Professional Documents
Culture Documents
Integral Equations: Karlstad University Division For Engineering Science, Physics and Mathematics
Integral Equations: Karlstad University Division For Engineering Science, Physics and Mathematics
Integral Equations
A compendium
Karlstad 2002
Contents
1 Preface 4
2 Notion and examples of integral equations (IEs). Fredholm IEs of the first
and second kind 5
2.1 Primitive function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5 Unique solvability of the Fredholm IE of the 2nd kind using the contraction
mapping. Neumann series 12
5.1 Linear Fredholm IE of the 2nd kind . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.2 Nonlinear Fredholm IE of the 2nd kind . . . . . . . . . . . . . . . . . . . . . . . 13
5.3 Linear Volterra IE of the 2nd kind . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.4 Neumann series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.5 Resolvent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2
10.5 The HilbertSchmidt theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
19 Matrix representation 71
19.1 Matrix representation of an operator in the Hilbert space . . . . . . . . . . . . . 71
19.2 Summation operators in the spaces of sequences . . . . . . . . . . . . . . . . . . 72
3
1 Preface
This compendium focuses on fundamental notions and statements of the theory of linear oper-
ators and integral operators. The Fredholm theory is set forth both in the Banach and Hilbert
spaces. The elements of the theory of harmonic functions are presented, including the reduction
of boundary value problems to integral equations and potential theory, as well as some methods
of solution to integral equations. Some methods of numerical solution are also considered.
The compendium contains many examples that illustrate most important notions and the
solution techniques.
The material is largely based on the books Elements of the Theory of Functions and Func-
tional Analysis by A. Kolmogorov and S. Fomin (Dover, New York, 1999), Linear Integral
Equations by R. Kress (2nd edition, Springer, Berlin, 1999), and Logarithmic Integral Equa-
tions in Electromagnetics by Y. Shestopalov, Y. Smirnov, and E. Chernokozhin (VSP, Utrecht,
2000).
4
2 Notion and examples of integral equations (IEs). Fred-
holm IEs of the first and second kind
Consider an integral equation (IE)
Z b
f (x) = (x) + K(x, y)(y)dy. (1)
a
This is an IE of the 2nd kind. Here K(x, y) is a given function of two variables (the kernel)
defined in the square
= {(x, y) : a x b, a y b},
f (x) is a given function, (x) is a sought-for (unknown) function, and is a parameter.
We will often assume that f (x) is a continuous function.
The IE Z b
f (x) = K(x, y)(y)dy (2)
a
constitutes an example of an IE of the 1st kind.
We will say that equation (1) (resp. (2)) is the Fredholm IE of the 2nd kind (resp. 1st kind)
if the the kernel K(x, y) satisfies one of the following conditions:
(i) K(x, y) is continuous as a function of variables (x, y) in the square ;
(ii) K(x, y) is maybe discontinuous for some (x, y) but the double integral
Z bZ b
|K 2 (x, y)|dxdy < , (3)
a a
i.e., takes a finite value (converges in the square in the sense of the definition of convergent
Riemann integrals).
An IE Z x
f (x) = (x) + K(x, y)(y)dy, (4)
0
is called the Volterra IE of the 2nd kind. Here the kernel K(x, y) may be continuous in the
square = {(x, y) : a x b, a y b} for a certain b > a or discontinuous and satisfying
condition (3); f (x) is a given function; and is a parameter,
where the kernel K(x, y) = (x2 + y 2 ) is defined and continuous in the square 1 = {(x, y) : 0
x 1, 0 y 1}, f (x) = x2 is a given function, and the parameter = 1.
5
were the kernel K(x, y) = ln |x y| is discontinuous (unbounded) in the square 1 along the
line x = y, f (x) is a given function, and the parameter = 1. Equation (6) a Fredholm IE
of the 2nd kind because the double integral
Z 1Z 1
ln2 |x y|dxdy < , (7)
0 0
(the antiderivative) is differentiable on [a, b] and F 0 (x) = f (x) for each x [a, b] .
Example 3 F0 (x) = arctan x is a primitive function for the function
1
f (x) =
2 x(1 + x)
because
dF0 1
F00 (x) = = = f (x). (11)
dx 2 x(1 + x)
where the kernel K(x, y) = exy , f (x) is a given continuously differentiable function, and is
a parameter. Differentiating (12) we obtain
Z x
f (x) = (x) exy (y)dy
0
Z x
0 0
f (x) = (x) [(x) + exy (y)dy].
