![MathJax Logo](/templates/jsp/_style2/_tandf/pb2/images/math-jax.gif)
Abstract
We consider in a two dimensional absorbing and scattering medium, an inverse source problem in the stationary radiative transport, where the source is linearly anisotropic. The medium has an anisotropic scattering property that is neither negligible nor large enough for the diffusion approximation to hold. The attenuating and scattering properties of the medium are assumed known. For scattering kernels of finite Fourier content in the angular variable, we show how to recover the anisotropic radiative sources from boundary measurements. The approach is based on the Cauchy problem for a Beltrami-like equation associated with A-analytic maps. As an application, we determine necessary and sufficient conditions for the data coming from two different sources to be mistaken for each other.
2010 Mathematics Subject Classifications:
1. Introduction
In this work, we consider an inverse source problem for stationary radiative transfer (transport) [Citation1,Citation2], in a two-dimensional bounded, strictly convex domain , with boundary Γ. The stationary radiative transport models the linear transport of particles through a medium and includes absorption and scattering phenomena. In the steady state case, when generated solely by a linearly anisotropic source f inside Ω, the density
of particles at z traveling in the direction
solves the stationary radiative transport boundary value problem
(1)
(1) In boundary value problem (Equation1
(1)
(1) ), the function
is the medium capability of absorption per unit path-length at z moving in the direction
called the attenuation coefficient, the function
is the scattering coefficient which accounts for particles from an arbitrary direction
which scatter in the direction
at a point z, and
is the incoming unit tangent sub-bundle of the boundary, with
being the outer unit normal at
. The attenuation and scattering coefficients are assumed known real valued functions. The boundary condition in (Equation1
(1)
(1) ) indicates that no radiation is coming from outside the domain. Throughout, the measure
on the unit sphere
is normalized to
.
The (forward) boundary value problem (Equation1(1)
(1) ) is known to be well-posed under various assumptions, e.g in [Citation3–6], with a general result in [Citation7] showing that, for an open and dense set of coefficients
and
, the boundary value problem (Equation1
(1)
(1) ) has a unique solution
for any
. In [Citation8], it is shown that for attenuation merely once differentiable,
and
, the boundary value problem (Equation1
(1)
(1) ) has a unique solution
for any
, p>1. Moreover, uniqueness result of the forward problem (Equation1
(1)
(1) ) are also establish in weighted
spaces in [Citation9], and in [Citation10,Citation11] using Carleman estimates.
In our reconstruction method here, some of our arguments require solutions ,
. We revisit the arguments in [Citation7,Citation8] and show that such a regularity can be achieved for sources
, p>4; see Theorem A.2 (iii) below.
For a given medium, i.e. a and k both known, we consider the inverse problem of determining the linear anisotropic source ; in particular, recovering the isotropic scalar field
and the vector field
from measurements
of exiting radiation on Γ,
(2)
(2) where
is the outgoing unit tangent sub-bundle of the boundary.
For anisotropic sources the problem has non-uniqueness [Citation4,Citation12–14]. One of our main result, Theorem 3.1 shows that from boundary measurement data , one can only recover the part of the linear anisotropic source
; in particular, only the solenoidal part
of the vector field
is recovered inside the domain. However, in Theorem 3.2, if one know apriori that the source
is divergence-free, then from the data
, one can recover both isotropic field
and the vector field
inside the domain. Moreover, instead of apriori information of the divergence-free source
, if one has the additional data
information along with the data
, then in Theorem 3.3, one can recover both sources
and
under subcritical assumption of the medium. One of the main crux in our reconstruction method is the observation that any finite Fourier content in the angular variable of the scattering kernel splits the problem into an infinite system of non-scattering case and a boundary value problem for a finite elliptic system. The role of the finite Fourier content has been independently recognized in [Citation15,Citation16].
The inverse source problem above has applications in medical imaging: In a non-scattering (k = 0) and non-attenuating (a = 0) medium the problem is mathematically equivalent to the one occurring in classical computerized X-ray tomography (e.g. [Citation17,Citation18]). In the absorbing non-scattering medium, such a problem (with only isotropic source ), appears in Positron/Single Photon Emission Tomography [Citation18,Citation19], and
with
, appears in Doppler Tomography [Citation18–20]. For applications in scattering media the inverse source problem formulated here is the two dimensional version of the corresponding three dimensional problem occurring in imaging techniques such as Bioluminescence tomography and Optical Molecular Imaging, see [Citation21–23] and references therein.
