Open Mathematics 2024; 22: 20240049

Research Article

Yu Qiao and Xiangtuan Xiong*

A quasi-reversibility method for solving

nonhomogeneous sideways heat equation
https://doi.org/10.1515/math-2024-0049
received September 25, 2023; accepted July 25, 2024

Abstract: We consider solving a severely ill-posed problem for determining the surface temperature and heat
flux distribution of the nonhomogeneous sideways heat equation by a quasi-reversibility regularization
method. An analytical solution is deduced based on the Fourier transform, and then, the logarithmic-type
error estimate for the regularized solution is achieved. Finally, three numerical examples, including smooth
and non-smooth functions, validate the feasibility and effectiveness of the proposed approach.

Keywords: ill-posed problem, nonhomogeneous sideways heat equation, a quasi-reversibility method, error
estimate

MSC 2020: 35R25, 35R30, 65J20, 65M30

1 Introduction
The sideways heat equation (SHE), which is a Cauchy problem for the heat equation, has attracted lots of
researchers in the last few decades. It is a model of a problem where one wants to determine the temperature
or heat flux on both sides of a thick wall, but where one side is inaccessible to measurements. Under these
circumstances, the missing information is compensated for by additional observations made through mea-
surement on the complementary accessible parts of the boundary of the body. SHE has many applications,
such as the accurate determination of the temperature spike in a cannon at firing [1], continuous data
monitoring [2], and re-entry vehicles [3].
The homogeneous SHE is formulated as follows:

⎧ut − uxx = 0, x > 0, t > 0,

⎪u(x , 0) = 0, x ≥ 0,
⎨ (1.1)
u(1, t ) = h(t ) , t ≥ 0,
⎪
⎩u(x , t )∣x →∞ bounded.

The goal here is to seek temperature distribution for 0 ≤ x < 1 from the measurement data h(t ). In general, this
problem is ill-posed in the sense of Hadamard [4], i.e., any small errors in the observation data will cause large
deviations in the estimate quantities. As a consequence, its solution does not meet the general requirement of
existence, uniqueness, and stability under small perturbation to the input data. Then, regularization techni-
ques [5] are required to make it a well-posed problem, e.g., Tikhonov method [6], spectral method [7], quasi-
reversibility method [8–11], quasi-boundary value method and iteration method [12], wavelet method [13–16],
multi-resolution method [17], and wavelet-Galerkin method [18,19]. Besides, it has been shown that difference


* Corresponding author: Xiangtuan Xiong, Department of Mathematics, Northwest Normal University, Lanzhou, P. R. China,
e-mail: [email protected]
Yu Qiao: Department of Mathematics, Northwest Normal University, Lanzhou, P. R. China

Open Access. © 2024 the author(s), published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0
International License.
2  Yu Qiao and Xiangtuan Xiong

schemes have a regularizing effect [20], so Eldén and Guo et al. [21–23] employed the finite difference method
to deal with this ill-posed problem.
Still in the homogeneous situation, Xiong et al. addressed the difference regularization method [24]
and the Fourier method [25] to discuss the non-standard inverse heat conduction problem
⎧ut + ux = uxx , x > 0, t > 0,
⎪u(x , 0) = 0, x ≥ 0,
⎨ (1.2)
u(1, t ) = φ(t ) , t ≥ 0,
⎪
⎩u(x , t )∣x →∞ bounded.
This article acquired stable estimates of temperature and flux distribution and used numerical example
to show that the computational effect of their methods is satisfactory. Furthermore, Xiong and Fu [26] inves-
tigated a more general problem
⎧ut (x , t ) = a (x )uxx + b(x )ux + c (x ) , x > 0, t > 0,
⎪u(x , 0) = 0, x ≥ 0,
⎨ (1.3)
u(1, t ) = ψ(t ) , t ≥ 0,
⎪
⎩u(x , t )∣x →∞ bounded.
In this work, the authors adopted the wavelet dual least-squares method to determine surface temperature
and heat flux and proved the error estimates between the approximate solution and the exact one are order
optimal. Subsequent to [26], Liu and Deng et al. [27–29] used the method of iteration to compute the approx-
imate solution of (1.3) and showed that their solution matched the exact solution well. Furthermore, there are
also some articles (see, e.g., [30–34] and references therein) that study the sideways problems in two-dimen-
sional and higher-dimensional space.
However, severe ill-posedness and the existence of the nonhomogeneous term f (x , t ) make the use of a
priori bound E more troublesome. Therefore, it is much more difficult to solve SHE in the nonhomogeneous
case than in the homogeneous case. In particular, the determination of heat flux is scarcely discussed. To the
knowledge of the author, there are still very few results on the Cauchy problem of the heat conduction
equation in the nonhomogeneous case. Liu and Chang [35] solved a three-dimensional nonhomogeneous
SHE in a cuboid by a Fourier sine series method, and the analysis of the regularization parameter and the
stability of the solution were worked out. This study points out that their method is quite accurate but does not
give an error estimate of the regularized solution. Luan and Khanh [36] investigated the two-dimensional
inhomogeneous heat equation in the presence of a general source term and proposed a kernel regularization
scheme to recover the temperature and thermal flux distribution from the measured data. However, an error
estimate at the boundary is not given. In the present study, we are devoted to extending the work [10] and [11]
to a more general case, i.e., the following nonhomogeneous system:
⎧ut − uxx = f (x , t ) , x > 0, t > 0,
⎪u(x , 0) = 0, x ≥ 0,
⎨ (1.4)
u(1, t ) = g (t ) , t ≥ 0,
⎪
⎩ u ( x , t )∣x →∞ bounded,
where the function f (x , t ) is the heat source density. There are many ways to solve partial differential
equations [37,38]. Here, we aim at obtaining an analytical solution to (1.4) via Fourier transform techniques.
Due to the ill-posedness of the problem, a fourth-order modified method is derived for the stable reconstruc-
tion of the solution. Moreover, under some appropriate a priori assumptions, the logarithmic-type error
estimate for the temperature and flux regularized solutions in the whole domain is obtained, which proves
the usefulness of the quasi-reversibility regularization method. The quasi-reversibility method was first intro-
duced by Lattès and Lions [39], and since then, the idea of this method has been successfully used for solving
various types of ill-posed problems. Throughout this article, we assume that fˆ ∈ L2 (0, 1; L2 ()) satisfies
1

∫∣fˆ (s, ξ )∣2 ds < e−3 ∣ξ ∣ , ∀ξ ∈  . (1.5)

0
 Quasi-reversibility method for nonhomogeneous sideways heat equation  3

2 Mathematical analysis and quasi-reversibility method

In order to apply the Fourier transform, we extend all the functions in this article to the whole line −∞ < t < ∞
by making them zero outside the original domain, and our theoretical analysis will be performed in L2 ().
Let ĝ denote the Fourier transform of g (t ) defined by
∞
1
gˆ (ξ ) ≔
2π
∫ g (t )e−iξtdt.
−∞

We also define the following norms:

1
∞
⎡ ⎤2
∫
‖u(0, ⋅ )‖p ≔ ⎢ (1 + ξ 2 ) p∣uˆ (0, ξ )∣2 dξ ⎥ ,
⎣−∞ ⎦
1
⎛1 ⎞2
‖f ‖L2(0,1; Hp()) ∫
≔ ⎜ ‖f (x , ⋅ )‖2p dx ⎟ .
⎝0 ⎠

When p = 0, ‖ ⋅ ‖p = ‖ ⋅ ‖ denotes the L2 () norm.

Applying the Fourier transform with respect to t to both sides of (1.4), we obtain the following second-
order ordinary differential equation in the frequency space:

⎧uˆxx (x , ξ ) − iξuˆ (x , ξ ) = − fˆ (x , ξ ) , ξ ∈  ,
⎪
⎨uˆ (1, ξ ) = gˆ (ξ ) , ξ ∈ , (2.1)
⎪
⎩uˆ (x , ξ )∣x →∞ bounded, ξ ∈ .

The solution of (2.1) is given by

1
fˆ (s , ξ ) sinh(θ (ξ )(s − x ))
uˆ (x , ξ ) = e θ(ξ )(1 − x )gˆ (ξ ) + ∫ θ (ξ )
ds , 0 ≤ x < 1, (2.2)
x

and equivalently,
∞ 1
1 ⎛ fˆ (s , ξ ) sinh(θ (ξ )(s − x )) ⎞ iξt
u (x , t ) =
2π
∫ ⎜e
θ (ξ )(1 − x )gˆ (ξ ) +
∫ θ (ξ )
ds ⎟e dξ , 0 ≤ x < 1, (2.3)
−∞ ⎝ x ⎠

where

∣ξ∣
θ (ξ ) ≔ iξ = (1 + σi ) , σ = sign(ξ ) , ∀ξ ∈  . (2.4)
2

Remark 2.1. When f = 0, (2.2) and (2.3) give the following solution:

uˆ (x , ξ ) = e θ(ξ )(1 − x )gˆ (ξ ) , 0 ≤ x < 1,

∞
1
u (x , t ) =
2π
∫ eθ(ξ )(1−x )gˆ (ξ )eiξtdξ , 0 ≤ x < 1,
−∞

which is the solution of problem (1.1) [40]. Thus, the solution of nonhomogeneous problem is consistent with
its homogeneous system.

sinh(θ )
Remark 2.2. In (2.2), since θ is an exponentially increasing function, in order to obtain a valid estimate, we
need to assume that fˆ is a rapidly decreasing function. Condition (1.5) is therefore reasonable and will be used
frequently afterward.
4  Yu Qiao and Xiangtuan Xiong

Note that the real part of θ (ξ ) is an increasing positive function of ξ . Therefore, the terms e θ(ξ )(1 − x ) and
sinh(θ (ξ )(s − x ))
θ (ξ )
are unbounded as ∣ξ ∣ → ∞. This means small errors in the data can blow up and completely
destroy the solution for 0 ≤ x < 1 . Thus, the Cauchy problem of the nonhomogeneous SHE is severely ill-posed.
Next, we apply a quasi-reversibility regularization method to construct an approximate solution for (1.4).
Here, the quasi-reversibility method is to find the solution of the following problem:

⎧utα, δ = uxxα, δ α, δ
− α2uxxtt + f δ (x , t ) , x > 0, t > 0,
⎪uα, δ (x , 0) = 0, x ≥ 0,
⎨ α, δ (2.5)
δ
⎪u (1, t ) = g (t ) , t ≥ 0,
α , δ
⎩u (x , t )∣x →∞ bounded,

where α > 0 is a regularization parameter and the measured data (g δ , f δ ) ∈ L2 () × L2 ((0, 1) ; L2 ()) satisfy

max{‖g δ − g‖, ‖f δ − f ‖L2(0,1; L2()) } ≤ δ , (2.6)

where δ > 0 is called an error level. Similarly, applying the Fourier transform with respect to t to both sides
of (2.5), the formal solution of problem (2.5) in the frequency space can be obtained as follows:
1 δ
fˆ (s , ξ ) sinh(θ˜(ξ )(s − x ))
uˆ α, δ (x , ξ) = e θ˜(ξ )(1 − x )gˆ δ (ξ ) + ∫ θ˜(ξ )
ds , 0 ≤ x < 1, (2.7)
x

where
1
θ˜(ξ ) ≔ θ (ξ )θ 1 and θ 1 = ≤ 1. (2.8)
1 + α2ξ 2

Lemma 2.1. [8] For arbitrary β ≥ 0, z1 , z2 ∈  , the following inequalities hold:

1 − e −β ≤ β , (2.9)

∣1 − e ±βi∣ ≤ β , (2.10)

∣sinh(z1) − sinh(z2 )∣ ≤ ∣e z1 − e z2∣ . (2.11)

Lemma 2.2. [11] The function

∣ξ ∣ j
h j ( ∣ξ∣ ) = e − 2 x ∣ξ∣2 ,
j
= ⎛ xe ⎞ at ∣ξ∣ =
j j 2 j 2
obtains the maximum h max , here j = 4, 5, 6.
⎝ ⎠ x

Lemma 2.3. [36] For arbitrary z ∈  , x ∈ (0, 1], and η ∈ [0, x ), we have
sinh((x − η)z)
≤ (x − η)e (x − η) ∣ z ∣, (2.12)
z

∣cosh(xz)∣ ≤ e x R(z) ≤ e x ∣ z ∣, (2.13)

where R(z) denotes the real parts of z.

 Quasi-reversibility method for nonhomogeneous sideways heat equation  5

3 Determination of temperature structure and error estimate

So as to acquire a more sharp convergence, we use the following a priori bound on the exact solution:
‖u(0, ⋅ )‖ ≤ E , (3.1)

and the regularization parameter α is chosen as

1
α= 2.
(3.2)
2⎛ ln δ ⎞
E
⎝ ⎠

1 fˆ sinh(θs )
Lemma 3.1. If conditions (1.5) and (3.1) hold, let A(ξ ) = uˆ (0, ξ ) − ∫0 θ
ds , then
‖A(ξ )‖ ≤ E + M1,

where M1 is a constant.

Proof. Using successively the triangle inequality, (3.1), the Cauchy-Schwarz integral inequality, (2.12) and (1.5),
we obtain
1
fˆ sinh(θs )
‖A(ξ )‖ ≤ ‖uˆ (0, ξ )‖ + ∫ θ
ds
0
1
⎡ ∞ ⎛ 1 sinh(θs ) 2 1
⎞ ⎤
2

≤E+⎢ ⎜ ∫∫ θ
∫
ds ∣fˆ ∣2 ds ⎟dξ ⎥
⎣−∞⎝ 0 0 ⎠ ⎦
1
⎡∞⎛1 1
⎞ ⎤
2

∫∫ s ∣ θ ∣ 2 ˆ 2
∫
≤ E + ⎢ ⎜ ∣se ∣ ds ∣f ∣ ds ⎟dξ ⎥
⎣−∞⎝ 0 0 ⎠ ⎦
1
∞
⎛ 1 ⎞2
≤E+⎜ ∫ e− ∣ ξ ∣ dξ .
⎟
⎝−∞ 2 ∣ξ∣ ⎠

It is easy to know that the generalized integral on the right-hand side of the last inequality converges,
here we introduce the notation
1
∞
⎛ 1 ⎞2
M1 ≔ ⎜ ∫ e− ∣ ξ ∣ dξ .
⎟
⎝−∞ 2 ∣ξ∣ ⎠

Therefore,
‖A(ξ )‖ ≤ E + M1,

where M1 is a constant. □

Theorem 3.1. Let uˆ (x , ξ ) given by (2.2) be the exact solution of problem (1.4) in the frequency space, uˆ α, δ (x , ξ )
given by (2.7) be the exact solution of problem (2.5) in the frequency space. Conditions (1.5), (2.6), and (3.1) hold,
and the regularization parameter α is chosen by (3.2). Then, for a fixed x ∈ (0, 1), we have

⎛ ⎞
1 C1 C2
‖uα, δ (x , ⋅ ) − u(x , ⋅ )‖ ≤ ⎜ + 1⎟E 1 − x δ x + 4 (E + M1 ) + 4,
⎜ E ⎟ ⎛ E⎞ ⎛ E⎞
⎝ 2 ln δ ⎠ ln ln
⎝ δ⎠ ⎝ δ⎠

where C1 and C2 are some constants that only depend on x and M1 is a constant.
6  Yu Qiao and Xiangtuan Xiong

Proof. By the triangle inequality, we have

‖uα, δ (x , ⋅ ) − u(x , ⋅ )‖ ≤ ‖uα, δ (x , ⋅ ) − uα,0(x , ⋅ )‖ + ‖uα,0(x , ⋅ ) − u(x , ⋅ )‖ ,
  (3.3)
1 2

where uα,0(x , t ) is the exact solution of problem (2.5) in the frequency space with noise-free data. Next,
we divide the argument into two steps.
Step 1. Estimate the term 1 in (3.3). It follows immediately from Parseval’s equality and the triangle
inequality that
1
sinh(θ˜(s − x )) ˆ δ ˆ
1 = ‖uˆ α, δ (x , ⋅ ) − uˆ α,0(x , ⋅ )‖ ≤ ‖e θ˜(1 − x )(gˆ δ − gˆ )‖ +
   ∫ θ˜
( f − f )ds .
x
¯1 
¯2

By (2.6), (3.2), and note that

iξ 1
∣θ˜∣ = ≤ , (3.4)
1+ α2ξ 2 2α

we have
1
¯1 ≤ δ sup∣e θ˜(1 − x )∣ ≤ δe ∣ θ˜∣ (1 − x ) ≤ δe 2α
(1 − x )
= E 1−xδ x .
ξ ∈

Using the Cauchy-Schwarz integral inequality, (2.12), (3.4), and (2.6) yields
1
⎡ ∞ ⎛ 1 sinh(θ˜(s − x )) 2 ⎞⎛ 1 δ ⎞ ⎤
2

¯2 ≤ ⎢ ⎜ ∫∫ θ˜
∫
ˆ ˆ 2
ds ⎟⎜ ∣f − f ∣ ⎟dξ ⎥
⎣−∞⎝ x ⎠⎝ x ⎠ ⎦
1
⎡∞⎛1 1 ⎞⎛ 1
⎞ ⎤
2
2(s − x ) δ
≤⎢ ⎜ e ∫∫ 2α ds ⎟⎜ ∫
∣fˆ − fˆ ∣2 ds ⎟dξ ⎥
⎣−∞⎝ x ⎠⎝ 0 ⎠ ⎦
1
⎛ 2α 2(1 − x ) 1 ⎞2 1
≤⎜ e 2α ⎟ δ = E 1−xδ x .
⎝ 2 ⎠ E
2 ln δ

Hence,
⎛ ⎞
1
1 ≤ ⎜ + 1⎟E 1 − x δ x . (3.5)
⎜ E ⎟
⎝ 2 ln δ ⎠

Step 2. Estimate the term 2 in (3.3). Again, in view of Parseval’s identity and the triangle inequality,
1
⎡ sinh(θ˜(s − x )) sinh(θ (s − x )) ⎤ ˆ
2 = ‖uˆ α,0(x , ⋅ ) − uˆ (x , ⋅ )‖ ≤ ∥(e θ˜(1 − x )
−

+ e θ(1 − x ))gˆ ∥ ∫⎢⎣ ˜
θ
−
θ ⎥⎦
f ds .
x
˜1   
˜2

We start by estimating the first term above. Let

A˜1 (ξ ) = ∣e θ˜(1 − x ) − θ − e −θx ∣ ,

by (2.2), we obtain
1
⎛ fˆ sinh(θs ) ⎞
˜1 = (e θ˜(1 − x ) − θ − e −θx ) ⎜uˆ (0, ξ ) − ∫ θ
ds ⎟ ≤ sup A˜1 (ξ )‖A(ξ )‖.
ξ ∈
⎝ 0 ⎠
 Quasi-reversibility method for nonhomogeneous sideways heat equation  7

Next, we estimate A˜1 (ξ ), let

∣ξ∣ ⎛ 1 ⎞
γ = θ − θ˜ = γ1(1 + σi ) , γ1 = ⎜1 − ⎟ ≥ 0, (3.6)
2 ⎝ 1+ α2ξ 2 ⎠

by the triangle inequality, (2.9), and (2.10), we obtain

A˜1 (ξ ) = ∣e −θx ∣∣1 − e −γ(1 − x )∣

∣ξ ∣
= e− 2 x (∣1 − e −γ1(1 − x )∣ + ∣e −γ1(1 − x ) − e −γ1(1 − x )(1 + iσ )∣)
∣ξ ∣
≤ e− 2 x (γ (1 − x ) + e −γ1(1 − x )∣1 − e −γ1(1 − x )iσ ∣)
1
∣ξ ∣
≤ 2γ1(1 − x )e − 2 x.