0
6
Subtracting termwise we obtain an ordinary differential equatin (ODE) of the first order
with respect to the (same) unknown function (x). Setting x = 0 in (12) we obtain the initial
condition for the unknown function (x)
Integrating (15) subject to the initial condition (16) we obtain the Volterra IE (12), which
is therefore equivalent to the initial value problem (15) and (16).
Solve (12) by the method of successive approximations. Set
K(x, y) = exy 0 y x 1,
K(x, y) = 0 0 x y 1.
where Z 1
(x y)2 xy
K3 (x, y) = K(x, t)K2 (t, y)dt = e . (23)
0 2!
7
The general relationship has the form
n
X Z 1
m
n (x) = f (x) + Km (x, y)f (y)dy, (24)
m=1 0
where
Z 1
(x y)m1 xy
K1 (x, y) = K(x, y), Km (x, y) = K(x, t)Km1 (t, y)dt = e . (25)
0 (m 1)!
Assuming that successive approximations (24) converge we obtain the (unique) solution to IE
(12) by passing to the limit in (24)
X Z 1
(x) = f (x) + m Km (x, y)f (y)dy. (26)
m=1 0
and changing the order of summation and integration in (26) we obtain the solution in the
compact form
Z 1
(x) = f (x) + (x, y, )f (y)dy. (28)
0
8
with the initial condition
y(x0 ) = y0 , (32)
where f (x, y) is defined and continuous in a two-dimensional domain G which contains the
point (x0 , y0 ).
Integrating (31) subject to (32) we obtain
Z x
(x) = y0 + f (t, (t))dt, (33)
x0
which is called the Volterra integral equation of the second kind with respect to the unknown
function (t). This equation is equivalent to the initial value problem (31) and (32). Note that
this is generally a nonlinear integral equation with respect to (t).
Consider now a linear ODE of the second order with variable coefficients
y 00 + A(x)y 0 + B(x)y = g(x) (34)
with the initial condition
y(x0 ) = y0 , y 0 (x0 ) = y1 , (35)
where A(x) and B(x) are given functions continuous in an interval G which contains the point
x0 . Integrating y 00 in (34) we obtain
Z x Z x Z x
0 0
y (x) = A(t)y (x)dx B(x)y(x)dx + g(x)dx + y1 . (36)
x0 x0 x0
Integrating the first integral on the right-hand side in (36) by parts yields
Z x Z x
y 0 (x) = A(x)y(x) (B(x) A0 (x))y(x)dx + g(x)dx + A(x0 )y0 + y1 . (37)
x0 x0
9
Example 5 Consider a homogeneous linear ODE of the second order with constant coefficients
y 00 + 2 y = 0 (44)
K(x, t) = 2 (t x),
f (x) = x,
we find that the IE (43), equivalent to the initial value problem (44) and (45), takes the form
Z x
y(x) = x + (t x)y(t)dt. (46)
0
is complete.
10
Example 7 The space C[a, b] of continuous functions defined in a closed interval a x b
with the metric
(1 , 2 ) = max |1 (x) 2 (x)|
axb
Definition 2 Let R be an arbitrary metric space. A mapping A of the space B into itself is
said to be a contraction if there exists a number < 1 such that
Theorem 2 Every contraction mapping defined in a complete metric space R has one and only
one fixed point; that is the equation Ax = x has only one solution.
We will use the principle of contraction mappings to prove Picards theorem.
Then on a some interval x0 d < x < x0 + d there exists a unique solution y = (x) to the
initial value problem
dy
= f (x, y), y(x0 ) = y0 . (52)
dx
Proof. Since function f (x, y) is continuous we have |f (x, y)| k in a region G0 G which
contains (x0 , y0 ). Now choose a number d such that
11
Let us prove that this is a contraction mapping of the complete set C = C[x0 d, x0 + d] of
continuous functions defined in a closed interval x0 d x x0 + d (see Example 7). We have
Z x Z x Z x
|(x) y0 | = f (t, (t))dt
|f (t, (t))| dt k dt = k(x x0 ) kd. (54)
x0 x0 x0
Z x
|1 (x) 2 (x)| |f (t, 1 (t)) f (t, 2 (t))| dt M d max |1 (x)| 2 (x). (55)
x0 x0 dxx0 +d
Since function f(t, y) is continuous we have ||f (x, y)|| K in a region G0 G which contains
(x0 , y0 ). Now choose a number d such that
(x, y) G0 if |x0 x| d, ||
y0 y|| Kd
M d < 1.
and prove that this is a contraction mapping of the complete space C = C[x
0 d, x0 + d] of
continuous (componentwise) vector-functions into itself. The proof is a termwise repetitiion of
the reasoning in the scalar case above.
