Integral Equation Methods for the Morse-Ingard Equations

2025-05-05 0 0 1.27MB 22 页 10玖币
侵权投诉
Integral Equation Methods for the Morse-Ingard Equations
Xiaoyu Weia, Andreas Klöcknera,, Robert C. Kirbyb
aDepartment of Computer Science, University of Illinois at Urbana-Champaign,
Champaign, Illinois, United States
bDepartment of Mathematics, Baylor University, Waco, Texas, United States
Abstract
We present two (a decoupled and a coupled) integral-equation-based methods
for the Morse-Ingard equations subject to Neumann boundary conditions on
the exterior domain. Both methods are based on second-kind integral equation
(SKIE) formulations. The coupled method is well-conditioned and can achieve
high accuracy. The decoupled method has lower computational cost and more
flexibility in dealing with the boundary layer; however, it is prone to the ill-
conditioning of the decoupling transform and cannot achieve as high accuracy
as the coupled method. We show numerical examples using a Nyström method
based on quadrature-by-expansion (QBX) with fast-multipole acceleration.
We demonstrate the accuracy and efficiency of the solvers in both two and
three dimensions with complex geometry.
Keywords: The Morse-Ingard Equations, Fast Multipole Method, Integral
Equation Method, Quadrature-by-Expansion
1. Introduction
The Morse-Ingard equations are time-harmonic, steady-state equations
derived from the linearized Navier-Stokes equations [
4
,
12
]. They model the
pressure and the temperature variations of a fluid due to a heat source inside
the fluid, in a regime where both the acoustic wavelength and the thermal
boundary layer thickness are of interest. Specifically, we consider the following
nondimensionalized Morse-Ingard equations from [
4
]: in the fluid domain
Dc
Corresponding author
Email addresses: xywei@illinois.edu (Xiaoyu Wei), andreask@illinois.edu
(Andreas Klöckner), Robert_Kirby@baylor.edu (Robert C. Kirby)
Preprint submitted to Elsevier April 24, 2023
arXiv:2210.12542v2 [math.NA] 21 Apr 2023
with DRdbounded,
2T+iT iγ1
γP=S,
(1 Λ)2P+γ1Λ
+Λ
Pγ1Λ
T=Λ
S,
(1)
where
P
is the pressure field,
T
is the temperature field,
S
is the heat source,
and ,Λ, γ are dimensionless parameters.
The application that motivated this work is to model trace gas sensors
that utilize optothermal and photoacoustic and effects to aid designing such
sensors. The
q
uartz-
e
nhanced
p
hoto
a
coustic
s
pectroscopy (QEPAS) sensor,
for example, employs a quartz tuning fork to detect via the piezoelectric effect
the acoustic pressure waves that are generated when optical radiation from a
laser is periodically absorbed by molecules of a trace gas [
6
,
8
]. The
r
esonant
o
pto
t
hermo
a
coustic
de
tection (ROTADE) sensor, on the other hand, uses
the same tuning fork to detect the thermal diffusion wave via the indirect
pyroelectric effect [
7
]. An efficient and accurate solver for the Morse-Ingard
equations on the exterior domain can be a key tool for the modeling and
design of both QEPAS and ROTADE sensors [4, 14, 17].
To suit this application, we consider the thermoacoustic scattering
problem, where the scattered waves obey the homogeneous Morse-Ingard
equations on the exterior domain (
S
= 0), coupled with sound-hard boundary
conditions in pressure and continuity in heat flux, leading to Neumann
boundary conditions in both Tand P,
T
n D
=gT,
P
n D
=gP,
(2)
where the volumetric source
S
in
(1)
is accounted for by the incoming waves
it induces in the boundary conditions above.
For solution uniqueness, we also require
T(x)<and P(x)<,when |x|→∞.(3)
As shown in [
17
],
(3)
ensures solution uniqueness by ruling out unphysical
waves from the infinity, similar to what the Sommerfeld radiation condition
does for the Helmholtz equation. In our applications, the wave number
has positive imaginary part. Therefore, boundedness at infinity is sufficient.
2
For general wave numbers, we also provide a direct generalization of the
Sommerfeld condition in Section 2.3.
In this paper, we propose two integral-equation-based methods for (1)
to (3). For both methods, we derive representations of the solution in terms
of layer potentials involving unknown densities, leading to integral equations
that can be reduced to the form of a Fredholm integral equation of the second
kind
(IA)ρ=f, (4)
where
A
is a compact integral operator; therefore, our methods are all based
on second-kind integral equation (SKIE) formulations. SKIEs are attractive
because both the condition number and the number of iterations required for
iterative solvers like GMRES [
16
] are bounded by constant when refining the
mesh. In fact, [
11
] gives rigorous superlineary GMRES convergence estimates
in this case. The first method is based on a direct SKIE formulation to the
original equations using the free-space Green’s functions. While deriving
the analytic formula for the free-space Green’s functions, we also obtain a
decoupling transform that converts the problem into two decoupled Helmholtz
