PARTIAL DIFFERENTIAL EQUATIONS

BANACH CENTER PUBLICATIONS, VOLUME 27

INSTITUTE OF MATHEMATICS
POLISH ACADEMY OF SCIENCES
WARSZAWA 1992

TOWARDS THE CAUCHY PROBLEM FOR

THE LAPLACE EQUATION

D I N H N H O H À O, T R A N D U C V A N

Hanoi Institute of Mathematics

P.O. Box 631 Bo Ho, Hanoi, Vietnam

RUDOLF GORENFLO

FB Mathematik, Institut für Mathematik I, Freie Universität Berlin

Arnimallee 2-6, D-1000 Berlin 33, Germany

1. Introduction. The Cauchy problem for elliptic equations occurs in the

study of many practical problems. For example, in investigations of a gravita-
tional (electric, magnetic) field it is often necessary to determine the potential
of the field in a domain outside the mass (charge, current) creating the field,
from the values of the potential in a part of this domain. A related problem is
that of extending an analytic vector field in R3 or a curve in R2 to a harmonic
vector field outside with determination of its singularities. Such problems play a
role, e.g., in the construction of magnetohydrostatic equilibria [24, 25, 54]. The
solution of this problem is equivalent to the solution of a Cauchy problem for the
Laplace equation that is ill-posed in the sense of Hadamard [3, 27, 46, 81]. For
an introduction to the literature of such problems the reader is referred to [3, 27,
34, 46, 64, 81].
The solution of the Cauchy problem for the Laplace equation will exist only
if strong compatibility or smoothness conditions are imposed on the initial data.
It was Hadamard who showed that unless a certain compatibility relation holds
among the Cauchy data no global solution can exist. He further showed that even
if the data are such that a classical solution exists, this solution will not depend
continuously on the data [26, 27].
In this paper we consider the following problems A, B and C:
Problem A. Find a function u(y, x) which satisfies
(1.1) uyy (y, x) + ∆u(y, x) = 0, (y, x) ∈ RN
+
+1
= {(y, x) | y > 0, x ∈ RN },

[111]
112 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

(1.2) u(0, x) = f (x) , x ∈ RN ,

(1.3) uy (0, x) = g(x) , x ∈ RN .
Problem B. Find a function u(y, x) which satisfies the equations
(1.4) uyy (y, x) + uxx (y, x) = 0, (y, x) ∈ DY = {(y, x) | 0 < y, −π < x < π},
(1.5) u(0, x) = f (x) , uy (0, x) = g(x) , −π < x < π .
Problem C. Find a function u(y, x) which satisfies
(1.6) uyy (y, x) + ∆u(y, x) = 0 , y ∈ (0, Y ) , x ∈ (−π, π)N =: Ω ,
(1.7) u(0, x) = f (x) , uy (0, x) = g(x) , x ∈ (−π, π)N ,
(1.8) u(y, x) |∂Ω = 0 , y ∈ (0, Y ) .
These problems are well known to be ill-posed in the sense of Hadamard,
and many attempts, many investigations from various aspects, as existence–uni-
queness theorems [1, 5, 13–15, 17, 19, 20, 26, 27, 29–31, 37, 39–42, 47, 48, 51, 52,
57–59, 61, 62–67, 69, 73–79, 83, 94–99, 100, 101, 104, 106], stability estimates [3,
4, 7, 8, 21, 33, 35, 38, 40, 43–45, 48, 53, 55], regularization, least square methods
[21, 28, 32, 48, 56, 81, 92, . . .] for such problems have been discussed.
We shall use the theory of pseudodifferential operators with real analytic sym-
bols (Ψ DOAS) [18, 84–89] and the approximation theory of functions to regularize
the above problems. We observe that every function in Lp (RN ) can be approxi-
mated by a sequence of entire functions of exponential type on CN , the restrictions
of which to RN belong to Lp (RN ) [60] (class Mνp ), and in these classes our prob-
lems are well-posed. Therefore, we should first consider our problems with data
in Mνp and then approximate “improper” data in Lp by Mνp functions and sug-
gest the solutions with these approximation data as a regularization. We call this
method the mollification method. Tran Duc Van in his joint work with other au-
thors [87] has used this idea to approximate the solution of the well-posed Cauchy
problem for the wave equation. Dinh Nho Hào in his joint work with Gorenflo [16]
has observed that this method with a modification still works for many ill-posed
problems. We hope to present an abstract version of this method in a later work.
In Section II we give a short survey on the Cauchy problem for elliptic equa-
tions. Section III is devoted to Problems A, B and C. Numerical experiments will
be described in a succeeding work.
This paper was written during the stay of the first and second authors at the
Free University of Berlin (West). This research stay was supported by DAAD
(German Academic Exchange Service) and by the Alexander von Humboldt-
Stiftung. The authors are members of the research group “Regularization”.

2. A short survey
2.1. Existence and uniqueness theorems. The first result on the Cauchy prob-
lem for the Laplace equation is that by Hadamard [26]. He proved, for a particular
case of Problem B, the following
 CAUCHY PROBLEM 113

Theorem 2.1. A necessary and sufficient condition for the existence of a

function u(y, x) which satisfies uyy + uxx = 0 in a neighborhood of the interval
−π < x < π of the x-axis, where y > 0 and u(0, x) = f (x), uy (0, x) = g(x),
−π < x < π, is that
1 Rπ
(2.1) H(x) = f (x) − g(ξ) ln |x − ξ| dξ
π
−π

be an analytic function of x for −π < x < π. Here f is of class C 1 and g is of

class C 0 in (−π, π).
An analogous result for Problem C has been obtained by Payne and Sather
in [67]. Furthermore, these authors have generalized the result of Hadamard to a
class of degenerate elliptic equations.
Theorem 2.2. A necessary and sufficient condition for the existence of a
function u(y, x) which satisfies uyy + y α uxx = 0 (α > 0) in a neighborhood of the
interval (−π, π) of the x-axis, where y > 0, and u(0, x) = f (x), uy (0, x) = g(x),
−π < x < π, is that
Rπ g(ξ) 
2

