See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/346035831

Scattering of acoustic waves by a nonlinear resonant bubbly screen

Article in Journal of Fluid Mechanics · November 2020

CITATIONS READS

7 317

5 authors, including:

Kim Pham Jean-François Mercier

ENSTA Paris ENSTA ParisTech
62 PUBLICATIONS 1,874 CITATIONS 82 PUBLICATIONS 981 CITATIONS

SEE PROFILE SEE PROFILE

Daniel Fuster J.-J. Marigo

Sorbonne Université École Polytechnique
80 PUBLICATIONS 2,331 CITATIONS 235 PUBLICATIONS 14,624 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Agnès Maurel on 20 November 2020.

The user has requested enhancement of the downloaded file.

Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

J. Fluid Mech. (2021), vol. 906, A19. © The Author(s), 2020. 906 A19-1
Published by Cambridge University Press
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence
(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and
reproduction in any medium, provided the original work is properly cited.
doi:10.1017/jfm.2020.799

Scattering of acoustic waves by a nonlinear

resonant bubbly screen

Kim Pham1, †, Jean-Francois Mercier2 , Daniel Fuster3 , Jean-Jacques Marigo4

and Agnès Maurel5
1 IMSIA, CNRS, EDF, CEA, ENSTA Paris, Institut Polytechnique de Paris, 828 Bd des Maréchaux,
91732 Palaiseau, France
2 Poems, CNRS, ENSTA Paris, INRIA, 828 Bd des Maréchaux, 91762 Palaiseau, France
3 Institut d’Alembert, CNRS, Univ. Paris 6, 4 place Jussieu, 75005 Paris, France
4 Laboratoire de Mécanique des Solides, CNRS, Ecole Polytechnique, 91120 Palaiseau, France
5 Institut Langevin, ESPCI Paris, PSL University, CNRS, 1 rue Jussieu, 75005 Paris, France

(Received 14 June 2020; revised 11 September 2020; accepted 15 September 2020)

Our study concerns the propagation of acoustic waves through a thin screen made of a
periodic arrangement of air bubbles in water. The bubbles are oscillators of the Minnaert
type whose dynamics is modified by the containment. This nonlinear dynamics is obtained
in the time domain using asymptotic analysis and a homogenization technique involving
three scales, those being the scale of a bubble, that of the array and eventually that of the
wavelength. The resulting effective model is set in the water (the screen has disappeared)
and it encapsulates the effect of the screen in a jump of the normal acoustic velocity.
The jump is linked to the continuous version of the bubble radius which satisfies an
equation of the Rayleigh–Plesset type. This allows us to highlight two important effects.
Firstly, a bubble within the array has a much larger radiative damping than an isolated
bubble. Secondly it perceives a pressure which differs from the acoustic pressure imposed
by the source due to bubble–bubble interactions; it results in a term of mass correction
deduced from the Green’s function for a Laplace problem which accounts for the bubble
arrangement. Our findings are exemplified by numerical experiments of the scattering of
a short pulse in the linear and nonlinear regimes.
Key words: bubble dynamics

1. Introduction
In 1985, Caflisch and co-workers analysed the wave propagation in bubbly liquids
(Caflisch et al. 1985a,b) offering a rigorous mathematical framework to the former
analysis developed by Van Wijngaarden (1968). Using asymptotic analysis, they derived
an effective wave equation involving a continuous version of the bubble radius satisfying

† Email address for correspondence: [email protected]

Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-2 K. Pham and others

2R

r
x

F IGURE 1. The actual problem: a bubbly screen in the x = 0 plane with spacing h in both
directions.

an equation of the Rayleigh–Plesset type. Once linearized in the harmonic regime, they
exhibited an effective speed of sound with strong variations in the vicinity of the Minnaert
frequency.
In some practical situations, the region of the bubbles is reduced to one or a few
layers (figure 1). The most famous examples are the anechoic tiles, codenamed Alberich,
developed during the Second World War by the German Navy; these tiles are the building
blocks of a rubber net containing bubbles used to block the signals of sonars (Gaunaurd
1977; Hladky-Hennion & Decarpigny 1991). Other examples are the bubble nets used
to protect underwater structures from damage by underwater explosions (Domenico
1982) or those created by some of the marine mammals to catch fish (Leighton 2004).
Recently, these bubbly screens with subwavelength thickness have been revisited in the
context of metamaterials. Leroy et al. (2009) realized controlled experiments to study
the transmission of ultrasound by a single layer of bubbles in water; Bretagne, Tourin &
Leroy (2011) realized a bubble raft close to an interface with air or sandwiched in air and
analysed the effects of such interfaces; Leroy et al. (2015) demonstrate the ability of the
metascreens to produce superabsorption and Lanoy et al. (2018) obtain perfect absorption
by critical coupling. Eventually an original metasurface has been recently proposed by
Schnitzer, Brandão & Yariv (2019) with cylindrical bubbles trapped in the grooves of
a microstuctured hydrophobic wall which is shown to be capable of supporting guided
waves analogues to spoof plasmon waves.
From a theoretical point of view, the first study on the interaction of acoustic waves
with an array of bubbles dates from 1966 (Weston 1966). Although Weston concludes that
‘a plane array behaves as a plane screen of gas – there is no resonance at all’, his study
contains almost all the elements of the model of Leroy et al. (2009). In both cases, the
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-3

analysis is based on multiple scattering theory in the linear regime but Weston consider an
over-simplified scattering function of a single bubble. In contrast, the model of Leroy et al.
(2009) shows that bubbles within a screen experience a radiative damping much larger
than an isolated bubble does; also that the resonant frequency of the screen is higher than
that of an isolated bubble due to bubble–bubble interactions. A different approach has
been proposed by Ng & Ting (1986) and Miksis & Ting (1989) who use the effective
wave equation derived in Caflisch et al. (1985a,b) for a bubbly liquid. By considering
the limiting case of a thin layer, they derive transmission conditions through an effective
screen in the nonlinear regime and in the time domain. This model recovers the large
damping of the screen but misses the shift of the resonant frequency. Eventually, Ammari
et al. (2017a) provides almost exact analytical solution of the scattering coefficients in
the linear harmonic regime within a rigorous mathematical framework when the screen is
placed in the vicinity of a Dirichlet wall; their work applies mutatis mutandis to an isolated
screen.
In parallel to these works, one approach has become popular to describe the collective
dynamics of a cluster of bubbles. It consists of enriching the Rayleigh–Plesset (RP)
equation with a term encapsulating the effect of bubble–bubble interaction. Starting with
Harkin, Kaper & Nadim (2001) and Doinikov (2001), who consider the interaction of two
bubbles, the model has been further extended heuristically to a system of N bubbles. With
neglect of the surface tension and liquid viscosity, this heuristic RP equation reads
    3γ 
d  R2j Ṙj
N
3 2 Req
ρ Ri R̈i + Ṙi + peq 1 − = −pa − ρ , (1.1)
2 Ri dt j=1,j =/ i hij

where ρ is the mass density of the liquid, Ri is the radius of one bubble within the cluster
and Req its value at equilibrium, peq the pressure at equilibrium and pa the forcing acoustic
pressure (γ stands for the adiabatic index). In this equation, the sum in the right-hand side
term is assumed to account for the effect of the (N − 1) bubbles with radii Rj at distances
hij of the ith bubble, see e.g. Bremond et al. (2006), Guédra, Cornu & Inserra (2017). One

obvious drawback of (1.1) is that 1/hij diverges for large N while R2j Ṙj is bounded. The
reason is that this sum is the contribution of the bubbles j to the pressure seen by the
bubble i; as such, it should contain the term eikhij with k the wavenumber, as in Weston
(1966) and Leroy et al. (2009). Hence, (1.1) holds for a cluster extension much smaller
than the typical wavelength.
In this study, we adapt the asymptotic analysis of Miksis & Ting (1989) to the case of
a thin bubbly screen. The calculations are performed in the time domain and preserving
the nonlinear dynamics of the bubble oscillations. To do so, we focus on the limit of
sparse arrays with spacing much larger than the bubble radius, which allows for the
resonances to take place; in the other limit, a dense array would behave basically as
a wall. As the typical wavelength remains larger than the spacing, three scales can be
defined. In § 2, we discuss the meanings of the problems obtained at these three scales
and their matchings; the technical calculations have been collected in appendix A. The
whole analysis results in effective transmission conditions with a jump of the normal
acoustic velocity dictated by a modified RP equation, see forthcoming (2.5) (figure 2).
In § 3 results of the model are discussed in the linear and nonlinear regimes for a short
incident pulse. In particular, it is shown that for one-dimensional propagation, a different
(although equivalent) formulation of the effective problem provides a simple physical
interpretation of the mechanisms of interaction, see forthcoming (3.2); our modified RP
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-4 K. Pham and others

r
x

F IGURE 2. The effective problem: an equivalent screen at x = 0 across which jump conditions
(2.5) apply, with R(r, t) being a continuous version of the bubble radius.

equation is similar although not equivalent to (1.1) and it generalizes to the nonlinear
regime the result of Leroy et al. (2009). The properties of the model in terms of energy
conservation are discussed in § 4. Eventually, concluding remarks and perspectives are
given in § 5.
The numerical results reported throughout the paper use the following parameters:
⎫
ρg c2g ⎪
ρ = 10 kg m ,
3 −3
c = 1500 m s ,−1
peq = = 0.225 atm.,⎬
γ (1.2a–e)
⎪
⎭
γ = 1.4, Req = 10 μm,

with ρg, and cg, the mass density and speed of sound in the gas and in the liquid at
equilibrium. Different spacings h from 0.1 to 2 mm will be considered. The separation of
the three scales is verified with a wavelength λM  1 cm at the Minnaert resonance

3γ peq
ωM = . (1.3)
ρ R2eq

2. The actual problem and the effective screen problem

The effective screen problem aims to simplify the actual problem of the interaction of
acoustic waves with an array of gas bubbles. In the actual problem, the array is located
at x = 0 and it has the periodicity h in both directions (figure 1). We assume a weak
compressibility of the liquid, an irrotational flow and the effects of the viscosity and
surface tension are disregarded. Let ρ, p and u be the density, the acoustic pressure and the
acoustic velocity which are functions of the time t and space x = xex + r (with r · ex = 0).
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-5

In the liquid and in the gas, (ρ, p, u) satisfy the Euler equations
 
∂ρ ∂u
+ div(ρu) = 0, ρ + (u · ∇)u = −∇p, ∇ ∧ u = 0, (2.1a–c)
∂t ∂t
where p = ( ptot − peq ) with ptot the total pressure. At the equilibrium ptot = peq , u = 0 and
the mass densities in the liquid and in the gas are ρ,g . For a weakly compressible liquid
and a perfect gas under adiabatic transformations (see e.g. Bergamasco & Fuster (2017)
for a discussion on the validity of the adiabatic regime), the equations of state are written
in the form
⎫
ptot = peq + c2 (ρ − ρ ), in the liquid, ⎪
⎪
⎬
 γ (2.2)
ptot ρ γ peq ⎪
= , in the gas, with c2g = .⎪
⎭
peq ρg ρg

For an acoustic wavelength that is large compared to the bubble radii, each bubble
with index i is set in forced radial oscillations Ri (t). Thus, at the interface gas/liquid, the
pressure and the velocity are continuous and

u · n = Ṙi , (2.3)

where the dot means the time derivative.