equations, with decoupled Neumann data. Our second method takes advantage
of this transform and uses SKIEs for the Helmholtz equations to obtain the
solution. As such we refer to the first method the coupled method and the
second the decoupled method. We solve the SKIEs using a Nyström method
with GMRES, which requires evaluating layer potentials at the boundary
using a suitable singular quadrature scheme. For our numerical experiments,
we use quadrature-by-expansion (QBX) with fast-multipole acceleration
[
5
,
18
]. Consequently, both methods have linear complexity with respect to
the number of degrees of freedoms. It is noteworthy that our methods are
agnostic of the singular quadrature scheme. Our methods are also capable
of using high order discretization and can handle complex geometries. We
demonstrate this through numerical examples in two and three dimensions.
The rest of this paper is organized as follows. We first present the deriva-
tion for the free-space Green’s functions and the decoupling transform in
Section 2. Then we give the SKIE formulations for both the decoupled and
coupled methods in Section 3, and present some details of our numerical
implementation in Section 4. After that, we present the numerical results in
Section 5, and give some concluding discussion in Section 6.
3
2. Analysis of the problem
2.1. Thermal and acoustic modes
In order to obtain an integral equation formulation, we require the free-
space Green’s function for the Morse-Ingard equations (1). Following the
derivation for the analytic solution to the Morse-Ingard equations in a cylin-
drically symmetric geometry in [
4
], we first identify the eigenmodes by looking
for particular solutions where
T
is an eigenfunction of
2
, s.t.
2T
=
k2T
.
Substituting into the first equation of (1) yields
P=γ
γ1[(1 + ik2)T+iS].(5)
Consider the homogeneous case by letting
S
= 0, then
P
=
mT
is also an
eigenfunction of
2
, where the constant
m
=
γ
γ1
(1 +
i
k2
). Substituting
(5) and P=mT into (1) yields
(1 Λ)(k2mT ) + γ1Λ
+Λ
(mT )γ1Λ
T= 0.(6)
For (6) to have nontrivial solution, the coefficient of Tmust vanish, so that
(iΩ + γΩΛ)k4+ (1 iΛ)k21=0.(7)
Let
Q
to be a complex constant such that
Q2
= 4(
i
+
γ
ΩΛ)+(1
i
Λ)
2
.
Based on physical interpretation, we classify the roots of
(7)
into two groups:
1. ktcorresponding to the thermal modes that attenuate rapidly:
k2
t=i
2Ω 1iΛ + Q
1Λ, mt:= γ
γ1(1 + ik2
t).(8)
2. kpcorresponding to the acoustic modes that attenuate slowly:
k2
p=i
2Ω 1iΛQ
1Λ, mp:= γ
γ1(1 + ik2
p).(9)
Since the eigenfunctions of
2
form a basis of
H1
, we obtain the funda-
mental set of solutions to the homogeneous problem under radial symmetry,
denoted Ud,(d= 2,3),
T(r)
P(r)Ud.
4
In two dimensions (d= 2),
U2= span (J0(kpr)
mpJ0(kpr),"H(1)
0(kpr)
mpH(1)
0(kpr)#,J0(ktr)
mtJ0(ktr),"H(1)
0(ktr)
mtH(1)
0(ktr)#),
(10)
where
J0
is the Bessel function of the first kind of order zero,
H(1)
0
is the
Hankel function of the kind of order zero. Similarly, in three dimensions
(d= 3),
U3= span (j0(kpr)
mpj0(kpr),"h(1)
0(kpr)
mph(1)
0(kpr)#,j0(ktr)
mtj0(ktr),"h(1)
0(ktr)
mth(1)
0(ktr)#),
(11)
where
j0
is the spherical Bessel function of the first kind of order zero,
h(1)
0
is
the spherical Bessel function of the third kind of order zero. We also note
that h(1)
0has simple closed form
h(1)
0(r) = j0(r) + iy0(r) = sin r
ricos r
r=i
reir.(12)
Note that the choice of basis is not unique. We deliberately chose the above
basis functions so that the condition in (3) can be easily enforced.
2.2. The decoupled equations
From the above radial symmetry solutions, it is obvious that the Morse-
Ingard equations are a linear superposition of two Helmholtz-type equations.
To get the actual change of variables that decouples the PDE system, we
solve for
tC
such that the sum of the first equation and
t
times the second
equation in (1) reduces to a scalar Helmholtz-type PDE
a1(t)2T+a2(t)2P+a3(t)T+a4(t)P=a5(t)S, (13)
where
a1
= Ω,
a2
= (1
Λ)
t
,
a3
=
iγ1Λ
t
,
a4
=
iγ1
γ
+
γ1Λ
+Λ
t
,
a5
= 1
Λ
t
are all linear functions of
t
. The condi-
tion under which
(13)
becomes a scalar Helmholtz-type PDE is
a1a4
=
a2a3
,
which is a quadratic equation of tand admits two roots
t±=(2ΛγΛγ+i)Ω iQ
2γΩ)(iΛγ1) .(14)
Letting
Vt
= Ω
T
+
t+
(1
Λ)
P
,
Vp
= Ω
T
+
t
(1
Λ)
P
, we find the
decoupled scalar PDEs,
2Vt+k2
tVt=a5(t+)S,
2Vp+k2
pVp=a5(t)S, (15)
5
摘要:

IntegralEquationMethodsfortheMorse-IngardEquationsXiaoyuWeia,AndreasKlöcknera,,RobertC.KirbybaDepartmentofComputerScience,UniversityofIllinoisatUrbana-Champaign,Champaign,Illinois,UnitedStatesbDepartmentofMathematics,BaylorUniversity,Waco,Texas,UnitedStatesAbstractWepresenttwo(adecoupledandacoupled...

展开>> 收起<<
Integral Equation Methods for the Morse-Ingard Equations.pdf

共22页,预览5页

还剩页未读, 继续阅读

声明:本站为文档C2C交易模式,即用户上传的文档直接被用户下载,本站只是中间服务平台,本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私,请立即通知玖贝云文库,我们立即给予删除!
分类:图书资源 价格:10玖币 属性:22 页 大小:1.27MB 格式:PDF 时间:2025-05-05

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务
/ 22
客服
关注