12
where the kernel K(x, y) is a continuous function in the square
= {(x, y) : a x b, a y b},
If satisfies (63) we can prove in the same manner that the mapping g = Af defined by
Z b
g(x) = (x) + K(x, y, f (y))dy, (67)
a
of the complete space C[a, b] into itself is a contraction because for this mapping, the inequality
(61) holds.
13
5.3 Linear Volterra IE of the 2nd kind
Consider a Volterra IE of the 2nd kind
Z x
f (x) = (x) + K(x, y)(y)dy, (68)
a
where the kernel K(x, y) is a continuous function in the square for some b > a, so that
|K(x, y)| M for (x, y) .
Formulate a generalization of the principle of contraction mappings.
Theorem 4 If A is a continuous mapping of a complete metric space R into itself such that
the mapping An is a contraction for some n, then the equation Ax = x has one and only one
solution.
In fact if we take an arbitrary point x R and consider the sequence Akn x, k = 0, 1, 2, . . ., the
repetition of the proof of the classical principle of contraction mappings yields the convergence
of this sequence. Let x0 = limk Akn x, then Ax0 = x0 . In fact Ax0 = limk Akn Ax. Since
the mapping An is a contraction, we have
Consequently,
lim (Akn Ax, Akn x) = 0,
k
i.e., Ax0 = x0 .
Consider the mapping g = Af defined by
Z x
g(x) = (x) + K(x, y)(y)dy, (69)
a
and take the right-hand side f (x) as the zero approximation, setting
Substitute the zero approximation into the right-hand side of (73) to obtain the first approxi-
mation Z b
1 (x) = f (x) + K(x, y)0 (y)dy, (73)
a
14
and so on, obtaining for the (n + 1)st approximation
Z b
n+1 (x) = f (x) + K(x, y)n (y)dy. (74)
a
Assume that the kernel K(x, y) is a bounded function in the square = {(x, y) : a x
b, a y b}, so that
|K(x, y)| M, (x, y) .
or even that there exists a constant C1 such that
Z b
|K(x, y)|2 dy C1 x [a, b], (75)
a
Theorem 5 Assume that condition (75) holds Then the sequence n of successive approxima-
tions (74) converges uniformly for all satisfying
Z bZ b
1
, B= |K(x, y)|2 dxdy. (76)
B a a
where Z b
K3 (x, y) = K(x, t)K2 (t, y)dt. (83)
a
15
The general relationship has the form
n
X Z b
m
n (x) = f (x) + Km (x, y)f (y)dy, (84)
m=1 a
where Z b
K1 (x, y) = K(x, y), Km (x, y) = K(x, t)Km1 (t, y)dt. (85)
a
and Km (x, y) is called the mth iterated kernel. One can easliy prove that the iterated kernels
satisfy a more general relationship
Z b
Km (x, y) = Kr (x, t)Kmr (t, y)dt, r = 1, 2, . . . , m 1 (m = 2, 3, . . .). (86)
a
Assuming that successive approximations (84) converge we obtain the (unique) solution to IE
(70) by passing to the limit in (84)
X Z b
m
(x) = f (x) + Km (x, y)f (y)dy. (87)
m=1 a
which gives
Cm B 2 Cm1 , (92)
and finally the required estimate
Cm B 2m2 C1 . (93)
Denoting s
Z b
D= |f (y)|2 dy (94)
a
16
Thus the common term of the series (87) is estimated by the quantity
q
D C1 ||m B m1 , (96)
so that the series converges faster than the progression with the denominator ||B, which proves
the theorem.
If we take the first n terms in the series (87) then the resulting error will be not greater
than q ||n+1 B n
D C1 . (97)
1 ||B
5.5 Resolvent
Introduce the notation for the resolvent
X Z b
m1
(x, y, ) = Km (x, y) (98)
m=1 a
Changing the order of summation and integration in (87) we obtain the solution in the compact
form
Z b
(x) = f (x) + (x, y, )f (y)dy. (99)
a
is a normed and complete space. Indeed, every fundamental (convergent) functional sequence
{n (x)} of continuous functions converges to a continuous function in the C[a, b]-metric.
Definition 4 A set M in the metric space R is said to be compact if every sequence of elements
in M contains a subsequence that converges to some x R.
17
Definition 5 A family of functions {} defined on the closed interval [a, b] is said to be uni-
formly bounded if there exists a number M > 0 such that
|(x)| < M
Definition 6 A family of functions {} defined on the closed interval [a, b] is said to be equicon-
tinuous if for every > 0 there is a > 0 such that
for all x1 , x2 such that |x1 ) x2 | < and for all in the given family.
f = + A. (102)
and considered in the (Banach) space C[a, b] of continuous functions in a closed interval a
x b represents an extensive class of completely continuous linear operators.