2.1. Formulation of the effective screen problem

The effective problem is set in the liquid only; because of the weak compressibility of the
liquid and because of the low Mach number, the propagation is linear, hence ( p, u) satisfy
the Euler equations
∂p ∂u
+ ρ c2 div u = 0, ρ = −∇p, ∇ ∧ u = 0. (2.4a–c)
∂t ∂t
The bubbly screen has disappeared, its effects being now encapsulated in a jump of the
normal velocity, specifically
⎫
4πR2 ⎪
[[p]] = 0, [[ux ]] = Ṙ, ⎪
⎪
h2 ⎬
    3γ  (2.5)
3 ρg cg
2
Req ρ ∂  ⎪
ρ RR̈ + Ṙ2 + 1− −δ R2 Ṙ = −p(x = 0, r, t),⎪
⎪
⎭
2 γ R h ∂t

where R(r, t) is a continuous version of the bubble radius. We have defined [[p]] =
p(0+ , r, t) − p(0− , r, t) and [[ux ]] = ux (0+ , r, t) − ux (0− , r, t) the jumps of the acoustic
pressure and normal velocity across the effective screen (figure 2). In (2.5), the RP
equation contains two important contributions. Firstly, the forcing term is the acoustic
pressure p(0, r, t) at the equivalent interface; from (2.4a–c) and the form of [[ux ]] in
(2.5), p(0, r, t) has a contribution ρ c (R2 Ṙ/h2 ) corresponding to the radiative damping
of the screen. It is termed superradiative damping in Leroy et al. (2015) since it is much
greater than the radiative damping of an isolated bubble considered in, for example, Keller
& Miksis (1980). Next, the term δ (ρ /h) ∂t (R2 Ṙ) is attributable to the bubble–bubble
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-6 K. Pham and others

interaction, with δ a constant depending on the arrangement of the bubbles within the
array; it will be discussed further in the following section. For the time being, we can
notice that the RP equation in (2.5) differs from (1.1); it applies to an infinite number of
bubbles since δ is finite with a sign a priori free while δ is negative by construction in
(1.1).

2.2. Results of the asymptotic analysis at the three scales

As previously said, the effective screen problem is obtained using asymptotic analysis
combined with homogenization. The underlying assumptions are (i) a low frequency
regime, (ii) a dilute screen and (iii) a low Mach number, specifically

ω Req
ε= h  1, = O(ε), Ma = O(ε 4 ), (2.6a–c)
c h

where ω is the typical frequency imposed by the source. These scalings produce a
separation of three scales, the scale of a single bubble, that of the screen spacing and
that of the typical wavelength λ = 2π(c /ω); each scale is associated with a problem
much simpler than the complete actual one and which can be solved at the dominant order
and then iteratively at higher orders in ε. At each order, the analysis aims to put those
three problems together using matching conditions. Our analysis has been conducted at
the dominant, order 0 in ε, see appendix A.2 and at order 1 in ε, see appendix A.3; the
results at the two first orders are then gathered to form a unique problem. Below, we
summarize the results at the three scales.
The microscopic problem results from a zoom in on a single bubble of centre r n . It
corresponds to the classical problem of an isolated bubble submitted to a pressure at
infinity p∞ (r n , t), primarily considered by Rayleigh (1917) in the incompressible case,
see also Plesset & Prosperetti (1977). Due to the low Mach number, the pressure and mass
density are uniform within the bubble, with ∂t ρg + ρg div u = 0. In the liquid, the flow
satisfies the incompressible nonlinear Euler equation ρ (∂t u + u · ∇u) = −∇p. We get
⎫
at the microscopic scale:   ⎪
⎪
3γ ⎪
⎪
Req Ṙ|x − r n | ⎪
⎪
in the bubble, p(t) = peq −1 , u(x, t) = er , ⎪
⎬
R R
  ⎪
⎪
∂t (R2 Ṙ) R4 Ṙ2 ṘR2 ⎪
⎪
⎪
⎪
in the liquid, p(x, t) = p∞ (r n , t) + ρ − , u(x, t) = er . ⎪
⎭
|x − r n | 2|x − r n |4 |x − r n |2
(2.7)

A classical form of the Rayleigh–Plesset equation is obtained of the form ρ (RR̈ + 32 Ṙ2 ) =
pn − p∞ , where pn (t) is the pressure in the liquid at the interface |x − r n | = R, and in the
absence of surface tension pn (t) = p(t). However, at this scale p∞ (t), being the pressure
at infinity seen by a single bubble, is unknown. In figure 3, we report an illustration of the
patterns of the radial velocity u · er and of the projection u · ex for a bubble at r n = 0 and
t = 0. We have considered a plane wave at incidence 45◦ in the linear regime pinc (x, t) =
Δp exp(ik · x − iωt) with Δp = 0.1peq and ω = 0.7ωM ; the array spacing is h = 50Req .
The velocity fields (2.7) in the bubble and in the liquid are easily obtained owing to the
resolution of the linearized version of (2.5), see appendix B. The order of magnitude of
the velocity of 0.1 m s−1 is given by Ṙ ∼ ωΔp/ρ Req (ωM 2
− ω2 ).
The mesoscopic problem corresponds to the scale of the array spacing, intermediate
between the scale of the bubbles and that of the wavelength (figure 4). At this scale, the
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-7

(a) Microscopic problem (b) (c)
0 0.1
–0.1 u · er u · ex
– 0.1

0.1

rn.
–0.1

–5 5 –5 5
x/R x/R
F IGURE 3. Solution of the microscopic problem – a single bubble with pressure p∞ at
infinity (a). Radial velocity (b) and velocity along the x-axis (b) in m s−1 for an incident plane
wave at oblique incidence in the linear regime with amplitude Δp = 0.1peq , see main text.

bubbles are reduced to points at r n . The flow in the liquid is still incompressible but now
it is linear. Specifically, for x ∈ Yn the periodic cell containing r n , we have
⎫
at the mesoscopic scale (in the liquid): ⎪
⎪
⎪
⎪
∂t (R Ṙ)
2 ⎪
⎬
p(x, t) = p∞ (r n , t) − ρ (G(x − r n ) − δ) , (2.8)
h ⎪
⎪
⎪
R2 Ṙ ⎪
⎪
u(x, t) = u∞ (r n , t) + ∇G(x − r n ), ⎭
h
where ρ ∂t u∞ = −∇p∞ and G(x) is the Green’s function for the Laplace problem set in
Yn with
h
G(x) ∼ − + δ, (2.9)
|x|→0 |x|
and G(x) ∼|x|→∞ 2π/h|x| (i.e. without constant at infinity), see also (A 34). For a square
array, the Green’s function can be calculated explicitly using the decomposition
2π 1
G(x) = |x| + S(x), with S(x) = − exp(−(2π/h)|x|) exp(2iπ/h(nr2 + mr3 )),
h 

√ (2.10)
with  = n 2 + m2 and (n, m) ∈ Z2 \(0, 0). The Green’s function is singular at x = 0
since S(x) ∼ x→0 −(h/|x|) + δ. We find δ = 3.9 in excellent agreement with Weston
(1966) and Leroy et al. (2009). In both references the authors evaluate the pressure seen
by a single bubble within the array using multiple scattering theory in the linear harmonic
regime. They introduce a cutoff distance b as the effective inter-bubble distance; in Weston
(1966),√b = h log γ with log γ = 0.58 the Euler constant and in Leroy et al. (2009),
b = h/ π; the cutoff is linked to δ through δ = 2π(b/h) which allows us to conclude
on the agreement. Interestingly, both authors justify this value by making reference to
a problem similar to the Green’s function problem (2.9). However, this problem is set
for the Helmholtz problem while ours is set for the Laplace problem since k → 0 at
this intermediate scale. An illustration of the mesoscopic solution is shown in figure 4.
From (2.8), the order of magnitude of the velocity (u − u∞ ) · ex when x → ±∞, of
approximately 0.5 × 10−3 m s−1 , is given by (2πR2 /h2 )Ṙ ∼ 2πωReq Δp/ρ h2 (ωM 2
− ω2 )
and it diverges at x = r n due to the singularity of the Green’s function (see appendix B).
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-8 K. Pham and others

(a) Mesoscopic problem (b) (×10−3)
1
–1 (u – u∞) · ex (×10−3)
rn 1
ex
Yn
–1
–4 x/h 4

F IGURE 4. Solution of the mesoscopic problem – the bubble is reduced to a singularity at r n

within a cell Yn infinite along ex and periodic in the two other directions (a). Velocity (u − u∞ ) ·
ex along the x-axis from (2.8) resulting from the microscopic solution in figure 3. The profile
along x is taken at a distance 0.15 h from the centreline; the profile at the centreline is singular
at x = 0 according to (2.9) (b).

(a) (b) (c)

Micro Meso
0.5
0.1 ux micro
ux meso 2Req
– 0.1 h
–0.5
– 0.1 0.1
x/h
–0.1 0.1 – 0.1 0.1
x/h x/h
F IGURE 5. Micro-to-meso matching – the solution at the mesoscopic scale is the limit of that
at the microscopic scale when 2Req /h vanishes. Two-dimensional fields of the ux velocity at the
microscopic scale (a) and at the mesoscopic scale (b); the dotted white lines show the bubble
boundary; velocity profiles along the centreline (c).