(2.2) Hα (x) = f (x) + mβ dξ β =
|x − ξ|1−β α+2
−π

be an analytic function of x for −π < x < π. Here f is of class C 1 and g is of

class C 0 in (−π, π), and
β 1−β Γ ( 21 (3 − β))
(2.3) mβ = √ .
π(1 − β)Γ ( 12 (2 − β))
An interesting result on existence and uniqueness for Problem A has been
given by B. H. Li and Y. Q. Li [49]:
Theorem 2.3. Let f, g ∈ D0 (RN ) and let (1.2), (1.3) be satisfied in the sense
of D0 . Then Problem A is solvable if and only if
(2.4) D0 -lim f y (y, x) − g
y&0

is an entire function, where f is a harmonic function such that

D0 -lim f (y, x) = f (x) .
y&0

In [85], as a direct consequence of the result of Tran Duc Van in his recent
work on Ψ DOAS [84], we have also obtained an existence and uniqueness theorem
for Problem A. Before outlining this result we need some definitions.
Let x ∈ RN N
x , N ≥ 1, and ξ ∈ Rξ be real variables, D
α αN
= D1α1 . . . DN ,
Dj = −i∂/∂xj , j = 1, . . . , N , α = (α1 , . . . , αN ), |α| = α1 + . . . + αN .
Definition 1. The space W +∞ (RN N
x ) is the set of functions f : Rx → C,
satisfying the following conditions: f admits analytic continuation to an entire
114 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

function on CN and for each ε > 0 there exist constants r < ∞ and Cε , possibly
depending on f , such that
|f (x + iy)| ≤ Cε exp(r|y| + ε|x|) , x + iy ∈ CN .
We list here some classes of functions which belong to W +∞ (RN x ): all functions
N
f ∈ L2 (Rx ) the support of whose Fourier transform f (ξ) is compact (band-
b
limited functions) [18], all functions in Mνp , 1 ≤ p ≤ +∞, ν < ∞ [60], all
functions in W +∞ (RN ) [88, 89]. From the Paley–Wiener theorem it follows that
f ∈ W +∞ (RN x ) if and only if its analytic continuation f (z) is the Fourier–Laplace
transform of an analytic functional with compact support.
Definition 2. A sequence of functions fn ∈ W +∞ (RN x ) is said to converge
to f ∈ W +∞ (RN x ) if and only if for each ε > 0 there exists a constant r < ∞
such that (with z = x + iy, x ∈ RN , y ∈ RN )
sup |fn (z) − f (z)| exp(−r|y| − ε|x|) → 0 , n → ∞.
z∈CN

Let the entire function A(ξ), ξ ∈ RN

ξ , be expanded in the Taylor series
∞
X
A(ξ) = aα ξ α , aα = (iD)α A(0)/α! .
|α|=0
−∞
Denote by W (RN
x )
the space of all continuous linear functionals defined on
+∞ N −∞
W (Rx ). Let h ∈ W (RNx ) and
hA(D)h, ϕi := hh, A(−D)ϕi , ∀ ϕ ∈ W +∞ (RN
x ).
−∞
Theorem 2.4. Let f, g ∈ W +∞ (RNx ) (W (RNx )). Then there exists a unique
2 1 +∞ −∞
solution of Problem A in C (R+ ; W (Rx )) (C 2 (RN
N
+;W (RN
x )), and
∞ ∞
X (−1)n y 2n ∆n f (x) X (−1)n−1 y 2n−1 ∆n−1 g(x)
(2.5) u(y, x) = + .
n=0
(2n)! n=1
(2n − 1)!
For any y > 0, the right side of (2.5) converges in the sense of W +∞ (RN
x )
(W −∞ (RN
x )).
Medeiros in [52] has given another existence and uniqueness theorem for Prob-
lem C.
Uniqueness results have received quite extensive treatment [1, 5, 9, 13–15, 20,
26, 29, 30, 31, 37, 39, 40–47, 49, 51–53, 57–59, 61–67, 69–79, 85, 94–99, 100–
102, 106]. We note here again important works by Carleman [9], Hörmander [30,
31], Lavrent’ev [43, 44], Lavrent’ev, Romanov and Shishat-skiı̆ [46], Müller [59],
Heinz [29], Cordes [15], Landis [40–42] and by Payne and his collaborators [37,
62–67]. Uniqueness results for elliptic equations with multiple characteristics were
considered in [94–99, 106].
2.2. Stable methods for solutions. The first results concerning stable methods
to solve the Cauchy problems for the Laplace equation are contained in the works
 CAUCHY PROBLEM 115

of Lavrent’ev [43, 45] and Pucci [70, 71] (see also John [35]). These authors estab-
lished stability estimates and proposed some stable methods. For further works
see [51, 53]. Other approaches have been tried, such as: establishing stability es-
timates [3, 4, 7, 8, 21, 22, 33–35, 46, 53], regularization by various methods, e.g.,
least squares, Carleman estimates, stable summation of Fourier series, finding a
Carleman function, Tikhonov regularization . . . [6, 7, 20, 21, 22, 28, 55, 56, 81,
90–93, 104, . . .]. For a function-theoretic method see [11, 12].

3. A mollification method
3.1. Problem A. Suppose that instead of the exact data f and g (which are
supposed to belong to Lp (RN ), 1 ≤ p ≤ ∞) we only have the measured data fε
and gε such that
(3.1) kfε − f k ≤ ε , kgε − gk ≤ ε .
We introduce mollification operators Mν which map a function f ∈ Lp to a
function fν in Mνp , and we require that fν → f in Lp as ν = (ν1 , . . . , νN ) → ∞.
For simplicity of presentation, we first describe results only for the case N = 1.
We need some notations. Let ∆h f = ∆h f (x) = f (x + h) − f (x), where h is
any real number. Then
k  
l+k k
X
k k−1
∆h f (x) = ∆h ∆h f (x) = (−1) f (x + lh) , k ∈ {1, 2, . . .},
l
l=0
∆0h =f.
Let
ω k (δ) = ω k (f, δ)p = sup k∆kh f (·)kp , ω(δ) = ω(f, δ)p = ω 1 (f, δ) .
|h|≤δ