In this work, except for the results in the appendix, the attenuation coefficient are assumed isotropic , and that the scattering kernel
depends polynomially on the angle between the directions. Moreover, the functions a, k and the source f are assumed real valued.
In Section 2, we recall some basic properties of A-analytic theory, and in Section 3 we provide the reconstruction method for the full (part) of the linearly anisotropic source. Our approach is based on the Cauchy problem for a Beltrami-like equation associated with A-analytic maps in the sense of Bukhgeim [Citation17]. The A-analytic approach developed in [Citation17] treats the non-attenuating case, and the absorbing but non-scattering case is treated in [Citation24]. The original idea of Bukhgeim from the absorbing non-scattering media [Citation17,Citation24] to the absorbing and scattering media has been extended in [Citation8,Citation15]. In here we extend the results in [Citation8,Citation15] to linear anisotropic sources.
In Section 4, the method used will explain when the data coming from two different linear anisotropic field sources can be mistaken for each other.
In the appendix, we revisit the arguments in [Citation7,Citation8] and remark on the existence and regularity of the forward boundary value problem. The results in the appendix consider both attenuation coefficient and scattering kernel in general setting.
2. Ingredients from A-analytic theory
In this section, we briefly introduce the properties of A-analytic maps needed later, and introduce notation. We recall some of the existing results and concepts used in our reconstruction method.
For , p = 1, 2, we consider the Banach spaces:
(3)
(3) where
. Similarly, we consider
, and
.
For , we consider the Cauchy-Riemann operators
A sequence valued map
in
is called
-analytic (in the sense of Bukhgeim [Citation17]), if
(4)
(4) where L is the left shift operator
and
.
Bukhgeim's original theory [Citation17] shows that solutions of (Equation4(4)
(4) ), satisfy a Cauchy-like integral formula,
(5)
(5) where
is the Bukhgeim-Cauchy operator acting on
. We use the formula in [Citation25], where
is defined component-wise for
by
(6)
(6) Similar to the analytic maps, the traces of
-analytic maps on the boundary must satisfy some constraints, which can be expressed in terms of a corresponding Hilbert-like transform introduced in [Citation26]. More precisely, the Bukhgeim-Hilbert transform
is defined component-wise for
by
(7)
(7) and we refer to [Citation26] for its mapping properties.
Another ingredient, in addition to -analytic maps, consists in the one-to-one relation between solutions
satisfying
(8)
(8) and the
-analytic map
satisfying (Equation4
(4)
(4) ), via a special function h, see [Citation27, Lemma 4.2] for details. The function h is defined as
where
is the Radon transform of the attenuation a, and
is the classical Hilbert transform [Citation28]. The function h has vanishing negative Fourier modes yielding the expansions
for
Using the Fourier coefficients of
, define the sequence valued maps
and define the operators
component-wise for each
, by
(9)
(9) Note the commutating property
.
Lemma 2.1
[Citation29, Lemma 4.2]
Let ,
, and
be operators as defined in (Equation9
(9)
(9) ).
If
solves
, then
solves
.
Conversely, if
solves
, then
solves
.
3. Reconstruction of a sufficiently smooth linearly anisotropic source
For an isotropic real valued vector field and real map
, recall the boundary value problem (Equation1
(1)
(1) ):
(10)
(10) with an isotropic attenuation
, and with the scattering kernel
depending polynomially on the angle between the directions,
(11)
(11) for some fixed integer
. Note that, since
is both real valued and even in θ, the coefficient
is the
Fourier coefficient of
. Moreover
is real valued, and
.
For the real vector field , let
(12)
(12) and for
, a calculation shows that the linear anisotropic source
(13)
(13) We assume that the coefficients
are such that the forward problem (Equation10
(10)
(10) ) has a unique solution
for any
, p>1, see Theorem A.1. Moreover, we assume also an unknown source of a priori regularity
, p>4, and by Theorem A.2 part (iii), the solution
. Furthermore, the functions a, k and source f are assumed real valued, so that the solution u is also real valued.
