. The confidence intervals for cc further involve the resolution of the frequency spectra from which ω\omega and ωsat\omega_{sat} are determined. See table 4 for the coordinates of the points where uθ′u'_{\theta} is recorded.

Conclusion

We studied the stability of a laminar Eckart streaming jet flowing in a closed cylindrical cavity. A plane circular transducer placed at one end of the cavity radiates an axisymmetric

diffracting sound beam, whose attenuation drives a steady axisymmetric jet impinging the cavity wall facing the transducer. To our knowledge, this work is the first stability analysis of a streaming jet forced by a realistic sound field that includes the effects of both diffraction and attenuation on its spatial structure. As such, this study provides a generic framework to understand the conditions in which realistic acoustic streaming jets are stable, and their path to unsteadiness when they are not.

We assessed the effect of the cavity size on the streaming flow stability by either varying the cavity length with respect to the attenuation length (N=1N = 1 and 0.250.25), and by changing the ratio CRC_R between the cavity radius and the beam radius at the downstream cavity wall (CR∈[1,6]C_R \in[1,6]). For each case, we (a) identified the destabilisation mechanism by analysis of the flow topology and its critical points, (b) computed the critical forcing magnitude GraccGr_{ac}^{c} above which the flow is linearly unstable (Linear Stability Analysis), and (c) determined the nature of the bifurcation by means of Stuart-Landau analyses of data obtained from nonlinear 3D-3C (three dimensions, three components) unsteady simulations. The main outcomes of this work are summarised below:

  1. For all cavity sizes, recirculating structures originating from the jet impingement are formed near the downstream wall. The primary bifurcations result from the destabilisation of these structures, rather than from the destabilisation of the jet itself.

  1. Reducing the cavity size has a stabilising effect (Figure 21): the critical GraccGr_{ac}^{c} for the onset of instability ranges from Gracc=2191Gr_{ac}^{c} = 2191 for (N,CR)=(1,6)(N,C_R) = (1,6) (less-confined case) to Gracc=15310Gr_{ac}^{c} = 15310 for (N,CR)=(0.25,1)(N,C_R) = (0.25,1) (most confined case). For 4≤CR≤64 \le C_R \le6, the leading perturbation is non-oscillatory (Branch I). This unstable mode displays large perturbation velocities near the jet impingement. The first significant effect of confinement occurs for CR=3C_R = 3: the leading mode remains non-oscillatory, but its topology changes (Branch II). The reason for the change of topology is that the mode corresponding to Branch I with an m=2m = 2 azimuthal wavenumber becomes radially so confined that it becomes azimuthally stretched to the point of breaking up. At this point, an m=4m = 4 mode from Branch I emerges. Flows in further confined settings are destabilised by oscillatory perturbations travelling in the azimuthal direction (Branch IV for CR=1C_R = 1). For CR=2C_R = 2, these perturbations also propagate backwards, i.e., towards the source (Branch III). Overall, reducing CRC_R (a) increases the shear in a layer separating counter-rotating structures in the bulk of the flow, and (b) enhances an adverse pressure gradient slowing down the jet. As a consequence, reducing CRC_R moves

Figure 21. Variations of the critical Grashof number GraccGr_{ac}^{c} with the radial confinement ratio CRC_R defined by equation (9). The N=0.25N = 0.25 cases are represented by filled discs and are interpolated by cubic splines. The GraccGr_{ac}^{c} obtained for (N,CR)=(1,6)(N,C_R) = (1,6) is also reported (black square) and illustrates the destabilising effect of increasing NN. The dashed vertical lines are arbitrarily placed in between cases having different primary bifurcation types.

the locus of instability from the jet impingement to the shear layer between the jet and the recirculating structures.

(iii) Finally, changing the flow confinement significantly affects the nature of the primary bifurcations. Whilst the unstable modes of Branches I and IV respectively trigger supercritical pitchfork and Hopf bifurcations, the bifurcations associated with Branch II and III modes are subcritical.