ω k (δ) is called the modulus of continuity of order k of the function f in the Lp

metric. It is well known that if f ∈ Lp (R), 1 ≤ p < ∞, then limδ→0 ω(δ) = 0. For
p = ∞ this property does not hold in general. However, it is satisfied in a trivial
way if f is uniformly continuous on every compact subset of R.
Suppose that K(ξ) is a nonnegative even function of one variable, of exponen-
tial type 1, satisfying the condition
R∞
(3.2) K(t) dt = 1 ,
−∞

and K is chosen so that the integral

R∞
(3.3) K(ξ)|ξ|l dξ ,
−∞

where l is a fixed natural number, is finite. For K we may choose a function of

116 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

the form
 λ
sin(ξ/λ)
(3.4) µ ,
ξ
where λ ≥ l + 2 is an even number and µ is a positive constant such that (3.2)
holds.
For f ∈ Lp (R) and ν ∈ R+ the function
R∞
(3.5) fν (x) = Mν f (x) = Kν (ξ − x)f (ξ) dξ ,
−∞

where
l    
X l ν ξν
(3.6) Kν (ξ) = (−1)j−1 K
j=1
j j j

is well defined. Since K is an entire function of one variable, of exponential type

1, fν is an entire function of spherical type ν, lying in Lp (R) ([60], p. 186). Now
we have the following
Lemma 1 ([60], §5.2). If f has derivatives of order m lying in Lp , k = l − m,
0 ≤ m ≤ l, then fν is an entire function of spherical type ν and
c(m) k (m)
(3.7) kfν − f kp ≤ ω (f , 1/ν)p ,
νm
where
R∞
(3.8) c(m) = K(|ξ|)|ξ|m (1 + |ξ|)k dξ .
−∞

If s is a nonnegative integer with s ≤ m then

c
(3.9) kfν(s) − f (s) kp ≤ m−s ω k (f (s) , 1/ν)p , ν > 0,
ν
where
R∞
(3.10) c= K(ξ)(1 + |ξ|)l dξ .
0

Now instead of (1.1)–(1.3) with the measured data fε and gε we consider the
mollified problem: Find uεν (y, x) such that
∂ 2 uεν (y, x) ∂ 2 uεν (y, x)
(3.11) + = 0, (y, x) ∈ R+ × R ,
∂y 2 ∂x2
(3.12) uεν (0, x) = fνε (x) , x ∈ R,
∂uεν (0, x)
(3.13) = gνε (x) , x ∈ R.
∂y
 CAUCHY PROBLEM 117

Theorem 3.1. Problem (3.11)–(3.13) is well-posed in C 2 (R; W +∞ (RN

x )) and
∞ ∞
X (−1)n y 2n (fνε )(2n) (x) X (−1)n−1 y 2n−1 (gνε )(2n−1) (x)
(3.14) uεν (y, x) = + .
n=0
(2n)! n=0
(2n − 1)!
Furthermore, for every m, 0 ≤ m ≤ l, k = l − m,
(3.15) kuεν (y, ·) − u(y, ·)kp
c(m) k ∂ m u(y, ·) 1
   
sinh(yν)
≤ m ω , + cl ε cosh(yν) + ,
ν ∂xm ν p ν
where cl ≤ 2l − 1 does not depend on ν.
P r o o f. Formula (3.14) is a consequence of Theorem 2.4. We have
uεν (y, x) − u(y, x) = cos(yd/dx)fνε + [sin(yd/dx)/(d/dx)]gνε
− cos(yd/dx)f + [sin(yd/dx)/(d/dx)]g
= {cos(yd/dx)fνε − cos(yd/dx)fν
+ [sin(yd/dx)/(d/dx)]gνε − [sin(yd/dx)/(d/dx)]gν }
+ {cos(yd/dx)fν − cos(yd/dx)f
+ [sin(yd/dx)/(d/dx)]gν − [sin(yd/dx)/(d/dx)]g}
=: Σ1 + Σ2 .
From the fact that fν , gν ∈ Mνp and (3.5) it can be verified that
cos(yd/dx)fν + [sin(yd/dx)/(d/dx)]gν
= Mν (cos(yd/dx)f + [sin(yd/dx)/(d/dx)]g) = Mν u(y, x) .
Thus,
c(m) k ∂ m u(y, ·) 1
 
kΣ2 kp = kMν (u(y, ·)) − u(y, ·)kp ≤ m ω , .
ν ∂xm ν p
On the other hand, since fνε , fν , gνε , gν belong to Mνp , we obtain from the
Bernstein–Nikol’skiı̆ inequalities ([60], p. 115)
sinh(yν) ε
kΣ1 kp ≤ cosh(yν)kfνε − fν kp + kgν − gν kp .
ν
But
R∞
kfνε − fν kp = Kν (ξ − x)(f ε (x) − f (x)) dx

p
−∞
∞
R
≤ |Kν (ξ)| dξ kf ε − f kp ≤ cl ε
−∞
l
where cl ≤ 2 − 1 does not depend on ν ([60], p. 190). Analogously,
kgνε − gν kp ≤ cl ε .
118 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

Thus,
 
sinh(yν)
kΣ1 kp ≤ cl ε cosh(yν) + ,
ν
and (3.15) is proved.
Theorem 3.2. Let
(3.16) k∂ l u(y, ·)/∂xl kp ≤ M for y ∈ [0, Y ] ,
where m is a nonnegative integer. Then, with m = l − 1,
c(m)M
(3.17) kuεν (y, ·) − u(y, ·)kp ≤ + 2cl εeyν .
νl
Furthermore, for y > 0 let
 
c(m)M yν
(3.18) Ω(ε) = inf + 2cl εe .
ν>0 νl
Then
 −1/l
1 lc(m)M
(3.19) Ω(ε) = c(m)M ln (1/l)+l +o(1) as ε → 0.
y 
1 lc(m)M
2ycl ε ln
y 2ycl ε
The infimum on the right hand side of (3.18) is attained if ν is the solution ν(ε)
of the equation
2
c(m)M l e−yν
(3.20) ε= ,
2ycl ν l+2
which can be written in the form
 −1/l
1 lc(m)M
(3.21) ν(ε) = ln (1/l)+l + o(1) as ε → 0 .
y 
1 lc(m)M
2ycl ε ln
y 2ycl ε
P r o o f. The estimate (3.17) is a direct consequence of (3.15) and (3.16). To
prove (3.19)–(3.21) we need the following result:
Lemma 2 ([16]). Let
(3.22) Ωc (ε) = inf (δ + c(δ)ε) .
δ>0
−η
If c(δ) = c0 exp(sδ ), η > 0, then
 −η
1 1
(3.23) Ωc (ε) = ln (η+1)/η + o(1) as ε → 0 .
s 
1 1
sc0 ηε ln
s sc0 ηε
The infimum is attained if δ is the solution δ(ε) of the equation
1 η+1 −sδ−η
(3.24) ε= δ e ,
c0 sη
 CAUCHY PROBLEM 119