Let be the formal Fourier series representation of the solution of (Equation10
(10)
(10) ) in the angular variable
. Since u is real valued, the Fourier modes
occurs in complex-conjugate pairs
, and the angular dependence is completely determined by the sequence of its nonpositive Fourier modes
(14)
(14) For the derivatives
in the spatial variable, the advection operator
. By identifying the Fourier coefficients of the same order, Equation (Equation10
(10)
(10) ) reduces to the system:
(15)
(15)
(16)
(16)
(17)
(17)
(18)
(18) where
as in (Equation12
(12)
(12) ).
By Hodge decomposition [Citation13], any vector field decomposes into a gradient field and a divergence-free (solenoidal) field :
(19)
(19) where
and
.
Note that for in (Equation12
(12)
(12) ), we have
(20)
(20) Using
, we have
and
Moreover, for
, the Hodge decomposition (Equation19
(19)
(19) ) can be rewritten as
(21)
(21) The following result show that from the knowledge of boundary data, one can only recover the part of the linear anisotropic source f; in particular, only the solenoidal part
of the vector field source
can be recovered inside Ω.
Theorem 3.1
Let be a strictly convex bounded domain, and Γ be its boundary. Consider the boundary value problem (Equation10
(10)
(10) ) for some known real valued
such that (Equation10
(10)
(10) ) is well-posed. If scalar and vector field sources
and
are real valued,
and
-regular, respectively, with p>4, then the data
defined in (Equation2
(2)
(2) ), uniquely determine the solenoidal part
in Ω. Moreover,
is also uniquely determined in Ω, where u is the solution of (Equation10
(10)
(10) ) and
is the zeroth Fourier mode of u in the angular variable.
Proof.
Let u be the solution of the boundary value problem (Equation10(10)
(10) ) and let
be the sequence valued map of its non-positive Fourier modes. Since the scalar field
, and isotropic vector field
, then the anisotropic source
belong to
for p>4. By applying Theorem A.2 (iii), we have
. Moreover, by the Sobolev embedding [Citation30],
with
, we have
, and thus, by [Citation26, Proposition 4.1 (ii)], the sequence valued map
.
Since , then by compact imbedding of Sobolev spaces [Citation30],
. By Hodge decomposition (Equation19
(19)
(19) ), field
with
and
.
We note from (Equation18(18)
(18) ) that the shifted sequence valued map
solves
(22)
(22) Let
. By Lemma 2.1, the system (Equation22
(22)
(22) ) becomes
, i.e
is
analytic.
By (Equation2(2)
(2) ), the data
determines the sequence valued map
on Γ. By Proposition Equation9
(9)
(9) (iii), and the convolution formula (Equation9
(9)
(9) ), traces
determines the traces
on Γ.
Since is the boundary value of an
-analytic function in Ω, then [Citation26, Theorem 3.2 (i)] yields
(23)
(23) where
is the Bukhgeim-Hilbert transform in (Equation7
(7)
(7) ).
From on Γ, we use the Bukhgeim-Cauchy Integral formula (Equation6
(6)
(6) ) to construct the sequence valued map
inside Ω:
(24)
(24) By [Citation29, Proposition 2.3] and [Citation26, Theorem 3.2 (ii)], the constructed sequence valued map
is
-analytic in Ω.
From the convolution formula (Equation9(9)
(9) ), we construct the sequence valued map
(25)
(25) Thus, determining
inside Ω for
. In particular, we recover modes
.
Recall that the modes satisfy
(26a)
(26a)
(26b)
(26b) By applying
to (Equation26a
(26a)
(26a) ), the mode
(for
) is then the solution to the Dirichlet problem for the Poisson equation
(27a)
(27a)
(27b)
(27b) where the right hand side of (Equation27a
(27a)
(27a) ) is known. We solve repeatedly (Equation27a
(27a)
(27a) ) for
in (Equation26a
(26a)
(26a) ), to recover the modes
(28)
(28) From determined
in (Equation25
(25)
(25) ) and modes
in (Equation28
(28)
(28) ), the sequence
is determined in Ω. Thus
is determined in Ω.
Since , we can take
on both sides of Equation (Equation16
(16)
(16) ) to get
(29)
(29) where in the last equality we use (Equation20
(20)
(20) ).