The subcritical nature of bifurcations for the CR∈{2,3}C_R \in\{2,3\} cases raises the question of whether these bifurcations could be triggered Grac<GraccGr_{ac} < Gr_{ac}^{c}. The nonlinear 3D-3C unsteady simulations carried out between 0.95 Gracc0.95\,Gr_{ac}^{c} and 0.99 Gracc0.99\,Gr_{ac}^{c} revealed that white-noise-based perturbations could not ignite subcritical transitions towards another state. Whether this scenario could occur remains uncertain at this stage and shall require further analysis, for instance by finding the initial disturbances maximising the energy growth over a finite time range (Schmid [62]). Other examples of convective of shear flows exhibit similar subcritical bifurcations where transition away from the steady state is not easily triggered in the subcritical regime Kumar & Pothérat [42]; Camobreco et al. (2020, 2021, 2023); Huang et al. [37]. The stability and transition to turbulence in these flows are very different to the classical shear flows such as Couette or Poiseuille flows, where the subcritical nature of the bifurcation favours transition to turbulence via mechanisms that radically differ from those of the LSA. Camobreco et al. (2023) and Huang et al. [37] showed that transition to turbulence in quasi-2D and 2D shear flows could be triggered only in mildly subcritical regimes and that the mode mediating the transition was in fact the leading LSA eigenmode. In acoustically-driven jets, the instability mainly originates from shear, but the base flow is significantly more complex than in these examples. Hence an interesting question for further work is whether when the transition is weakly subcritical, the leading eigenmodes underpin any transition to unsteadiness and possibly to turbulence, as they do in 2D shear flows.

For CR∈{2,3}C_R \in\{2,3\}, nonlinear 3D-3C unsteady simulations at supercritical GraccGr_{ac}^{c} showed that the growth and saturation of the initial white-noise perturbation were followed by the development of a secondary instability. For CR=2C_R = 2 in particular, that secondary instability triggered a transition towards an unsteady state. This raises the question of whether this transition is the trace of an edge state, i.e., if there exists a threshold for the initial perturbation energy below which the saturated state transitions back to the initial steady axisymmetric state. Answering these questions would require further analysis.

Finally, the obtained GraccGr_{ac}^{c} correspond to relatively low forcing magnitudes, as experiments usually operate at higher GracGr_{ac} to ease the measurement of streaming velocities (Kamakura et al. [39]; Mitome [48]; Moudjed et al. [51]). Our study thus sheds light on the rather unstable nature of acoustic streaming jets, and confirms the relevance of numerical simulations and LSA to identify the primary bifurcation of an Eckart streaming jet that is otherwise delicate to observe experimentally. This work is thus an important milestone towards the thorough understanding of the rich dynamics of streaming flows (Cambonie et al. [8]; Launay et al. [44]) and their transition to turbulence. Conversely, the strong variation of GraccGr_{ac}^{c} with confinement offers a convenient way to manipulate the nature of the flow forced in specific configurations: where turbulence is desired for mixing or other purposes, immersing the acoustic transducer with a plate cutting the jet in an open region of the flow would favour the instabilities conducive to turbulence. If, by contrast, a well-controlled, laminar flow is sought, the transducer could be housed at one end of a thin cylinder open towards the forcing region at the other end: such a configuration would suppress the shear instability at the edge of the jet whilst still injecting the required amount of momentum into the flow. The analysis carried out in this paper offers predictive tools that may be easily adapted to design such a device.

Supplementary data. Supplementary movies are available at INSERT LINK TO SUPPLEMENTARY MOVIES

Acknowledgements. The support from the PMCS2I (Ecole Centrale de Lyon) and from the High Performance Computing Centre of the Faculty of Engineering, Environment and Computing (Coventry University) for the numerical calculations is gratefully acknowledged. The authors would like to thank Laurent Pouilloux (Ecole Centrale de Lyon) and Alex Pedcenko (Coventry University) for their availability and for providing help at any stage of the project. The authors also wish to thank Hugh M. Blackburn for the support on the use of the spectral-element code Semtex.

For the purpose of Open Access, a CC-BY public copyright licence has been applied by the authors to the present document and will be applied to all subsequent versions up to the Author Accepted Manuscript arising from this submission.

Funding. This work was carried out as part of the BRASSOA project supported by the Institut Carnot Ingénierie@Lyon and by a Royal Society International Exchange grant (Ref. IES\R2\202212). AP acknowledges support from EPSRC through grant No. EP/X010937/1.

Declaration of interests. The authors report no conflict of interest.

Author ORCIDs. B. Vincent, https://orcid.org/0000-0002-9937-7494; A. Kumar, https://orcid.org/0000-0002-6026-5727; D. Henry, https://orcid.org/0000-0002-7231-7918; S. Miralles, https://orcid.org/0000-0002-4701-8609; V. Botton, https://orcid.org/0000-0002-0180-3089; A. Potherat, https://orcid.org/0000-0001-8691-5241