which can be written in the form

 −η
1 1
(3.25) δ(ε) = ln (η+1)/η + o(1) as ε → 0 .
s 
1 1
sc0 ηε ln
s sc0 ηε
Our estimates (3.19)–(3.21) now follow directly from this lemma if we take
c0 = cl /c(m)M, s = y, η = 2/l.
When N > 1, we can also approximate the measured data f ε and g ε by
functions of classes Mνp , ν = (ν1 , . . . , νN ), by the above process, but since the
description is somewhat lengthy and no new idea is offered, we do not write it
down here.
3.2. Problem B. Instead of the exact data f and g in (1.5), let there be given
measured data f ε and g ε such that
(3.26) kf − f ε kLp (−π,π) ≤ ε , kg − g ε kLp (−π,π) ≤ ε .
The following lemma helps us to treat this problem.
Lemma 3. If f and g can be extended to entire functions in R of exponential
type, then the solution of Problem B can be written in the form
∞ ∞
X (−1)n y 2n f (2n) (x) X (−1)n−1 y 2n−1 g (2n−1) (x)
(3.27) u(y, x) = + .
n=0
(2n)! n=1
(2n − 1)!
Furthermore, this series converges in the sense of W ∞ , for fixed y ∈ (0, Y ),
uniformly in x for |x| < π.
P r o o f. Denote the extensions of f and g to all of R by the same notations f
and g, respectively. Now we consider Problem A with these data. By Theorem 2.4
the solution of Problem A exists uniquely in W ∞ (Rx ) and
∞ ∞
X (−1)n y 2n f (2n) (x) X (−1)n−1 y 2n−1 g (2n−1) (x)
u(y, x) = + .
n=0
(2n)! n=1
(2n − 1)!
For fixed y ∈ (0, Y ) this series converges uniformly in every compact subset of R.
It is clear that the function u(y, x) satisfies (1.4)–(1.5). ROn the other hand,
π
since g(x) is an entire function of exponential type, so is −π g(ξ) ln |x − ξ| dξ
1
R π
([60], §3.6). Thus f (x) − 2π −π
g(ξ) ln |x − ξ| dξ is also an entire function, or
Problem B has a unique solution (Theorem 2.1). The lemma is proved.
Now for ν ∈ R+ we define
Rπ
fνε (x) = Mν∗ f ε (x) = Kν (ξ − x)f ε (ξ) dξ ,
−π
(3.28) π
R
gνε (x) = Mν∗ g ε (x) = Kν (ξ − x)g ε (ξ) dξ .
−π
120 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

By Lemma 1, fνε and gνε are elements of Mνp . Therefore, from Lemma 3,
∞ ∞
X (−1)n y 2n (fνε )(2n) (x) X (−1)n−1 y 2n−1 (gνε )(2n−1) (x)
(3.29) uεν (y, x) = +
n=0
(2n)! n=1
(2n − 1)!
is the unique solution of Problem B with the data fνε instead of f , and gνε instead
of g. We hope that uεν can serve as an approximation to u(y, x). Indeed, the
following theorems will show how well uεν (y, x) approximates u(y, x).
Theorem 3.3. For every m, 0 ≤ m ≤ l, k = l − m, we have
kuεν (y, ·) − u(y, ·)kLp (−π,π)
c(m) k ∂ m u(y, ·) 1
   
sinh(yν)
≤ m ω , + cl ε cosh(yν) + ,
ν ∂xm ν p ν
where cl does not depend on ν.
Theorem 3.4. Let
k∂ l u(y, ·)/∂y l kLp (−π,π) ≤ M for any y > 0 ,
where l is a positive integer. Then the estimates (3.17)–(3.21) are valid , with the
norm in (3.17) being that of Lp (−π, π).
The proofs are similar to those of Theorems 3.1 and 3.2.
3.3. Problem C. We can repeat, of course, the method 3.2 for Problem C.
This means that when the data are not precisely given, we approximate them by
entire functions of exponential type, and therefore all the above results remain
valid. But since the data, and so the solution, of Problem C are periodic, we can
approximate the data by trigonometric polynomials.
Let τν (t) (ν = 0, 1, 2, . . .) be a trigonometric polynomial of order not higher
than ν, having the following properties:
Rπ
(3.30) τν (t) dt = 1 ,
−π
Rπ
(3.31) |τν (t)| dt ≤ c (ν = 1, 2, . . .) ,
−π

where c is a constant not depending on ν. Obviously

(3.32) τ0 (t) = 1/(2π) .
For ν > 0 the polynomials τν are defined nonuniquely. One may obtain such
polynomials, for example, by means of the formula
 2σ
1 sin(λt/2)
(3.33) dν (t) = ,
cν sin(t/2)
 CAUCHY PROBLEM 121

where σ is a positive integer not depending on λ, λ ∈ N+ (then ν = (λ − 1)σ).

Here
Rπ  sin(λt/2) 2σ
(3.34) cν = dt ∼ λ2σ−1 , λ = 1, 2, . . .
sin(t/2)
−π

([60], p. 87). Define

l l

k−1
(−1)k−1
 
X X t + 2sπ
(3.35) Kν∗ (t) := k
τν .
k s=0
k
k=1

Then Kν∗ (t) is a trigonometric polynomial of order not higher than ν. Further-
more, if
ν
X
(3.36) τν (t) = aλ eiλt (aλ = a−λ ) ,
λ=−ν

then
k−1   ν k−1
X t + 2sπ X X
(3.37) τν = aλ eiλ(t+2sπ)/k ,
s=0
k s=0
λ=−ν

and therefore
l l

k−1 h ν k−1 ν
X (−1)k−1 X X X i X
(3.38) Kν∗ (t) = k
aλ eiλ(t+2sπ)/k = kλ∗ eiλt .
k s=0 s=0
k=1 λ=−ν λ=−ν

Now let
Rπ
(3.39) fν (x) = Kν∗ (ξ)f (x + ξ) dξ .
−π

Then fν (t) is a trigonometric polynomial of order not higher than ν. Furthermore,

as an analogue to Lemma 1 we have
Lemma 4 ([60, §5.3]). Suppose f ∈ Lp (−π, π) and f has a generalized deriva-
tive f (m) . Furthermore, suppose that the even nonnegative trigonometric polyno-
mials τν (t) of order ν satisfy along with the condition (3.30) the further condition
Rπ am
(3.40) τν (t)tm dt ≤ ,
(ν + 1)m
0

where the constant am does not depend on ν = 0, 1, 2, . . . Then the function fν (t)
defined by (3.39) approximates f in the metric of Lp (−π, π) with the following
estimate:
π
ωpk (f (m) , ν+1 )
(3.41) kf − fν kLp (−π,π) ≤ bm for ν = 0, 1, 2, . . . ,
(ν + 1)m
where bm = 2(π m + 2al /π), k = l − m.
122 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

R e m a r k. The function dν (t) in (3.33) with 2σ −m ≥ 3 satisfies the condition

of Lemma 4.
Now, suppose that instead of the exact f and g in (1.7), we only have the
measured f ε and g ε such that
(3.42) kf − f ε kLp (−π,π) ≤ ε , kg − g ε kLp (−π,π) ≤ ε ,
(3.43) f ε (−π) = f ε (π) = g ε (−π) = g ε (π) = 0 .
For a positive integer ν, let
Rπ
(3.44) fνε (x) = Kν∗ f ε (x + ξ) dξ ,
−π
Rπ
(3.45) gνε (x) = Kν∗ g ε (x + ξ) dξ .
−π

It is not hard to see that

∞ ∞
X (−1)n y 2n (fνε )(2n) (x) X (−1)n−1 y 2n−1 (gνε )(2n−1) (x)
(3.46) uεν (y, x) = +
n=0
(2n)! n=0
(2n − 1)!
is the unique solution of Problem C with the data fνε and gνε instead of f and g,
respectively. On the other hand, from (3.38) and (3.44), (3.45) we have
ν
X ν
X
fνε (x) = (fνε )λ eiλx , gνε (x) = (gνε )λ eiλx ,
λ=−ν λ=−ν

where
∗
Rπ
(3.47) (fνε )λ = k−λ f ε (x)e−iλx dx ,
−π
Rπ
∗
(3.48) (gνε )λ = k−λ g ε (x)e−iλx dx .
−π

Thus,
ν ν
X X sinh(λy) iλx
(3.49) uεν (y, x) = (fνε )λ cosh(λy)eiλx + (gνε )λ e
λ
λ=−ν λ=−ν

is a trigonometric polynomial of order not higher than ν.

The same estimates as in Theorems 3.1, 3.2 can be established, but we do not
write them down here again.
R e m a r k. Although our estimates for the regularizing solutions are not of
Hölder continuity type, they are stronger than any logarithmic continuity. An
important feature is that our estimates are uniform for y with 0 < y ≤ Y .
This has not been established in earlier works (see, e.g, [38], [43], [46], . . .). In
[6] Cannon and DuChateau have given a direct process for solving Problem B.
 CAUCHY PROBLEM 123

Their estimates are also not of Hölder continuity type, but are stronger than any
logarithmic continuity. They are not valid for all y ∈ (0, Y ], as ours are, but only
for y ∈ (0, Y /3].

References

[1] N. A. A r o n s z a j n, A unique continuation theorem for solutions of elliptic partial differ-

ential equations or inequalities of second order , J. Math. Pures Appl. 36 (1957), 235–249.
[2] M. A. A t a k h o d z h a e v and U. E. K o b i l o v, The Cauchy problem for the Laplace equa-
tion in an infinite three-dimensional layer , in: Direct and Inverse Problems for Partial
Differential Equations and their Applications 186, Fan, Tashkent 1978, 62–75 (in Rus-
sian).
[3] J. B a u m e i s t e r, Stable Solution of Inverse Problems, Vieweg & Sohn, Braunschweig
1987.
[4] P. S. B o n d a r e n k o and A. V. R i l o v, The estimation of the modulus of continuity of the
inverse Laplace transform in the solution of the Cauchy problem for Laplace’s equation,
in: Approximate and Qualitative Methods of the Theory of Differential Equations, Inst.
Mat. Akad. Nauk Ukrain. SSR, Kiev 1971, 167–174 (in Russian).
[5] A. P. C a l d e r ó n, Uniqueness in the Cauchy problem for partial differential equations,
Amer. J. Math. 80 (1958), 16–36.
[6] J. R. C a n n o n and P. D u C h a t e a u, Approximating the solution to the Cauchy problem
for Laplace’s equation, SIAM J. Numer. Anal. 14 (1977), 473–483.
[7] J. R. C a n n o n and J. D o u g l a s, J r ., The approximation of harmonic and parabolic
functions on half-spaces from interior data, in: Numerical Analysis of Partial Differential
Equations (C.I.M.E. 20 Ciclo, Ispra 1967), Edizioni Cremonese, Rome 1968, 193–230.
[8] J. R. C a n n o n and K. M i l l e r, Some problems in numerical analytic continuation, SIAM
J. Numer. Anal. 2 (1965), 87–98.
[9] T. C a r l e m a n, Sur un problème d’unicité pour les systèmes d’équations aux dérivées
partielles à deux variables indépendantes, Ark. Mat. Astr. Fys. B 26 (1939), 1–9.
[10] L. A. C h u d o v, Difference methods for solving Cauchy problem for Laplace’s equation,
Dokl. Akad. Nauk SSSR 143 (1962), 789–801; English transl.: Soviet Math. Dokl. 3
(1962), 499–503.
[11] D. L. C o l t o n, Improperly posed initial value problem for self-adjoint hyperbolic and
elliptic equations, SIAM J. Math. Anal. 4 (1973), 42–51.
[12] D. L. C o l t o n, Partial Differential Equations in the Complex Domain, Pitman, London
1976.
[13] J. C o n l a n and R. P. G i l b e r t, Non-linear initial data for second and higher order semi-
linear elliptic equations, J. Reine Angew. Math. 276 (1975), 1–14.
[14] J. C o n l a n and G. N. T r y t t e n, Pointwise bounds in the Cauchy problem for elliptic
systems of partial differential equations, Arch. Rational Mech. Anal. 22 (1966), 143–152.
[15] H. O. C o r d e s, Über die Bestimmheit der Lösungen elliptischer Differentialgleichungen
durch Anfangsvorgaben, Nachr. Akad. Wiss. Göttingen Math.-Phys. IIa 11 (1956), 239–
258.
[16] D i n h N h o H à o and R. G o r e n f l o, A noncharacteristic Cauchy problem for the heat
equation, Acta Appl. Math. 24 (1991), 1–27. (See also: D i n h N h o H à o and R. G o r e n f l o,
An ill-posed problem for the heat equation, Z. Angew. Math. Mech. 71 (1991), T763-
T766).
[17] A. D o u g l i s, Uniqueness in Cauchy problems for elliptic systems of equations, Comm.
Pure Appl. Math. 13 (1960), 593–608.
124 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

[18] Yu. A. D u b i n s k iı̆, The algebra of pseudodifferential operators with analytic symbols and
its applications to mathematical physics, Uspekhi Mat. Nauk 37 (1982), 97–137; English
transl.: Russian Math. Surveys 37 (1982), 107–153.
[19] G. A. D z h a f a r l i, A uniqueness theorem for the solutions of an elliptic system in the
half-plane, Dokl. Akad. Nauk Azerbaidzhan. SSR 26 (1970), 12–14 (in Russian).
[20] Kh. Sh. D z h u r a e v, On a solution of the Cauchy problem for the Laplace equation, Dokl.
Akad. Nauk Tadzhik. SSR 29 (1986), 506–509 (in Russian).
[21] R. S. F a l k and P. B. M o n k, Logarithmic convexity for discrete harmonic functions and
the approximation of the Cauchy problem for the Poisson’s equation, Math. Comp. 47
(1986), 135–149.
[22] Yu. L. G a p a n e n k o, Stability of the solution of the Cauchy problem for the Laplace
equation on a weak compactum, in: Numerical Mathematics and Sofware, Moskov. Gos.
Univ., 1985, 44–50 (in Russian).
[23] V. I. G o r b aı̆ c h u k, Estimates of the accuracy of an approximate solution for the Cauchy
problem for the Laplace equation in an infinite strip, Ukrain. Mat. Zh. 32 (1980), 731–736;
English transl.: Ukrainian Math. J. 32 (1980), 489–494.
[24] R. G o r e n f l o, Funktionentheoretische Bestimmung des Aussenfeldes zu einer zweidimen-
sionalen magnetohydrostatischen Konfiguration, Z. Angew. Math. Phys. 16 (1965), 279–
290.
[25] —, Behandlung ebener magnetohydrostatischer Gleichgewichtsprobleme mittels komplexer
Analysis, in: Tagung Freie Randwertaufgaben, E. Grafarend und N. Weck (eds.), Mit-
teilungen aus dem Institut für Theoretische Geodäsie der Universität Bonn, N.4, 1972,
2–18.
[26] J. H a d a m a r d, Sur les problèmes aux derivées partielles et leur signification physique,
Bull. Univ. Princeton 13 (1902), 49–32.
[27] —, Lectures on Cauchy’s Problem in Linear Partial Differential Equations, Yale Univ.
Press, 1923.
[28] H. D. H a n, The finite element method in the family of improperly posed problems, Math.
Comp. 38 (1982), 55–65.
[29] E. H e i n z, Über die Eindeutigkeit beim Cauchyschen Anfangswertproblem einer ellipti-
schen Differentialgleichung zweiter Ordnung, Nachr. Akad. Wiss. Göttingen Math.-Phys.
IIa 1 (1955), 1–12.
[30] L. H ö r m a n d e r, Linear Partial Differential Operators, Springer, Berlin 1976.
[31] —, Uniqueness theorems for second order elliptic differential equations, Comm. Partial
Differential Equations 8 (1983), 21–64.
[32] V. K. I v a n o v, The Cauchy problem for the Laplace equation in an infinite strip, Differ-
entsial’nye Uravneniya 1 (1965), 131–136 (in Russian).
[33] F. J o h n, A note on “improper” problems in partial differential equations, Comm. Pure
Appl. Math. 8 (1955), 494–495.
[34] —, Differential Equations with Approximate and Improper Data, New York University,
1955.
[35] —, Continuous dependence on the data for solutions of partial differential equations with
a predescribed bound , Comm. Pure Appl. 13 (1960), 551–586.
[36] L. A. K n i z h n e r m a n, Numerical solution of the Cauchy problem for the Laplace equation
by means of an expansion in Fourier–Chebyshev series, Izv. Akad. Nauk SSSR Ser. Fiz.
Zemli 1984 (10), 76–81 (in Russian).
[37] R. J. K n o p s and L. E. P a y n e, Uniqueness and continuous dependence of the null so-
lution in the Cauchy problem for a nonlinear elliptic system, in: Inverse and Improperly
Posed Problems in Differential Equations, G. Anger (ed.), Akademie-Verlag, Berlin 1979,
151–160.
 CAUCHY PROBLEM 125