Moreover, since is real valued and
, by equating the real part in (Equation29
(29)
(29) ) yields the boundary value problem:
(30a)
(30a)
(30b)
(30b) where the right hand side of (Equation30a
(30a)
(30a) ) is known. Thus, real valued function
is recovered in Ω, by solving the Dirichlet problem for the above Poisson Equation (Equation30a
(30a)
(30a) ).
Even though is not determined, the function
is uniquely determined in Ω. Moreover, modes
and
are also uniquely determined in Ω. Furthermore, using expression of
from (Equation16
(16)
(16) ) and
from (Equation21
(21)
(21) ), we define
(31)
(31) with
satisfying
Thus, the solenoidal part , of the vector field
is recovered in Ω.
If we know apriori that the vector field is incompressible (i.e divergenceless), then we can reconstruct both scalar field source
and vector field source
in Ω.
Theorem 3.2
Let be a strictly convex bounded domain, and Γ be its boundary. Consider the boundary value problem (Equation10
(10)
(10) ) for some known real valued
such that (Equation10
(10)
(10) ) is well-posed. If the unknown scalar field source
and divergenceless vector field sources
are real valued,
and
-regular, respectively, with p>4, then the data
defined in (Equation2
(2)
(2) ) uniquely determine both
and
in Ω.
Proof.
Let u be the solution of the boundary value problem (Equation10(10)
(10) ) and let
be the sequence valued map of its non-positive Fourier modes, Since the isotropic scalar and vector field
, and
respectively for p>4, then the anistropic source
and by applying Theorem A.2 (iii), we have
. By the Sobolev embedding [Citation30],
with
, we have
, and thus, by [Citation26, Proposition 4.1 (ii)],
.
Since , then by compact imbedding of Sobolev spaces [Citation30],
. By Hodge decomposition (Equation19
(19)
(19) ), field
with
and
.
If we know apriori that the vector field is incompressible (i.e divergenceless
). Then
and
implies
inside Ω. Thus, vector field
inside Ω.
By Theorem 3.1, the data uniquely determine the solenoidal field
in Ω by Equation (Equation31
(31)
(31) ) with
, and the sequence valued map
in Ω. Moreover, the real valued mode
is also then recovered (with
) in Ω, by solving the Dirichlet problem for the Poisson Equation (Equation30a
(30a)
(30a) ).
Thus, from modes and
, the scalar field
is also recovered in Ω by
(32)
(32)
In the radiative transport literature, the attenuation coefficient , where
represents pure loss due to absorption and
is the isotropic part of scattering kernel. We consider the subcritical region:
(33)
(33)
Remark 3.1
In addition to the hypothesis to Theorem 3.1, if we assume that coefficients satisfies (Equation33
(33)
(33) ), then in the region
, one can recover explicitly the entire vector field
. Indeed, Equation (Equation15
(15)
(15) ) gives
and, following (Equation16
(16)
(16) ), the vector field
can be recovered by the formula
(34)
(34)
Next, we show that one can also determine both scalar field and vector field
, if one has the additional data
(or
) information, instead of
being incompressible as in Theorem 3.2.
Theorem 3.3
Let be a strictly convex bounded domain, and Γ be its boundary. Consider the boundary value problem (Equation10
(10)
(10) ) for some known real valued
such that (Equation10
(10)
(10) ) is well-posed. If the unknown scalar field source
and vector field source
are real valued,
and
-regular, respectively, with p>4, and coefficients
satisfying (Equation33
(33)
(33) ), then the data
and
defined in (Equation2
(2)
(2) ) uniquely determine both
and
in Ω.
Proof.
Let u be the solution of the boundary value problem (Equation10(10)
(10) ) and let
be the sequence valued map of its non-positive Fourier modes. Since the scalar field
, and isotropic vector field
, then the anisotropic source
belong to
for p>4. By applying Theorem A.2 (iii), we have
. Moreover, by the Sobolev embedding,
with
, we have
, and thus, by [Citation26, Proposition 4.1 (ii)], the sequence valued map
.
We consider the boundary value problems
(35)
(35)
(36)
(36) Then u = v + w satisfy the boundary value problem (Equation10
(10)
(10) ).
We consider first the boundary value problem (Equation35(35)
(35) ), and reconstruct the scalar field
from the given boundary data
as follows.
