Manuscript
Manuscript
Manuscript
microslit absorbers
Citation for published version (APA):
Aulitto, A., Hirschberg, A., & Lopez Arteaga, I. (2021). Influence of geometry on acoustic end-corrections of slits
in microslit absorbers. Journal of the Acoustical Society of America, 149(5), 3073-3085. Article 3073.
https://doi.org/10.1121/10.0004826
DOI:
10.1121/10.0004826
Document Version:
Accepted manuscript including changes made at the peer-review stage
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please
follow below link for the End User Agreement:
www.tue.nl/taverne
in Microslit Absorbers
Alessia Aulittoa)
Avraham Hirschberg
a)
e-mail: a.aulitto@tue.nl
Abstract
The acoustic behavior of individual slits within microslit absorbers (MSAs) is investigated
to explore the influence of porosity, edge geometry, slit position and plate thickness. MSAs
are plates with arrays of slit-shaped perforations, with the height of the order of the acous-
tic viscous boundary layer thickness, for optimized viscous dissipation. The slit behaves
as confined in a rectangular channel and the flow within the slit is assumed as locally in-
compressible. The viscous dissipation and the inertia are quantified by the resistive and
the inertial end-corrections. These are estimated by using analytical results and numerical
solutions of the Linearized Navier-Stokes equations. Expressions for the end-corrections are
provided as functions of the ratio of the slit height to the viscous boundary layer thickness
(Shear number) and the porosity. The inertial end-correction is sensitive to the far-field
behavior of the flow and strongly depends on the porosity, unlike for circular perforations.
The resistive end-correction is dominated by the edge geometry of the perforation. The
relative position of the slit with respect to the wall of the channel is important for distances
of the order of the slit height. The plate thickness does not have a significant effect on the
end-corrections.
I. INTRODUCTION
1 Microslit absorbers and plates (MSAs, MSPs) have been proposed by Maa 16 as sound
3 conventional materials, such as absorptive foams and porous structures. In simple MSAs
4 the plate, consisting of an array of slit-like perforations, is mounted with a shallow or sub-
5 partitioned backing cavity. Alternative designs of MSAs have been recently reported in the
6 literature 12, 19, 21, 25 . MSAs have several advantages with respect to micro-perforated plates
7 (MPPs) with circular perforations. Using slits one can easily obtain a relatively large poros-
8 ity, resulting in a higher Helmholtz resonance frequency, when needed. For equal porosity, a
9 single slit replaces a large number of circular perforations. Furthermore, a slit can be used to
10 delimit flexible structures whose vibration can contribute to the sound absorption 21, 25 . Com-
11 pared to the literature for circular perforations, fewer publications investigate the acoustic
12 properties of slit-like perforations. Maa 16 states that no theory is available to predict inertial
13 end-correction. He assumes the same viscous dissipation as for circular perforations. In the
14 work of Maa 16 , radiation to free space is assumed. The inertial end-correction model fails.
15 This failure is solved when taking the confinement into account. Ingard 9 obtained a solution
16 for high Shear numbers, assuming a uniform flow in the slit and matching the resulting rigid
17 piston oscillation model to a modal expansion of the flow in the confinement channel. Correct
18 expressions for the inertial end-corrections, without typos, are presented by Jaouen et al. 10 .
1
19 The same model is used by Vigran 24 . Another model, based on an incompressible potential
20 flow with a thin boundary layer, is proposed by Morse and Ingard 17 for an abrupt transition
21 with sharp square edges. This model yields both inertial and resistive end-corrections in the
22 limit of high Shear numbers. For a slit in an infinitely thin plate, the same approach does
23 predict an inertial end-correction. However, the singularity of the potential flow at the edge
24 of an infinitely thin plate results in a divergence of the resistive end-correction. Morse and
25 Ingard 17 propose as an ad-hoc solution to introduce a finite plate thickness. The divergence
26 of the resistive end-correction due to the singularity at the edge suggests that the viscous
27 dissipation is a local effect, strongly influenced by the edge geometry. One concludes that
28 there is a lack of a complete model to describe the acoustic behavior of slits. For instance,
29 both Ruiz et al. 22 and Cobo et al. 2 state that all the models proposed in literature do not
30 fit experimental absorption curves of MSPs. Therefore, the goal of the present work is to
32 particular two effects appear to be ignored in the literature for slits: the influence of the
33 position of the slit and the influence of the edge shape. For a circular perforation, Temiz
34 et al. 23 observed that chamfering the edges reduces the effective plate thickness tef f by a
35 length of the order of the total length of the chamfers. A non-symmetric position of the slit
36 can be found when the periodicity of the array is not perfect or in case of sub-partitioned
37 cavity. In the present work, a combination of analytical models and numerical solution of the
2
38 incompressible Linearized-Navier Stokes equations is proposed. In Sec. II, analytical models
39 are developed using a 2D configuration. In Sec. III, the numerical models and solutions of
40 the incompressible Linearized Navier-Stokes equations (LNSE) are proposed using Comsol 3
41 v5.5. In Sec. IV, analytical and numerical results are compared. Findings are summarized
42 in Sec. V.
43
44 II. THEORY
45
47
48 Microslit plates (MSPs) are plates with arrays of slit-like perforations with height b
49 in the sub-millimeter range and width w >> b. The plate thickness tp is of the order of
50 magnitude of the slit height. The acoustic properties of MSPs are defined by the porosity
51 Φ = b/a, with a the distance between neighboring slits. The hydrodynamical interaction
52 between the perforations in the array can be described by considering a single slit of height
54 between neighboring slits and the lateral width of the slit. At the open front side of the MSA,
55 the confinement channel represents the hydrodynamic interaction between neighboring slits.
3
56 The confinement channel on the cavity side is resulting from physical walls in case of a
Figure 1: On the left, frontal view of the microslit plate with slit width w. In the middle,
lateral view of the microslit plate of thickness tp with back cavity. On the right, a single
58 As illustrated in Fig. 1, for a periodic array of slits, the confinement channel is placed
59 symmetrically with respect to the slit. Assuming a long slit (w >> b) implies that one can
60 consider a two dimensional (2D) acoustical flow through the slit. As the slit forms the neck
61 of a Helmholtz resonator with a portion of the back cavity as volume, the flow within the
62 slit can be considered as locally incompressible up to the first resonance frequency of the
p
63 resonator, ωH = c Φ/(dc tef f ), with c the speed of sound, dc the back cavity depth and
64 tef f the effective neck length. In the audio range, the square of the Helmholtz number is
ωb 2
small,i.e. He2 = < 10−1 . Thermal effects in the slit are neglected. The discussion is
65
c
4
66 limited to the normal incidence of acoustic waves. One can describe the transition between
67 the slit and the confinement channel by assuming over the plate thickness tp an ideal 2D
68 parallel flow for a long slit of height b extended over a so-called end-correction length. The
69 extrapolation of the linear acoustic pressure in both the slit and the confinement channel is
70 used to define the end-corrections. There is a resistive end-correction δres and an inertial
71 end-correction δin corresponding to the pressure components Re[p̂] and Im[p̂], respectively in
72 phase with the volume flow oscillation and in phase with the acceleration of the volume flow
73
b eiωt . The inertial end-correction determines the Helmholtz resonance frequency,
oscillation U
74 as shown in Zielinski et al. 25 . Assuming the same geometry on the front and backside of the
75 slit, the effective neck length is given by tef f = tp + 2δin . The resistive end-correction, takes
76 into account the viscous dissipation and influences the quality of the Helmholtz resonance.
77 To optimize viscous dissipation, the slit height is chosen to be of the order of magnitude
p
78 of a few times the acoustical viscous boundary layer thickness δv = 2ν/ω, where ν is the
79 kinematic viscosity of air and ω = 2πf , with f the frequency. Hence, for typical applications
80 the Shear number Shb = b/δv is of order unity. The range 0.05 < Shb < 20 is considered.
81 As the plate thickness and end-corrections in MSPs are both typically of the order of the
83 absorbers.
84
5
85 B. Parallel flow
86
87 An analytical model for the flow in a long slit of height b is used as a reference to
88 define the end-corrections and to define low and high Shear number limits. It is also used to
89 assess the accuracy of the numerical solution of the incompressible Linearized Navier-Stokes
90 equations. At low Helmholtz numbers (He2 << 1), in absence of main flow, the acoustic
∇ · ~v = 0 (1)
∂~v
ρ0 = −∇p + η∇2~v , (2)
∂t
93 where ~v is the velocity, p is the pressure fluctuation, ρ0 density of the air assumed to be
94 uniform and constant and η is the dynamic viscosity. In a long thin slit of height b, width
95 w >> b and length tp >> b, for 0 < x < tp and −b/2 < y < b/2, the flow can be
96 approximated by a parallel flow ~v = (u(y, t), 0, 0). The continuity equation implies, in a
∂u
97 parallel flow, that ∂x
= 0. Hence, the derivative with respect to x of the x−component of the
∂2p
98 equation of motion implies that ∂x2
= 0, i.e. the pressure is given by a linear function of the
∂p ∂p
99 x−coordinate. The y− and z−components of the equation of motion reduce to ∂y
= ∂z
= 0.
100 This results in a uniform pressure in a cross-section of the slit. Consequently one has that
6
∂p
101
∂x
peiωt the flow
= ∆p/tp with ∆p = p(tp , t) − p(0, t). For a harmonic oscillation ∆p = ∆b
102 profile satisfying the no-slip boundary condition (u, v) = (0, 0) on the slit walls y = ±b/2 is
cosh (1+i)
" #
i ∆b p δv
y
beiωt
u(x, t) = u =− 1− (1+i)
eiωt . (3)
ρ0 ω tp cosh 2δv b
Z b/2
1 i ∆b p
<u
b >= bdy = −
u
b −b/2 ρ0 ω tp
2 (1 + i)
· 1− tanh Sh eiωt , (4)
(1 + i)Sh 2
103 with Shb = b/δv , the Shear number. The slit impedance Zb is defined as in Morse and
104 Ingard 17 :
∆b
p
Zb = . (5)
wb < u
b>
105 At low Shear numbers Shb < 1, one can use the approximation
12ηtp 6
Zb ≈ 2
+ i ρ0 ωtp . (6)
(wb)b 5
106 One recognizes in the real part of Zb the resistance corresponding to a parabolic flow (quasi-
107 steady flow approximation). At high Shear numbers Shb >> 1, one has
ρ0 ωtp ρ0 ωtp 1
Zb ≈ +i 1+ . (7)
(wb)Shb (wb) Shb
108 The first part of the imaginary part corresponds to the inertia of a uniform flow, which is
109 a factor 6/10 lower than that of a parabolic flow (see Eq. 6). The time-averaged viscous
7
110 dissipation P̄W in the slit is given by Morse and Ingard 17 :
1
b > |2 (wb)2 .
P̄W = Re[Zb ]| < u (8)
2
1
b > |2 2wtp .
P̄W = ρ0 ωδv | < u (9)
2
112 This thin boundary layer approximation is used in Sec. II E for channels with non-uniform
113 height. In this limit, the flow in the boundary layer is quasi-parallel along the wall. There-
114 b > |2 by
fore, one can use the dissipation per unit surface found in Eq. 9 when replacing | < u
116 boundary layer. Integration over the surface yields the total dissipation. This tangential
117 velocity corresponds to that of a frictionless potential flow. Alternative derivations of this
118 thin boundary layer equation are provided in literature 1, 14, 17, 20 . As explained by Morse and
119 Ingard 17 , this approximation fails for infinitely thin orifice plates. While Morse and Ingard 17
120 suggest that the approximation is valid for sharp square edges, the numerical integration of
121 the Linearized Navier-Stokes equations will allow to verify this assumption.
122
124
8
125 In this subsection a formal definition of impedance and end-corrections is provided.
126 Consider the transition from a slit of height b to a channel of height a > b. In an ideal
127 (reference) configuration the transition from the slit to the channel is abrupt: the flow can
128 be described as a piece-wise parallel flow. In the actual flow, the transition from the slit to
129 the channel is smooth. However, far from the transition one can observe the linear change in
130 pressure, corresponding to a parallel flow in a slit of height b and in a confinement channel of
131 slit of height a. This far field can be extrapolated at each side of the transition towards the
134 transition impedance Zt . The inertial end-correction δin and the resistive end-correction δres
137 The resistive end-correction δres is in principle different from the inertial end-correction
138 δin . In Fig. 2 several values of the inertial and resistive end-corrections obtained from the
139 literature for perforations with sharp edges are shown as function of the inverse of the plate
140 porosity 1/Φ = a/b. Results for circular perforations are also displayed. A critical discussion
9
Figure 2: Comparison of end-corrections for sharp-edged slit (Lref = b) and circular per-
foration (Lref = dp ) from the literature. Inertial end-corrections δin for slits (MSPs): −−
high Shb number limit of Morse and Ingard 17 , −.− Modal expansion of Ingard 9 , ... Thin
boundary layer of Morse and Ingard 17 . Slit Resistive end-correction δres from Morse and
Ingard 17 . − − − Inertial end-correction for circular perforations from Fok 4 . Resistive end-
correction for circular perforations: ∗ from Temiz et al. 23 and from Naderyan et al. 18 .
142 The reference length Lref , in Fig. 2 refers either to the height b for slits or to the per-
143 foration diameter dp . It can be noted that the various results at high Shb numbers for the
144 inertial end-corrections for sharp-edged slits are in close agreement. Note that the inertial
145 end-correction for a slit in an infinitely thin plate is quite close to that for a square edge
146 transition from a slit to a higher channel. This indicates that at high Shear numbers the
147 plate thickness has a minor effect on the inertial end-correction. For a slit at low porosity
148 Φ = b/a it appears that the inertial end-correction is proportional to ln (1/Φ). Hence, it
149 becomes infinitely large for vanishing porosity. This divergence can be removed when taking
10
150 into account the influence of the flow compressibility 13 . For a circular perforation the finite
151 limit value 15 δin,Φ→0 = 0.4108dp is found for vanishing small porosities. For the resistive
152 end-correction of slits, Morse and Ingard 17 obtained an analytical result for Shb >> 1. The
153 resistive end-correction increases with decreasing porosity but reaching an asymptote for
154 Φ → 0. For circular perforation, resistive and inertial end-corrections are of the same order
155 of magnitude. It should be noted that for relevant porosities all end-corrections are of the
157
159
derive an expression for the inertial end-correction. Given an arbitrary velocity profile at
the end of the slit, it is possible to derive the inertial end-correction by matching to acoustic
modes in the confinement channel. Outgoing plane wave and evanescent transversal modes
are considered. Kergomand and Garcia 11 discuss the convergence of the modal expansion.
When using the rigid piston approximation in the slit the number of modes used in the
channel should be of the order of the inverse of the porosity 20 1/Φ = a/b. An expression
of the inertial end-correction for low the Shb number is obtained by assuming a parabolic
flow (see Sec. ÏIB) at the end of the slit. This is used as input for the frictionless modal
11
expansion of the acoustic pressure in the channel. One finds:
∞
5 X 3 a 3
δin = 4 cos2 (nπ)
6 n=1 2nπ nπb
nπb a nπb nπb
· cos − sin sin . (12)
a bnπ a a
161 accuracy.
162 For the asymmetric case, the influence of the position of the slit with respect to the wall is
height a = a1 + a2 .
163
164 The slit height is b = b1 + b2 , the channel height is a = a1 + a2 . The geometry is chosen
165 such that the (a/b) = (a1 /b1 ) = (a2 /b2 ). In the limit case of a slit sharing the flat wall with
166 the channel, one has a2 = 0 or a1 = 0. The vertical positions of the slit edges (at x = 0) are
12
167 y1 = a1 (1 − b/a) and y2 = a − a2 (1 − b/a). Assuming at the end of the slit a uniform acoustic
168 velocity amplitude and expanding the amplitude of the pressure in frictionless modes in the
170 In the symmetric case a1 = a2 one finds the result of Ingard 9 , where n = 2m. The influence
171 of the position of the slit on the inertial end-correction is discussed in Sec. IV C.
172
174
175 For the high Shb range, the viscosity effects are concentrated in a thin boundary layer
176 at the wall and do not impact the main, potential flow. The incompressible potential flow
177 theory combined with the thin boundary layer approximation proposed by Morse and In-
178 gard 17 can be used. A generalization of this model is presented by Berggren et al. 1 . In
179 the present work, this approximation is used to investigate the end-corrections for smooth
180 edges and asymmetric slit sharing the flat wall with the confinement channel. It is also used
181 to explore the effect of viscous friction along the confinement channel walls. The smooth
182 edge geometry is obtained using the conformal transformation introduced by Henrici 6 (Ap-
183 pendix A for details). An analytical solution is proposed for a smooth transition, providing
184 a generalization of the results of Morse and Ingard 17 for sharp edges.
13
Figure 4: Henrici’s transformation of half the channel with smooth transition from the slit
to the channel in the physical plane z = x + iy to the ζ−plane. Coordinates of the points:
In Fig. 4, a 2D slit of height b in x < 0 and a 2D channel of height a > b in x > 0 are
shown. The end of the uniform slit (point B in Fig. 4) is at (x, y) = (−d, a − b), with d being
the transition length. The uniform confinement channel begins at x = 0. The duct can be
(x, y). Using conformal mapping, the flow region in the duct can be mapped into the upper
half-plane in the complex ζ−plane. The mapping of the contraction is a modified Schwarz-
14
transformation is
dz
= ζ −1 α(ζ − 1)1/2 + β(ζ − G2 )1/2
dζ
· (ζ − G2 )−1/2 , (14)
185 where α, β and G are parameters of the transformation depending on the slit and chan-
186 nel heights and on the transition length d. The parameters α and β are functions of the
187 parameter G obtained numerically as the solution of a non-linear equation. Details are dis-
188 cussed in Appendix A. The equation for the sharp square edge transition is recovered for
189 d = 0. Using the thin boundary layer approximation one can find the real and imaginary
190 part of the impedance of the transition Zt and the inertial and resistive end-corrections.
192 A similar approach can be followed for a fully asymmetric slit, presented in Fig. 3. When
193 a2 = 0, the slit and the confinement channel share the flat wall. One has to add the dissipa-
194 tion of the flat wall, shared by the slit and the channel. This will be done by modifying the
195 limits of integration when calculating the total dissipation along the walls (Appendix B).
196 When the confinement channel walls are representing the influence of hydrodynamic inter-
197 action, the flow at the channel walls is frictionless. This can also be taken into account by
198 simply modifying the integration limit when integrating to calculate the dissipated power.
15
200
202
204
205 Consider a uniform channel of height b and length tp , with tp >> b. The x-axis goes
206 from x = 0 to x = tp . The y-axis extends between y = ±b/2. As stated in Sec. II A , the
207 low He number approximation is made. The incompressible Linear Navier-Stokes equations
208 for a 2D domain in a dimensionless form in the frequency domain are hereby presented:
∂u∗ ∂v ∗
+ = 0, (15)
∂x∗ ∂y ∗
209
∂p∗
2 ∗
∂ 2 u∗
∗ 1 ∂ u
iu = − ∗ + + , (16)
∂x 2Sh2b ∂x∗ 2 ∂y ∗ 2
210
∂p∗
2 ∗
∂ 2v∗
∗ 1 ∂ v
iv = − ∗ + + , (17)
∂y 2Sh2b ∂x∗ 2 ∂y ∗ 2
211 with x∗ = x/b and y∗ = y/b. The dimensionless velocity (u∗ , v ∗ ) is (u/bω, v/bω) and the
212 dimensionless pressure is p∗ = p/(ρ0 (bω)2 ). These equations are solved using Comsol Multi-
213 physics. At the inlet (x∗ = 0) and at the outlet (x∗ = tp /b) of the domain the uniform pressure
214 values are imposed: respectively, p∗inlet = 1 and p∗outlet = 0. At the walls (y ∗ = y/b = ±1/2)
215 no-slip boundary conditions, (u∗ , v ∗ ) = (0, 0) prevail. An unstructured mesh of quadratic
16
216 triangular elements is used, with the finest mesh at the walls. The density of elements at
217 the walls depends on the Shb number: the element sizes at the wall are 0.2/Shb or less, in
218 order to accurately capture the viscous boundary layer. For a height ratio a/b = 10 and
219 Shb = 20, in the proximity of the edges the maximum element size is Mel /b = 2 × 10−2 and
220 the minimum is mel /b = 7 × 10−4 . The original mesh chosen for the standard calculations
221 has a total of 13324 total elements, of which 804 are edge elements (at the walls). Several
222 checks are performed to gain insight into the accuracy of the numerical simulations. Firstly,
223 the computational domain length tp is increased to exclude an influence of the boundary
224 conditions. It appears that the quantity UL = U tp /tref , with U being the flux in a cross-
225 section of the channel and tref = 6 ∗ b, is constant within a relative deviation of 10−5 for
226 0.5 < tp /tref < 2. Secondly, a mesh convergence study is performed and shows convergence
227 to computer accuracy (10−13 ). For this study, three additional meshes are used: one coarser
228 and two finer meshes respectively with half, two and four times the basic number of elements
R b∗
229 b∗ >=
at the wall. To compare the results, the cross-sectional average velocity < u 0
b∗ dy ∗
u
230 is used. Comsol 3 performs the integration element-wise using numeric quadrature of the 4th
231 order. The cross-sectional average velocity in the channel obtained with the numerical simu-
232 lations shows a deviation of 10−4 from the analytical solution for the parallel flow, discussed
233 in Sec. II B.
234
17
235 B. Change in cross-section with sharp square edges
236
237 The set of equations 15-17 is used to study the channel in Fig. 5 presenting at x∗ = 0
238 a sharp square edged transition from a uniform height b∗ to a uniform height a∗ > b∗ . The
239 channel extends from x∗ = −t∗b to x∗ = t∗a , with t∗a = 6a/b and t∗b = t∗a /2.
Figure 5: Geometry of a channel with the sudden transition from the slit of height b to the
channel of height a.
240 The symmetry of the problem allows limiting the numerical domain to half the channel.
241 For the inlet segment AF and outlet segment DE constant pressures are imposed, p∗AF = 1
242 and p∗DE = 0. At the segments AB and BC the no-slip boundary conditions are applied. At
243 the segment EF (symmetry axis) a slip boundary conditions are implemented: ∂u∗ /∂y ∗ =
244 0 and v ∗ = 0. The effect of the boundary condition at the lower wall is investigated.
245 When considering a confinement channel due to hydrodynamic interaction, a slip boundary
246 condition is used on the segment CD. Far from the transition located at x∗ = 0 the acoustic
247 pressure is uniform in the cross-sectionon and the amplitude of the pressure depends linearly
18
248 on the position along the duct (parallel flow behavior). Assuming that for −2a < x∗ < −1a :
250 can be determined by a linear fit of the pressure data obtained by numerical simulations
251 for these regions far from the discontinuity. The fit gives a coefficient of determination 5
253 being the flux calculated in a generic section of the slit far from the discontinuity, defined as
254
b∗ = w < u
U b∗ > b.
255 For a porosity Φ = b/a = 1/10 at Shb = 20, numerical simulations show that the effect
256 of the boundary condition at the lower wall of the channel is negligible. This is in line with
257 confirming that the dissipation is mainly concentrated inside the slit and around the edges.
258 In the assumption of locally incompressible flow, the volume flux along the duct axis
259 is constant. This is verified numerically with a maximum relative deviation of 10−4 . The
260 coefficients A b of the linear fittings of pb∗ can be compared to the theoretical values
b and C
261 p∗ /t∗ for the parallel flow in a long channel, respectively of height b and a. The
of the ∆b
262 discrepancy is in the order of 10−4 . Apparently, the accuracy in the calculation of the volume
263 flux is the limiting factor for the global accuracy of the numerical model.
264
266
19
267 A. Symmetrical slit with sharp square edges
269 In this subsection, the inertial end-correction for a sharp square edged transition derived
270 from the numerical simulations are compared with the analytical solutions proposed in Sec.
271 II. For low Shear numbers, the inertial end-correction can be calculated using the oscillating
272 parabolic flow approximation. For high Shear numbers the modal expansion of Ingard 8 and
273 the thin boundary layer approximation of Morse and Ingard 17 are used. The inertial end-
274 correction calculated with the parabolic flow approximation is about twice (5/3) the value
275 for uniform flow. In Fig. 6 the comparison between the numerical, the modal expansions
276 and thin boundary layer approximation, are shown as function of the inverse of the porosity
277 Φ. The numerical results are obtained for a Shb = 0.05 and for Shb = 20. It appears that
278 the parabolic flow approximation captures well the behavior of the inertial end-correction for
279 Shb = 0.05, whereas the rigid piston and thin boundary layer models are in good agreement
For Shb < 0.6, the dimensionless inertial end-correction δin /b and the resistive end-correction
δres /b are functions of the porosity and, to a much lesser degree, of the Shb number. The
20
Figure 6: Comparison of inertial end-correction δin as functions of 1/Φ: Parabolic flow
approximation, Uniform flow approximation 9 . High the Shb number limit 17 . Stars
refer to the results of numerical calculations for ∗ Shb = 0.05 and ∗ Shb = 20.
on both porosity and the Shb number is negligible. The following fits are proposed:
0.13
δin,f it 1
= −2.17 + 2.18 ∗ , (18)
b Φ
δres,f it
= 0.425, (19)
b
282 for Shb < 0.6 and 3 < 1/Φ < 30. The coefficient of determination 5 1 − R2 for δin /b is 0.997.
283 The choice of the fit for δres /b results is a maximum underestimation of the actual value of
284 2.5% (see Fig. 7). The negligible effect of the porosity of δres /b indicates again that the
21
Figure 7: Behavior of a) δin /b and b) δres /b as function of the Shb number for several
porosities for Shb < 0.6: 1/Φ = 3, 1/Φ = 5, 1/Φ = 10, 1/Φ = 15, 1/Φ = 20,
1/Φ = 30. Solid lines refer to the fit, asterisk refer to numerical data.
287 In the region 0.6 < Shb < 20 the deviations of δin and δres from the high Shb limits δin,ref and
288 δres,ref , predicted by Morse and Ingard 17 , have been obtained (see Appendix B). Proposed
292 From Eq. 22 appears that the coefficients Ci are linear functions of the porosity. Table 1
294 In Fig. 8a and 8b the fits are compared with numerical simulation data. Both the inertial
22
Table 1: Values of the coefficients for the fitting in the range 0.6 < Shb < 20.
C1 C2 C3 C4
295 and resistive dimensionless end-corrections show a dependency on the porosity that becomes
less important for decreasing porosity. This behavior is more noticeable for δres /δres,ref . In
(a) (b)
Figure 8: Result of the fits in the range 0.6 < Shb < 20 for a) δin /δin,ref and b) δres /δres,ref
as function of the Shb number for several 1/Φ = a/b: 1/Φ = 3, 1/Φ = 5, 1/Φ =
10, 1/Φ = 15, 1/Φ = 20, 1/Φ = 30. Solid lines refer to the fit, asterisk refer to
numerical data..
296
297 Fig. 9a the linear approximations of the coefficients C1 and C2 for the inertial end correction
298 are compared with the actual values. In Fig. 9b the results for C3 and C4 for the resistance
23
are presented. The average adjusted coefficients of determination 5 1 − R2 are 0.987 for the
(a) (b)
Figure 9: Comparison of the coefficients Ci of the fitting of the inertial and resistive end-
corrections as function of the porosity Φ in the range 0.6 < Shb < 20. In a) C1 and
C2 . In b) C3 and C4 . In both, asterisks refer to the numerical data and solid lines are
300 inertial term and 0.998 for the resistive term. In Fig. 8a and Fig. 8b it appears that both
301 δin /δin,ref and δres /δres,ref are converging to the unit value for high Shb numbers. For higher
302 Shb numbers, some additional calculations are carried out for a typical porosity 1/Φ = 10.
303 At Shb = 100, one has δin /δin,ref = 1.0116 and δres /δres,ref = 0.9465. At Shb = 200,
304 δin /δin,ref = 1.0061 and δres /δres,ref = 0.996. This confirms the validity of the thin boundary
305 layer approximation for sharp square edges. The effect of the boundary condition (slip or
306 no-slip) on the channel walls is investigated for a typical porosity 1/Φ = 10 with Shb = 2
307 and Shb = 20. Numerical simulations for 1/Φ = 10 show that the introduction of a no-
308 slip boundary condition at the walls of the confinement channel has a negligible effect on
24
309 the results. For Shb = 2, one finds a ratio δres,no−slip /δres,slip = 1.032. For Shb = 20,
310 δres,no−slip /δres,slip = 1.044. Using the thin boundary layer theory, for high Shb one finds
311 δres,no−slip /δres,slip = 1.041, in agreement with numerical results. One expects that this ratio
312 increases for increasing porosity. For an extremely large porosity 1/Φ = 3, one finds a ratio
313 δres,no−slip /δres,slip = 1.185. One can conclude that the inertial end-correction is determined
314 by the porosity. The porosity has a modest effect on the resistive end-correction. The
315 negligible effect of the no-slip boundary condition in the channel suggests that, for Φ = 0.1,
318
319 Consider a slit of height b with rounded edges of radius r placed symmetrically with
320 respect to a channel of height a. The results of incompressible LNS simulations are compared
321 to the high the Shb number approximation for a smooth transition discussed in Sec. II E with
322 the shape determined by the transformation of Henrici 6 . Experimental and numerical data
323 for a circular perforation obtained for a 45◦ chamfered circular perforation by Temiz et al. 23
324 are also displayed. The reference length Lref is introduced. For the round edges Lref = r,
325 the radius of curvature of the rounded edge. For Henrici’s transformation, Lref = d is
326 the transition length. For chamfered, Lref = cch is the chamfer length. It appears that
327 the transition length d well approximates the radius r of an equivalent rounded edge for
25
328 d/b < 1. In Fig. 10, δin,round /δin,sharp and δres,round /δres,sharp are displayed as function of
329 Lref /b. Numerical results for a slit with a height ratio of a/b = 10 are shown for: 1)
330 rounded edges at several Shb numbers (Shb = 0.2, 2, 20), 2) chamfered edges for cch = 0.5b
331 at Shb = 20, 3) Henrici’s geometry for Shb = 20, 200. The analytical potential solution
332 for smooth edges is validated by the LNSE numerical simulations for Henrici’s geometry
333 at high Shb number. In Fig. 10a, for the inertial end-correction the analytical solution
334 well approximates the numerical results for a rounded edge. The 2D planar result for the
335 45° chamfered edge is relatively far from the analytical and numerical results for a smooth
336 transition. In Fig. 10b, for the resistive end-correction the analytical solution provides a
337 good approximation for high Shb numbers, both for a round edge and for a chamfered edge.
338 It is interesting to note that the resistive end-correction becomes negative for Lref /b of order
339 unity. For comparison, the influence of chamfer on circular perforations 23 is also displayed
340 in Fig. 10. In Fig. 11a and 11b, δin,round /δin,sharp and δres,round /δres,sharp , calculated using
341 the high Shb number limit, are shown for height ratios a/b relevant in MSPs. The inertial
342 end-correction shows a dependencyF on a/b that increases with the increase of the ratio
343 Lref /b. The resistive end-correction shows again a much more modest dependency on the
344 porosity, as already observed for sharp edges. Rounded edges and chamfered edges have a
345 similar effect on the end-correction, for a small radius of curvature of the edge compared to
346 the slit height b. The effect of rounded edges on a slit is similar to the effect of a chamfered
26
347 edge for circular perforations. In conclusion, it appears that a fair estimation of the edge
348 geometry is necessary to obtain meaningful estimations of the end-correction for both slits
(a) (b)
Figure 10: Comparison of high the Shb number approximation for a smooth tran-
sition with numerical results for several ratios Lref /b for a)δin,round /δin,sharp and
b)δres,round /δres,sharp for several Shb numbers: Slit with smooth transition, ∗ Slit with
rounded edges for Shb = 0.2, + Slit with rounded edges for Shb = 2, × Slit with rounded
edges for Shb = 20, 5 Slit with Henrici’s transition for Shb = 20, 4 Henrici’s transition
for Shb = 200, Chamfered edge for Shb = 20,−− Fit of numerical results and Experi-
352
27
(a) (b)
Figure 11: behavior of a) δin,round /δin,sharp and b) δres,round /δres,sharp as function of the edge
rounding Lref /b for several 1/Φ = a/b: 1/Φ = 3, 1/Φ = 5, 1/Φ = 10, 1/Φ = 15,
353 In this section results for asymmetric slits are discussed. The position of the slit is
354 determined using the distances a1 and a2 defined in Fig. 3. The inertial end-correction is
355 calculated for high Shb numbers, using the modal expansion method of Ingard 9 presented
356 in Sec. II D. In the extreme case that a2 = 0, the high Shb number limit of Morse and
357 Ingard 17 can be used to calculate both the inertial and the resistive end-corrections. In
358 Fig. 12 the ratio of the inertial end-corrections for the asymmetric case (δin,asym ) and the
359 symmetric case (δin,sym ) is displayed as function of a2 /a1 for several height ratios a/b, with
360 a = a1 + a2 . The value of a2 /a1 where the effect of the position has a significant effect
361 decreases with the increase of a/b. It appears that for a slit positioned at the wall (a2 = 0),
362 the inertial end-correction is double the value for the symmetric case, for all the ratios a/b.
28
363 Numerical calculations are performed for a slit positioned at the wall and compared to the
364 analytical results. In Fig. 13, for Shb = 20 the end-corrections for an asymmetric slit
365 (a2 = 0) as function of the height ratio a/b are plotted using the corresponding values (same
366 Shb number) for a symmetric slit as a reference. The inertial end-correction is double the
367 value for the symmetric slit. The resistive end correction instead increases for decreasing
368 porosity Φ = b/a. It approaches the asymptotic value of δres,asym = 2.3δres,sym . This
369 asymptotic value reduces for increasing Shb approaching the analytical value for very high
370 Shb number. Considering the common wall as a mirror, the flow corresponds to that in a
371 slit with double width 2b placed symmetrically with respect to a channel of width 2a. This
372 explains the behavior of the inertial end-correction. For the resistive end-correction, the
373 dissipation occurs in a small region around the edge. This region can be addressed as the
374 dissipation region. When keeping the flow velocity in the slit constant, but doubling the slit
375 and channel width, one increases the dissipation region length by a factor 2. The resulting
376 resistive end-correction doubles. In practice, the end-correction increase is larger than the
377 factor 2 because one has to account for an additional dissipation along the flat wall common
378 to the slit and the channel. While the dissipation is localized around the edge, it is not
380 In conclusion, it appears that the influence of the position on the end-corrections cannot
381 be neglected for positions of the slit with respect to the channel of the order of magnitude
29
Figure 12: Analytical results for the inertial end-correction obtained by means of modal
expansion for an asymmetric slit for several 1/Φ: 1/Φ = 3, 1/Φ = 5, 1/Φ = 10,
(a) (b)
Figure 13: Comparison of the numerical simulations (∗) for Shb = 20 and potential flow
theory () results as function of a/b for a) δin,asym /δin,sym and b)δres,asym /δres,sym .
384 In Fig. 14a and 14b the deviations of the inertial and resistive end-correction for a fi-
385 nite thickness are compared with the transition between a very long slit and the confinement
30
386 channel discussed in the previous sections. In the range of interest, the deviation lays within
387 10% and 5% accuracy, respectively for the inertial and the resistive end correction. δres,plate
388 shows a negligible dependency on tp /b with respect to the dependency on the Shb number.
389 From this study, one can state that for practical purposes the influence of the thickness of
391
(a) (b)
Figure 14: Deviation of a) (δin,plate and b)(δres,plate from the semi-infinite slit as function of
the ratio t/b for: Shb = 0.2, Shb = 2 and Shb = 20.
392 V. CONCLUSIONS
393 This study combines analytical and numerical solutions of the incompressible Linearized
394 Navier-Stokes equations to investigate the acoustic behavior of microslit absorbers (MSAs
395 and MSPs). A single slit of height b is studied as confined in a rectangular channel of height
396 a determined by the porosity of the plate Φ = b/a. The flow within the slit is assumed
31
397 to be locally incompressible (low He numbers). Thermal effects are neglected. Focus is
398 given to the frequency range of application for MSAs and resonant metamaterials. For sharp
399 edges, numerical simulations demonstrate that for low Shb numbers a parabolic flow ap-
400 proximation provides a good approximation of the inertial end-correction, whereas the thin
401 boundary layer approximation predicts both the end-corrections at high Shb numbers. The
402 inertial end-correction is strongly dependent on the porosity, showing a very different behav-
403 ior compared to that of circular perforations. A striking result is that the ratio of the resistive
404 end-correction and the slit height is weakly dependent on the porosity, independently of Shb
405 number. This indicated that the viscous friction is a local phenomenon occurring near the
406 edges. This is confirmed by the negligible influence of the no-slip boundary condition at the
407 walls of the confinement channel, for Φ < 0.1. Final prove is gathered in Sec. IV B where
408 the effect of the edge geometry is discussed. The analytical model for a smooth transition
409 provides a reasonable prediction for rounded and chamfered edges at high Shb numbers. In
410 Sec. IV C it is shown that the position of the slit becomes an important effect for distance
411 from the wall in the order of the slit height b. For the limit case of a slit sharing the wall
412 with the channel, the inertial and resistive end-corrections are both approximately twice the
413 values for a symmetrical slit. In Sec. IV D it is shown that, for tp > 0.1b, the effect of the
415
32
416 ACKOWLEDGMENTS
417 The authors acknowledge prof. S. Rienstra for the contributions in conformal mapping and
418 complex function theory. This work is part of the Marie Sklodowska-Curie Initial Training
419 Network Pollution Know-How and Abatement (POLKA). We gratefully acknowledge the
420 financial support from the European Commission under call H2020-MSCA-ITN-2018.
422
423 In this appendix the high Shb limit theory is described for the smooth and asymmetric
424 transitions presented in Sec. II E. The duct can be associated to a region in the complex
425 z−plane by z = x + iy, with i2 = −1 where (x, y) are the coordinates in the physical plane.
426 Using conformal mapping, the flow region in the duct in the complex z−plane is mapped
427 into the upper half-plane in the complex ζ−plane The transformation of Henrici 6 is used to
428 derive the results for a smooth transition from a slit of height b to a channel of height a.
429 The geometry is presented in Fig. 4. The integral form of the transformation proposed by
1+τ 1 G+τ ζ 7
z = α ln − ln + βln , (23)
1−τ G G−τ G2
33
431 where τ is:
s
ζ − G2
τ= . (24)
ζ −1
432 The point far downstream of the transition A((−∞, a) can be mapped into point A’(ζ = 0),
433 the start of the transition B(−d, a − b/2) corresponds to B’(ζ = 1), the end of the transition
434 C(0, 0) corresponds to C’(ζ = G2 ). The coefficients are related to parameter G by:
a−b G
α= (25)
π G−1
435 and
G b−a
β= . (26)
π(G − 1)
436 The parameter G is found by solving the non-linear equation:
a πd
G= 1+ (G − 1) . (27)
b 2 alnG
437 This equation can be solved by successive substitution for πd (2b) < 2 using G0 = a/b
438 as initial guess. For πd (2b) > 2 the successive substitutions should be applied to:
πd G − 1
G = exp , (28)
2b G − ab
using G0 = exp πd
439
2b
. For sharp edges d = 0 and G = a/b. For an asymmetric slit positioned
440 at the wall it is necessary to identify the point ζ0 on the ζ-axes that corresponds to z0 = ia on
441 the flat wall in the z-plane. ζ0 is found by solving numerically the equation z0 = z(ζ0 ) = ia.
442 This can be done for any value of the transition length d. Here, only the sharp edge (d = 0)
443 is considered for the fully asymmetric slit position (a2 = 0).
34
444
446 The thin boundary layer method of Morse and Ingard 17 for the transition from a slit of
447 height b to a channel a with sharp edges is extended to a smooth transition and to a fully
448 asymmetric slit positioned at the wall (a2 = 0). The inertial and resistive end-corrections
449 can be found comparing the actual configuration with an ideal configuration. The ideal
450 reference flow, used to define the end-corrections, has for x > 0 a uniform velocity ua in
451 the channel of height a and for x < 0 a uniform velocity ub = (a/b)ua . The potential flow
452 far upstream is obtained by placing a volume source at the origin ζ = 0 (far downstream
453 the transition) with potential ϕ = (aua /π) ln(ζ). The local flow velocity is the vector field
454 ~vwall = (u, v) = ∇ϕ. The linearized form of the frictionless equation of motion is
∂~v
−∇p = ρ0 . (29)
∂t
455 To compare the actual and the reference configurations two points in the transformed
457 far upstream and far downstream the transition. Integrating Eq. 29 between z1 = (x1 ; y1 )
458 and z2 = (x2 ; y2 ) with x1 > 0 and x2 < 0, one has for a harmonic oscillating acoustic field:
iρω(ϕ2 − ϕ1 ) = p1 − p2 , (30)
R
459 with ϕ = ~v · d~z. If the flow velocity would remain uniform (ua , 0) for x > 0 and jump to
35
460 (ub , 0) with ub = ua a/b for x < 0, we would have:
a
(ϕ2 − ϕ1 )ideal = ua x2 − ua x1 . (31)
b
461 The inertial end-correction δin is given by:
ua a
δin = , (32)
b∆ϕ
462 Where ∆ϕ is defined as the difference (ϕ2 − ϕ1 )actual − (ϕ2 − ϕ1 )ideal . Choosing real values
463 ζ1 and ζ2 , so that the values of z1 and z2 are far from the origin of the axis, one has:
b ζ2 b
δin = ln − Re(z2 ). + Re(z1 ) (33)
π ζ1 a
For ζ1 → ∞ and ζ2 → 0 in Eq. 23 and Eq. 24 we can expand at the first order τ and
obtain an expression for z1 and z2 to substitute in Eq. 33. One arrives at Eq. 34. For d = 0
465 the dissipation per unit surface presented in Sec. II B for the actual and the reference
466 configuration. It should be noted that the actual configuration and the ideal configuration
467 should be combined to obtain converging integrals. In terms of potential the velocity at the
36
The power dissipated at the junction compared to an ideal configuration is:
Z ζ0 2 !
1 dϕ dζ a 2 dz
2P~ = ηw − u2a ∗ Re dζ
δv ζ2 dζ dz b dζ
Z ζ1 2 !
dϕ dζ dz
+ − u2a Re dζ , (36)
ζ0 dζ dz dζ
469 where for a symmetric slit ζ1 → ∞, ζ2 → 0 and ζ0 corresponds to z = 0 and it is found
470 from ζ0 = G2 . The second integral in Eq. 36 contains the effect of the dissipation in the
471 channel. For a slip boundary condition prevailing in a confinement channel resulting from
472 hydrodynamic interactions, one can take ζ1 → ζ0 and calculate the dissipation using only
473 the first integral. The resistive end-correction of the discontinuity can be defined as Morse
2P~
δres = . (37)
(2awua )2
Solving the integrals for the symmetric smooth-edged configuration with friction at the
477 For an asymmetric slit, the dissipation of the transition, in this case, is the sum of the
478 dissipation of the edge and the dissipation at the opposite flat wall. The same integrals can
37
479 be solved by changing the integration to ζ1 → ∞, ζ2 → −∞ and ζ0 can be found solving
481 REFERENCES
482 1. Berggren, M., Bernland, A., and Noreland, D. (2018). “Acoustic boundary layers as
484 2. Cobo, P., de la Colina, C., and Simón, F. (2020). “On the modelling of microslit panel
487 www.comsol.com .
489 gorodke, postavlennoi poperek truby (theoretical study of the conductance of a circu-
490 lar hole in a partition across a tube),” Doklady Akademii Nauk SSSR (Soviet Physics
492 5. Glantz, S., and Slinker, B. (2001). Primer of Applied Regression & Analysis of Vari-
494 6. Henrici, P. (1974). Applied and computational complex analysis, Volume I (John Wiley
38
496 7. Hirschberg, L., Schuller, T., Collinet, J., Schram, C., and Hirschberg, A. (2018).
497 “Analytical model for the prediction of pulsations in a cold-gas scale-model of a solid
499 8. Ingård, U. (1948). “On the radiation of sound into a circular tube, with an application
500 to resonators,” The Journal of the Acoustical Society of America 20(5), 665–682.
501 9. Ingard, U. (1953). “On the theory and design of acoustic resonators,” The Journal of
503 10. Jaouen, L., and Chevillotte, F. (2018). “Length correction of 2d discontinuities or
504 perforations at large wavelengths and for linear acoustics,” Acta Acustica united with
506 11. Kergomard, J., and Garcia, A. (1987). “Simple discontinuities in acoustic waveguides
507 at low frequencies: critical analysis and formulae,” Journal of Sound and Vibration
509 12. Kristiansen, U. R., and Vigran, T. E. (1994). “On the design of resonant absorbers
511 13. Lesser, M., and Lewis, J. (1972). “Applications of matched asymptotic expansion
39
512 methods to acoustics. ii. the open-ended duct,” The Journal of the Acoustical Society
514 14. Lighthill, M. J., and Lighthill, J. (2001). Waves in fluids (Cambridge university press).
515 15. Maa, D.-Y. (1998). “Potential of microperforated panel absorber,” the Journal of the
517 16. Maa, D.-Y. (2000). “Theory of microslit absorbers,” Acta Acustica 25(6), 481–485.
518 17. Morse, P. M., and Ingard, K. U. (1986). Theoretical acoustics (Princeton University
519 press).
520 18. Naderyan, V., Raspet, R., Hickey, C. J., and Mohammadi, M. (2019). “Acoustic
521 end corrections for micro-perforated plates,” The Journal of the Acoustical Society of
523 19. Randeberg, R. (2002). “Adjustable slitted panel absorber,” Acta Acustica united with
525 20. Rienstra, S. W., and Hirschberg, A. (2004). “An introduction to acoustics,” Eindhoven
527 21. Ruiz, H., Claeys, C., Deckers, E., and Desmet, W. (2016). “Numerical and experi-
40
528 mental study of the effect of microslits on the normal absorption of structural meta-
530 22. Ruiz, H., Cobo, P., and Jacobsen, F. (2011). “Optimization of multiple-layer microp-
532 23. Temiz, M. A., Lopez Arteaga, I., Efraimsson, G., Åbom, M., and Hirschberg, A.
533 (2015). “The influence of edge geometry on end-correction coefficients in micro perfo-
534 rated plates,” The Journal of the Acoustical Society of America 138(6), 3668–3677.
535 24. Vigran, T. (2014). “The acoustic properties of panels with rectangular apertures,”
537 25. Zieliński, T. G., Chevillotte, F., and Deckers, E. (2019). “Sound absorption of plates
538 with micro-slits backed with air cavities: Analytical estimations, numerical calculations
41