The singularity of the mesoscopic solution results from the micro-to-meso matching.
Indeed, the mesoscopic solution in (2.8) when x/h → 0 has to coincide with the
microscopic solution (2.7) when |x − r n |/R → ∞. Hence, from (2.7), the singular
behaviour of the pressure has to be of the form p ∼ p∞ (r n , t) + ρ (∂t (R2 Ṙ)/|x − r n |), and
that of u is given by ρ ∂t u = −∇p; this is what we recover with (2.8). This matching is
illustrated in figure 5. Reporting the fields of both solutions for a small but non-zero Req /h
shows that the mesoscopic solution coincides with the microscopic one in the liquid that
it extends up to 0. At this stage p∞ (r n , t) is still unknown.
The compressibility of the liquid appears in the macroscopic problem. At this
wavelength scale, the discrete set of points r n is seen as a continuum hence p∞ (r n , t)
can be replaced by p∞ (r, t), the same for u∞ . Now, we aim to know the meanings of p∞
and u∞ and this is done by considering the limits x/h → ±∞ of the mesoscopic solution
which provide the limits x/λ → 0± of the macroscopic solution. Owing to the behaviour
of G in (2.8), the macroscopic solution has to satisfy
⎫
∂t (R2 Ṙ) ∂t (2πR2 Ṙ) ⎪
±
p(x → 0 , r, t) = p∞ (r, t) + ρ δ ∓ ρ x,⎪
⎬
h h2
(2.11)
2πR2 Ṙ ⎪
⎪
±
u(x → 0 , r, t) = u∞ (r, t) ± ex . ⎭
h2
The meaning of the relations for u is unambiguous: u∞ = 12 (u|x=0+ + u|x=0− ) is the mean
value of the velocity at x = 0 and u experiences a jump (u|x=0+ − u|x=0− ) = 4πR2 Ṙ/h2 ex
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-9

(a) (×10–3) (b)

2 ux meso
h
ux macro λ
–2

(×10–3)
3

–3

–1 1
x/λ
F IGURE 6. Meso-to-macro matchings – the solution at the macroscopic scale for x/λ → 0 is the
limit of that of the mesoscopic scale for large x/h; the solution at the mesoscopic scale accounts
for the jump in the velocity (2.5). Profiles and fields of the solutions (a); panel (b) shows a zoom
near the screen. The incident wave is a plane wave at incidence 45◦ and ω = 0.7ωM .

as announced in (2.5). Owing to this result, and with ρ ∂t u = −∇p which applies for
( p, u) and ( p∞ , u∞ ), the relations for p appear as the Taylor expansions of the pressure
p being continuous at x = 0 with a discontinuous first derivative with respect to x.
Specifically, the relations on p in (2.11) can be written as
⎫
∂p ± ⎪
p(x → 0± , r, t) = p(0, r, t) + (0 , r, t)x, ⎪
⎬
∂x
(2.12)
∂t (R2 Ṙ) ∂p ± ∂t (2πR2 Ṙ) ⎪
⎪
⎭
with p(0, r, t) = p∞ (r, t) + ρ δ, (0 , r, t) = ∓ρ .
h ∂x h2
This tells us that [[p]] = 0, as announced in (2.5), and provides the meaning of p∞ . The
pressure p∞ (r, t) = p(0, r, t) − ρ (∂t (R2 Ṙ)/h)δ seen by a single bubble has a contribution
due to the arrangement of the array and which has been captured by the Green’s function
at the mesoscopic scale; again, as the Green’s function is that for the Laplace problem, δ
is independent of the frequency. It is now sufficient to replace p∞ in the RP equation that
we recover from the microscopic solution (2.7) to get the final RP equation in (2.5).
The meso-to-macro matching is illustrated in figures 6 (for ω = 0.7ωM ) and 7 (for
ω = ωM ). We have represented the macroscopic solution in reflection and transmission
for |x|/λ > h/λ and the mesoscopic ones for |x|/λ < h/λ; this choice of the x-range
is somehow arbitrary, it simply means the close vicinity of the array. It is visible
in both cases that the mesoscopic velocities are basically reduced to constant values
u∞ (r n , t) ± (2πR2 Ṙ/h2 )ex from (2.8); also visible is the fact that these values coincide
with the macroscopic velocities u(x → 0± , r ∈ Yn , t).

3. Analysis of the effective model

In this section we shall analyse the effective model (2.5) thanks to numerical
simulations. The one-dimensional compressible Euler equations in the liquid are solved
using the finite element method presented in Fuster, Dopazo & Hauke (2011) and the
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-10 K. Pham and others

(a) (×10–3) (b)

5 ux meso

ux macro
–5

(×10–3)
3

–3

–1 1
x/λ
F IGURE 7. Same representation as in figure 6 for ω = ωM resulting in almost perfect reflection.

transmission condition is accounted for by using dual nodes at the location of the bubbly
screen, each node representing the two side solution ux (0± , t). Next, R is obtained by
integrating the RP equation in (2.5) with a classical Runge–Kutta method. Following Doc
et al. (2016), we consider an initial Gaussian pulse of the form
pinc (x, t) = Δp exp(−(x − c t)2 /σ 2 ), (3.1)
and we use σ = λM /10. Hence, with a duration T/10, and T = λM /c ∼ 6 μs, the pulse
acts almost as a delta Dirac function and excites a wide frequency band around ωM .
Results are reported in figure 8 against h/Req in the linear and nonlinear regimes. The
two time scales, the duration of pulse T/10 and the free oscillation period T, are visible
in the response R(t) of the bubbly screen and in the resulting acoustic pressure p(0, t).
Next, two limiting cases are observed. For large inter-bubble distances h/Req , each bubble
behaves as if it were on its own. The radiative damping is small, which allows for long
time oscillations but as a counterpart, the screen has a weak interaction with the liquid
(p(0, t) ∼ pinc (0, t)). In the opposite limit of small inter-bubble distances, the radiative
damping is large, resulting in over-damped oscillations and strong interaction with the
liquid (p(0, t) significantly departs from pinc (0, t)). This is consistent with the intuitive
idea that dense arrays of bubbles act basically as perfect shields while sparse ones are
transparent. From figure 8, the intermediate regime is for h/Req equal to a few dozen.
The figure 9 show snapshots of the acoustic pulse p(x, t) in the liquid in this
intermediate regime (h/Req = 20) for increasing Δp. While the propagation in the liquid
is linear, increasing Δp affects the response of the nonlinear oscillator (the screen). It is
visible that the signal emitted by the screen, with nonlinear shape, is transmitted to the
liquid, in reflection and also, but less visible, in transmission. What is also visible is the
fact that increasing the nonlinearities weakens the interaction of the screen with the liquid.

3.1. Analysis in a one-dimensional problem

We shall now simplify the screen model (2.5) to get physical insights into the interaction
mechanisms. Restricting ourselves to one-dimensional problem allows for the emergence
of a modified RP equation which contains explicitly the radiative damping term and which
is independent of the propagation in the liquid; this later will appear as a simple byproduct
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-11

R(t) – Req p(0, t)
(a)
R p
1 1
0
−1 0
160 160
80 80
h 40 h 40
Req 20 6 Req 20 6
4 4
10 2 t 10 2 t
0 0
T T
(b)
0.2 1
0
− 0.2 0
160 160
80 80
h 40 h 40
Req 20 6 Req 20 6
4 4
10 2 t 10 2 t
0 0
T T
F IGURE 8. Time variations of the radius R(t) and acoustic pressure p(0, t) in (2.5) for increasing
array spacing h/Req (a) in the linear regime, Δp/peq = 10−3 and (b) in the nonlinear regime,
Δp/peq = 15; ΔR = (Δp/3γ peq )Req is the amplitude of linear undamped oscillations.

of the RP solution R(t). We shall after this consider the harmonic regime for a direct
comparison with the model of Leroy et al. (2009).

3.1.1. Interpretation in terms of pressure radiated by the screen

One-dimensional propagation involves acoustic pressure of the form p(x, t) =
pinc (x, t) + f (t ∓ x/c ) for x ∈ R± . The same function f is involved for x ∈ R± since in
our problem (2.5) the pressure at x = 0 is continuous; whence f (t) = p(0, t) − pinc (0, t).
Next, the propagation in the liquid is governed by the Euler equations (2.4a–c) which
implies that ux (x, t) = 1/ρ c ( pinc (x, t) ± f (t ∓ x/c )) for x ∈ R± . Eventually the jump
in the velocity in (2.5) provides f (t) = 2πρ c (R2 /h2 )Ṙ. It follows that the problem
(2.4a–c) and (2.5) can be written as
    3γ   ⎫
3 2 ρg c2g Req 2πρ c 2 δρ d R2 Ṙ ⎪
⎪
ρ RR̈ + Ṙ + 1− + R Ṙ − = −pinc (0, t),⎪
⎪
⎬
2 γ R h 2 h dt
  ⎪
x R2 (t)Ṙ(t) ⎪
⎪
p(x, t) = p (x, t) + p
inc rad
t∓ , with prad (t) = 2πρ c , ⎪
⎭
c h 2

(3.2)
for x ∈ R± (figure 10). This formulation provides a different although equivalent
interpretation of the mechanism of interaction. In a first step, the incident pulse
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-12 K. Pham and others

(a) p (b) p
1 peq = 1 1 peq = 4
0 0
–1 –1
0.8 0.8

t 0.4 t 0.4
T T
0 10 0 10
0 x 0 x
−10 σ −10 σ

(c) (d)
1
p 1
p
peq = 20 peq = 50
0 0
–1 –1
0.8 0.8

t 0.4 t 0.4
T T
0 10 0 10
0 x 0 x
−10 σ –10 σ
F IGURE 9. Snapshots of the acoustic pressure in the liquid p(x, t)/Δp for the bubbly screen in
a over-damped regime and increasing nonlinearities Δp/peq (h/Req = 20).

excites the screen; pinc (0, t) is the forcing term in the RP equation, (3.2), and it is worth
noting that the superradiative damping is now explicitly accounted for by the term in R2 Ṙ.
In a second step, the screen radiates pressure fluctuations in the liquid, symmetrically
on the right and on the left. These fluctuations prad (t ∓ x/c ), initially generated through
a nonlinear mechanism at x = 0, propagate linearly in the liquid; they contain the two
time scales, that of the forcing term pinc and that of the natural oscillation (under- or
over-damped). For our short incident pulse pinc the radiated pressure lasts much longer
than the pulse as sketched in figure 10. Incidentally, and from a practical point of view,
the resolution of (3.2) is simpler than that of (2.4a–c) and (2.5). Indeed, the RP equation
is set in time only, hence it can be solved once and for all, and afterwards the solution in
the liquid is explicitly known. In figure 10(b), the two numerical solutions of (2.4a–c) and
(2.5) and of (3.2) are shown to coincide.

3.1.2. One-dimensional problem in the linear regime

In the linear regime, with R = Req + r, r  Req , (3.2) simplifies further to
  ⎫
δReq 2πc pinc (0, t) ⎪
1− r̈ + 2 Req ṙ + ωM 2
r=− , ⎪
⎪
h h ρ Req ⎬
  ⎪
x R2eq ⎪
p(x, t) = p (x, t) + p
inc rad
t∓ , with prad (t) = 2πρ c 2 ṙ, for x ∈ R± .⎪
⎭
c h
(3.3)
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-13

p(x, t)
(a) (b)
p
pinc (x, t)
0.5 1.0
x
0

0
x p p
peq = 10
–3
peq = 10
x
prad t − c – 1.0 – 0.5
x –10 0 10 –10 0 10
prad t + c 
 x x
σ σ
F IGURE 10. (a) In the one-dimensional problem (3.2), the screen radiates pressure wave trains
symmetrically on the right and on the left after the incident signal has passed through it. The
solution is the sum of the two contributions p = pinc + prad from (3.2). (b) Resulting pressure
profiles p in the linear and nonlinear regimes, blue lines from (3.2) and dotted red line from the
resolution of (2.5).

The analysis can be conducted  inc frequency by frequency by simple Fourier

transform.
 rad We use p inc
(x, t) = p̂ (ω) exp(−i(ωt − kx)) dω and prad
(t ± x/c) =
p̂ (ω) exp(−i(ωt ± kx)) dω, with k = ω/c the wavenumber. The reflection R and
transmission T coefficients read

p̂rad (ω) = R(ω)p̂inc (ω) = (T (ω) − 1)p̂inc (ω), (3.4)

and [ωM 2
− ω2 (1 − δReq /h + iKReq )]r̂ = −p̂inc /ρ Req , p̂rad = −iKρ R2eq ω2 r̂ with K =
2π/kh2 , hence
iKReq
R(ω) =  2 , T (ω) = 1 + R(ω), (3.5a,b)
ωM δReq
−1+ − iKReq
ω h
with the radiative damping contained in KReq and a mass correction contained in
δ(Req /h). (Obviously, the same result for (R, T ) is found by linearizing (2.5) with ux =
−i/(ρ ω)∂x p from (2.4a–c).) As previously said, the damping of the array is much larger
than that of an isolated bubble; this latter is given by kReq and we have k/K = (kh)2  1.
The mass correction is the product of δ with Req /h. The constant δ depends on the
arrangement of the bubbles within the array and it is independent of Req and h. Hence, for
the same arrangement, the mass correction vanishes for sparse arrays Req /h → 0. From
(3.5a,b), the terms of bubble–bubble interaction and of radiative damping vanish for large
h resulting in R  0 (the screen is transparent for the acoustic waves). In the opposite limit
of dense arrays Req /h ∼ 1, hence KReq 1, resulting in R  −1; the screen becomes
equivalent to a wall of air with vanishing
 acoustic pressure. In the intermediate regime, R
has a maximum at a frequency ωM / 1 − δReq /h shifted above the Minnaert frequency as
observed in experiments (Leroy et al. 2009, 2015).
It is worth noting that nonlinearities also produce a shift of the resonance frequency.
This is illustrated in figure 11 where we report the Fourier transforms of the reflected and
transmitted pressure fields for increasing Δp/peq ∈ (1, 50) in our Gaussian pulse (3.1).
By analogy with the linear regime, the nonlinear resonance frequency is identified at
the maximum in the reflected signal. Expectedly, the frequency shift in the linear regime
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-14 K. Pham and others

(a) FFT [ p(x < 0, t)] (b) FFT [ p(x > 0, t)]

50

30

20

15
p
eq

10
Increasing p

10–1 100 101 102 10–1 100 101 102

ω ω
ωM ωM

F IGURE 11. Fourier transforms of the reflected p̂(x < 0, ω) and transmitted p̂(x > 0, ω)
pressures for increasing nonlinearities Δp/peq (h/Req = 40).

(Δp/peq = 1) for h/Req = 40 is small and we recover (3.4). In the nonlinear regime, the
screen imparts to the liquid acoustic pressure fluctuations with complex frequency content.
It is in particular visible that (i) for moderate nonlinearities (up to Δp/peq ∼ 10) the shift of
the resonance to higher frequency is significantly increased from ωM to approximately 3ωM
in the reported case, (ii) strong nonlinearities produce the appearance of harmonics and
multiple local maxima and minima in the reflection spectrum; in this case, the fundamental
resonance is shifted back to ωM .

4. Energy equation
4.1. Conservation of the energy in the screen model
In the absence of viscous loss, the sum of the energy in the liquid and of that in the
bubbles is conserved in time. This has to be true in the effective model (2.4a–c) and (2.5)
too. To check that this is indeed the case, we consider the equation of energy conservation
associated with the Euler equations in the liquid

  
d 1 p2
E + Φ = 0, with E = + ρ u dr dx,
2
(4.1)
dt 2 Ω ρ c2

for any bounded domain Ω; E is the acoustic energy and Φ = ∂Ω p u · n dS the flux of the
Poynting vector. However, since x = 0 is a surface of discontinuity, Φ has a contribution
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-15


Φ scr = − p[[ux ]]dr which from (2.5) reads
Γ

      3γ  
3 2 ρg c2g Req δρ ∂  2 dr
Φ = 4π
scr
ρ RR̈ + Ṙ + 1− − R Ṙ R Ṙ 2 .
2
x=0 2 γ R h ∂t h
(4.2)
It is easy to see that this flux is the time derivative of an effective energy Φ scr = (d/dt)E scr
with
 ⎫
 dr
E =
scr
ep + ec 2 , ⎪
⎪
h ⎪
⎬
x=0
   3(γ −1)  (4.3)
2
Mg cg R 3
1 Req γ 1 ⎪
2 ⎪
ep = + − , ec = Mrad Ṙ ,⎭ ⎪
γ Req γ −1 R γ −1 2

where ep = 0 at equilibrium (note that, to derive (4.3), we used notably that

d/dt(R3/2 Ṙ)2 = 2R2 Ṙ(RR̈ + 32 Ṙ2 )), and
 
4π R
Mg = ρg Req , Mrad = 4πρ R 1 − δ
3 3
. (4.4a,b)
3 h

Here, Mg is the (constant) mass of a bubble and Mrad , sometimes termed the ‘radiation
mass’, is the effective, or dynamic, mass obtained in the linear regime in (3.3).
Expectedly, the effective energy is similar to the classical energy of a single bubble in
an incompressible liquid, with
 R   
Req 3γ
ep = − p(4πR dR ), with p = peq
2
−1 (4.5)
Req R

the work necessary to change the bubble radius from Req to R. Next, ec is the kinetic
energy invested in a (incompressible) liquid during bubble oscillations; it takes the form
∞
ec = 12 R u2 (4πρ r2 dr) for an isolated bubble. For the array, it can be written
 h/δ
1 R2 Ṙ
ec = u2 (4πρ r2 dr), with u = er , (4.6)
2 R r2
where we recover the cutoff distance h/δ introduced by Weston (1966) and Leroy et al.
(2009). Recent works discuss the conservation of energy for a compressible liquid, see e.g.
Devaud et al. (2008) and Wang (2016), and the effects of confinement (Leighton 2011). As
in these references, we find that the energy which is conserved is not the ‘local energy’
E scr of the bubbles; the liquid and the bubbles can exchange energy hence, in the absence
of viscous losses, only (E scr + E ) is conserved.
We report in figure 12(a) an example of the time variations of E scr (t) along with those
of ER (reflected, i.e. computed in x < 0) and ET (transmitted, i.e. computed in x > 0)
in the linear regime. In (4.1) E = ER + ET is the energy in the liquid and the energy
conservation applies to (E + E scr ). In agreement with the snapshots in the figure 8(a),
the screen takes a significant part of the acoustic energy during the transit of the incident
pulse, up to 40 % of the total energy in the reported case. It then releases this energy over
a time which scales with T. In the present case, the regime is over-damped and the energy
release varies as ∼ exp(−5.6(t/T)), in agreement with (3.3) (δReq /h  0.2, KReq  2.4).
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-16 K. Pham and others

(a) (b) (c)
1
E R (t) ET (∞)
ET (∞)

ET (t)
E scr(t)
E R (∞)
E R (∞)
0
–1 0 1 10 102 1 10
t/T h/Req p/peq

F IGURE 12. Normalized energies. (a) Example of time variations of ER and ET (energies in the
liquid, computed for x < 0 and x > 0, with E = ER + ET in (4.1)), and E scr from (4.3); grey
dashed line shows (E + E scr )  1 (h/Req = 20, Δp/peq = 10−3 ); (b) ER and ET against h/Req
in the linear regime (Δp/peq = 10−3 ); (c) ER and ET against Δp/peq (h/Req = 20).

Panels (b,c) further quantify the influence of the array spacing and of the nonlinearities.
We have considered a sufficiently large time so that almost all the energy is in the liquid. As
previously said, going from sparse to dense array increasing h/Req produces a transition
from a perfect shield (total reflection) to a transparent screen (total transmission). Next,
incident signals with high nonlinearities are less sensitive to the screen; for Δp/peq larger
than approximately 50, the reflected energy ER is negligible.

5. Concluding remarks
We have derived an effective model which aims to reproduce the scattering properties
of a bubbly screen with resonances of the Minnaert type. The model enables to reduce the
effect of the screen to a jump of the acoustic velocity in the liquid coupled to a modified RP
equation. The RP equation contains the super-radiative damping of the array being much
larger than the classical damping of an isolated bubble. Next, a more significant result is
the derivation of the term of bubble–bubble interaction in the RP equation dictating the
nonlinear dynamics of the array. This contribution was absent from the analysis of Caflisch
et al. (1985a,b) for a bubbly liquid and its extension to a bubbly screen (Ng & Ting 1986;
Miksis & Ting 1989); indeed its derivation requires the analysis to be conducted one order
further than the dominant one considered in these references. Our result generalizes the
findings of Leroy et al. (2009) to the nonlinear regime and to the time domain. It was
shown that this contribution was able to describe the shift in the resonance frequency
to the higher frequency observed experimentally (Leroy et al. 2009, 2015; Lanoy et al.
2018); we have stressed that this tendency can be significantly accentuated in the weakly
nonlinear regime and inverted with the occurrence of harmonics.
Some extensions of the present work are easy, some others are more tricky. We have
neglected the effects of surface tension and viscosity; as in Schnitzer et al. (2019), this
is partially justified by the large, dominant, radiative damping of the array. However,
there are no particular difficulties in including them, as has been done in Caflisch et al.
(1985a,b), and there is little doubt that this will affect the Rayleigh–Plesset equation only.
This may become necessary, for instance to explain the perfect absorption by critical
coupling (a balance between viscous loss and radiative damping) studied by Lanoy et al.
(2018). It is also straightforward to account for different arrangements within the array
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-17

which would affect the mesoscopic solution by modifying the cross-section of the unit
cell for the Green’s function in (2.9). Next, accounting for different shape of the bubbles
would affect the microscopic solution, namely the response of a single bubble e.g. a
disk with monopolar resonance (Calvo et al. 2015). Following Ammari et al. (2017a,
2018, 2019), it should be possible to extend our present work to any geometry. The case
of arrays of cylindrical bubbles involved in anechoic tiles, see e.g. Hladky-Hennion &
Decarpigny (1991) and Sharma et al. (2017), with a long or infinite length in one direction
is slightly more involved but of interest; it requires us to use a different scaling adapted
to the two-dimensional problem, as done in Schnitzer et al. (2019). For this application,
accounting in addition for a viscoelastic matrix rather than for a liquid may be quite
involved depending on the considered viscoelastic model. Next, a different extension
concerns the case of bubbly liquids. Pursuing the analysis of Caflisch et al. (1985a,b), and
using the mathematical framework of Ammari & Zhang (2017), Ammari et al. (2017b),
would provide the bubble–bubble interaction in these media by use of a Green’s function
that is periodic in the three directions; this would allow us to test the validity of (1.1)
and in particular to inspect under which circumstances the dynamic mass Mrad in (4.4a,b)
is higher or lower than the actual bubble mass (δ positive or negative). Eventually, we
have assumed that the transformations undergone by the gas are adiabatic and the result
holds for isothermal transformations by simply setting γ = 1. A much more challenging
extension is to account for the thermal exchanges between the gas and the liquid since the
whole asymptotic analysis has to be reconsidered including the complete thermodynamics.

Acknowledgements
K.P., J.-F.M. and A.M. thank V. Leroy for fruitful discussions. K.P. thanks the support of
the Agence de l’Innovation de Défense (AID) from the Direction Générale de l’Armement
(DGA), under grant 2019 65 0070.

Declaration of interests
The authors report no conflict of interest.

Appendix A. Asymptotic analysis

In this appendix, we shall use non-dimensional variables t → ωt and x → (ω/c )x
(with x = (x, r)), where ω is the typical frequency imposed by the source. Denoting by
Ua the amplitude of the acoustic velocity, we have, from (2.1a–c) and (2.2), u ∼ Ua ,
p ∼ ρ c Ua , (ρ − ρ ) ∼ (ρ /c )Ua . With Ma = Ua /c the Mach number, we shall use
dimensionless quantities

1 1  1 R
p→ p, ρ→ ρ − ρ,g , u→ u, R→ . (A 1a–d)
ρ c2 Ma ρ Ma c Ma Req

Note that the bubble radius is not assumed to scale with the Mach number. This choice
prevents a simple linear response where the solution would be proportional to Ma. In
the low frequency regime, the maximum wavelength imposed by the source is much
larger than h and Req ; this is accounted for with ε = (ω/c )h  1. Next, we impose the
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-18 K. Pham and others

Microscopic scale Mesoscopic scale Macroscopic scale

1
rm r
1 xm x
xμ 1
Y

1 x
xμ = xm = x x
reqR E 2 E

F IGURE 13. The separation of the scales – at the microscopic scale, the problem is that of a
single bubble with a radius of unity in a liquid with x μ ∈ R3 . At the mesoscopic scale, the
bubble is reduced to a singularity within a periodic cell Y with x m = xm ex + r m , xm ∈ R and
r m ∈ Y , Y = (− 12 , 12 )2 . At the final, macroscopic scale the whole bubbly screen is reduced to an
effective screen at x = 0 across which jump conditions apply (with x = xex + r, x ∈ R2 ).

following scalings:
ω ρg ρg c2g
Req = r eq ε 2 , = αε 4 , = βε 4 , Ma = mε 4 . (A 2a–d)
c ρ ρ c2
Doing so, we assume that the bubbly screen is dilute since Req /h = O(ε). Such separation
of scales was already used in Caflisch et al. (1985b) for bubbly liquid, and in this case,
the separation was even more demanding with Req /h = O(ε 2 ). The Minnaert resonance
is dictated by the contrast in the mass densities only, hence, we set ρg /ρ = O(ε 4 )
and cg /c = O(1). The scalings are such that the resonance takes place in the low
frequency regime, specifically ωM h/c = O(ε). Eventually, the scaling for the Mach
number produces a linear propagation of waves at large scale in the liquid while keeping
nonlinear responses of the bubbles, see Lombard, Barrière & Leroy (2015). The asymptotic
analysis requires the use of the rescaled dimensionless coordinates x at the macroscopic
scale of the wavelength, x m at the mesoscopic scale of the array spacing and x μ at the
microscopic scale of the bubble radius (figure 13), with
⎫
x, x = (x, r) ∈ R3 , ⎪
⎪
 2 ⎪ ⎪
⎪
⎪
x 1 1 ⎬
x m = , x m = (xm , r m ) ∈ Y , Y = {xm ∈ R, r m ∈ Y } , Y = − , ,
ε 2 2 ⎪
(A 3)
⎪
⎪
1 x ⎪
⎪
xμ = , x μ ∈ R3
. ⎪
⎭
r eq R(r, t) ε 2

A.1. Non-dimensional form of the problem

The problem (2.1a–c) and (2.2) now reads as
⎧ ∂ρ γm  m γ
⎪
⎪ + ε 4
(α div u + m div(ρu)) = 0, 1 + p = 1 + ρ ,
⎨ ∂t β α
in the gaz:  
⎪
⎪ ∂u
⎩ ε 4 (α + mρ) + ε m(u · ∇)u = −∇p, ∇ ∧ u = 0,
4
∂t
(A 4)
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-19

⎧
⎪ ∂ρ
⎪
⎨ ∂t + div u + mε div(ρu) = 0, p = ρ,
4

in the liquid:   (A 5)
⎪
⎪  ∂u
⎩ 1 + ε 4 mρ + ε m(u · ∇)u = −∇p,
4
∇ ∧ u = 0,
∂t
with initial conditions (i.c.) and boundary conditions at the bubble interface (b.c.)
⎫
i.c. p(x, 0) = 0, ρ(x, 0) = 0, u(x, 0) = 0, ⎬
1 r eq ⎭ (A 6)
b.c. p, u · n are continuous and u · n = 2 Ṙ.
ε m

A.2. Derivation of effective transmission conditions at the dominant order

The separation of the scales requires the asymptotic analysis to be performed at three
scales, the microscopic scale being that of a single bubble, the mesoscopic scale of the
array where the bubbles are reduced to points and eventually the macroscopic scale of the
waves, the scale at which the whole screen is reduced to an interface x = 0 (figure 2).

A.2.1. Order 0 – resolution at the microscopic scale: the scale of the bubble
The microscopic scale is the scale corresponding to a zoom on a single bubble at a given
time t. At this scale, the spherical coordinate x μ = (1/r eq R(r, t))x/ε 2 (see (A 3)) where,
following Caflisch et al. (1985b), the discrete version of the bubble radius Ri (see e.g.
(2.3)) has been replaced by a continuous counterpart R(r, t). Doing so, the bubble has a
unitary radius and the nearby bubbles have been sent to infinity. We consider the following
expansions of the solution in the gas:
⎫
1 −2 1 −1 ⎬
u= û (x μ , r, t) + û (x μ , r, t) · · · , p = p̂0
(x μ , r, t) + ε p̂1
(x μ , r, t) + · · · ,
ε2 μ ε μ μ μ
⎭
ρ = ρ̂μ0 (x μ , r, t) + ε ρ̂μ1 (x μ , r, t) + · · · , R = R0 (r, t) + εR1 (r, t) + · · · ,
(A 7)
and in the liquid (where p = ρ)

1 −2 1
u= uμ (x μ , r, t) + u−1 (x μ , r, t) + · · · ,
p = p0μ (x μ , r, t) + εp1μ (x μ , r, t) + · · · .
ε 2 ε μ
(A 8a,b)
Accordingly, denoting ∇ μ = ∂/∂ x μ and ∇ r = ∂/∂r, the differential operators read

1 ∇μ ∇r R ∂ ∂ Ṙ
∇→ + ∇r − x μ · ∇μ and → − x μ · ∇μ . (A 9)
r eq R ε 2 R ∂t ∂t R

Resolution inside the bubble – using the expansions (A 7) in (A 4) along with (A 9),
the problem at the dominant order gives uniform pressure and mass density with
∇ μ p̂0μ = 0 and

∂ ρ̂μ0 α + mρ̂μ0 m 0  m γ
+ divμ û−2
μ = 0, ∇ ∧ û−2
μ = 0, 1+γ p̂μ = 1 + ρ̂μ0 ,
∂t r eq R0 β α
(A 10a–c)
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-20 K. Pham and others

and initial and boundary conditions, from (A 6), of the form

⎫
i.c. at t = 0, p̂0μ = ρ̂μ0 = 0, û−2
μ = 0, R0 = 1,⎬
r eq (A 11)
b.c. for |x μ | = 1, p̂0μ = p0μ , û−2
μ · er = Ṙ0 , ⎭
m

where er = x μ /|x μ |. Now, we shall express p̂0μ (r, t) and ρ̂μ0 (r, t) as a function of R0 (r, t).
By integrating the balance of mass over the  gas bubble (first equation in (A 10a–c)) we get
(4π/3)(∂ ρ̂μ0 /∂t) + (α + mρ̂μ0 )(1/r eq R0 ) |x μ |=1 û−2
μ · er ds = 0. This differential equation
on ρ̂μ0 is solved using in (A 11) (i) the boundary condition on û−2 μ and (ii) the initial
conditions ρ̂μ = 0, R = 1 at t = 0. Also, making use of the equation of state in (A 10a–c),
0 0

we obtain the pressure p̂0μ . Eventually, the velocity û−2

μ is a byproduct of ρ̂μ in (A 10a–c).
0

The result is
    ⎫
α 1 β 1 ⎪
ρ̂μ0 (r, t) = −1 , p̂0μ (r, t) = − 1 ,⎪
⎬
m (R0 )3 γm (R0 )3γ
(A 12a–c)
r eq ⎪
⎪
û−2
μ (x μ , r, t) = Ṙ |x μ | er ,
0 ⎭
m

given in dimensional form in (2.7).

Resolution in the liquid – in the liquid, we repeat the exercise, using the expansions
(A 8) in (A 5) and (A 6) along with (A 9). We get at the dominant order

⎫
divμ u−2 ∇ μ ∧ u−2
μ = 0, μ = 0, ⎬
∂u−2
μ m Ṙ0
1 (A 13)
+ (u−2 −2 −2
μ · ∇ μ )uμ − 0 x μ · ∇ μ uμ = − ∇ μ p0μ , ⎭
∂t r eq R0 R r eq R0

with initial and boundary conditions

⎫
i.c. at t = 0, p0μ = 0, u−2
μ = 0, ⎬
r eq
b.c. for |x μ | = 1, u−2
μ · er = Ṙ0 and lim p0μ (x μ , r, t) = P0 (r, t).⎭
m |x μ |→+∞
(A 14)

For the time being, the continuity of the pressure at |x μ | = 1 is not needed, but we have
introduced the (unknown) pressure field P0 (r, t); its meaning will appear later on. We have
also assumed that the bubble remains spherical.
From (A 13), the flow is incompressible and irrotational. Hence it is associated
with a velocity potential φ and, the bubble remaining spherical, φ = (A/|x μ |) + B.
Next, the boundary condition in (A 14) for u−2 μ = ∇ μ φ provides A = −r eq Ṙ /m.
0
−2
Using uμ in (A 13), the pressure pμ is found to satisfy a Bernoulli equation
0

∇ μ ( p0μ + (r 2eq /m) F(R0 , Ṙ0 , R̈0 , |x μ |)) = 0, that we integrate using p0μ →|x μ |→+∞ P0 (r, t)
from (A 14). Doing so, the fields (u−2 μ , pμ ) are found to be functions of R and of their
0 0
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-21

time derivatives
 
r eq
Ṙ0 r 2eq R0 R̈0 + 2(Ṙ0 )2 (Ṙ0 )2
u−2
μ (x μ , r, t) = er , p0μ (x μ , r, t) =P + 0
− ,
m |x μ |2 m |x μ | 2|x μ |4
(A 15a,b)
given in dimensional form in (2.7). At this stage, once R0 (r, t) is known, the solution in the
bubble is known from (A 12a–c), and the solution in the liquid is known up to the pressure
field P0 (r, t) from (A 15a,b).
An equation of the Rayleigh–Plesset type – we shall establish that R0 (r, t) satisfies
a continuous version of the Rayleigh–Plesset equation. To do so, we use p0μ = p̂0μ for
|x μ | = 1, with p̂0μ in (A 12a–c) and p0μ in (A 15a,b). We get that R0 (r, t) is solution of
 
3 β 1 m
R0 R̈0 + (Ṙ0 )2 + 2 1 − 0 3γ = − 2 P0 (r, t) (A 16)
2 γ r eq (R ) r eq

and at the equilibrium P0 = 0, the bubbles and the liquid are at rest and the acoustic
pressures vanish (R0 = 1 in (A 16) hence, u−2
μ = 0 and p̂μ = pμ = 0 in (A 15a,b)).
0 0

A.2.2. Order 0 – resolution at the mesoscopic scale: the scale of the screen spacing
From (A 12a–c), (A 15a,b) and (A 16), the solution at the microscopic scale is known
once P0 (r, t) is known, and P0 will be provided by the mesoscopic problem. This scale is
that of the array spacing, intermediate between the bubble size and the wavelength. It is
associated with x m = x/ε, x m = (xm , r m ) defined in (A 3), and we use the expansions

u = u0m (x m , r, t) + εu1m (x m , r, t) + · · · ,
p = p0m (x m , r, t) + εp1m (x m , r, t) + · · · , ρ = ρm0 (x m , r, t) + ερm1 (x m , r, t) + · · · ,
(A 17)
where the (uim , pim , ρmi ), i = 0, 1, . . . are periodic with respect to r m ∈ Y . In x m -coordinate,
denoting ∇ m = ∂/∂ x m (and ∇ r = ∂/∂r as in (A 9)), the differential operators read
1 ∂ ∂
∇→ ∇m + ∇r , → . (A 18a,b)
ε ∂t ∂t
Matching conditions – the matching conditions tell us that the solution of the
microscopic problem has to coincide with that of the mesoscopic problem in some
intermediate region, when |x μ | → +∞ and x m → 0. Since |x m | = O(|x|/ε) √ and |x μ | =
O(|x|/ε 2 ), this intermediate region√ can be sought√at distance ε ε of the bubble,
in x coordinate; hence, |x μ | ∼ 1/ ε and |x m | ∼ ε. The matching conditions are
obtained by identifying the expansions (A 7) and (A 17) for x μ = x m /ε r eq R and x m → 0.
At the dominant order, anticipating that p0m does not depend on x m , we get
   
xm 1 −2 xm
pm (r, t)+ · · · ∼ pμ
0 0
, r, t + · · · , um (x m , r, t) + · · · ∼ 2 uμ
0
, r, t + · · ·
r eq εR ε r eq εR
(A 19a,b)
for x m → 0, hence
r 3eq xm
lim p0μ (x μ , r, t) = p0m (r, t) = P0 (r, t), u0m (x m , r, t) ∼ (R0 )2 Ṙ0 .
|x μ |→+∞ x m →0 m |x m |3
(A 20a,b)
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-22 K. Pham and others

We have used u−2 μ in (A 15a,b) and we have assumed that the terms uμ , when increasing
n
−2
n > −2, decrease faster than uμ ∝ 1/|x μ | , which is consistent with non-diverging mass
2

flux through a spherical cap. It follows that the main results provided by the mesoscopic
problem are: the pressure P̃0 (r, t) introduced in (A 14) is the pressure p0m at |x m | = 0 (at
the mesoscopic scale, the bubbles are reduced to points), the velocity u0m is singular at the
origin, and it encapsulates the effect of the oscillating bubble point (if Ṙ0 = 0, the bubble
is not seen and u0m is regular). In dimensional form, these results contribute to the solution
given in (2.8) at the dominant order; they will be complemented with the solution found
at the following order.
The mesoscopic problem – using (A 17) in (A 5) along with (A 18a,b), we get that
∇ m p0m = 0. At this scale again, the pressure p0m is uniform as announced in (A 19a,b)
resulting in (A 20a,b). It follows that the problem at the dominant order is set on (u0m , p1m )
with
⎫
∂u0m 1 ⎪
divm um = 0, ∇ ∧ um = 0,
0 0
= −∇ r P (r, t) − ∇ m pm ,⎪
0
⎪
⎪
∂t ⎬
i.c. pm = 0, um = 0 at t = 0,
1 0 (A 21)
⎪
⎪
⎪
b.c. lim u0 = U0± (r, t), p1 , u0 periodic on ∂ Y , ⎪
m m m
⎭
xm →±∞

and u0m is singular at x m = 0 from (A 20a,b). We have three unknown macroscopic fields:
P0 that we reach back from the microscopic scale (see (A 14)) and U0± in (A 21). Note that
we have assumed that U0± do not depend on r m ; this will be justified in the forthcoming
section.

A.2.3. Order 0 – derivation of the effective transmission conditions

We now dive into the problem at the macroscopic scale, where only the coordinate x is
needed, and we consider the following expansions:

u = u0 (x, t) + εu1 (x, t) + · · · , p = p0 (x, t) + εp1 (x, t) + · · · ,
(A 22a–c)
ρ = ρ 0 (x, t) + ερ 1 (x, t) + · · · .
At this scale, the bubbly screen is reduced to an interface but the boundary conditions
when x → 0± are still missing. These are the conditions we are looking for and they
are provided by matching the solutions of the mesoscopic and macroscopic √ problems.
The intermediate
√ region where both solutions are valid corresponds to x ∼ ± ε hence
xm ∼ ±1/ ε. Hence, at the dominant order, the matchings read
p0 (0± , r, t) = p0m (r, t) = P0 (r, t), u0 (0± , r, t) = lim u0m (x m , r, t) = U 0± (r, t),
xm →±∞
(A 23a,b)
where we used (A 20a,b) and (A 21). In (A 23a,b), U 0± appear as the limits of the acoustic
velocity on both sides of the array; this justifies a posteriori that U 0± do not depend on r m .
Eventually, from (A 5) written at the dominant order, ( p0 , u0 ) satisfy the linearized Euler
equations in an irrotational flow
∂p0 ∂u0
+ div x u0 = 0, = −∇ x p0 , ∇ ∧ u0 = 0, (A 24a–c)
∂t ∂t
with the boundary conditions (A 23a,b) and the initial conditions p0 (x, 0) = 0,
u0 (x, 0) = 0. From (A 23a,b), the pressure p0 is continuous at x = 0. This is not
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-23

the case for the normal velocity, and to show this result, we have to come back to
the problem satisfied by u0m . The equation of incompressibility, divm u0m = 0 in (A 21),
is integrated in the domain Y minus a ball Br of radius r (in order to avoid the
singularity of u0m ). Using the boundary conditions in (A 21) and the singular behaviour
of u0m in (A 20a,b), we get 0 = Y /Br divm u0m dx m from which u0x (0+ , r, t) − u0x (0− , r, t) =

(r 3 /m)((R0 )2 Ṙ0 /|x m |2 ) dx m . The integral over ∂Br gives a finite but non-zero
∂Br eq
contribution (independent of the radius of Br ), and eventually we obtain a non-trivial
transmission condition

4πr 3eq
[[p0 ]] = 0, [[u0x ]] = (R0 )2 Ṙ0 . (A 25a,b)
m

Conclusion of the dominant order – at this stage the dominant order provides a model in
which the bubbly screen is reduced to an effective screen across which the normal velocity
experiences a jump dictated by R0 , and R0 satisfies (A 16) with P0 (r, t) = p0 (0, r, t). The
result is identical to that obtained in Miksis & Ting (1989). We shall see that conducting
the analysis at the following order makes the contribution of the bubble–bubble interaction
appear.

A.3. A result at the following order: the bubble–bubble interaction

The analysis at order 1 is conducted following exactly the same steps as at order 0. For
most of the results, we simply recover the same prediction in terms of Taylor expansions
of the fields with respect to the small bubble radius, see e.g. (A 12a–c) and forthcoming
(A 28a–c); this will be specified whenever it happens. The single but significant difference
concerns the pressure seen by a single bubble at infinity. It is P0 in (A 15a,b) and we have
established that P0 (r, t) = p0 (0, r, t) the macroscopic pressure. It will be P̃1 in (A 30) and
we shall establish that P̃1 (r, t) = p1 (0, r, t)+ a ‘δ-contribution’ due to the presence of the
other bubbles, with δ defined by (A 34) and (A 36). This result is given in (A 49); it results
in the modified Rayleigh–Plesset equation (A 52) whose dimensional form is (2.5).
Incidentally, solving at order 1 allows us to obtain the explicit form of the solution at
the mesoscopic scale (A 33), which coincides with (2.8) in dimensional form.

A.3.1. Order 1 – resolution at the microscopic scale

Resolution inside the bubble – at the next order, we get from (A 4) that the pressure and
mass density are still uniform, with ∇ μ p̂1μ = 0 and

    ⎫
∂ ρ̂μ1 α + mρ̂μ0 mρ̂μ1 α + mρ̂μ0 R1 ⎪
+ divμ û−1
μ + − divμ û−2
μ = 0,⎪
⎪
⎬
∂t r eq R0 r eq R0 r eq (R0 )2
(A 26)
β ⎪
⎪
⎪
⎭
∇∧ û−1
μ = 0, p̂1μ = (R0 )3(1−γ ) ρ̂μ1 ,
α

where (α + mρ̂μ0 ) = α/(R0 )3 and divμ û−2 μ = 3r eq Ṙ /m from (A 12a–c). The system in
0

(û−1
μ , p̂μ , ρ̂μ ) is complemented with initial and boundary conditions coming from (A 6),
1 1
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-24 K. Pham and others

of the form ⎫
i.c. at t = 0, p̂1μ = ρ̂μ1 = 0, û−1
μ = 0, R0 = 0,⎬
r eq (A 27)
b.c. for |x μ | = 1, p̂1μ = p1μ , û−1
μ · er = Ṙ1 . ⎭
m
By integrating the balance of mass in (A 26) along with (A 27), as we have done at order
0 ((A 10a–c) and (A 11) providing (A 12a–c)) we get an ordinary differential equation on
ρ̂μ1 which can be solved explicitly and leads to

3α R1 3β R1 r eq
ρ̂μ1 (r, t) = − , p̂1μ (r, t) = − , û−1
μ (x μ , r, t) = Ṙ1 |x μ | er .
m (R0 )4 m (R0 )3γ +1 m
(A 28a–c)
We remark that (A 12a–c) along with (A 28a–c) correspond to the Taylor expansions
up to O(ε 2 ) of ρ̂μ = α/m(1/R3 − 1), p̂μ = β/γ m(1/R3γ − 1) and ûμ = (r eq /m)Ṙ|x μ | er .
Hence, order 1 does not modify the form of the solution obtained at order 0, and given in
dimensional form in (2.7).
Resolution in the liquid – next, in the liquid, we have from (A 5)
⎫
divμ u−1
μ = 0, ∇ μ ∧ u−1μ = 0, ⎪
⎪
⎪
  ⎪
⎪
∂uμ −1
m −2 −1 −1 −2 R 1
−2 −2
⎪
⎬
+ (uμ · ∇ μ )u μ + (u μ · ∇ μ )u μ − (u μ · ∇ μ )u μ . . .
∂t r eq R0 R 0
  ⎪⎪
⎪
⎪
Ṙ0
−1 R 1 0
Ṙ − R 0 1
Ṙ −2 1 R 1
⎪
⎪
. . . − 0 x μ · ∇ μ uμ + x μ · ∇ u
μ μ = − ∇ p
μ μ
0
− ∇ p
μ μ
1
, ⎭
R (R )
0 2 r eq R 0 R 0

(A 29)

where (u−2 −1
μ , pμ ) are known from (A 15a,b), hence the above system is set on (uμ , pμ ). The
0 1

first two equations in (A 29) together with the boundary condition (A 27) provide the same
form of the velocity as that obtained at order 0. Using further u−1 μ in the third equation
of (A 29), we obtain a differential equation in p1μ which can be integrated in space. This
provides
 
−1 r eq Ṙ1 r 2eq R1 R̈0 + R0 R̈1 + 4Ṙ0 Ṙ1 Ṙ0 Ṙ1
uμ (x μ , r, t) = er , pμ (x μ , r, t) = P̃ +
1 1
− .
m |x μ |2 m |x μ | |x μ |4
(A 30a,b)
Again, we recognize the second contribution of the Taylor expansions of the pressure
and of the velocity in (A 15a,b). However, we shall see that P̃1 (r, t) in (A 30),
being the equivalent of P0 (r, t) = p0 (0, r, t) in (A 15a,b), differs from p1 (0, r, t);
specifically, it contains a contribution due to the arrangement of the bubbles within the
array. Once P̃1 is known, the Rayleigh–Plesset equation will be obtained by equating the
pressure p1μ in (A 30) and p̂1μ in (A 28a–c) at the interface |x μ | = 1; this requires some
work and it will be done in the forthcoming (A 50).

A.3.2. Order 1 – resolution at the mesoscopic scale

The mesoscopic problem associated with ( p1m , u0m ) is given by (A 21) but now,
we have established that P0 (r, t) = p0 (0, r, t) and U 0± (r, t) = u0 (0± , r, t), (A 23a,b).
It is convenient to denote u0 = u0x ex + u0r (with u0r · ex = 0). From (A 24a–c) and
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-25

(A 25a,b), u0x is discontinuous at x = 0 while u0r is continuous with ∂t u0r = −∇ r p0
(and p0 is continuous). Besides, from (A 20a,b) and (A 25a,b), we know that u0m ∼ x m →0
1/4π[[u0x ]] x m /|x m |3 . Hence denoting

u0 (0± , r, t) = u0x (0± , r, t)ex + u0r (0, r, t) and u0x (0± , r, t) = u¯0x (0, r, t) ± 12 [[u0x ]],
(A 31a,b)
the problem (A 21) can be written
⎫
∂u0m ∂u0r 1 ⎪
divm u0m = 0, ∇ ∧ u0m
= 0, = (0, r, t) − ∇ m pm ,⎪
⎪
⎪
∂t ∂t ⎪
⎪
  ⎪
⎪
⎬
1
b.c. lim um = u¯0x (0, r, t) ± [[ux ]] ex + ur (0, r, t),
0 0 0 (A 32)
xm →±∞ 2 ⎪
⎪
⎪
⎪
⎪
⎪
0
[[ux ]] x m ⎪
⎪
( pm , um ) periodic on ∂ Y , um ∼
1 0 0
. ⎭
x m →0 4π |x m |3

The above problem set in (∂t (u0m − u0r (0, r, t)), p1m ) is linear with respect to the
macroscopic fields ∂t u¯0x (0, r, t) and ∂t [[u0x ]]. It follows that ( p1m , u0m ) read
⎫
∂ u¯0 1 ∂[[u0x ]] ⎪
⎪
p1m (x m , r, t)= − x (0, r, t)xm − G(x m ) + P1 (r, t), ⎬
∂t 4π ∂t (A 33)
[[u0 ]] ⎪
⎪
u0m (x m , r, t) = u0r (0, r, t) + u¯0x (0, r, t)ex + x ∇ m G(x m ), ⎭
4π

where P1 (r, t) is an unknown (at this stage) constant at the mesoscopic scale and where
the field G is the Green’s function of the periodic cell Y satisfying G, ∇ m G periodic in
∂ Y and
1
m G = 0, with lim (G − 2π|xm |) = 0, G ∼ − . (A 34)
xm →±∞ x m →0 |x m |

It is worth noting that (A 33) leads to

lim u0m (x m , r, t) = u0 (0± , r, t), (A 35)

xm →±∞

as imposed by the macro-to-meso matching conditions (A 23a,b). Eventually, we shall see

that an important effective parameter resulting from the Green’s function problem is
 
1
δ = lim G(x m ) + . (A 36)
x m →0 |x m |

A.3.3. Order 1 – derivation of the effective transmission conditions

Jump condition on the pressure – equipped with the expression (A 33) of p1m , the jump
conditions at order 1 in the pressure can now be written. The macro-to-meso matching
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-26 K. Pham and others

condition at order 1 in the pressure is given by

p0 (x, t) + εp1 (x, t) ∼ p0m (x m , r, t) + εp1m (x m , r, t), x = εx m . (A 37a,b)

x→0± ,xm →±∞

Making the Taylor expansion of the left-hand side in (A 37) and passing to the limit when
xm → ±∞ gives the order-0 matching condition (A 23a,b) and at order 1, we get
 
1 ± ∂p0 ±
p (0 , r, t) = lim pm (x m , r, t) − xm
1
(0 , r, t) . (A 38)
xm →±∞ ∂x
At the macroscopic scale, the Euler equation (A 24a–c) imposes

∂p0 ± ∂ u¯0 1 ∂[[u0x ]]

(0 , r, t) = − x (0, r, t) ∓ . (A 39)
∂x ∂t 2 ∂t
By using the above relation in (A 38) along with (A 33) and (A 34), we get the continuity
of the pressure at the next order

p1 (0± , r, t) = P1 (r, t), ⇒ [[p1 ]] = 0. (A 40)

We also obtain P1 , which contributes to p1m in the liquid (A 33), and we shall see now that
P1 differs from P̃1 , which contributes to p1μ in the bubble (A 30), as previously said, this is
the important result at this order. To do so, we have to inspect the meso-to-micro matching
conditions.
Jump condition on the velocity – we use the meso-to-micro matching condition on the
velocity which formally reads
1 −2 1
u0m (x m , r, t) + εu1m (x m , r, t) + · · · ∼ uμ (x μ , r, t) + u−1 (x μ , r, t) + · · · ,
x m →0,|x μ |→∞ ε 2 ε μ
(A 41)

where up to the second order in ε, x m ∼ ε r eq (R0 + εR1 )x μ , see (A 3). Using u−2μ in
(A 15a,b) and u−1μ in (A 30), we recover the singular behaviour of u 0
m in (A 20a,b) but
also that of um which reads
1

r 3eq  xm
u1m (x m , r, t) ∼ 2R0 R1 Ṙ0 + (R0 )2 Ṙ1 . (A 42)
x m →0 m |x m |3
We are now able to derive the jump conditions on the velocity at order 1. As we have done
at order 0 to get (A 25a,b), we start with the balance of mass at order 0 at the mesoscopic
scale. From (A 4) and (A 18a,b), it reads
∂p0m
+ divm u1m + divr u0m = 0, (A 43)
∂t
that we shall integrate on Yr = Y ∗ /Br where Y ∗ = (−xm∗ , xm∗ ) × (− 21 , 12 )2 and Br a
ball of radius r then pass to the limit r → 0 and xm∗ → +∞. Firstly, we know that
p0m = p0 (0, r, t) from (A 23a,b) while u0m is given by (A 33) hence divr u0m =
divr u0r (0, r, t) + (1/4π)∇ r [[u0x ]] · ∇ m G(x m ). From (A 34), G is an even function of x m ·
e y and of x m · ez hence Yr ∇ r [[u0x ]] · ∇ m G(x m ) dx m = 0. Using (A 24a–c), it results that
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-27


Yr
(∂t p0m + divr u0m ) dx m = −V (Yr )∂u0x /∂ x(0, r, t) with V (Yr ) = (2xm∗ − (4π/3)r3 ). The
result of the integration of (A 43) reads
  
∗ ∂ux
0
um · ex |xm∗ − xm
1
(0, r, t) dr m
∂x
   
∗ ∂ux
0
− um · ex |−xm∗ + xm
1
(0, r, t) dr m = · · · u1m (x m , r, t) · er dx m . (A 44)
∂x ∂Br

It is now sufficient to use the matching condition written as

 
∗ ∂um
0
1 ± ±
u (0 , r, t) = ∗lim um|±xm∗ ∓ xm
1
(0 , r, t) , (A 45)
xm →+∞ ∂x

and the singular behaviour (A 42) of u1m yielding ∂Br u1m · er = 4πr 3eq /m(2R0 R1 Ṙ0 +
(R0 )2 Ṙ1 ) to get
4πr 3eq 
[[u1x ]] = 2R0 R1 Ṙ0 + (R0 )2 Ṙ1 . (A 46)
m
Again, we recognize a Taylor expansion with respect to the bubble radius of the jump on
the normal velocity at the dominant order, see (A 25a,b).