If is the Fourier series expansion in the angular variable
of a solution v of boundary value problem (Equation35
(35)
(35) ), then, by identifying the Fourier modes of the same order, (Equation35
(35)
(35) ) reduces to the system:
(37)
(37)
(38)
(38)
(39)
(39) Let
be the sequence valued map of its non-positive Fourier modes. By Theorem 3.1, the data
, uniquely determine the sequence
in Ω. Moreover, as (Equation38
(38)
(38) ) holds also for n = 0 (
in this case), the mode
is also determined in Ω by solving the Dirichlet problem for the Poisson equation
(40a)
(40a)
(40b)
(40b) where the right hand side of (Equation40a
(40a)
(40a) ) is known. Thus, using modes
and
, the isotropic scalar source
is recovered in Ω by
(41)
(41) Next, we consider the boundary value problem (Equation36
(36)
(36) ), and reconstruct the vector field
from the given boundary data
as follows.
If is the Fourier series expansion in the angular variable
of a solution w of the boundary value problem (Equation36
(36)
(36) ), then (Equation36
(36)
(36) ) reduces to the system:
(42)
(42)
(43)
(43)
(44)
(44)
(45)
(45) Let
be the sequence valued map of its non-positive Fourier modes. By Theorem 3.1, the data
, uniquely determine the sequence
in Ω.
Using the subcriticality condition (Equation33(33)
(33) ):
we define via (Equation42
(42)
(42) ):
(46)
(46) Using determined modes
in
and mode
from (Equation42
(42)
(42) ), the real valued vector field
is recovered in Ω by
(47)
(47)
4. When can the data coming from two sources be mistaken for each other ?
In this section, we show when the data coming from two different linear anisotropic field sources can be mistaken for each other.
In Theorem 4.1 below the data are assuming the same attenuation a and scattering coefficient k.
Theorem 4.1
Let
,
be real valued, with
, and
, p>4 be real valued with
. Then
is a real valued vector field such that the data
coming from the linear anisotropic source
, is the same as data
coming from a different linear anisotropic source
Let
be real valued with
. Assume that there are real valued linear anisotropic sources
and
, with isotropic fields
, p>4, and vector fields
, p>4. If the data
of the linear anisotropic source f equals the data
of the linear anisotropic source
. Then
Proof.
(i) Assume is the data of some real valued anisotropic source
, i.e. it is the trace on
of solution w to the stationary transport boundary value problem:
(48)
(48) where the operator
, for
and
.
Using the subcriticality condition (Equation33(33)
(33) ):
with
, and isotropic real valued functions ψ and
, we note:
(49)
(49) where second equality use the fact that both ψ and
are angularly independent functions.
Let and
. Then
where the second equality uses the linearty of K and (Equation49
(49)
(49) ), the third equality uses (Equation48
(48)
(48) ), and the last equality uses the definition of
. Moreover, since
vanishes on Γ, we get
(ii) For isotropic scalar fields
, p>4, and vector fields
, p>4, the real valued linear anisotropic sources
, and
, p>4.
Let the data equals data
i.e.
Consider the corresponding boundary value problems
(50a)
(50a)
(50b)
(50b) respectively, subject to
(50c)
(50c) Since
, Theorem A.2 (iii), yields solutions
. Moreover, by the Sobolev embedding,
with
.
The corresponding sequences of non-positive Fourier modes of u satisfy
(51)
(51)
(52)
(52)
(53)
(53)
(54)
(54) whereas the non-positive Fourier modes
of w satisfy
(55)
(55)
(56)
(56)
(57)
(57)
(58)
(58) Since the boundary data g is the same
, we also have the sequence valued map
(59)
(59) Moreover, by [Citation26, Proposition 4.1 (ii)], the sequence
with
.
Claim 4.1
The above systems subject to boundary condition (Equation59(59)
(59) ) yields
(60)
(60) inside Ω.
Proof of Claim 4.1.
We first show that the systems (Equation54(54)
(54) ) and (Equation58
(58)
(58) ) subject to (Equation59
(59)
(59) ) yields
(61)
(61) From (Equation54
(54)
(54) ) and (Equation58
(58)
(58) ), the shifted sequence valued maps
and
, respectively, solves
(62)
(62) From
in (Equation59
(59)
(59) ), we use the Bukhgeim-Cauchy Integral formula (Equation6
(6)
(6) ) to construct the sequence valued map
and
inside Ω:
(63)
(63) where
is the operator in (Equation9
(9)
(9) ).