Note that
∣ξ∣ 1 2 5
γ1 ≤ ( 1 + α2ξ 2 − 1) ≤ α ∣ξ ∣2 , (3.7)
2 2 2
by Lemma 2.2 and (3.2), we have
5
1 − x 2⎛ 5 2 ⎞ C1
A˜1 (ξ ) ≤ α⎜ ⎟ = 4,
2 ⎝ xe ⎠ ⎛ ln E ⎞
⎝ δ⎠
where
5
1 − x⎛5 2 ⎞
C1 = ⎜ ⎟ . (3.8)
4 2 ⎝ xe ⎠

Combining with Lemma 3.1, we obtain

C1
˜1 ≤ 4 (E + M1) .
⎛ ln E ⎞ (3.9)
⎝ δ⎠
Let
A˜2 (ξ ) = ∣e θ˜(s − x ) − e θ(s − x )∣ ,
sinh(θ˜(s − x )) sinh(θ (s − x ))
A˜3 (ξ ) = − .
θ˜ θ

To estimate ˜2 , we first estimate A˜2 (ξ ) and A˜3 (ξ ). Using (3.6), the triangle inequality, (2.9), (2.10), and (3.7)
in order, for ∀s ∈ (x, 1), we have

A˜2 (ξ ) = ∣e (s − x )θ∣∣1 − e −(s − x )γ∣

∣ξ ∣
≤ e (s − x ) 2 (∣1 − e −(s − x )γ1σi∣ + ∣e −(s − x )γ1σi − e −(s − x )γ1(1 + σi )∣)
∣ξ ∣ (3.10)
≤ 2e (s − x ) 2 (s − x )γ1
2 ∣ξ ∣ 5
≤ e 2 (s − x )∣ξ∣2 α2.
2
1
Using successively the triangle inequality, (2.12), (2.11), note that 1 − θ 1 ≤ 2 α2 ∣ξ∣2 and ∣θ˜∣ ≤ ∣θ∣, then for
∀s ∈ (x, 1), we obtain

sinh(θ˜(s − x )) − θ 1sinh(θ (s − x ))
A˜3 (ξ ) =
θ˜
sinh(θ˜(s − x )) sinh(θ˜(s − x )) − sinh(θ (s − x ))
≤ (1 − θ 1 ) +
θ˜ θ
8  Yu Qiao and Xiangtuan Xiong

1 A˜2 (ξ )
≤ α2 ∣ξ∣2 (s − x )e ∣ ξ ∣ (s − x ) +
2 ∣ξ∣
≤ 2 (s − x )e ∣ξ ∣ ∣ξ∣2 α2.

Now, we estimate ˜2. Using the Cauchy-Schwarz integral inequality yields

1
⎡∞⎛1 ⎞⎛ 1 sinh(θ˜(s − x )) sinh(θ (s − x ))
2 ⎞ ⎤
2

∫∫ ˆ
2 ≤ ⎢ ⎜ ∣f ∣ ds ⎟⎜
˜ 2
∫ ˜
θ
−
θ
ds ⎟dξ ⎥
⎣−∞⎝ x ⎠⎝ x ⎠ ⎦
1
⎡∞ 1 ⎤2
∫∫
≤ ⎢ ( 2 (s − x )e ∣ξ ∣ ∣ξ∣2 α2)2dse −3 ∣ ξ ∣ dξ
⎥
⎣−∞ x ⎦
1
∞
6 3 ⎛ ⎞2
≤
3
∫
(1 − x )2 α2⎜ ∣ξ∣4 e − ∣ ξ ∣ dξ .
⎟
⎝ −∞ ⎠

Similarly, since the generalized integral on the right-hand side of the last inequality converges, we introduce
the notation
1
∞
⎛ ⎞2
M2 ≔ ⎜ ∣ξ∣4 e − ∫ ∣ ξ ∣ dξ .
⎟
⎝−∞ ⎠

By (3.2), we obtain
C2
˜2 ≤ 4,
⎛ ln E ⎞ (3.11)
⎝ δ⎠
where
3
(1 − x ) 2
C2 = M2 . (3.12)
2 6
Substituting (3.9) and (3.11) into 2 , we have
C1 C2
2 ≤ 4 (E + M1) + 4,
⎛ ln E ⎞ ⎛ ln E ⎞ (3.13)
⎝ δ⎠ ⎝ δ⎠

where C1 and C2 are given by (3.8) and (3.12), respectively. Theorem 3.1 follows by combining (3.3), (3.5),
and (3.13). □

Remark 3.1. In this article, we consider that both the data g (t ) and the source term f (x , t ) are given approxi-
mately. From the proof of Theorem 3.1, it follows that the total error ‖uα, δ (x , ⋅ ) − u(x , ⋅ )‖ can be split into the
noise propagation error 1 and the approximation error 2 . Clearly, as α tends to 0, 1 tends to infinity, yet 2
tends to 0. Therefore, we need to choose α = α (δ ) dependent on δ to keep the total error as small as possible.

Note that the asymptotics of the regularized solution become progressively lower as x → 0. To obtain
the continuous dependence of the solution at x = 0, we have to introduce stronger assumptions
‖u(0, ⋅ )‖p ≤ E , p > 0, (3.14)
1

(1 + ξ 2) p ∫∣fˆ (s, ξ )∣2 ds < e−3 ∣ξ ∣ , ∀ξ ∈  , (3.15)

0
 Quasi-reversibility method for nonhomogeneous sideways heat equation  9

and the regularization parameter α is chosen as

1
α= 2.
⎛ ⎛E −2p ⎞
⎞ (3.16)
2⎜ ln ⎜ δ ⎛ ln δ ⎞ ⎟⎟
E

⎝ ⎝ ⎝ ⎠ ⎠⎠

p 1 fˆ sinh(θs )
Lemma 3.2. Let conditions (3.14) and (3.15) hold, A˜ (ξ ) = (1 + ξ 2 )2 ⎡⎣uˆ (0, ξ ) − ∫0 θ
ds⎤⎦, then

‖A˜ (ξ )‖ ≤ E + M1,

where M1 is a constant.

The proof is similar to the proof of Lemma 3.1, and we omit it here.

Theorem 3.2. Let uˆ (x , ξ ) given by (2.2) be the exact solution of problem (1.4) in the frequency space and uˆ α, δ (x , ξ )
given by (2.7) be the exact solution of problem (2.5) in the frequency space. Conditions (2.6), (3.14), (3.15) hold,
and the regularization parameter α is chosen by (3.16). Then, for p > 0, we have

⎡ ⎤
⎢ ⎥
α , δ ⎛ E ⎞−2p ⎢ 1 ⎥
‖u (0, ⋅ ) − u(0, ⋅ )‖ ≤ E ln + 1 + ε1(E + M1) + ε2M3,
⎝ δ ⎠ ⎢⎛ −2p ⎞ ⎥
⎛E ⎞
⎢ ⎜2 ln ⎜ δ ⎛ ln δ ⎞ ⎟ ⎟
E
⎥
⎣⎝ ⎝ ⎝ ⎠ ⎠
⎠ ⎦

⎧ 4 2 ⎫ ⎧ 4 ⎫
where M1 and M3 are some constants, and ε1 ≕ max ⎨2α 5 p, 2
α2 ⎬, ε2 ≕ max ⎨2α 5 p, 2 α2 ⎬.
⎩ ⎭ ⎩ ⎭
Proof. By the triangle inequality, we have
‖uα, δ (0, ⋅ ) − u(0, ⋅ )‖ ≤ ‖uα, δ (0, ⋅ ) − uα,0(0, ⋅ )‖ + ‖uα,0(0, ⋅ ) − u(0, ⋅ )‖ .
  (3.17)
3 4

Next, we divide the argument into two steps.

Step 1. Estimate the term 3 in (3.17). Taking a similar procedure of the estimate of 1 , and by (3.16),
we obtain
1
˜ ) δ
sinh(θs
3 ≤ ‖e θ˜(gˆ δ − gˆ )‖ + ∫ θ˜
( fˆ − fˆ )ds
0
1
1 ⎛ 2α 2 ⎞2
≤ δe 2α + δ⎜ e 2α ⎟

⎝ 2 ⎠
(3.18)
⎡ ⎤
⎢ ⎥
⎛ E ⎞−2p ⎢ 1 ⎥
= E ln 1 + 1 .
⎝ δ⎠ ⎢ − 2p 2 ⎥
⎛ ⎛ E ⎛ E ⎞ ⎞⎞
⎢ ⎜2 ln ⎜ δ ln δ ⎟⎟ ⎥
⎣⎝ ⎝ ⎝ ⎠ ⎠⎠ ⎦

Step 2. Estimate the term 4 in (3.17).

1
˜ )
⎛ sinh(θs sinh(θs ) ⎞ ˆ
4 ≤ ‖(e θ˜ − e θ )gˆ ‖ +
  
⎜
∫⎝θ˜
−
θ
⎟f ds

⎠
.
(3.19)
0
˜3 
˜4
10  Yu Qiao and Xiangtuan Xiong

We first estimate ˜3 in (3.19). Let

p
A˜ 4 (ξ ) = ∣e θ˜− θ − 1∣(1 + ξ 2 )− 2 .

By (2.2), we obtain
1
⎛ sinh(θs ) ˆ ⎞
˜3 = (e θ˜− θ − 1)⎜uˆ (0, ξ ) − ∫ θ
f ds ⎟ ≤ sup A˜ 4 (ξ )‖A˜ (ξ )‖.
ξ ∈
⎝ 0 ⎠
4
Let ξ0 = α − 5 , we discuss in the following three situations to estimate A˜ 4 (ξ ).
Case 1: ∣ξ ∣ ≥ ξ0. By the triangle inequality, (3.6), we have
p 4
A˜ 4 (ξ ) ≤ (∣e −γ∣ + 1)(1 + ξ 2 )− 2 ≤ 2 ∣ξ ∣−p ≤ 2α 5 p.

Case 2: 1 ≤ ∣ξ∣ < ξ0. Note that A˜1 (ξ ) ≤ 2γ1 when x = 0, combine with (3.7), we obtain
p 1 2 5 −p
A˜ 4 (ξ ) ≤ 2γ1(1 + ξ 2 )− 2 ≤ α ∣ξ∣2 .
2
5
If 0 < p ≤ 2 , we have
1 2 − 4 5 −p 4
A˜ 4 (ξ ) ≤ α (α 5 )2 ≤ 2α 5 p.
2
5
If p > 2 , we obtain

2 2
A˜ 4 (ξ ) ≤ α.
2
Case 3: ∣ξ ∣ < 1.
p 1 2 5 2 2
A˜ 4 (ξ ) ≤ 2γ1(1 + ξ 2 )− 2 ≤ α ∣ξ∣2 ≤ α.
2 2

In conclusion,

⎧ 4 2 2⎫
A˜ 4 (ξ ) ≤ max ⎨2α 5 p, α ⎬ ≕ ε1. (3.20)
⎩ 2 ⎭

By Lemma 3.2, we obtain

˜3 ≤ ε1(E + M1) . (3.21)

Now, we estimate ˜4 in (3.19). Let

˜ )
sinh(θs sinh(θs ) p
A˜5 (ξ ) = − (1 + ξ 2 )− 2 e − ∣ξ ∣ .
θ˜ θ

Similarly, we discuss A˜5 (ξ ) in three cases.