Theorem 7 Formula (103) defines a completely continuous linear operator in the (Banach)
space C[a, b] if the function K(x, y) is to be bounded in the square
= {(x, y) : a x b, a y b},
and all points of discontinuity of the function K(x, y) lie on the finite number of curves
y = k (x), k = 1, 2, . . . n, (104)
18
Proof. To prove the theorem, it is sufficient to show (according to Arzelas theorem) that the
functions (103) defined on the closed interval [a, b] (i) are continuous and the family of these
functions is (ii) uniformly bounded and (iii) equicontinuous. Below we will prove statements
(i), (ii), and (iii).
Under the conditions of the theorem the integral in (103) exists for arbitrary x [a, b]. Set
and denote by G the set of points (x, y) for which the condition
|y k (x)| < , k = 1, 2, . . . n, (105)
12M n
holds. Denote by F the complement of G with respect to . Since F is a closed set and K(x, y)
is continuous on F , there exists a > 0 such that for points (x0 , y) and (x00 , y) in F for which
|x0 x00 | < the inequality
1
|K(x0 , y) K(x00 , y)| < (b a) (106)
3
holds. Now let x0 and x00 be such that |x0 x00 | < holds. Then
Z b
0 00
|f (x ) f (x )| |K(x0 , y) K(x00 , y)||(y)|dy (107)
a
19
6.3 Properties of completely continuous operators
Theorem 8 If {An } is a sequence of completely continuous operators on a Banach space E
which converges in (operator) norm to an operator A then the operator A is completely contin-
uous.
Proof. Consider an arbitrary bounded sequence {xn } of elements in a Banach space E such
that ||xn || < c. It is necessary to prove that sequence {Axn } contains a convergent subsequence.
The operator A1 is completely continuous and therefore we can select a convergent subse-
quence from the elements {A1 xn }. Let
(1) (1)
x1 , x2 , . . . , x(1)
n , ..., (112)
be the inverse images of the members of this convergent subsequence. We apply operator A2 to
each members of subsequence (112)). The operator A2 is completely continuous; therefore we
can again select a convergent subsequence from the elements {A2 x(1)
n }, denoting by
(2) (2)
x1 , x2 , . . . , x(2)
n , ..., (113)
the inverse images of the members of this convergent subsequence. The operator A3 is also com-
pletely continuous; therefore we can again select a convergent subsequence from the elements
{A3 x(2)
n }, denoting by
(3) (3)
x1 , x2 , . . . , x(3)
n , ..., (114)
the inverse images of the members of this convergent subsequence; and so on. Now we can form
a diagonal subsequence
(1) (2)
x1 , x2 , . . . , x(n)
n , .... (115)
Each of the operators An transforms this diagonal subsequence into a convergent sequence.
If we show that operator A also transforms (115)) into a convergent sequence, then this will
establish its complete continuity. To this end, apply the Cauchy criterion (48) and evaluate the
norm
||Ax(n) (m)
n Axm ||. (116)
We have the chain of inequalities
Choose k so that ||A Ak || < 2c
and N such that
||Ak x(n) (m)
n Ak xm || <
2
holds for arbitrary n > N and m > N (which is possible because the sequence ||Ak x(n)
n ||
converges). As a result,
||Axn(n) Ax(m)
m || < , (117)
i.e., the sequence ||Ax(n)
n || is fundamental.
20
Theorem 9 (Superposition principle). If A is a completely continuous operator and B is a
bounded operator then the operators AB and BA are completely continuous.
Proof. If M is a bounded set in a Banach space E, then the set BM is also bounded. Conse-
quently, the set ABM is compact and this means that the operator AB is completely continuous.
Further, if M is a bounded set in E, then AM is compact. Then, because B is a continuous
operator, the set BAM is also compact, and this completes the proof.
Proof. In the contrary case, the identity operator I = AA1 would be completely continuous
in E which is impossible.
Definition 8 The number is said to be a characteristic value of the operator A if there exists
a nonzero vector x (characteristic vector) such that Ax = x.
21
remainder of the spectrum: those for which A I has a (bounded) inverse whose domain is
not dense in E, this part is called the residual spectrum.
If the point is regular, i.e., if the inverse of the operator A I exists, then for sufficiently
small the operator A ( + )I also has the inverse, i.e., the point + is regular also. Thus
the regular points form an open set. Consequently, the spectrum, which is a complement, is a
closed set.
Theorem 11 The inverse of the operator A I exists for arbitrary for which || > ||A||.