By [Citation29, Proposition 2.3] and [Citation26, Theorem 3.2 (ii)], are
-analytic in Ω:
(64)
(64) and also coincide at the boundary Γ. By uniqueness of
-analytic functions with a given trace, they coincide inside:
(65)
(65) Using the operator
in (Equation9
(9)
(9) ), we construct the sequence valued map
(66)
(66) and conclude that (Equation61
(61)
(61) ) holds.
Moreover, by Lemma 2.1, the sequences and
in (Equation66
(66)
(66) ) satisfies (Equation62
(62)
(62) ).
Next, we show that the systems (Equation53(53)
(53) ) and (Equation57
(57)
(57) ) subject to boundary condition (Equation59
(59)
(59) ) yield
(67)
(67) inside Ω.
Define the function
(68)
(68) Since the boundary data g is the same, Equation (Equation50c
(50c)
(50c) ) yields
(69)
(69) From (Equation61
(61)
(61) ), we note that
(70)
(70) By subtracting system (Equation57
(57)
(57) ) from (Equation53
(53)
(53) ), and using (Equation68
(68)
(68) ) and (Equation69
(69)
(69) ), yields the boundary value problem
(71a)
(71a)
(71b)
(71b)
Note that for j = 1 in (Equation71a(71a)
(71a) ), the right hand side of (Equation71a
(71a)
(71a) ) contains modes
and
which are both zero inside Ω by (Equation70
(70)
(70) ). Thus, for j = 1, the mode
satisfy the Cauchy problem for the
-equation,
(72a)
(72a)
(72b)
(72b) The unique solution of the above Cauchy problem is
. Therefore, resulting
.
We then solve repeatedly (Equation71a(71a)
(71a) ) starting for
, where the right hand side of (Equation71a
(71a)
(71a) ) in each step is zero, yielding the Cauchy problem (Equation72a
(72a)
(72a) ) for each subsequent modes, and thus, resulting in the recovering of the modes
in Ω. Hence, establishing (Equation67
(67)
(67) ).
From (Equation61(61)
(61) ) and (Equation67
(67)
(67) ), we establish (Equation60
(60)
(60) ), and thus Claim 4.1.
By subtracting (Equation55(55)
(55) ) from (Equation51
(51)
(51) ), and using (Equation60
(60)
(60) ) yields
Using
with
, yields
(73)
(73) Moreover, by subtracting (Equation56
(56)
(56) ) from (Equation52
(52)
(52) ), and using (Equation60
(60)
(60) ) yields
Since both
and
are real valued we have from (Equation73
(73)
(73) ):
Remark 4.1
Note that in Theorem 4.1(i), the assumption on scattering kernels of finite Fourier content in the angular variable is not assumed, and the result holds for a general scattering kernels which depends polynomially on the angle between the directions.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Additional information
Funding
References
- Cercignani C. The boltzmann equation and its applications. Berlin: Springer-Verlag; 1988.
- Chandrasekhar S. Radiative transfer. New York: Dover Publ; 1960.
- Anikonov DS, Kovtanyuk AE, Prokhorov IV. Transport equation and tomography. Brill Academics; 2002. (Inverse and Ill-Posed Problems Series; 30).
- Choulli M, Stefanov P. An inverse boundary value problem for the stationary transport equation. Osaka J Math. 1999;36:87–104. https://projecteuclid.org/journals/osaka-journal-of-mathematics/volume-36/issue-1/An-inverse-boundary-value-problem-for-the-stationary-transport-equation/ojm/1200788449.pdf
- Dautray R, Lions J-L. Mathematical analysis and numerical methods for science and technology. Vol. 4. Berlin: Springer-Verlag; 1990.
- Mokhtar-Kharroubi M. Mathematical topics in neutron transport theory. Singapore: World Scientific; 1997.
- Stefanov P, Uhlmann G. An inverse source problem in optical molecular imaging. Anal PDE. 2008;1(1):115–126. doi: 10.2140/apde
- Fujiwara H, Sadiq K, Tamasan A. A two dimensional source reconstruction method in radiative transport using boundary data measured on an arc. Inverse Probl. 2021;37:115005. doi: 10.1088/1361-6420/ac2d75
- Egger H, Schlottbom M. An Lp theory for stationary radiative transfer. Appl Anal. 2014;93(6):1283–1296. doi: 10.1080/00036811.2013.826798
- Klibanov MV, Li J, Nguyen LH, et al. Convexification numerical method for a coefficient inverse problem for the radiative transport equation. SIAM J Img Sci Comput. 2023;16:35–63. doi: 10.1137/22M1509837
- Smirnov AV, Klibanov MV, Nguyen LH. On an inverse source problem for the full radiative transfer equation with incomplete data. SIAM J Sci Comput. 2019;41(5):B929–B952. doi: 10.1137/19M1253605
- Bal G, Tamasan A. Inverse source problems in transport equations. SIAM J Math Anal. 2007;39:57–76. doi: 10.1137/050647177
- Sharafutdinov VA. Integral geometry of tensor fields. Utrecht: VSP; 1994.
- Tamasan A. Tomographic reconstruction of vector fields in variable background media. Inverse Probl. 2007;23:2197–2205. doi: 10.1088/0266-5611/23/5/022
- Fujiwara H, Sadiq K, Tamasan A. A Fourier approach to the inverse source problem in an absorbing and anisotropic scattering medium. Inverse Probl. 2019;36(1):015005. doi: 10.1088/1361-6420/ab4d98
- Monard F, Bal G. Inverse source problems in transport via attenuated tensor tomography, arXiv:1908.06508v1.
- Bukhgeim AL. Inversion formulas in inverse problems, chapter in linear operators and Ill-posed problems by Lavrentiev MM, and Savalev LY. New York: Plenum; 1995.
- Natterer F. The mathematics of computerized tomography. New York: Wiley; 1986.
- Natterer F, Wübbeling F. Mathematical methods in image reconstruction. SIAM monographs on mathematical modeling and computation. Philadelphia, PA: SIAM; 2001.
- Sparr G, Stråhlén K, Lindström K, et al. Doppler tomography for vector fields. Inverse Probl. 1995;11:1051–1061. doi: 10.1088/0266-5611/11/5/009
- Han W, Eichholz JA, Huang J. RTE-based bioluminescence tomography: a theoretical study. Inverse Probl Sci Eng. 2011;19(4):435–459. doi: 10.1080/17415977.2010.500383
- Klose AD, Ntziachristos V, Hielscher AH. The inverse source problem based on the radiative transfer equation in optical molecular imaging. J Comput Phys. 2005;202:323–345. doi: 10.1016/j.jcp.2004.07.008
- Yi HC, Sanchez R, McCormick NJ. Bioluminescence estimation from ocean in situ irradiances. Appl Opt. 1992;31:822–830. doi: 10.1364/AO.31.000822
- Arbuzov EV, Bukhgeim AL, Kazantsev SG. Two-dimensional tomography problems and the theory of A-analytic functions. Siberian Adv Math. 1998;8:1–20.
- Finch DV. The attenuated x-ray transform: recent developments, In: Inside out: inverse problems and applications, Math. Sci. Res. Inst. Publ., 47, Cambridge: Cambridge Univ. Press; 2003. p. 47–66.
- Sadiq K, Tamasan A. On the range of the attenuated radon transform in strictly convex sets. Trans Amer Math Soc. 2015;367(8):5375–5398. doi: 10.1090/tran/2015-367-08
- Sadiq K, Scherzer O, Tamasan A. On the X-ray transform of planar symmetric 2-tensors. J Math Anal Appl. 2016;442(1):31–49. doi: 10.1016/j.jmaa.2016.04.018
- Muskhelishvili NI. Singular integral equations. New York: Dover; 2008.
- Sadiq K, Tamasan A. On the range characterization of the two dimensional attenuated Doppler transform. SIAM J Math Anal. 2015;47(3):2001–2021. doi: 10.1137/140984282
- Adam R. Sobolev spaces. New York: Academic Press; 1975.
Appendix A.
Some remarks on the regularity of the forward problem
In this section, we revisit the arguments in [Citation7,Citation8], and remark on the well posedness in of the boundary value problem (Equation1
(1)
(1) ).