References

  1. [1]Absar, S., Pasumarthi, P. & Choi, H. 2017 Numerical and experimental studies about the effect of acoustic streaming on ultrasonic processing of metal matrix nanocomposites (MMNCs). J. Manuf. Process. 28, 515–522.DOI
  2. [2]Alghane, M., Fu, Y. Q., Chen, B. X., Li, Y., Desmulliez, M. P. Y. & Walton, A. J. 2012 Frequency effect on streaming phenomenon induced by Rayleigh surface acoustic wave in microdroplets. J. Appl. Phys. 112.DOI
  3. [3]Barkley, D., Blackburn, H. M. & Sherwin, S. J. 2008 Direct optimal growth analysis for timesteppers. Intl J. Numer. Meth. Fluids 57, 1435–1458.DOI
  4. [4]Ben Hadid, H., Dridi, W., Botton, V., Moudjed, B. & Henry, D. 2012 Instabilities in the Rayleigh-Bénard–Eckart problem. Phys. Rev. E 86, 016312.pubmed.ncbi.nlm.nih.gov/23005530
  5. [5]Blackburn, H.M., Lee, D., Albrecht, T. & Singh, J. 2019 Semtex: a spectral element–Fourier solver for the incompressible Navier–Stokes equations in cylindrical or cartesian coordinates. Comput. Phys. Commun. 245, 106804.
  6. [6]Blackburn, H.M. & Sherwin, S.J. 2004 Formulation of a Galerkin spectral element–Fourier method for three-dimensional incompressible flows in cylindrical geometries. J. Comput. Phys. 197, 759–778.DOI
  7. [7]Blackstock, D.T. 2000 Fundamentals of Physical Acoustics. John Wiley & Sons, Inc.
  8. [8]Cambonie, T., Moudjed, B., Botton, V., Henry, D. & Ben Hadid, H. 2017 From flying wheel to square flow: Dynamics of a flow driven by acoustic forcing. Phys. Rev. Fluids 2, 123901.DOI
  9. [12]Carmo, B. S., Sherwin, S. J., Bearman, P. W. & Willden, R. H. J. 2008 Wake transition in the flow around two circular cylinders in staggered arrangements. J. Fluid Mech. 597, 1–29.DOI
  10. [13]Charrier-Mojtabi, M-C., Fontaine, A. & Mojtabi, A. 2012 Influence of acoustic streaming on thermodiffusion in a binary mixture under microgravity. Intl. J. Heat Mass Transfer 55, 5992–5999.DOI
  11. [14]Charrier-Mojtabi, M-C., Jacob, X., Dochy, T. & Mojtabi, A. 2019 Species separation of a binary mixture under acoustic streaming. Eur. Phys. J. E 42, 64.DOI
  12. [15]Chong, M. S., Perry, A. E. & Cantwell, B. J. 1990 A general classification of three-dimensional flow fields. Phys. Fluids A 2, 765–777.DOI
  13. [16]Cleve, S., Guédra, M., Mauger, C., Inserra, C. & Blanc-Benon, P. 2019 Microstreaming induced by acoustically trapped, non-spherically oscillating microbubbles. J. Fluid Mech. 875, 597–621.DOI
  14. [17]Clune, T. & Knobloch, E. 1993 Pattern selection in rotating convection with experimental boundary conditions. Phys. Rev. E 47, 2536–2550.pubmed.ncbi.nlm.nih.gov/9960284
  15. [18]Délery, J. M. 2001 Robert Legendre and Henri Werlé: Toward the elucidation of three-dimensional separation. Ann. Rev. Fluid Mech. 33, 129–154.DOI
  16. [19]Dentry, M. B., Yeo, L. Y. & Friend, J. R. 2014 Frequency effects on the scale and behavior of acoustic streaming. Phys. Rev. E 89, 013203.pubmed.ncbi.nlm.nih.gov/24580352
  17. [21]Dousset, V. & Pothérat, A. 2010 Formation mechanism of hairpin vortices in the wake of a truncated cylinder in a duct. J. Fluid Mech. 653, 519–536.DOI
  18. [22]Dousset, V. & Pothérat, A. 2012 Characterisation of the wake of a truncated cylinder in a duct under a strong axial magnetic field. J. Fluid Mech. 691, 341–367.
  19. [23]Dridi, W., Henry, D. & Ben Hadid, H. 2008a Influence of acoustic streaming on the stability of a laterally heated three-dimensional cavity. Phys. Rev. E 77, 046311.pubmed.ncbi.nlm.nih.gov/18517736
  20. [24]Dridi, W., Henry, D. & Ben Hadid, H. 2008b Influence of acoustic streaming on the stability of melt flows in horizontal Bridgman configurations. J. Cryst. Growth 310, 1546–1551.DOI