[38] T. I. K o r o l y u k, The Cauchy problem for the Laplace equation, Izv. Vyssh. Uchebn.
Zaved. Mat. 130 (1973), 53–55 (in Russian).
[39] H. K u m a n o-g o, On the uniqueness of the solution of the Cauchy problem and the unique
continuation theorem for elliptic equations, Osaka Math. J. 14 (1962), 182–212.
[40] E. M. L a n d i s, Certain properties of equations of elliptic type, Dokl. Akad. Nauk SSSR
107 (1956), 640–643 (in Russian).
[41] —, Some questions in the qualitative theory of elliptic and parabolic equations, Uspekhi
Mat. Nauk 14 (1) (1959), 21–85; English transl. in Amer. Math. Soc. Transl. (2) 20
(1962), 173–238.
[42] —, Some questions in the qualitative theory of second-order elliptic equations (case of
several independent variables), Uspekhi Mat. Nauk 18 (1) (1963), 3–62; English transl.
in Russian Math. Surveys 18 (1963), 1–62.
[43] M. M. L a v r e n t ’ e v, On the Cauchy problem for the Laplace equation, Izv. Akad. Nauk
SSSR Ser. Mat. 120 (1956), 819–842 (in Russian).
[44] —, On the Cauchy problem for linear elliptic equations of second order , Dokl. Akad.
Nauk SSSR 112 (1957), 195–197 (in Russian).
[45] —, Some Improperly Posed Problems in Mathematical Physics, Springer, New York 1967.
[46] M. M. L a v r e n t ’ e v, V. G. R o m a n o v and S. P. S h i s h a t-s k iı̆, Ill-posed Problems of
Mathematical Physics and Analysis, Transl. Math. Monographs, Amer. Math. Soc., Prov-
idence, R.I., 1986.
[47] N. L e r n e r, Unicité du problème de Cauchy pour des opérateurs elliptiques, Ann. Sci.
École Norm. Sup. (4) 17 (1984), 469–505.
[48] H. A. L e v i n e and S. V e s s e l l a, Estimates and regularization for solutions of some
ill-posed problems of elliptic and parabolic type, Rend. Circ. Mat. Palermo 123 (1986),
161–183.
[49] B. H. L i and Y. Q. L i, On the initial value problem of the Laplace equation, J. Systems
Sci. Math. Sci. 7 (1987), 1–6.
[50] O. A. L i s k o v e t s, A solution of the Cauchy problem for the Laplace equation by a gen-
eralized method of the summability of series, Vestsi Akad. Navuk BSSR Ser. Fiz.-Mat.
Navuk 1970 (4), 68–74 (in Russian).
[51] V. G. M a z y a and V. P. K h a v i n, On the solutions of the Cauchy problem for Laplace’s
equation (solvability, normality, approximation), Trans. Moscow Math. Soc. 30 (1974),
65–117.
[52] L. A. M e d e i r o s, Remarks on a non well-posed problem, Proc. Roy. Soc. Edinburgh 102A
(1986), 131–140.
[53] S. M. M e r g e l y a n, Harmonic approximation and approximate solution of the Cauchy
problem for the Laplace equation, Uspekhi Mat. Nauk 11 (1956), 3–26 (in Russian).
[54] F. M e y e r und H. U. S c h m i d t, Torusartige Plasmakonfigurationen ohne Gesamtstrom
durch ihren Querschnitt im Gleichgewicht mit einem Magnetfeld , Z. Naturforsch. 13a
(1958), 1005–1015.
[55] K. M i l l e r, Three circle theorems in partial differential equations and applications to
improperly posed problems, Arch. Rational Mech. Anal. 16 (1964), 126–154.
[56] —, Least squares methods for ill-posed problems with a prescribed bound , SIAM J. Math.
Anal. 1 (1970), 52–74.
[57] S. M i z o h a t a, Unicité dans les problèmes de Cauchy pour quelques équations différen-
tielles elliptiques, Mem. Coll. Sci. Univ. Kyoto Ser. A Math. 31 (1958), 121–128.
[58] —, Unicité du prolongement des solutions des équations elliptiques du quatrième ordre,
Proc. Japan Acad. 34 (1958), 687–692.
[59] C. M ü l l e r, On the behavior of the solutions of the differential equation ∆u = F (x, u) in
the neighborhood of a point, Comm. Pure Appl. Math. 7 (1954), 505–514.
126 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

[60] S. M. N i k o l ’ s k iı̆, Approximation of Functions of Several Variables and Imbedding The-

orems, Springer, Berlin 1975 (translated from the Russian).
[61] L. N i r e n b e r g, Uniqueness in Cauchy problems for differential equations with constant
leading coefficients, Comm. Pure Appl. Math. 10 (1957), 89–105.
[62] L. E. P a y n e, Bounds in the Cauchy problem for the Laplace equation, Arch. Rational
Mech. Anal. 5 (1960), 35–45.
[63] —, On a priori bounds in the Cauchy problem for elliptic equations, SIAM J. Math. Anal.
1 (1970), 82–89.
[64] —, Improperly Posed Problems in Partial Differential Equations, SIAM, Philadelphia,
Penn., 1975.
[65] L. E. P a y n e and D. S a t h e r, On some improperly posed problems for quasilinear equa-
tions of mixed type, Trans. Amer. Math. Soc. 128 (1967), 135–141.
[66] —, —, On some improperly posed problems for the Chaplygin equation, J. Math. Anal.
Appl. 19 (1967), 67–77.
[67] —, —, On an initial-boundary value problem for a class of degenerate elliptic operators,
Ann. Mat. Pura Appl. 78 (1968), 323–337.
[68] A. P. P o d d u b n y a k and V. E. E m e t s, The Cauchy problem for the Laplace equation
in an infinite n + 1-dimensional layer , Mat. Metody i Fiz.-Mekh. Polya 15 (1982), 13–15
(in Russian).
[69] M. H. P r o t t e r, Unique continuation for elliptic equations, Trans. Amer. Math. Soc. 95
(1960), 81–91.
[70] C. P u c c i, Sui problemi di Cauchy non “ben posti”, Rend. Accad. Naz. Lincei 8 (18)
(1955), 473–477.
[71] —, Discussione del problema di Cauchy per le equazioni di tipo ellittico, Ann. Mat. Pura
Appl. 46 (1958), 131–153.
[72] —, Some Topics in Parabolic and Elliptic Equations, Lecture Series No. 36, Institute for
Fluid Mechanics and Applied Mathematics, University of Maryland, 1958.
[73] A. R o m a n o v i c h and A. V. F u r s i k o v, On the question of the solvability of the Cauchy
problem for the Laplace operator , Vestnik Moskov. Gos. Univ. Ser. I Mat. Mekh. 1987
(3), 78–80.
[74] J.-C. S a u t et B. S c h e u r e r, Sur l’unicité du problème de Cauchy et le prolongement
unique pour des équations elliptiques à coefficients non localement bornés, J. Differential
Equations 43 (1982), 28–43.
[75] —, —, Unique combination and uniqueness of the Cauchy problem for elliptic equations
with unbounded coefficients, in: Nonlinear Partial Differential Equations and their Ap-
plications, Collège de France Séminaire, Vol. V (Paris 1981/ 1982), Res. Notes in Math.
93, Pitman, Boston, Mass., 1983, 260–275.
[76] P. W. S c h a e f e r, On the Cauchy problem for an elliptic system, Arch. Rational Mech.
Anal. 20 (1965), 391–412.
[77] P. W. S c h a e f e r, Pointwise bounds in the Cauchy problem for an elliptic system, SIAM
J. Appl. Math. 15 (1967), 665–677.
[78] —, Improvable estimates in some non-well-posed problems for a system of elliptic equa-
tions, SIAM J. Math. Anal. 4 (1973), 447–455.
[79] —, On uniqueness, stability, and pointwise estimates in the Cauchy problem for coupled
elliptic equations, Quart. Appl. Math. 31 (1973), 321–328.
[80] S. D. S h a l a g i n o v, The Cauchy problem for the Laplace equation in a complex space,
Differentsial’nye Uravneniya 16 (1980), 947–949 (in Russian).
[81] A. N. T i k h o n o v and V. Y. A r s e n i n, Solution of Ill-posed Problems, Wiley, New York
1977 (Russian third ed., 1986).
[82] N. V. T k a c h e n k o, Solution of a certain Cauchy problem for the Laplace equation, in:
Math. Physics, Leningrad Gos. Ped. Inst., Leningrad 1987, 27–32 (in Russian).
 CAUCHY PROBLEM 127