The results in appendix consider both attenuation coefficient and scattering kernel in general setting. Adopting the notation in [Citation7,Citation8], we consider the operators
where the intervening functions are extended by 0 outside Ω.
Using the above operators, the boundary value problem (Equation1(1)
(1) ) can be rewritten as
(A1)
(A1) If the operator
is invertible, then the problem (EquationA1
(A1)
(A1) ) is uniquely solvable, and has the form
. By using the formal expansion
(A2)
(A2) We recall some results in [Citation8].
Proposition A.1
[Citation8, Proposition 2.1]
Let and
. Then the operator
(A3)
(A3)
Theorem A.1
[Citation8, Theorem 2.1]
Let p>1, , and
. At least one of the following statements is true.
is invertible in
.
there exists
such that
is invertible in
, for any
.
For our main Theorems, we require , p>4 and such a regularity can be achieved for sources
, p>4; see Theorem A.2 (iii) below. We refer to [Citation8, Theorem 2.2] for part (i) and (ii) of Theorem A.2, and we include the proof here.
The regularity of the solution u of (Equation1(1)
(1) ) increases with the regularity of f as follows.
Theorem A.2
Consider the boundary value problem (Equation1(1)
(1) ) with
. For p>1, let
be such that
is invertible in
, and let
in (EquationA2
(A2)
(A2) ) be the solution of (Equation1
(1)
(1) ).
If
, then
.
If
, then
.
If
, then
.
Proof.
(i) We consider the regularity of the solution u of (Equation1(1)
(1) ) term by term as in (EquationA2
(A2)
(A2) ). It is easy to see that the operator
preserve the space
, and also the operator K preserve the space
, so that the first two terms,
and
, both belong to
. Moreover,
, and now, by using Proposition A.1, the last term is also belong in
.
(ii) We define the following operators
(A4)
(A4) where
and
for i, j = 1, 2.
It is easy to see that and
preserve
.
By evaluating the radiative transport equation in (Equation1(1)
(1) ) at
and integrating in t from
to 0, the boundary value problem (Equation1
(1)
(1) ) with zero incoming fluxes is equivalent to the integral equation:
(A5)
(A5) According to part (i), for
, the solution
, and so
. In particular
solves the integral equation:
(A6)
(A6) Moreover, since
,
, and
, the right-hand-side of (EquationA6
(A6)
(A6) ) lies in
. By applying part (i) above, we get that the unique solution to (EquationA6
(A6)
(A6) )
(A7)
(A7) For
, also according to part (i),
. In particular
is the unique solution of the integral equation
(A8)
(A8) which is of the type (EquationA5
(A5)
(A5) ) with K = 0. Moreover, since
, and, according to (EquationA7
(A7)
(A7) ),
, the right-hand-side of (EquationA8
(A8)
(A8) ) lies in
. Again, by applying part (i), we get
Thus,
.
(iii) For , according to part (ii),
, and
. In particular
is the unique solution of the integral equation
(A9)
(A9) Moreover, since
,
, and
, the right-hand-side of (EquationA9
(A9)
(A9) ) lies in
. By applying part (i) above, we get that the unique solution to (EquationA9
(A9)
(A9) )
(A10)
(A10) For
, also according to part (ii),
, and
. In particular
is the unique solution of the integral equation
(A11)
(A11) which is of the type (EquationA5
(A5)
(A5) ) with K = 0.
Moreover, since , and, according to (EquationA10
(A10)
(A10) ),
, the right-hand-side of (EquationA11
(A11)
(A11) ) lies in
. Again, by applying part (i), we get
(A12)
(A12) For
, also according to part (ii),
. In particular
is the unique solution of the integral equation
(A13)
(A13) which is of the type (EquationA5
(A5)
(A5) ). Moreover, since
,
, and
, the right-hand-side of (EquationA13
(A13)
(A13) ) lies in
. Again, by applying part (i), we get
(A14)
(A14) From (EquationA10
(A10)
(A10) ), (EquationA12
(A12)
(A12) ), and (EquationA14
(A14)
(A14) ), we get
.
We remark that for Theorem A.2 part (i) we only need and
, and we only require
and
for Theorem A.2 part (ii). Moreover, in a similar fashion, one can show that under sufficiently increased regularity of a and k, the solution u of (Equation1
(1)
(1) ) belong to
for
, provided
.