  21. [25]Dridi, W., Henry, D. & Hadid, H. Ben 2010 Stability of buoyant convection in a layer submitted to acoustic streaming. Phys. Rev. E 81, 056309.pubmed.ncbi.nlm.nih.gov/20866324
  22. [26]Eckart, C. 1948 Acoustic resonances and streams caused by sound waves. Phys. Rev. 73, 68–76.
  23. [27]El Ghani, N., Miralles, S., Botton, V., Henry, D., Ben Hadid, H., Ter-Ovanessian, B. & Marcelin, S. 2021 Acoustic streaming enhanced mass transfer at a wall. Intl. J. Heat Mass Transfer 172.DOI
  24. [28]Fauconnier, M., Mauger, C., Béra, J.-C. & Inserra, C. 2022 Nonspherical dynamics and microstreaming of a wall-attached microbubble. J. Fluid Mech. 935, A22.DOI
  25. [29]Foss, J. F. 2004 Surface selections and topological constraint evaluations for flow field analyses. Exp. Fluids 37, 883–898.DOI
  26. [30]Friend, J. & Yeo, L. Y. 2011 Microscale acoustofluidics: microfluidics driven via acoustics and ultrasonics. Rev. Mod. Phys. 83, 647–704.DOI
  27. [31]Frommelt, T., Kostur, M., Wenzel-Schäfer, M., Talkner, P., Hänggi, P. & Wixforth, A. 2008 Microfluidic mixing via acoustically driven chaotic advection. Phys. Rev. Lett. 100, 034502.pubmed.ncbi.nlm.nih.gov/18232985
  28. [32]Green, A., Marshall, J. S., Ma, D. & Wu, J. 2016 Acoustic streaming and thermal instability of flow generated by ultrasound in a cylindrical container. Phys. Fluids 28, 104105.DOI
  29. [33]Hagsäter, S. M., Jensen, T. G., Bruus, H. & Kutter, J. P. 2007 Acoustic resonances in microfluidic chips: full-image micro-PIV experiments and numerical simulations. Lab Chip 7.pubmed.ncbi.nlm.nih.gov/17896019
  30. [34]Henderson, R. D. 1997 Nonlinear dynamics and pattern formation in turbulent wake transition. J. Fluid Mech. 352, 65–112.DOI
  31. [35]Henderson, R. D. & Barkley, D. 1996 Secondary instability in the wake of a circular cylinder. Phys. Fluids 8, 1683–1685.
  32. [36]Henry, D., Vincent, B., Miralles, S., Botton, V. & Hadid, H. Ben 2022 Steady and oscillatory flows generated by Eckart streaming in the two-dimensional Rayleigh–Bénard configuration. J. Fluid Mech. 952, A28.DOI
  33. [37]Huang, Z., Gao, R, Gao, Y.Y & Xi, G. 2024 Subcritical transitional flow in two-dimensional plane poiseuille flow. J. Fluid Mech. 994, A6.DOI
  34. [38]Hunt, J. C. R., Abell, C. J., Peterka, J. A. & Woo, H. 1978 Kinematical studies of the flows around free or surface-mounted obstacles; applying topology to flow visualization. J. Fluid Mech. 86, 179–200.DOI
  35. [39]Kamakura, T., Sudo, T., Matsuda, K. & Kumamoto, Y. 1996 Time evolution of acoustic streaming from a planar ultrasound source. J. Acoust. Soc. Am. 100, 132–138.DOI
  36. [40]Karniadakis, G. E., Israeli, M. & Orszag, S. A. 1991 High-order splitting methods for the incompressible Navier-Stokes equations. J. Comput. Phys. 97, 414–443.DOI
  37. [41]Kozhemyakin, G.N. 2003 Imaging of convection in a Czochralski crucible under ultrasound waves. J. Cryst. Growth 257 (3), 237–244.DOI
  38. [42]Kumar, A. & Pothérat, A. 2020 Mixed baroclinic convection in a cavity. J. Fluid Mech. 885, A40.DOI
  39. [43]Landau, L. D. & Lifschitz, E. M. 1987 Fluid Mechanics, 2nd edn. Pergamon Books Ltd.
  40. [44]Launay, G., Cambonie, T., Henry, D., Pothérat, A. & Botton, V. 2019 Transition to chaos in an acoustically driven cavity flow. Phys. Rev. Fluids 4, 044401.DOI
  41. [45]Lebon, G., Salloum-Abou-Jaoude, G., Eskin, D., Tzanakis, I., Pericleous, K. & Jarry, P. 2019 Numerical modelling of acoustic streaming during the ultrasonic melt treatment of direct-chill (DC) casting. Ultrason. Sonochem. 54, 171–182.DOI