[83] G. T r y t t e n, Pointwise bounds for solutions of the Cauchy problem for elliptic equations,
Arch. Rational Mech. Anal. 13 (1963), 222–244.
[84] T r a n D u c V a n, On the pseudodifferential operators with real analytic symbols and their
applications, J. Fac. Sci. Univ. Tokyo Sect. IA Math. 36 (1989), 803–825.
[85] T r a n D u c V a n, D i n h N h o H à o and R. G o r e n f l o, Approximating the solution to the
Cauchy problem and the boundary problem for the Laplace equation, to appear.
[86] T r a n D u c V a n, D i n h N h o H à o, T r i n h N g o c M i n h and R. G o r e n f l o, On the Cauchy
problems for systems of partial differential equations with a distinguished variable, Numer.
Funct. Anal. Optim. 12 (1&2) (1991), 213–236.
[87] T r a n D u c V a n, N g u y e n D u y T h a i S o n and D i n h Z u n g, Approximately solving
Cauchy problem for the wave equation by the method of differential operators of infinite
order , Acta Math. Vietnam. 13 (2) (1988), 127–136.
[88] T r i n h N g o c M i n h and T r a n D u c V a n, Cauchy problems for systems of partial dif-
ferential equations with a distinguished variable, Dokl. Akad. Nauk SSSR 284 (1985),
507–510; English transl.: Soviet Math. Dokl. 32 (1985), 562–565.
[89] T r i n h N g o c M i n h, Linear differential operators of infinite order and their applications,
Acta Math. Vietnam. 12 (1) (1987) 101–124.
[90] P. N. V a b i s h c h e v i c h, On the solution of the Cauchy problem for the Laplace equations
in a doubly connected domain, Dokl. Akad. Nauk SSSR 241 (1978), 1257–1260; English
transl.: Soviet Math. Dokl. 19 (1978), 976–980.
[91] —, Numerical solution of the Cauchy problem for elliptic equations and systems, Vestnik
Moskov. Univ. Ser. XV Vychisl. Mat. Kibernet. 1979 (3), 3–10 (in Russian).
[92] P. N. V a b i s h c h e v i c h, V. B. G l a s k o and Yu. A. K r i k s i n, Solution of the Hadamard
problem by a Tikhonov-regularizing algorithm, U.S.S.R. Comput. Math. and Math. Phys.
19 (1979), 103–112.
[93] P. N. V a b i s h c h e v i c h and P. A. P u l a t o v, A method of numerical solution of the
Cauchy problem for elliptic equations, Vestnik Moskov. Gos. Univ. Ser. XV Vychisl.
Mat. Kibernet. 1984, 3–8.
[94] K. W a t a n a b e, On the uniqueness of the Cauchy problem for certain elliptic equations
with triple characteristics, Tôhoku Math. J. (2) 23 (1971), 473–490.
[95] —, A unique continuation theorem for an elliptic operator of two independent variables
with nonsmooth double characteristics, Osaka J. Math. 10 (1973), 243–246.
[96] —, Remarque sur l’unicité dans les problèmes de Cauchy pour les opérateurs différentiels
elliptiques à caractéristiques de multiplicité constante et au plus triple, C. R. Acad. Sci.
Paris Sér. A–B 284 (1977), A1385–A1386.
[97] K. W a t a n a b e et C. Z u i l y, Unicité du problème de Cauchy pour une classe d’opérateurs
différentiels elliptiques à caractéristiques de multiplicité variable, C. R. Acad. Sci. Paris
Sér. A–B 283 (1976), A627–A630.
[98] —, —, Unicité du problème de Cauchy pour une classe d’opérateurs différentiels ellip-
tiques à caractéristiques de multiplicité variable, Sém. Goulaouic–Schwartz 1976/1977,
Équations aux dérivées partielles et analyse fonctionnelle, Exp. No. 15, 9 pp., Centre
Math., École Polytech., Palaiseau 1977.
[99] —, —, On the uniqueness of the Cauchy problem for elliptic differential operators with
smooth charateristics of variable multiplicity, Comm. Partial Differential Equations 2
(1977), 831–855.
[100] A. Y a n u s h a u s k a s, A Cauchy problem for Laplace’s equation and the multiplication
operation for harmonic functions, Dokl. Akad. Nauk SSSR 159 (1964), 286–289 (in Rus-
sian).
[101] —, Entire solutions of second order elliptic equations, Siberian Math. J. 11 (1970), 863–
869.
128 DINH NHO HÀO, TRAN DUC VAN AND R. GORENFLO

[102] A. Y a n u s h a u s k a s, On the Cauchy problem for the Laplace equation with three inde-
pendent variables, ibid. 16 (1975), 1040–1047.
[103] Sh. Y a r m u k h a m e d o v, On the Cauchy problem for Laplace’s equation, Dokl. Akad.
Nauk SSSR 235 (1977), 281–283, English transl.: Soviet Math. Dokl. 18 (1977), 939–882.
[104] —, The Cauchy problem for the Laplace equation in an infinite domain, Dokl. Akad.
Nauk UzSSR 1980 (12), 9–10 (in Russian).
[105] —, The Cauchy problem for the Laplace equation in M. M. Lavrent’ev’s formulation, in:
Ill-posed Problems of Mathematical Physics and Analysis, A. S. Alekseev (ed.), Nauka,
Sibirsk. Otdel., Novosibirsk 1984, 203–209 (in Russian).
[106] C. Z u i l y, Unicité du problème de Cauchy pour des opérateurs elliptiques à caractéris-
tiques de hautes multiplicités, Comm. Partial Differential Equations 10 (1985), 219–244.

Added in proof. Recently, there have been some new papers on the Cauchy problem for
elliptic equations:
1. D i n h N h o H à o, A mollification method for ill-posed problems, preprint A-92-35, FB
Mathematik, FU Berlin.
2. A. V. F u r s i k o v, The Cauchy problem for a second-order elliptic equation in a condition-
ally well-posed formulation, Trans. Moscow Math. Soc. 52 (1990), 139–176.
3. M. V. K l i b a n o v and F. S a n t o s a, A computational quasi-reversibility method for Cauchy
problems for Laplace’s equation, SIAM J. Appl. Math. 5 (1991), 1653–1675.