Case 1: ∣ξ ∣ ≥ ξ0. By the triangle inequality, (2.12), and note that ∣θ˜∣ ≤ ∣θ∣, then for ∀s ∈ (x, 1), we have

˜ )
⎛ sinh(θs sinh(θs ) ⎞ p 4
A˜5 (ξ ) ≤ ⎜ + ⎟(1 + ξ 2 )− 2 e − ∣ξ ∣ ≤ 2e ∣ θ ∣ ∣ξ∣−p e − ∣ξ ∣ ≤ 2α 5 p.
⎝ θ˜ θ ⎠

Case 2: 1 ≤ ∣ξ∣ < ξ0. Note that A˜3 (ξ ) ≤ 2 se ∣ξ ∣ ∣ξ∣2 α2 when x = 0, then for ∀s ∈ (x, 1), we obtain
p
A˜5 (ξ ) ≤ 2 ∣ξ∣2 α2(1 + ξ 2 )− 2 ≤ 2 α2 ∣ξ ∣2 − p .

If 0 < p ≤ 2, note that α ∈ (0, 1), we have

4 4 2 4
A˜5 (ξ ) ≤ 2 α2(α − 5 )2 − p ≤ 2α 5 p + 5 ≤ 2α 5 p.
 Quasi-reversibility method for nonhomogeneous sideways heat equation  11

If p > 2 , we obtain

A˜5 (ξ ) ≤ 2 α2.

Case 3: ∣ξ ∣ < 1.
p
A˜5 (ξ ) ≤ 2 ∣ξ∣2 α2(1 + ξ 2 )− 2 ≤ 2 α2 ∣ξ ∣2 ≤ 2 α2.

In summary,

⎧ 4 ⎫
A˜5 (ξ ) ≤ max ⎨2α 5 p, 2 α2 ⎬ ≕ ε2. (3.22)
⎩ ⎭

Hence, for ∀s ∈ (x, 1), we have

1
⎡∞⎛1 ⎞⎛ 1 sinh(θs
˜ ) sinh(θs )
2 ⎞ ⎤
2

∫∫
˜4 ≤ ⎢ ⎜ ∣fˆ ∣2 ds ⎟⎜ ∫
θ˜
−
θ
ds ⎟dξ ⎥
⎣−∞⎝ 0 ⎠⎝ 0 ⎠ ⎦
1
2
⎡ ∞ 1 ⎛ sinh(θs
˜ ) sinh(θs ) ⎞ − p ⎤2
≤⎢ ∫∫ ⎜

⎝ θ˜
−
θ
⎟e

⎠
∣ ξ ∣ (1 + ξ 2)− 2 dse − ∣ ξ ∣ dξ ⎥
⎣−∞ 0 ⎦
1
∞
⎛ ⎞2
∫
≤ ε2⎜ e − ∣ ξ ∣ dξ .
⎟
⎝−∞ ⎠

Likewise, since the generalized integral on the right-hand side of the last inequality converges, we introduce
the notation
1
∞
⎛ ⎞2
M3 ≔ ⎜ e −∫ ∣ ξ ∣ dξ .
⎟
⎝−∞ ⎠

Therefore,
˜4 ≤ ε2M3 . (3.23)

Substituting (3.21) and (3.23) into (3.19), we have

4 ≤ ε1(E + M1) + ε2M3 , (3.24)

where ε1 and ε2 are given by (3.20) and (3.22). We arrive at the final conclusion by (3.17), (3.18), and (3.24). □

Remark 3.2. Since the regularization parameter α → 0 (δ → 0), we can easily find that, for p > 0, ε1 , ε2 → 0
(δ → 0). Therefore,
lim ‖uα, δ (0, ⋅ ) − u(0, ⋅ )‖ = 0, p > 0.
δ→0

Remark 3.3. Theorems 3.1 and 3.2 are the main results of this section, which are not provided in other articles.

4 Determination of flux structure and error estimate

In this section, we seek heat flux structure ux (x , t ) for 0 ≤ x < 1 from the measurement data g (t ).
By (2.1) and (2.13), we obtain
1

uˆx (x , ξ ) = − θe θ(1 − x )gˆ (ξ ) ∫

− fˆ (s , ξ ) cosh(θ (s − x ))ds , 0 ≤ x < 1, (4.1)
x
12  Yu Qiao and Xiangtuan Xiong

and
1

uˆxα, δ (x , ˜ θ˜(1 − x )gˆ δ (ξ ) − fˆ δ (s , ξ ) cosh(θ˜(s − x ))ds ,

ξ ) = − θe ∫ 0 ≤ x < 1. (4.2)
x

Theorem 4.1. Let uˆ (x , ξ ) given by (2.2) be the exact solution of problem (1.4) in the frequency space, uˆ α, δ (x , ξ )
given by (2.7) be the exact solution of problem (2.5) in the frequency space. Conditions (1.5), (2.6), and (3.1) hold,
and the regularization parameter α is chosen by (3.2). Then, for a fixed x ∈ (0, 1), we have

⎛ ⎞
E 1 C3 C4
‖uxα, δ (x , ⋅ ) − ux (x , ⋅ )‖ ≤ ⎜ln + ⎟E 1 − x δ x + 4 (E + M1 ) + 4,
⎜ δ E ⎟ ⎛ E⎞ ⎛ E⎞
⎝ 2 ln δ ⎠ ln ln
⎝ δ⎠ ⎝ δ⎠

where C3 and C4 are some constants that only depend on x and M1 is a constant.

Proof. By the triangle inequality, we have

‖uxα, δ (x , ⋅ ) − ux (x , ⋅ )‖ ≤ ‖uxα, δ (x , ⋅ ) − uxα,0(x , ⋅ )‖ + ‖uxα,0(x , ⋅ ) − ux (x , ⋅ )‖ .

    (4.3)
1 2

Next, we divide the argument into two steps.

Step 1. Estimate the term 1 in (4.3). Using Parseval’s equality and the triangle inequality, we have
1

1 = ‖uˆxα, δ (x , ⋅) − uˆxα,0(x , ˜ θ˜(1 − x )(gˆ − gˆ δ )‖ +

⋅ )‖ ≤ ‖θe
 ∫( fˆ − fˆδ ) cosh(θ˜(s − x ))ds .
x

¯1 



¯2

By (2.6), (3.4), and (3.2), we obtain

 ˜ θ˜(1 − x )∣ ≤ δ∣θ˜∣e ∣ θ˜∣ (1 − x ) ≤ δ 1 e

¯ 1 ≤ δ sup∣θe
1
2α
(1 − x ) E
= E 1 − x δ x ln . (4.4)
ξ ∈ 2α δ

Using the Cauchy-Schwarz integral inequality, (2.13), and (3.2) yields

1
⎡∞⎛1 ⎞⎛ 1 ⎞ ⎤
2
δ 2
¯ ∫∫ ˜ ˆ ˆ
2 ≤ ⎢ ⎜ ∣cosh(θ (s − x )∣ )ds ⎟⎜ ∣f − f ∣ ds ⎟dξ ⎥
2
∫
⎣−∞⎝ x ⎠⎝ x ⎠ ⎦
1
⎡∞⎛1 ⎞⎛ 1 ⎞ ⎤
2

˜∣ (s − x ) ∣ 2 δ 2
∫∫
≤ ⎢ ⎜ ∣e ∣ θ
∫
ds ⎟⎜ ∣fˆ − fˆ ∣ ds ⎟dξ ⎥
⎣−∞⎝ x ⎠⎝ 0 ⎠ ⎦ (4.5)
1
2
⎛ 2α 2 (1 − x ) ⎞ ˆ ˆ δ
≤⎜ e 2α ⎟ ‖f − f ‖L2(0,1; L2())
⎝ 2 ⎠
1
≤ E 1−xδ x .
E
2 ln δ

Thus,

⎛ ⎞
E 1
1 ≤ ⎜ln + ⎟E 1 − x δ x . (4.6)
⎜ δ E ⎟
⎝ 2 ln δ ⎠

Step 2. Estimate the term 2 in (4.3). Again, using Parseval’s identity and the triangle inequality, we have
1

2 = ‖uˆxα,0(x , ⋅ ) − uˆx (x , ⋅ )‖ ≤ ˜ θ˜(1 − x ))gˆ ‖ +

‖(θe θ(1 − x )
− θe
 ∫fˆ [cosh(θ(s − x )) − cosh(θ˜(s − x ))]ds .
x

˜1 

˜2
 Quasi-reversibility method for nonhomogeneous sideways heat equation  13

Let
˜ θ˜(1 − x ) − θ∣ .
B˜1(ξ ) = ∣θe −θx − θe

By (2.2), we obtain
1
⎛ fˆ sinh(θs ) ⎞
˜ 1 = (θe −θx ˜ θ˜(1 − x ) − θ )⎜uˆ (0, ξ ) −
− θe ∫ θ
ds ⎟ ≤ sup B˜1(ξ )‖A(ξ )‖.
ξ ∈
⎝ 0 ⎠

Next, we estimate B˜1(ξ ). By (3.6), (2.9), (2.10), and (3.7), for ∀x ∈ (0, 1), we have
˜ −γ(1 − x )∣
B˜1(ξ ) = ∣e −θx ∣∣θ − θe
∣ξ ∣
≤ e− 2 x (∣θ ˜ −iγ1σ (1 − x )∣ + ∣θe
− θe −iγ1σ (1 − x )∣ + ∣θe −iγ1σ (1 − x ) − θe ˜ −iγ1σ (1 − x ) − θe
˜ −(1 + σi )γ1(1 − x )∣)
∣ξ ∣
≤ e− 2 x (∣θ∣γ (1 − x )∣ + 2 γ1 + ∣θ∣γ1(1 − x ))
1

⎛ 2∣ξ∣ 1⎞ 5 ∣ξ ∣
≤⎜ + ⎟α2 ∣ξ∣2 e − 2 x.
⎝ 2 2 ⎠
5 6
Note that h max < h max for x ∈ (0, 1), we attain

⎡ 2 ⎛ 6 2 ⎞6 1 ⎛ 5 2 ⎞ 5 ⎤ ⎛6 2 ⎞ 2 2 ⎛6 2 ⎞
6
1
6
˜
B 1 (ξ ) ≤ ⎢ ⎜ ⎟ + ⎜ ⎟ α2 ≤ 2 ⎜ ⎟ α = ⎜ ⎟ .
⎣ 2 ⎝ xe ⎠ 2 ⎝ xe ⎠ ⎥⎦ ⎝ xe ⎠ 4 ⎝ xe ⎠ ⎛ E ⎞4
ln δ
⎝ ⎠

By Lemma 3.1, we obtain

C3
˜ 1 ≤ 4 (E + M1) ,
⎛ ln E ⎞ (4.7)
⎝ δ⎠
6
2 ⎛6 2 ⎞
C3 = ⎜ ⎟ . (4.8)
4 ⎝ xe ⎠

Let

B˜2(ξ ) = ∣cosh(θ˜(s − x )) − cosh(θ (s − x ))∣ .