  42. [46]Lighthill, J. 1978 Acoustic streaming. J. Sound Vib. 61, 391–418.
  43. [47]Lyubimova, T. P. & Skuridin, R. V. 2019 The effect of acoustic wave on the stability of stationary convective flow of binary fluid in a horizontal layer subjected to horizontal temperature and concentration gradients. Intl. J. Heat Mass Transfer 132, 789–801.DOI
  44. [48]Mitome, H. 1998 The mechanism of generation of acoustic streaming. Electron. Commun. Japan 81.doi.org/10.1002/(sici)1520-6440(199810)81:10<1::aid-ecjc1>3.0.co;2-9
  45. [49]Moffatt, H. K. 1964 Viscous and resistive eddies near a sharp corner. J. Fluid Mech. 18, 1–18.
  46. [50]Mondal, R. & Mahbub Alam, Md. 2023 Blockage effect on wakes of various bluff bodies: A review of confined flow. Ocean Eng. 286, 115592.DOI
  47. [51]Moudjed, B., Botton, V., Henry, D., Ben Hadid, H. & Garandet, J.-P. 2014a Scaling and dimensional analysis of acoustic streaming jets. Phys. Fluids 26, 093602.DOI
  48. [52]Moudjed, B., Botton, V., Henry, D., Millet, S., Garandet, J.-P. & Ben Hadid, H. 2014b Oscillating acoustic streaming jet. Appl. Phys. Lett. 105.DOI
  49. [53]Moudjed, B., Botton, V., Henry, D., Millet, S., Garandet, J.-P. & Ben Hadid, H. 2015 Near-field acoustic streaming jet. Phys. Rev. E 91, 033011.pubmed.ncbi.nlm.nih.gov/25871206
  50. [54]Muller, P. B. & Bruus, H. 2015 Theoretical study of time-dependent, ultrasound-induced acoustic streaming in microchannels. Phys. Rev. E 92, 063018.DOI
  51. [55]Orosco, J. & Friend, J. 2022 Modeling fast acoustic streaming: steady-state and transient flow solutions. Phys. Rev. E 106, 045101.DOI
  52. [56]Pothérat, A. and Zhang, L. 2018 Dean flow and vortex shedding in a three-dimensional 180 degrees sharp bend. arXiv preprint arXiv:1807.10950 .arxiv.org/abs/1807.10950
  53. [58]Rayleigh, L. 1884 On the circulation of air observed in Kundt’s tubes , and on some allied acoustical problems. Phil. Trans. R. Soc. London 175, 1–21.
  54. [59]Riaud, A., Baudoin, M., Bou Matar, O., Thomas, J-L. & Brunet, P. 2017 On the influence of viscosity and caustics on acoustic streaming in sessile droplets: an experimental and a numerical study with a cost-effective method. J. Fluid Mech. 821, 384–420.DOI
  55. [60]Rudenko, O.V & Sukhorukov, A.A. 1998 Nonstationary Eckart Streaming and Pumping of Liquid in an Ultrasonic Field. Acoust. Phys. 44, 565–570.
  56. [61]Sapardi, A. M., Hussam, W. K., Pothérat, A. & Sheard, G. J. 2017 Linear stability of confined flow around a 180-degree sharp bend. J. Fluid Mech. 822, 813–847.DOI
  57. [62]Schmid, P. J. 2007 Nonmodal stability theory. Ann. Rev. Fluid Mech. 39, 129–162.DOI
  58. [63]Sheard, G. J., Thompson, M. C. & Hourigan, K. 2004 From spheres to circular cylinders: non-axisymmetric transitions in the flow past rings. J. Fluid Mech. 506, 45–78.DOI
  59. [64]Sherwin, S. J. & Blackburn, H. M. 2005 Three-dimensional instabilities of steady and pulsatile axisymmetric stenotic flows. J. Fluid Mech. 533, 297–327.DOI
  60. [65]Thompson, M.C., Leweke, T. & Provansal, M. 2001 Kinematics and dynamics of sphere wake transition. J. Fluids Struct. 15, 575–585.DOI
  61. [67]Vincent, B., Henry, D., Kumar, A., Botton, V., Pothérat, A. & Miralles, S. 2024a Phenomenology of laminar acoustic streaming jets, arXiv: 2409.15083.DOI
  62. [68]Vincent, B., Miralles, S., Henry, D., Botton, V. & Pothérat, A. 2024b Experimental study of a helical acoustic streaming flow. Phys. Rev. Fluids 9, 024101.

Paper details

vlnc:W_k9b37xev
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 3

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents