Home Search Collections Journals About Contact us My IOPscience

One path to acoustic cloaking

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2007 New J. Phys. 9 45

(http://iopscience.iop.org/1367-2630/9/3/045)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 146.164.138.48
This content was downloaded on 07/02/2015 at 04:08

Please note that terms and conditions apply.

 New Journal of Physics The open–access journal for physics

One path to acoustic cloaking

Steven A Cummer1 and David Schurig
Department of Electrical and Computer Engineering, Duke University,
P O Box 90291, Durham, NC 27708, USA
E-mail: cummer@ee.duke.edu and dschurig@duke.edu
New Journal of Physics 9 (2007) 45
Received 10 November 2006
Published 2 March 2007
Online at http://www.njp.org/
doi:10.1088/1367-2630/9/3/045

Abstract. A complete analysis of coordinate transformations in elastic media

by Milton et al has shown that, in general, the equations of motion are not
form invariant and thus do not admit transformation-type solutions of the type
discovered by Pendry et al for electromagnetics. However, in a two-dimensional
(2D) geometry, the acoustic equations in a fluid are identical in form to the
single polarization Maxwell equations via a variable exchange that also preserves
boundary conditions. We confirm the existence of transformation-type solutions
for the 2D acoustic equations with anisotropic mass via time harmonic simulations
of acoustic cloaking. We discuss the possibilities of experimentally demonstrating
acoustic cloaking and analyse why this special equivalence of acoustics and
electromagnetics occurs only in 2D.

The transformation-based solutions to the Maxwell equations reported by Pendry et al [1] yield a
general method for rendering arbitrarily sized and shaped objects electromagnetically invisible.
Other approaches for almost eliminating electromagnetic scattering in all directions from certain
classes of objects of limited size have been described [2]–[4], and Leonhardt [5, 6] has given a
recipe that can cloak an object in the short wavelength limit that thus applies to different classes
of waves, including electromagnetic and acoustic. Similar transformation-based approaches have
been analysed for the related problem of electrical impedance tomography [7, 8]. The solution
found by Pendry et al stands out for its theoretically perfect performance regardless of wavelength
and object size and shape.
It is undoubtedly of interest whether the transformation-based electromagnetic solutions,
which were simulated in [9] and experimentally confirmed in [10], can be extended to waves
in other systems. From one perspective, it seems that the answer could be yes. The concepts of
wavevector, wave impedance, and power flow are universal, and the manner in which permittivity
1
Author to whom any correspondence should be addressed.

New Journal of Physics 9 (2007) 45 PII: S1367-2630(07)37675-1

1367-2630/07/010045+8$30.00 © IOP Publishing Ltd and Deutsche Physikalische Gesellschaft
2 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

and permeability control these in electromagnetics is closely analogous to that by material

properties in other systems. On the other hand, the Maxwell equations possess a special symmetry
that is an important element of the equivalence between coordinate transformations and material
properties for electromagnetic waves. The field equations that describe waves in most other
systems do not have this same symmetry.
Milton et al [11] analysed the full equations of motion for a general elastic medium
under coordinate transformations and found that, in general, the form of the equations of
motion are not invariant to transformations. The possibility of engineering artificial elastic
media that obeyed these modified equations of motion was discussed, but no simple scheme
for realizing a transformation-based application, such as acoustic cloaking, was reported. There
is at least one special case, however, in which the equivalence between electromagnetics and
elastodynamics is complete. Moreover, this case is practically useful and can potentially be
demonstrated experimentally. It has long been known that in two dimensions (2D) acoustics and
electromagnetics in isotropic media are exactly equivalent [e.g., 12]. We show below that this
isomorphism holds for anistropic media as well. For an inviscid fluid with zero shear modulus, the
linearized equations of state for small amplitude perturbations from conservation of momentum,
conservation of mass, and linear relationship between pressure and density are [13]

∂v
ρ0 = −∇p, (1)
∂t

∂p
= −λ∇ · v, (2)
∂t
where p is scalar pressure, v is vector fluid velocity, ρ0 is the unperturbed fluid mass density, and
λ is the fluid bulk modulus. This set of equations admits the usual compressional wave solutions
in which fluid motion is parallel to the wavevector. Milton et al [11] showed that any coordinate
transformation on an elastic medium produces an anisotropic mass represented by a second rank
tensor. In cylindrical coordinates with z invariance, and letting the mass density be anisotropic
but diagonal in these coordinates, the time harmonic acoustic equations of state simplify to (the
exp (+jωt) convention is used throughout)
1 ∂p
jωρφ vφ = − , (3)
r ∂φ

∂p
jωρr vr = − , (4)
∂r

1 1 ∂(rvr ) 1 ∂vφ
jω p = − − . (5)
λ r ∂r r ∂φ
Now consider the z-invariant 2D Maxwell equations for transverse electric (TE) polarization in
cylindrical coordinates with anisotropic but diagonal permittivity and permeability tensors
1 ∂(−Ez )
jωµr (−Hr ) = − , (6)
r ∂φ
New Journal of Physics 9 (2007) 45 (http://www.njp.org/)
3 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

∂(−Ez )
jωµφ Hφ = − , (7)
∂r

1 ∂(rHφ ) 1 ∂(−Hr )
jωz (−Ez ) = − − . (8)
r ∂r r ∂φ
The above equations have been arranged to highlight the exact duality between these two sets of
equations under the variable exchange

[p, vr , vφ , ρr , ρφ , λ−1 ] ↔ [− Ez , Hφ , −Hr , µφ , µr , z ]. (9)

The boundary conditions are also preserved under this exchange. At a fluid–fluid interface, p and
the normal velocity component are continuous, while at an electromagnetic material interface,
Ez and the tangential H component are continuous. Thus at a material discontinuity in radius,
p and vr are continuous, which are equivalent to Ez and Hφ under this replacement, which in
turn are the electromagnetic quantities that must be continuous. Consequently, a solution to
the 2D cylindrical Maxwell equations above is also a solution to the 2D acoustic equations
under the above replacement. This further implies that any transformation-based medium [1] for
electromagnetics, when restricted to 2D and normal incidence, can also be realized acoustically.
To demonstrate this, we consider the cloaking shell described in [1]. This exact equivalence
implies that a cylindrical shell containing specific spatial distributions of mass density and bulk
modulus can smoothly bend any incident wavefield around the centre of the shell with essentially
zero scattering in any direction, including the forward direction. For a 2D acoustic cloak, the
required relative mass density elements and bulk modulus are thus [9]

2
r r − R1 −1 R2 r − R1
ρr = , ρφ = , λ = , (10)
r − R1 r R2 − R1 r
where R1 is the radius of the cloaked region and R2 is the outer edge of the cloaking shell. These
values are relative to the bulk modulus and mass density of the fluid outside the shell.
This equivalence implies that the 2D electromagnetic simulations of plane wave interaction
with a cloaking shell presented in [9] are also solutions to the analogous acoustic cloaking
problems. But, in order to independently and unambiguously demonstrate the validity of this
acoustic cloaking approach, a time harmonic 2D cartesian grid finite difference code was written
and used to solve the acoustic equations (1) and (2) with an anisotropic mass tensor. Note that the
mass density tensor is not diagonal in cartesian coordinates, and the off-diagonal elements are
included in the formulation. The computational domain is shown in figure 1, in which a gaussian
beam is launched towards an incompressible cylinder 1.5 wavelengths in diameter. All simulation
quantities (excitation frequency, bulk modulus and background fluid density) are normalized to
unity, and all domain boundaries are perfectly matched layers to prevent reflections.
Figure 2 shows the simulated pressure field for three scenarios. The left panel shows the
baseline pressure field with no scatterer. The middle panel shows the pressure field with the
uncloaked scatterer, in which strong scattering in all directions is clear, especially the forward
direction which produces the shadow region. In contrast, when a 0.75 wavelength thick cloaking
shell specified by (10) is placed around the rigid scatterer, the incident wave is smoothly directed
around the scatterer with significantly reduced scattering in all directions. 2D acoustic cloaking
is thus realizable in an exact analogue to the electromagnetic equivalent.
New Journal of Physics 9 (2007) 45 (http://www.njp.org/)
4 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Figure 1. The 2D computational domain, in which an acoustic gaussian beam is

incident on an cylindrical object.

Figure 2. Left panel: The acoustic pressure p in the absence of a scatterer. Middle:
The acoustic pressure with an incompressible cylindrical scatterer. Right panel:
The acoustic pressure with the same scatterer surrounded by the cloaking shell
described in the text.

Some scattering is seen which we attribute to numerical artifacts associated with the cartesian
grid approximation of circular surfaces and to the relatively coarse grid used to compute these
solutions (20 grid points per wavelength). Although more exact cloaking has been simulated
numerically using a conformal triangular mesh [9], clear and effective cloaking performance
can also be achieved on a much simpler grid. This confirms the general insensitivity of the ideal
cloaking solution to material and geometric imperfections reported in [9], and also suggests that
effective cloaks (either electromagnetic or acoustic) could be designed based on a square grid,
rather than a conformal one such as that used in [10], which may be simpler to fabricate.
Although anisotropic fluid mass is not a property of ordinary fluids, there is hope for an
experimental realization of this idea. Milton et al [11] describes conceptually how it can be
achieved with spring loaded masses in a manner essentially equivalent to how self resonant
electromagnetic metamaterial structures can be used to achieve anisotropic electromagnetic
permittivity or permeability of almost any value [14]. Acoustic cloaking realized in this way
will consequently be of limited bandwidth.

New Journal of Physics 9 (2007) 45 (http://www.njp.org/)

5 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

If imperfect cloaking performance is tolerable, then an additional degree of freedom in the

material properties can be obtained in a reduced 2D acoustic cloak that is an exact analogue of the
2D electromagnetic case [9]. In a reduced cloak, the wavenumber is correct but the impedance
is not and is thus an ideal cloak only in the short wavelength limit. This extra degree of freedom
can be used to eliminate the inhomogeneity of one component of the material, as was done by
Schurig et al [10] in their experimental demonstration of electromagnetic cloaking. By recasting
the acoustic equations (1) and (2), this reduced approach can also shift the needed anisotropy
from the mass to the bulk modulus, which may be simpler to realize.
Substantial recent progress has also been made in the experimental realization of acoustic
metamaterials. It was described theoretically how a fluid loaded with particles can be used to
realize almost any isotropic effective mass and bulk modulus [15], and it has been experimentally
demonstrated that 1D acoustic waveguide loaded with resonators can yield an effective bulk
modulus of any value, even negative [16]. To be sure, there will be challenges in realizing a
functioning approximation of a fluid with continuously variable bulk modulus and anisotropic
mass elements, but it appears that this recipe for a 2D acoustic cloaking shell could be physically
realizable.
It is interesting to consider why this limited acoustic-electromagnetic equivalence is not
evident in the complete elastodynamic analysis of [11]. Here, we derive the equivalence from
first principles and show that it is specific to a 2D field in a 3D space. The 2D acoustic fields are
described by

p = p(x, y), (11)

v = vx (x, y)x̂ + vy (x, y)ŷ, (12)

where p is the pressure and v is the fluid velocity. We further assume that neither the
transformation matrix nor the material property tensors mix x–y coordinates with z so that
the transformation matrix has the following form
 
1 1 1 2 0
(i i ) =  2 1 2 2 0  . (13)
0 0 3 3 

In this case, we can show that elastic fluid wave equations (1) and (2), when placed in time
harmonic form, are isomorphic with the Maxwell curl equations

∇ × E + jωB = 0, (14)

∇ × H − jωD = 0, (15)

and thus form invariant by making the following variable exchange

[E, D, H, B] ↔ [pẑ, λ−1 pẑ, v×ẑ, (ρv) ]. (16)

New Journal of Physics 9 (2007) 45 (http://www.njp.org/)

6 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

Note that the Poynting vector S = E × H = pv correctly represents power flow in both systems.
Making the above substitutions into (14) yields

∇ × (pẑ) + jω(ρv)×ẑ = 0 (17)

and applying a vector identity this becomes

∇p × ẑ + p∇ × ẑ + jω (ρv)×ẑ = 0, (18)

which reduces to (1) since ẑ is a constant, and neither ∇p nor ρv have z-components. Following
the same procedure with (15) yields

v∇ × ẑ − ẑ∇ × v + (ẑ × ∇) v − (v × ∇) ẑ − jωλ−1 pẑ = 0, (19)

which reduces to (2) since ẑ is a constant and v does not depend on z.

Since we know how µ and ε behave under transformations [1], we can determine how ρ
and λ must behave. Applying the substitution (16) to the constitutive relation B = µH yields

(ρv) × ẑ = µ(v × ẑ), (20)

and in component form for the primed and unprimed coordinate system

ij3 ρjk vk = µim mn3 vn , (21)

i j 3 ρj k vk = µi m m n 3 vn . (22)

We substitute the transformation properties of v, (a vector),

vk = k i vi , (23)

and µ, (a tensor density of weight +1),

µi m = det ()−1 i i m m µim , (24)

into (21) and (22) to obtain

i j 3 ρj k k p vp = det ()−1 i i m m µim m n 3 n q vq . (25)

Solving (21) and (25) we find the general transformation for the density

ρi j = det ()−1 (i k 3 k k ki3 )(j l 3 l l lj3 ) ρij . (26)

Similarly, we can apply (16) to the other constitutive relation D = εE to obtain

λ−1 pẑ = εpẑ, (27)

New Journal of Physics 9 (2007) 45 (http://www.njp.org/)

7 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

which reduces to
λ−1 = ε33 .
and

ε3 3 = det ()−1 3 i 3 j εij , (28)

since ε is also a tensor density of weight +1. From the form of the transformation matrix (13)
we find that the general transformation for the bulk modulus is

λ = det ()−2
3 3 λ. (29)

Through the above procedure, the 2D acoustic wave has been placed in exact correspondence
with one polarization (TE, although it could also have been transverse magnetic) of the 3D
electromagnetic equations. A more general equivalence is not possible for acoustics because we
need three field variable components for correspondence with each electromagnetic polarization,
and a fluid provides only four total with the velocity vector and the scalar pressure. Even if we
consider a more general elastic medium in which strain is anisotropic, an exact correspondence
to electromagnetics in 3D is not possible as can be seen by examining the energy density, u. For
isotropic and anisotropic elastic waves we have respectively
1 1
u = λσii σkk + ρmn vm vn , (30)
2 2

1 1
u = cijkl σij σkl + ρmn vm vn , (31)
2 2
neither of which is isomorphic with the electromagnetic energy expression
1 1
u = εij Ei Ej + µkl Hk Hl . (32)
2 2
There may be more special cases of electromagnetic and elastodynamic equivalence, but as
shown in [11] and above, a general 3D equivalence is not possible.
In summary, we have shown that the exact equivalence in 2D between fluid acoustics
and single polarization electromagnetics holds for anisotropic materials. This anisotropic
isomorphism enables all classes of 2D transformation-based solutions of the Maxwell equations
to be realized acoustically as well, including the cloaking solution first described in [1]. It was
shown numerically that a fluid with an inhomogeneous bulk modulus and with an anisotropic
(but diagonal in cylindrical coordinates) and inhomogeneous mass density can smoothly and
reflectionlessly direct compressional acoustic waves around a strong scatterer. Engineering a
material shell to have the required properties may be a challenge, but it is experimentally feasible
in principle.

References

[1] Pendry J B, Schurig D and Smith D R 2006 Science 312 1780

[2] Kerker M 1975 J. Opt. Soc. Am. 65 376
[3] Alu A and Engheta N 2005 Phys. Rev. E 72 016623

New Journal of Physics 9 (2007) 45 (http://www.njp.org/)

8 DEUTSCHE PHYSIKALISCHE GESELLSCHAFT

[4] Milton G W and Nicorovici N P 2006 Proc. R. Soc. A 462 3027

[5] Leonhardt U 2006 Science 312 1777
[6] Leonhardt U 2006 New J. Phys. 8 118
[7] Greenleaf A, Lassas M and Uhlmann G 2003 Physiol. Meas. 24 413
[8] Greenleaf A, Lassas M and Uhlmann G 2003 Math. Res. Lett. 10 685
[9] Cummer S A, Popa B-I, Schurig D, Smith D R and Pendry J 2006 Phys. Rev. E 74 036621
(Preprint physics/0607242)
[10] Schurig D, Mock J J, Justice B J, Cummer S A, Pendry J B, Starr A F and Smith D R 2006 Science 314 977
[11] Milton G W, Briane M and Willis J R 2006 New J. Phys. 8 248
[12] Kelders L, Allard J F and Lauriks W 1998 J. Acoust. Soc. Am. 103 2730
[13] Lighthill J 2005 Waves In Fluids (Cambridge: Cambridge University Press)
[14] Pendry J B, Holden A J, Robbins D J and Stewart W J 1999 IEEE Trans. Microwave Theory Tech. 47 2075
[15] Li J and Chan C T 2004 Phys. Rev. E 70 055602
[16] Fang N, Xi D, Xu J, Ambati M, Srituravanich W, Sun C and Zhang X 2006 Nat. Mater. 5 452

New Journal of Physics 9 (2007) 45 (http://www.njp.org/)