To estimate ˜ 2 , we first estimate B˜2(ξ ):

(e (s − x )θ − e (s − x )θ˜)(1 − e −(s − x )(θ˜+ θ ))

B˜2(ξ ) = ≤ A˜2 (ξ )B˜3(ξ ) ,
2
˜ ˜
where A˜2 (ξ ) = ∣e (s − x )θ − e (s − x )θ∣ , B˜3(ξ ) = ∣1 − e −(s − x )(θ + θ )∣.
Using (2.4), (2.8), (2.9), and (2.10), we obtain
∣ξ ∣
B˜3(ξ ) = ∣1 − e −(s − x )(1 + θ 1) 2 (1 + σi )∣

∣ξ ∣ ∣ξ ∣ ∣ξ ∣
≤ ∣1 − e −(s − x )(1 + θ 1) 2 σi∣ + ∣e −(s − x )(1 + θ 1) 2 σi − e −(s − x )(1 + θ 1) 2 (1 + σi )∣

≤ 2 2 (s − x ) ∣ξ∣ .

Hence,
∣ξ ∣
B˜2(ξ ) ≤ 2(s − x )2α2e 2 ∣ξ ∣3 . (4.9)
14  Yu Qiao and Xiangtuan Xiong

Then, using the Cauchy-Schwarz integral inequality, (1.5), and (4.9) yields
1
⎡∞⎛1 ⎞⎛ 1 ⎞ ⎤
2

∫∫ ˆ ∫ ˜
2 ≤ ⎢ ⎜ ∣f ∣ ds ⎟⎜ ∣cosh(θ (s − x )) − cosh(θ (s − x ))∣ ds ⎟dξ ⎥
˜ 2 2

⎣−∞⎝ x ⎠⎝ x ⎠ ⎦
1
⎡∞⎛1 ⎞ 1⎛ ∣ξ ∣ ⎞
2⎤2
∫∫ ∫
≤ ⎢ ⎜ ∣fˆ ∣2 ds ⎟ ⎜2(s − x )2α2e 2 3
∣ξ∣ ⎟ dsdξ ⎥
⎣−∞⎝ 0 ⎠ x⎝ ⎠ ⎦
1
∞
2 5 ⎛ ⎞2
≤
5
∫
(1 − x )2 α2⎜ ∣ξ∣6 e 2 ∣ ξ ∣ − 3 ∣ ξ ∣ dξ
⎟ .
⎝ −∞ ⎠

Introducing the notation,

1
∞
⎛ ⎞2
∫
M3 = ⎜ ∣ξ∣6 e 2 ∣ ξ ∣ − 3 ∣ ξ ∣ dξ
⎟ .
⎝−∞ ⎠

Therefore,
C4
˜ 2 ≤ 4,
⎛ ln E ⎞ (4.10)
⎝ δ⎠

where

5 5
C4 = (1 − x )2 M3 . (4.11)
10
Combining (4.7) and (4.10), we have
C3 C4
2 ≤ 4 (E + M1) + 4,
⎛ ln E ⎞ ⎛ ln E ⎞ (4.12)
⎝ δ⎠ ⎝ δ⎠

where C3 and C4 are given by (4.8) and (4.11), respectively. By substituting (4.6) and (4.12) into (4.3), we arrive
at the final conclusion. □

Remark 4.1. Note that the stability estimate in Theorem 4.1 only solves our problem for 0 < x < 1 and does not
give any useful information on the continuous dependence of the solution at x = 0. The following Theorem 4.2
will use assumptions (3.14) and (3.15) to give the convergence of this boundary.

Theorem 4.2. Let uˆ (x , ξ ) given by (2.2) be the exact solution of problem (1.4) in the frequency space, uˆ α, δ (x , ξ )
given by (2.7) be the exact solution of problem (2.5) in the frequency space. Conditions (2.6), (3.14), and (3.15) hold,
1
and the regularization parameter α is chosen by (3.16). Then, for p > 2 , we have

⎡ ⎤
⎢ ⎥
α, δ ⎛ E ⎞−2p ⎢ ⎛ E ⎛ E ⎞−2p ⎞ 1 ⎥ + ε3(E + M1 ) + ε4M4 ,
‖ux (0, ⋅ ) − ux (0, ⋅ )‖ ≤ E ln ln ⎜ ln ⎟ +
⎝ δ ⎠ ⎢ ⎝δ ⎝ δ ⎠ ⎠
⎛ E ⎛ E ⎞−2p ⎞ ⎥
⎢ 2 ln ⎜ δ ln δ ⎟

⎣ ⎝ ⎝ ⎠ ⎠ ⎥⎦

⎧ 2 1 2 ⎛α 53 + α2 ⎞⎫ ⎧ 2 ⎫
where M1 and M4 are some constants, ε3 ≔ max ⎨2(α 3 ) p − 2 , ⎬ and ε4 ≔ max ⎨2α 3 p, α2 ⎬.
⎩ 2 ⎝ ⎠⎭ ⎩ ⎭
 Quasi-reversibility method for nonhomogeneous sideways heat equation  15

Proof. By the triangle inequality, we have

‖uxα, δ (0, ⋅ ) − ux (0, ⋅ )‖ ≤ ‖uxα, δ (0, ⋅ ) − uxα,0(0, ⋅ )‖ + ‖uxα,0(0, ⋅ ) − ux (0, ⋅ )‖ .

    (4.13)
3 4

Next, we divide the argument into two steps.

Step 1. Estimate the term 3 in (4.13). Taking a similar procedure of the estimate of 1 , we obtain
1
˜ θ˜(gˆ − gˆ δ )‖ +
3 ≤ ‖θe ∫( fˆ − fˆδ ) cosh(θs˜ )ds
0
1
1 1 ⎛ 2α 2 ⎞2
≤δ e 2α + δ⎜ e 2α ⎟

2α ⎝ 2 ⎠
(4.14)
⎡ ⎤
⎢ ⎥
⎛ E ⎞ ⎢ ⎛ E ⎛ E ⎞−2p ⎞
−2p
1 ⎥.
= E ln ln ⎜ ln ⎟ +
⎝ δ ⎠ ⎢ ⎝δ ⎝ δ ⎠ ⎠ −2p ⎥
⎛E ⎞
2 ln ⎜ δ ⎛ ln δ ⎞ ⎟ ⎥
E
⎢ ⎝ ⎝ ⎠ ⎠⎦
⎣

Step 2. Estimate the term 4 in (4.13). From the triangle inequality, we obtain
1

4 ≤ ˜ θ˜)gˆ ‖ +
‖(θe θ
− θe
 ∫fˆ [cosh(θs˜ ) − cosh(θs)]ds .
0

˜3 



˜4

Let
p
˜ θ˜− θ∣(1 + ξ 2 )− 2 .
B˜4 (ξ ) = ∣θ − θe

By (2.2), we obtain
1
⎛ fˆ sinh(θs ) ⎞
˜ 3 = ˜ θ˜− θ )⎜uˆ (0, ξ ) −
(θ − θe ∫ θ
ds ⎟ ≤ sup B˜4 (ξ )‖A˜ (ξ )‖.
ξ ∈
⎝ 0 ⎠
2
Let ξ1 = α − 3 , we discuss in the following three situations to estimate B˜4 (ξ ).
Case 1: ∣ξ ∣ ≥ ξ1. By the triangle inequality, (3.6), we have
p 1 1 2 1
B˜4 (ξ ) ≤ (∣θ∣ + ∣θ˜∣∣e −γ∣)(1 + ξ 2 )− 2 ≤ ∣θ∣(1 + e −γ )∣ξ ∣−p ≤ 2 ∣ξ∣2 − p ≤ 2(α 3 ) p − 2 , p> .
2
2∣ξ∣ +1 2 5
Case 2: 1 ≤ ∣ξ∣ < ξ1. Note that B˜4 (ξ ) ≤ 2
α ∣ξ∣2 when x = 0, we obtain

2∣ξ∣ + 1 1 2⎛
α 2 ∣ξ ∣3 − p + ∣ξ ∣2 − p ⎞.
5 5
B˜4 (ξ ) ≤ α2 ∣ξ ∣2 − p ≤
2 2 ⎝ ⎠
1 5
If 2
< p ≤ 2 , note that α ∈ (0, 1), we have

1 2⎡ 2 2 5 2 ⎛ 2p
α 3 + α 3 p + 3 ⎞ ≤ 2α 3 ( p − 2 ).
2 1 2 1
B˜4 (ξ ) ≤ α 2 (α − 3 )3 − p + (α − 3 )2 − p⎤⎦ ≤
2 ⎣ 2 ⎝ ⎠
5 1
If 2
< p ≤ 3 , note that 3 − p ∈ [0, 2 ), we obtain

1 2⎡ 2 1 2 1 2 5
B˜4 (ξ ) ≤ α 2 (α − 3 )3 − p + 1⎤⎦ ≤ α2⎡⎣ 2 (α − 3 )2 + 1⎤⎦ ≤ (α 3 + α 2 ) .
2 ⎣ 2 2
If p > 3 , note that α ∈ (0, 1), we obtain

1 2 2 5
B˜4 (ξ ) ≤ α ( 2 + 1) ≤ (α 3 + α 2 ) .
2 2
16  Yu Qiao and Xiangtuan Xiong

Case 3: ∣ξ ∣ < 1.