A.3.4. The Rayleigh–Plesset equation at order 1

To derive the Rayleigh–Plesset equation at order 1, we use the meso-to-micro matching
conditions on the pressure. It formally reads
p0m (r, t) + εp1m (x m , r, t) + · · · ∼ p0μ (x μ , r, t) + εp1μ (x μ , r, t) + · · · , (A 47)
x m →0,|x μ |→+∞

where, up to the first order in ε, x m ∼ ε r eq R0 x μ , see (A 3). All the fields are known
from (A 15a,b), (A 20a,b), (A 23a,b), (A 30), (A 33) and (A 40), but in (A 30) P̃1 is still
unknown. Owing to x m ∼ ε r eq R0 x μ , it results that
 
r 2eq R0 R̈0 + 2(Ṙ0 )2 ∂t [[u0x ]] ∂t [[u0x ]]
+ ε P̃1 + · · · ∼ + ε p 1
|x=0 − δ + ··· .
m |x μ | |x μ |→+∞ 4πr eq R0 |x μ | 4π
(A 48)
With [[u0x ]] = 4πr 3eq /m(R0 )2 Ṙ0 from (A 25a,b), it is easy to see that the above matching is
consistent at the order 0 and at the order 1 it provides P̃1 of the form
δ r 3eq ∂ 
P̃1 = p1|x=0 − (R0 )2 Ṙ0 . (A 49)
m ∂t
Once P̃1 is known, we use the equality of the pressures in the gas (A 28a–c) and in the
liquid (A 30) at the microscale at |x μ | = 1. We deduce the Rayleigh–Plesset equation at
the order 1
β R1 ∂  m
R1 R̈0 + R0 R̈1 + 3Ṙ0 Ṙ1 + 3 − δ r eq (R0 )2 Ṙ0 = − 2 p1|x=0 . (A 50)
r eq (R )
2 0 1+3γ ∂t r eq

The new contribution of order 1 is visible in the above relation which appears to differ
from the simple Taylor expansion of (A 16) by the contribution of the term proportional
to δ.
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-28 K. Pham and others

A.4. A unique effective problem
The final jump conditions are obtained on a unique problem with macroscopic unknowns
(R, p, ux ) which are the Taylor expansion up to order 1 of the bubble radius, the pressure
and the normal velocity

R = R0 + εR1 , p = p0 + εp1 , ux = u0x + εu1x . (A 51a–c)

Using the Rayleigh–Plesset equations (A 16) and (A 50); the continuity of the macroscopic
pressures p0 and p1 given by (A 25a,b) and (A 40); the jump conditions on the normal
velocity (A 25a,b) and (A 46), we deduce that up to the second order in ε
⎫
4πr 3eq ⎪
⎪
[[p]] = 0, [[ux ]] = (R) Ṙ, 2 ⎪
⎬
m
  (A 52)
3 β 1 ∂  m ⎪
RR̈ + (Ṙ)2 + 2 1 − − εδ r eq R2 Ṙ = − 2 p(0, r, t),⎪
⎪
⎭
2 γ r eq (R) 3γ ∂t r eq

which provides (2.5) in dimensional form.

Appendix B. Reconstruction of the solutions at the three scales

In this section, with two-dimensional settings and a one-dimensional propagation
problem (along x), we denote by y = r · e y the spatial coordinate along the array, and
yn = nh the centres of the bubbles.
(i) The microscopic scale – from (3.3), we have

Δp exp(ik y nh − iωt)
R( yn , t) = Req −    , (B 1)
Req
ρ Req ωM − ω 1 − δ
2 2 + iKReq
h

with K = 2π/kh2 , and Ṙ = −iω(R − Req ). This is sufficient to get u in (2.7).

(ii) The mesoscopic scale – with u∞ = 12 (u|x=0+ + u|x=0− ), see (2.11), and making use
of (3.4) and (3.5a,b) in (2.8), we have

R2eq Ṙ kx Δp
u(x, t) = u∞ ( yn , t) − iω ∇G(x − r n ), u∞ ( yn , t) =
exp(ik y nh − iωt).
h ρ ω
(B 2a,b)
(iii) The macroscopic scale – solving the linearized version of (2.5) for an incident wave

pinc (x, t) = Δp exp(ikx − iωt), (B 3)

gives the transmitted and reflected velocity fields

kx Δp ikx x  kx Δp
ux (x, t) = e − R e−ikx x eiky y−iωt , ux (x, t) = T eikx x eiky y−iωt , (B 4a,b)
ρ ω ρ ω

for x < 0 and x > 0 and with (R, T ) in (3.5a,b), see figures 6 and 7 for |x|/λ > h/λ.
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

Resonant bubbly metascreen 906 A19-29

REFERENCES
A MMARI , H., F ITZPATRICK , B., G ONTIER , D., L EE , H. & Z HANG , H. 2017a A mathematical and
numerical framework for bubble meta-screens. SIAM J. Appl. Maths 77 (5), 1827–1850.
A MMARI , H., F ITZPATRICK , B., G ONTIER , D., L EE , H. & Z HANG , H. 2017b Sub-wavelength focusing
of acoustic waves in bubbly media. Proc. R. Soc. Lond. A 473 (2208), 20170469.
A MMARI , H., F ITZPATRICK , B., G ONTIER , D., L EE , H. & Z HANG , H. 2018 Minnaert resonances for
acoustic waves in bubbly media. In Annales de l’Institut Henri Poincaré C, Analyse non linéaire,
vol. 35, pp. 1975–1998. Elsevier.
A MMARI , H., F ITZPATRICK , B., H ILTUNEN , E. O., L EE , H. & Y U, S. 2019 Subwavelength resonances
of encapsulated bubbles. J. Differ. Equ. 267 (8), 4719–4744.
A MMARI , H. & Z HANG , H. 2017 Effective medium theory for acoustic waves in bubbly fluids near
Minnaert resonant frequency. SIAM J. Math. Anal. 49 (4), 3252–3276.
B ERGAMASCO , L. & F USTER , D. 2017 Oscillation regimes of gas/vapor bubbles. Intl J. Heat Mass
Transfer 112, 72–80.
B REMOND , N., A RORA , M., O HL , C.-D. & L OHSE , D. 2006 Controlled multibubble surface cavitation.
Phys. Rev. Lett. 96 (22), 224501.
B RETAGNE , A., T OURIN , A. & L EROY, V. 2011 Enhanced and reduced transmission of acoustic waves
with bubble meta-screens. App. Phys. Lett. 99 (22), 221906.
CAFLISCH , R. E., M IKSIS , M. J., PAPANICOLAOU, G. C. & T ING , L. 1985a Effective equations for wave
propagation in bubbly liquids. J. Fluid Mech. 153, 259–273.
CAFLISCH , R. E., M IKSIS , M. J., PAPANICOLAOU, G. C. & T ING , L. 1985b Wave propagation in bubbly
liquids at finite volume fraction. J. Fluid Mech. 160, 1–14.
C ALVO , D. C., T HANGAWNG , A. L., L AYMAN J R ., C. N., CASALINI , R. & O THMAN , S. F. 2015
Underwater sound transmission through arrays of disk cavities in a soft elastic medium. J. Acoust.
Soc. Am. 138 (4), 2537–2547.
D EVAUD , M., H OCQUET, T., BACRI , J.-C. & L EROY, V. 2008 The Minnaert bubble: an acoustic
approach. Eur. J. Phys. 29 (6), 1263.
D OC , J.-B., C ONOIR , J.-M., M ARCHIANO , R. & F USTER , D. 2016 Nonlinear acoustic propagation in
bubbly liquids: multiple scattering, softening and hardening phenomena. J. Acoust. Soc. Am. 139
(4), 1703–1712.
D OINIKOV, A. A. 2001 Translational motion of two interacting bubbles in a strong acoustic field. Phys.
Rev. E 64 (2), 026301.
D OMENICO , S. N. 1982 Acoustic wave propagation in air-bubble curtains in water—Part I: history and
theory. Geophysics 47 (3), 345–353.
F USTER , D., D OPAZO , C. & H AUKE , G. 2011 Liquid compressibility effects during the collapse of a
single cavitating bubble. J. Acoust. Soc. Am. 129 (1), 122–131.
GAUNAURD , G. 1977 One-dimensional model for acoustic absorption in a viscoelastic medium containing
short cylindrical cavities. J. Acoust. Soc. Am. 62 (2), 298–307.
G UÉDRA , M., C ORNU, C. & I NSERRA , C. 2017 A derivation of the stable cavitation threshold accounting
for bubble-bubble interactions. Ultrason. Sonochem. 38, 168–173.
H ARKIN , A., K APER , T. J. & NADIM , A. 2001 Coupled pulsation and translation of two gas bubbles in a
liquid. J. Fluid Mech. 445, 377–411.
H LADKY-H ENNION , A.-C. & D ECARPIGNY, J.-N. 1991 Analysis of the scattering of a plane acoustic
wave by a doubly periodic structure using the finite element method: Application to Alberich
anechoic coatings. J. Acoust. Soc. Am. 90 (6), 3356–3367.
K ELLER , J. B. & M IKSIS , M. 1980 Bubble oscillations of large amplitude. J. Acoust. Soc. Am. 68 (2),
628–633.
L ANOY, M., G UILLERMIC , R.-M., S TRYBULEVYCH , A. & PAGE , J. H. 2018 Broadband coherent perfect
absorption of acoustic waves with bubble metascreens. Appl. Phys. Lett. 113 (17), 171907.
L EIGHTON , T. G. 2011 The inertial terms in equations of motion for bubbles in tubular vessels or between
plates. J. Acoust. Soc. Am. 130 (5), 3333–3338.
L EIGHTON , T. G. 2004 Nonlinear bubble dynamics and the effects on propagation through near-surface
bubble layers. In AIP Conference Proceedings, vol. 728, pp. 180–193. AIP.
Downloaded from https://www.cambridge.org/core. IP address: 90.44.66.135, on 16 Nov 2020 at 16:51:13, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2020.799

906 A19-30 K. Pham and others

L EROY, V., S TRYBULEVYCH , A., L ANOY, M., L EMOULT, F., T OURIN , A. & PAGE , J. H. 2015
Superabsorption of acoustic waves with bubble metascreens. Phys. Rev. B 91 (2), 020301.
L EROY, V., S TRYBULEVYCH , A., S CANLON , M. G. & PAGE , J. H. 2009 Transmission of ultrasound
through a single layer of bubbles. Eur. Phys. J. E 29 (1), 123–130.
L OMBARD , O., BARRIÈRE , C. & L EROY, V. 2015 Nonlinear multiple scattering of acoustic waves by a
layer of bubbles. Europhys. Lett. 112 (2), 24002.
M IKSIS , M. J. & T ING , L. 1989 Effects of bubbly layers on wave propagation. J. Acoust. Soc. Am. 86 (6),
2349–2358.
N G , K. C. & T ING , L. 1986 Wave propagation through a thin bubbly layer. J. Acoust. Soc. Am. 79 (4),
924–926.
P LESSET, M. S. & P ROSPERETTI , A. 1977 Bubble dynamics and cavitation. Annu. Rev. Fluid Mech. 9 (1),
145–185.
R AYLEIGH , L ORD 1917 VIII. On the pressure developed in a liquid during the collapse of a spherical
cavity. Lond. Edinb. Dublin Phil. Mag. J. Sci. 34 (200), 94–98.
S CHNITZER , O., B RANDÃO , R. & YARIV, E. 2019 Acoustics of bubbles trapped in microgrooves: from
isolated subwavelength resonators to superhydrophobic metasurfaces. Phys. Rev. B 99 (19), 195155.
S HARMA , G. S., S KVORTSOV, A., M AC G ILLIVRAY, I. & K ESSISSOGLOU, N. 2017 Acoustic
performance of gratings of cylindrical voids in a soft elastic medium with a steel backing. J. Acoust.
Soc. Am. 141 (6), 4694–4704.
VAN W IJNGAARDEN , L. 1968 On the equations of motion for mixtures of liquid and gas bubbles. J. Fluid
Mech. 33 (3), 465–474.
WANG , Q. 2016 Local energy of a bubble system and its loss due to acoustic radiation. J. Fluid Mech. 797,
201–230.
W ESTON , D. E. 1966 Acoustic interaction effects in arrays of small spheres. J. Acoust. Soc. Am. 39 (2),
316–322.

View publication stats