1 2 5 1 2 5
B˜4 (ξ ) ≤ α ( 2∣ξ∣ + 1)∣ξ∣2 ≤ α2( 2 + 1) ≤ (α 3 + α 2 ) .
2 2 2
In conclusion,

⎧ 2 1 2 5 ⎫ 1
B˜4 (ξ ) ≤ max ⎨2(α 3 ) p − 2 , (α 3 + α2 )⎬ ≕ ε3, p> . (4.15)
⎩ 2 ⎭ 2

By Lemma 3.2, we obtain

1
˜ 3 ≤ ε3(E + M1) , p> . (4.16)
2
Let
p ∣ξ ∣
˜ ) − cosh(θs )∣(1 + ξ 2 )− 2 e −
B˜5(ξ ) = ∣cosh(θs 2 .

To estimate ˜ 4 , we first estimate B˜5(ξ ). Similarly, divided into three cases.

Case 1: ∣ξ ∣ ≥ ξ1. By the triangle inequality and (2.11), we have
p ∣ξ ∣ ∣ξ ∣ ∣ξ ∣ ∣ξ ∣ 2
˜
B˜5(ξ ) ≤ (∣e θs∣ + ∣e θs ∣)(1 + ξ 2 )− 2 e − 2 ≤ (e 2 +e 2 θ 1 )∣ξ ∣−p e − 2 ≤ 2 ∣ξ ∣−p ≤ 2α 3 p.

Case 2: 1 ≤ ∣ξ∣ < ξ1. By (4.9), for ∀s ∈ (x, 1), we obtain

p
B˜5(ξ ) ≤ ∣ξ∣3 α2(1 + ξ 2 )− 2 ≤ ∣ξ ∣3 − p α2.

If 0 < p ≤ 3, we have
2 2
A˜ 4 (ξ ) ≤ (α − 3 )3 − pα2 ≤ 2α 3 p.

If p > 3 , note that ∣ξ ∣ ≥ 1 , we obtain

A˜ 4 (ξ ) ≤ α2.

Case 3: ∣ξ ∣ < 1.
p
B˜5(ξ ) ≤ ∣ξ∣3 α2(1 + ξ 2 )− 2 ≤ ∣ξ ∣3 α2 ≤ α2.

In conclusion,

⎧ 2 ⎫ 1
B˜5(ξ ) ≤ max ⎨2α 3 p, α2 ⎬ ≕ ε4 , p> . (4.17)
⎩ ⎭ 2

We are now in a position to estimate ˜ 4 .

1
⎡∞⎛1 ⎞⎛ 1 ⎞ ⎤
2

∫∫ ∫ ˜ )∣ 2 ds ⎟dξ ⎥
˜ 4 ≤ ⎢ ⎜ ∣fˆ ∣2 ds ⎟⎜ ∣ cosh(θs ) − cosh(θs
⎣−∞⎝ 0 ⎠⎝ 0 ⎠ ⎦
1
⎡∞ ⎛1 ⎞⎛ 1 p 2 ⎞ ⎤
2

⎛ ⎞
∫ 2 p ˆ 2
∫ ⎝
2 −
∫ ˜
≤ ⎢ (1 + ξ ) ⎜ ∣f ∣ ds ⎟⎜ (1 + ξ ) 2 ∣cosh(θs ) − cosh(θs )∣ ds ⎟dξ ⎥
⎠
⎣−∞ ⎝0 ⎠⎝ 0 ⎠ ⎦
1
⎡∞ 1
⎛ ∣ξ ∣
2 ⎤2
˜ )∣⎞⎟ dsdξ ⎥
p
≤⎢ e∫ 2∣ξ ∣ −3 ∣ξ ∣
∫ ⎜(1 + ξ 2 )− 2 e

⎝
−
2 ∣cosh(θs ) − cosh(θs
⎠
⎣−∞ 0 ⎦
1
∞
⎛ ⎞2
≤ ε4⎜ e ∫ 2 ∣ ξ ∣ − 3 ∣ ξ ∣ dξ
⎟ .
⎝−∞ ⎠
 Quasi-reversibility method for nonhomogeneous sideways heat equation  17

Introducing the notation,

1
∞
⎛ ⎞2
M4 = ⎜ e ∫ 2 ∣ ξ ∣ − 3 ∣ ξ ∣ dξ
⎟ .
⎝−∞ ⎠
Then,
˜ 4 ≤ ε4M4 . (4.18)

Substituting (4.16) and (4.18) into 4 , we have

4 ≤ ε3(E + M1) + ε4M4 , (4.19)

where ε3 and ε4 are given by (4.15) and (4.17). Theorem 4.2 now follows from (4.13), (4.14), and (4.19). □

1
Remark 4.2. Since the regularization parameter α → 0 (δ → 0), we can easily find that, for p > 2 , ε3 ,
ε4 → 0 (δ → 0). Therefore,
1
lim ‖uxα, δ (0, ⋅ ) − ux (0, ⋅ )‖ = 0, p> .
δ→0 2

Remark 4.3. If conditions (3.1), (3.14), and (3.15) hold, and the regularization parameter α is chosen by (3.16)
in Theorems 3.1 and 4.1, then we can also obtain convergence results.

Remark 4.4. Theorems 4.1 and 4.2 are the main results of this section, which are not provided in other articles.

Remark 4.5. The proposed method can be easily extended to the two-dimensional case, i.e.,

⎧ut − uxx − uyy = f (x , y , t ) , x > 0, y > 0, t > 0,

⎪u(1, y , t ) = g ( y , t ) , y ≥ 0, t ≥ 0,
⎪⎪u(x , y , 0) = 0, x ≥ 0, y ≥ 0,
⎨ (4.20)
u(x , 0, t ) = 0, x ≥ 0, t ≥ 0,
⎪
⎪uy (x , 0, t ) = 0, x ≥ 0, t ≥ 0,
⎪
⎩u(x , y , t )∣x →∞ bounded.

In fact, we can obtain error estimates for temperature and flux over the entire region by solving the following
problem:

⎧utα, δ = uxx α, δ α, δ
+ u yy α, δ
− α2(uxxyy α, δ
+ uxxtt ) = f δ (x , y , t ) , x > 0, y > 0, t > 0,
⎪ α, δ
u (1, y , t ) = g δ ( y , t ) , y ≥ 0, t ≥ 0,
⎪
⎪uα, δ (x , y , 0) = 0, x ≥ 0, y ≥ 0,
⎨ α, δ (4.21)
⎪u (x , 0, t ) = 0, x ≥ 0, t ≥ 0,
α, δ
⎪u y (x , 0, t ) = 0, x ≥ 0, t ≥ 0,
⎪ α, δ
⎩u (x , y , t )∣x →∞ bounded.
18  Yu Qiao and Xiangtuan Xiong

5 Application
In this section, we present several examples to illustrate the performance of the proposed method. In parti-
cular, for the convenience of testing, we take f (x , t ) = 0 as an example for numerical experiments. First,
in order to construct the data g (t ), we consider the direct problem

⎧ut − uxx = 0, x > 0, t > 0,

⎪u(x , 0) = 0, x ≥ 0,
⎨ (5.1)
u(0, t ) = h(t ) , t ≥ 0,
⎪
⎩u(x , t )∣x →∞ bounded,

which is well posed and may be solved by the method presented in [41]. The solution of (5.1) at x = 1 is
∞
1
g (t ) ≔ u(1, t ) =
2π
∫ eiξte− iξ hˆ (ξ )dξ . (5.2)
−∞

In the numerical implementation, we give the data h(t ), sample on the equidistant grid and perform the
discrete Fourier transform. Then, according to (5.2), we obtain the data g (t ) by inverse discrete Fourier
transform. The noise data g δ is generated by
g δ = g + ε ⋅ randn(size(g )) ,

where ε denotes the error level, and the function randn(⋅ ) generates arrays of random numbers whose
elements are normally distributed with a mean of 0 and a variance of σ 2 = 1. The total noise δ can be measured
in the sense of root mean square error according to
M
1
δ = ‖g δ − g‖ = ∑ (g δ (ti ) − g (ti ))2 . (5.3)
M i=1

The approximate solutions of temperature and flux are calculated by (2.7) and (4.2), respectively,
and the regularization parameter α is selected by (3.16). To show the accuracy of numerical solutions, we compute
the approximate relative errors in l 2 norm, denoted by

‖uα, δ (x , t ) − u(x , t )‖l 2

er (u, ε ) = (5.4)
‖u(x , t )‖l 2

and

‖uxα, δ (x , t ) − ux (x , t )‖l 2
er (ux , ε ) = . (5.5)
‖ux (x , t )‖l 2

Application 5.1. Consider a smooth data function h(t ) = t sin( − 4t ).

2
In this example, we take p = 7 . Figure 1 gives the numerical results without using any regularization
method for the error level ε = 0.03 . It is easy to see from Figure 1 that the approximate solutions without using
any regularization method are unacceptable. Since problem (1.4) is ill-posed, we must apply some regulariza-
tion methods to it. Figures 2 and 3 show the approximate results of temperature and flux at different locations.
We can observe that the smaller the error level, the better the approximation. To further illustrate the
proposed method, we compare the quasi-reversibility method with the Fourier regularization method [42]
in Table 1. er1(u, ε ) and er1(ux , ε ) denote the relative error of the Fourier method, and er2(u, ε ) and er2(ux , ε )
denote the relative error of the quasi-reversibility method. As can be seen from Table 1, the proposed approach
generally performs well in the current context.
 Quasi-reversibility method for nonhomogeneous sideways heat equation  19

15 100
exact solution exact solution

exact solution and approximate solution of ux (x,t)

exact solution and approximate solution of u(x,t)

=0.03 =0.03
80

10 60

40

5 20

0 -20

-40

-5 -60

-80

-10 -100
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
t t

(a) (b)
Figure 1: Un-regularized solutions and exact solutions with ε = 0.03 for Application 5.1: (a) temperature and (b) heat flux.

3 2
exact solution exact solution
exact solution and approximate solution of u(x,t)

exact solution and approximate solution of u(x,t)

=0.03 =0.03
=0.01 1.5 =0.01
2

1
1
0.5

0 0

-1 -0.5

-1
-2
-1.5

-3
-2

-4 -2.5
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
t t

(a) (b)
1.5 0.8
exact solution exact solution
exact solution and approximate solution of u(x,t)

exact solution and approximate solution of u(x,t)

=0.03 =0.03
0.6
=0.01 =0.01
1
0.4

0.5 0.2

0
0
-0.2
-0.5
-0.4

-1 -0.6

-0.8
-1.5
-1

-2 -1.2
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
t t

(c) (d)
Figure 2: Numerical results of temperature for Application 5.1: (a) x = 0, (b) x = 0.3 , (c) x = 0.6, and (d) x = 0.9.
20  Yu Qiao and Xiangtuan Xiong

8 6
exact solution exact solution

exact solution and approximate solution of ux (x,t)

exact solution and approximate solution of ux (x,t)

=0.03 =0.03
5
6 =0.01 =0.01

4
4
3
2
2

0 1

0
-2
-1
-4
-2
-6
-3

-8 -4
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
t t

(a) (b)
4 2
exact solution exact solution
exact solution and approximate solution of ux (x,t)

exact solution and approximate solution of ux (x,t)

=0.03 =0.03
3 =0.01 1.5 =0.01

1
2

0.5
1
0
0
-0.5

-1
-1

-2 -1.5

-3 -2
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
t t

(c) (d)
Figure 3: Numerical results of heat flux for Application 5.1: (a) x = 0, (b) x = 0.3 , (c) x = 0.6, and (d) x = 0.9.

Table 1: Comparison of numerical results between the Fourier Method (er1 ) and the quasi-reversibility method (er2 ) in Application 5.1
with ε = 0.01

x 0.05 0.2 0.35 0.5 0.65 0.875 0.95

er1(u, 0.01) 0.2274 0.1821 0.1125 0.0798 0.0449 0.0282 0.0176

er2(u, 0.01) 0.1603 0.1244 0.0779 0.0613 0.0442 0.0298 0.0218
er1(ux , 0.01) 0.7835 0.5424 0.3073 0.2006 0.1181 0.0949 0.0548
er2(ux , 0.01) 0.3906 0.3063 0.2280 0.1652 0.1070 0.0839 0.0584

Application 5.2. Choose a non-smooth but continuous data function

⎧ 1
t, 0≤t≤ ,
⎪ 2
h (t ) = ⎨ 1
⎪ 1 − t , 2 < t ≤ 1.
⎩
3
Figures 4 and 5 and Table 2 give the numerical results of Application 5.2 with p = 7 . From Table 2, it can be
found that the closer x is to 1, the better the numerical accuracy, which is reasonable in view of the character
of the proposed problem. On the other hand, since Application 5.2 is a non-smooth function, its numerical
results are slightly worse than those of Application 5.1. The numerical results are acceptable because of the ill-
 Quasi-reversibility method for nonhomogeneous sideways heat equation  21

0.4 0.3
exact solution exact solution
exact solution and approximate solution of u(x,t)

exact solution and approximate solution of u(x,t)

=0.003 =0.003
=0.001 =0.001
0.35 0.28

0.3 0.26

0.25 0.24

0.2 0.22

0.15 0.2

0.1 0.18
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
t t

(a) (b)
Figure 4: Numerical results of temperature for Application 5.2: (a) x = 0.4 and (b) x = 0.8.

0.3 0.15
exact solution exact solution
exact solution and approximate solution of ux (x,t)

exact solution and approximate solution of ux (x,t)

=0.003 =0.003
=0.001 =0.001
0.2 0.1

0.1 0.05

0 0

-0.1 -0.05

-0.2 -0.1

-0.3 -0.15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
t t

(a) (b)
Figure 5: Numerical results of heat flux for Application 5.2: (a) x = 0.4 and (b) x = 0.8 .

3
Table 2: Numerical results of Application 5.2 for different x with p = 7

x 0 0.15 0.3 0.45 0.6 0.75 0.9

er (u, 0.003) 0.1008 0.0609 0.0422 0.0345 0.0223 0.0184 0.0142

er (u, 0.001) 0.0479 0.0349 0.0211 0.0154 0.0100 0.0069 0.0044
er (ux , 0.003) 0.2884 0.2162 0.1744 0.1705 0.1201 0.1151 0.1076
er (ux , 0.001) 0.1670 0.1435 0.0997 0.0908 0.0726 0.0695 0.0555

posedness of the problem. Figure 6 illustrates a plot of the error in temperature and heat flux as a function
of p, from which it can be seen that for large p, the results are not necessarily good.

Application 5.3. Let the exact data function be h( y , t ) = 4 cos(2πy ) cos(2πt ).

The numerical results of Application 5.3 are presented in Figures 7 and 8, from which it is easy to see that
the recovery of temperature is slightly better than the recovery of flux at the same location. On the other hand,
it can be observed that the approximation of this example is not as good as that of Applications 5.1 and 5.2,
due to the fact that problem (4.20) has more variables and is more ill-posed. In summary, the proposed
quasi-reversibility method can be used to deal with similar problems in the two-dimensional case.
22  Yu Qiao and Xiangtuan Xiong

0.06 1.6

1.4
0.05

Relative error of temperature

1.2

Relative error of heat flux

0.04
1

0.03 0.8

0.6
0.02

0.4

0.01
0.2

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
p p

(a) (b)
Figure 6: Numerical results for Application 5.2 with ε = 0.001 at x = 0.8: (a) temperature u(x , t ) and (b) heat flux ux (x , t ).

0.03 0.03

The approximate solution at x=0.8

0.04 0.04
0.02
The exact solution at x=0.8

0.02

0.02 0.02
0.01 0.01

0 0
0 0

-0.02 -0.02
-0.01 -0.01

-0.04 -0.04
1 -0.02 1
-0.02
1 1
0.5 0.5
0.5 -0.03 0.5 -0.03

y 0 0 t y 0 0 t

(a) (b)
Figure 7: Numerical results of temperature at x = 0.8 for Application 5.3: (a) exact u(x , y , t ) and (b) numerical uδ (x , y , t ).

0.2
0.2

0.15 0.15
The approximate solution at x=0.8

0.3 0.3
The exact solution at x=0.8

0.1 0.1
0.2 0.2

0.1 0.05 0.1 0.05

0 0
0 0
-0.1 -0.1
-0.05 -0.05
-0.2 -0.2

-0.3 -0.1 -0.3 -0.1

1 1
1 1
-0.15 -0.15
0.5 0.5
0.5 0.5
-0.2 -0.2
y 0 0 t y 0 0 t

(a) (b)
Figure 8: Numerical results of heat flux at x = 0.8 for Application 5.3: (a) exact ux (x , y , t ) and (b) numerical uxδ (x , y , t ).
 Quasi-reversibility method for nonhomogeneous sideways heat equation  23

6 Conclusion
SHE is very important in some engineering contexts and industrial applications, so many researchers have
been involved in this topic in recent decades. For the homogeneous setting, theoretical concepts and computa-
tional implementation are plentifully discussed and developed well. However, there are very few theoretical
stability and convergence estimates for the nonhomogeneous system.
In this article, we use a quasi-reversibility method to recover the surface temperature and heat flux of the
nonhomogeneous SHE. The convergence estimates of regularized solutions are obtained under some appro-
priate a prior conditions. Numerical verification of the efficiency and accuracy of the quasi-reversibility
method is performed by solving some numerical examples. Furthermore, we point out that the proposed
method can be extended to the two-dimensional case. In fact, our main contribution in this work is focused on
the generalization of the standard inverse problems for heat equations to inverse problems for nonhomoge-
neous heat equations involving the source term f (x , t ). These mathematical problems have become an
important tool in modeling many real-life problems. It is worth noting that the estimates obtained are
sufficient to prove the results, but most of them are quite rough and can be improved. Here, the source
term f (x , t ) is a function of x and t . However, the source function may also be related to u , i.e.,
f = f (x , t , u). Obviously, in this case, some new and strong techniques [3,43] are needed to estimate the errors.
Moreover, the homogeneous initial condition, u(x, 0) = 0, is not important. A problem with non-zero initial
data, u(x , 0) = ϕ(x ), can be reduced to (1.4) by using the linearity of the differential equation. And the modified
method can easily be extended to variable coefficient problems, which requires further exploration.

Acknowledgement: The authors thank the referees for their valuable comments and suggestions, which
improved the presentation of this manuscript.

Funding information: The research was supported by the Natural Science Foundation of China (No. 11661072)
and the Natural Science Foundation of Northwest Normal University (No. NWNU-LKQN-17-5).

Author contributions: Yu Qiao and Xiangtuan Xiong wrote jointly the manuscript and prepared all figures
(tables). All authors read and approved the final manuscript.

Conflict of interest: The authors state no conflict of interest.

Data availability statement: Data sharing is not applicable to this article as no datasets were generated or
analyzed during this study.

References
[1] T. I. Seidman and L. Eldén, An ’optimal filtering’ method for the sideways heat equation, Inverse Problems 6 (1990), no. 4, 681–696,
DOI: https://doi.org/10.1088/0266-5611/6/4/013.
[2] F. Berntsson, Sequential solution of the sideways heat equation by windowing of the data, Inverse Probl. Eng. 11 (2003), no. 2, 91–103,
DOI: https://doi.org/10.1080/1068276021000048564.
[3] N. H. Tuan, D. Lesnic, T. Q. Viet, and V. Van Au, Regularization of semilinear sideways heat equation, IMA J. Appl. Math. 84 (2019), no. 2,
258–291, DOI: https://doi.org/10.1093/imamat/hxy058.
[4] Z. Qian, Regularization methods for the sideways heat equation and the idea of modifying the “kernel” in the frequency domain, Commun.
Appl. Math. Comput. 26 (2012), no. 3, 298–311.
[5] H. W. Engl, M. Hanke, and A. Neubauer, Regularization of Inverse Problems, Kluwer Academic Publishers Group, Dordrecht, 1996.
[6] A. Carasso, Determining surface temperatures from interior observations, SIAM J. Appl. Math. 42 (1982), no. 3, 558–574,
DOI: https://doi.org/10.1137/0142040.
[7] F. Berntsson, A spectral method for solving the sideways heat equation, Inverse Problems 15 (1999), no. 4, 891–906,
DOI: https://doi.org/10.1088/0266-5611/15/4/305.
24  Yu Qiao and Xiangtuan Xiong

[8] J. C. Liu and T. Wei, A quasi-reversibility regularization method for an inverse heat conduction problem without initial data, Appl. Math.
Comput. 219 (2013), no. 23, 10866–10881, DOI: https://doi.org/10.1016/j.amc.2013.05.009.
[9] J. R. Dorroh and X. P. Ru, The application of the method of quasi-reversibility to the sideways heat equation, J. Math. Anal. Appl. 236
(1999), no. 2, 503–519, DOI: https://doi.org/10.1006/jmaa.1999.6462.
[10] L. Eldén, Approximations for a Cauchy problem for the heat equation, Inverse Problems 3 (1987), no. 2, 263–273, DOI: https://doi.org/
10.1088/0266-5611/3/2/009.
[11] Z. Qian, C. L. Fu, and X. T. Xiong, A modified method for determining the surface heat flux of IHCP, Inverse Probl. Sci. Eng. 15 (2007),
no. 3, 249–265, DOI: https://doi.org/10.1080/17415970600725128.
[12] H. T. Nguyen and V. C. H. Luu, Two new regularization methods for solving sideways heat equation, J. Inequal. Appl. 2015 (2015), 65,
DOI: https://doi.org/10.1186/s13660-015-0564-0.
[13] C. Y. Qiu, C. L. Fu, and Y. B. Zhu, Wavelets and regularization of the sideways heat equation, Comput. Math. Appl. 46 (2003), no. 5–6,
821–829, DOI: https://doi.org/10.1016/S0898-1221(03)90145-3.
[14] C. L. Fu and C. Y. Qiu, Wavelet and error estimation of surface heat flux, J. Comput. Appl. Math. 150 (2003), no. 1, 143–155,
DOI: https://doi.org/10.1016/S0377-0427(02)00657-X.
[15] J. R. Wang, Uniform convergence of wavelet solution to the sideways heat equation, Acta Math. Sin. 26 (2010), no. 10, 1981–1992,
DOI: https://doi.org/10.1007/s10114-010-7242-4.
[16] T. Regińska, Sideways heat equation and wavelets, J. Comput. Appl. Math. 63 (1995), no. 1–3, 209–214, DOI: https://doi.org/10.1016/
0377-0427(95)00073-9.
[17] J. R. Wang, The multi-resolution method applied to the sideways heat equation, J. Math. Anal. Appl. 309 (2005), no. 2, 661–673,
DOI: https://doi.org/10.1016/j.jmaa.2004.11.025.
[18] T. Regińska and L. Eldén, Solving the sideways heat equation by a wavelet-Galerkin method, Inverse Problems 13 (1997), no. 4,
1093–1106, DOI: https://doi.org/10.1088/0266-5611/13/4/014.
[19] T. Regińska and L. Eldén, Stability and convergence of the wavelet-Galerkin method for the sideways heat equation, J. Inverse Ill-Posed
Probl. 8 (2000), no. 1, 31–49, DOI: https://doi.org/10.1515/jiip.2000.8.1.31.
[20] A. S. Carasso, Slowly divergent space marching schemes in the inverse beat conduction problem, Numer. Heat Trans. B 23 (1993), no. 1,
111–126, DOI: https://doi.org/10.1080/10407799308914892.
[21] L. Eldén, Numerical solution of the sideways heat equation by difference approximation in time, Inverse Problems 11 (1995), no. 4,
913–923, DOI: https://doi.org/10.1088/0266-5611/11/4/017.
[22] L. Guo, D. A. Murio, and C. Roth, A mollified space marching finite differences algorithm for the inverse heat conduction problem with slab
symmetry, Comput. Math. Appl. 19 (1990), no. 7, 75–89, DOI: https://doi.org/10.1016/0898-1221(90)90196-Q.
[23] D. A. Murio and L. Guo, A stable space marching finite differences algorithm for the inverse heat conduction problem with no initial
filtering procedure, Comput. Math. Appl. 19 (1990), no. 10, 35–50, DOI: https://doi.org/10.1016/0898-1221(90)90356-O.
[24] X. T. Xiong, C. L. Fu, and H. F. Li, Central difference method of a non-standard inverse heat conduction problem for determining surface
heat flux from interior observations, Appl. Math. Comput. 173 (2006), no. 2, 1265–1287, DOI: https://doi.org/10.1016/j.amc.2005.
04.070.
[25] X. T. Xiong, C. L. Fu, and H. F. Li, Fourier regularization method of a sideways heat equation for determining surface heat flux, J. Math.
Anal. Appl. 317 (2006), no. 1, 331–348, DOI: https://doi.org/10.1016/j.jmaa.2005.12.010.
[26] X. T. Xiong and C. L. Fu, Determining surface temperature and heat flux by a wavelet dual least squares method, J. Comput. Appl. Math.
201 (2007), no. 1, 198–207, DOI: https://doi.org/10.1016/j.cam.2006.02.014.
[27] Y. J. Deng and Z. H. Liu, Iteration methods on sideways parabolic equations, Inverse Problems 25 (2009), no. 9, 095004,
DOI: https://doi.org/10.1088/0266-5611/25/9/095004.
[28] Y. J. Deng and Z. H. Liu, New fast iteration for determining surface temperature and heat flux of general sideways parabolic equation,
Nonlinear Anal. Real World Appl. 12 (2011), no. 1, 156–166, DOI: https://doi.org/10.1016/j.nonrwa.2010.06.005.
[29] J. B. Liu and Y. J. Deng, A modified Landweber iteration for general sideways parabolic equations, Acta Math. Appl. Sin. Engl. Ser. 27
(2011), no. 4, 727–738, DOI: https://doi.org/10.1007/s10255-011-0104-8.
[30] Z. Qian and C. L. Fu, Regularization strategies for a two-dimensional inverse heat conduction problem, Inverse Problems 23 (2007), no. 3,
1053–1068, DOI: https://doi.org/10.1088/0266-5611/23/3/013.
[31] S. S. Liu and L. X. Feng, An inverse problem for a two-dimensional time-fractional sideways heat equation, Math. Probl. Eng. 2020 (2020),
5865971, DOI: https://doi.org/10.1155/2020/5865971.
[32] X. T. Xiong and X. M. Xue, Fractional Tikhonov method for an inverse time-fractional diffusion problem in 2-dimensional space,
Bull. Malays. Math. Sci. Soc. 43 (2020), no. 1, 25–38, DOI: https://doi.org/10.1007/s40840-018-0662-5.
[33] X. T. Xiong, Q. Zhou, and Y. C. Hon, An inverse problem for fractional diffusion equation in 2-dimensional case: stability analysis
and regularization, J. Math. Anal. Appl. 393 (2012), no. 1, 185–199, DOI: https://doi.org/10.1016/j.jmaa.2012.03.013.
[34] S. S. Liu and L. X. Feng, A revised Tikhonov regularization method for a Cauchy problem of two-dimensional heat conduction equation,
Math. Probl. Eng. 2018 (2018), 1216357, DOI: https://doi.org/10.1155/2018/1216357.
[35] C. S. Liu and C. W. Chang, A spring-damping regularization of the Fourier sine series solution to the inverse Cauchy problem for a 3D
sideways heat equation, Inverse Probl. Sci. Eng. 29 (2021), no. 2, 196–219, DOI: https://doi.org/10.1080/17415977.2020.1782400.
[36] T. N. Luan and T. Q. Khanh, Determination of temperature distribution and thermal flux for two-dimensional inhomogeneous sideways
heat equations, Adv. Comput. Math. 46 (2020), 54, DOI: https://doi.org/10.1007/s10444-020-09796-w.
 Quasi-reversibility method for nonhomogeneous sideways heat equation  25

[37] E. Ata and I. O. Kıymaz, New generalized Mellin transform and applications to partial and fractional differential equations, Int. J. Math.
Comput. Eng. 1 (2023), no. 1, 45–66, DOI: https://doi.org/10.2478/ijmce-2023-0004.
[38] E. Bas and M. Karaoglan, Representation of the solution of the M-Sturm-Liouville problem with natural transform, Int. J. Math. Comput.
Eng. 1 (2023), no. 2, 243–252, DOI: https://doi.org/10.2478/ijmce-2023-0020.
[39] R. Lattès and J. L. Lions, The Method of Quasi-reversibility. Applications to Partial Differential Equations, American Elsevier,
New York, 1969.
[40] L. Eldén, F. Berntsson, and T. Regińska, Wavelet and Fourier methods for solving the sideways heat equation, SIAM J. Sci. Comput. 21
(2000), no. 6, 2187–2205, DOI: https://doi.org/10.1137/S1064827597331394.
[41] Ö. İlhan and G. Şahin, A numerical approach for an epidemic SIR model via Morgan-Voyce series, Int. J. Math. Comput. Eng. 2 (2024),
no. 1, 123–138, DOI: https://doi.org/10.2478/ijmce-2024-0010.
[42] C. L. Fu, X. T. Xiong, and P. Fu, Fourier regularization method for solving the surface heat flux from interior observations, Math. Comput.
Model. 42 (2005), no. 5–6, 489–498, DOI: https://doi.org/10.1016/j.mcm.2005.08.003.
[43] N. Anh Triet, D. O’Regan, D. Baleanu, N. Hoang Luc, and N. Can, A filter method for inverse nonlinear sideways heat equation,
Adv. Differential Equations 2020 (2020), 149, DOI: https://doi.org/10.1186/s13662-020-02601-4.
 © 2024. This work is published under
http://creativecommons.org/licenses/by/4.0 (the “License”). Notwithstanding
the ProQuest Terms and Conditions, you may use this content in accordance
with the terms of the License.