Proof. Since
1
A I = I A ,
the resolvent 1
1 A 1X Ak
R = (A I)1 = I = .
k=0 k
This operator series converges for ||A/|| < 1, i.e., the operator A I has the inverse.
Corollary. The spectrum of the operator A is contained in a circle of radius ||A|| with center
at zero.
Theorem 12 Every completely continuous operator A in the Banach space E has for arbitrary
> 0 only a finite number of linearly independent characteristic vectors which correspond to
the characteristic values whose absolute values are greater than .
Proof. Assume, ad abs, that there exist an infinite number of linearly independent characteristic
vectors xn , n = 1, 2, . . ., satisfying the conditions
x = 1 x1 + . . . + k xk , (119)
Ax = 1 1 x1 + . . . + k k xk , (120)
and, consequently,
A1 x = (1 /1 )x1 + . . . + (k /k )xk (121)
and
1
||A1 x|| < ||x||.
Therefore, the operator A1 defined for vectors of the form (119) by means of (120) can be ex-
tended by continuity (and, consequently, with preservation of norm) to all of E1 . The operator
22
A, being completely continuous in the Banach space E, is completely continuous in E1 . Accord-
ing to the corollary to Theorem (9), a completely continuous operator cannot have a bounded
inverse in infinite-dimensional space. The contradiction thus obtained proves the theorem.
Proof. Assume the contrary: there exists an element x1 6= 0 such that x1 Ax1 = T x1 = 0
Those elements x for which T x = 0 form a linear subspace E1 of the banach space E. Denote
by En the subspace of E consisting of elements x for which the powers T n x = 0.
It is clear that subspaces En form a nondecreasing subsequence
E1 E2 . . . En . . . .
We shall show that the equality sign cannot hold at any point of this chain of inclusions. In
fact, since x1 6= 0 and equation y = T x is solvable for arbitrary y, we can find a sequence of
elements x1 , x2 , . . . , xn , . . . distinct from zero such that
T x 2 = x1 ,
T x 3 = x2 ,
...,
T xn = xn1 ,
....
The element xn belongs to subspace En for each n but does not belong to subspace En1 . In
fact,
T n xn = T n1 xn1 = . . . = T x1 = 0,
but
T n1 xn = T n2 xn1 = . . . = T x2 = x1 6= 0.
23
All the subspaces En are linear and closed; therefore for arbitrary n there exists an element
yn+1 En+1 such that
1
||yn+1 || = 1 and (yn+1 , En ) ,
2
where (yn+1 , En ) denotes the distance from yn+1 to the space En :
Corollary 1. If the equation y = x Ax is solvable for arbitrary y, then this equation has
unique solution for each y, i.e., the operator I A has an inverse in this case.
It is easy to verify this statement by recalling that the operator adjoint to a completely
continuous operator is also completely continuous, and the space E conjugate to a Banach
space is itself a Banach space.
24
These conditions are equivalent to the following:
f1 (y1 ) 6= 0, f1 (x Ax) = 0 x.
In fact,
(f1 A f1 )(x) = f1 (x) A f1 (x) = f1 (x) f1 (Ax) = f1 (x Ax).
Let G0 be a subspace consisting of all elements of the form x Ax. Consider the subspace
{G0 , y1 } of the elements of the form z + y1 , z G0 and define the linear functional by setting
f1 (z + y1 ) = .
Extending the linear functional to the whole space E (which is possible by virtue of the Hahn
Banach theorem) we do obtain a linear functional satisfying the required conditions. This
completes the proof of the theorem.
Corollary If the equation f A f = 0 does not have nonzero solution, then the equation
x Ax = y is solvable for all y.
i.e., h(x) = 0 if x Ax = 0.
2. Now assume that h(x) = 0 for all x such that xAx = 0. We shall show that the equation
f A f = h is solvable. Consider the set F of all elements of the form y = x Ax. Construct
a functional f on F by setting
f (T x) = h(x).
This equation defines in fact a linear functional. Its value is defined uniquely for each y because
if T x1 = T x2 then h(x1 ) = h(x2 ). It is easy to verify the linearity of the functional and extend
it to the whole space E. We obtain
f (T x) = T f (x) = h(x).
i.e., the functional is a solution of equation f A f = h. This completes the proof of the
theorem.
Corollary If the equation x Ax = 0 does not have nonzero solution, then the equation
f A f = h is solvable for all h.
Theorem 16 If the equation x Ax = 0 has x = 0 for its only solution, then the equation
x Ax = y has a solution for all y.
25
Proof. If the equation x Ax = 0 has only one solution, then by virtue of the corollary to
the preceding theorem, the equation f A f = h is solvable for all h. Then, by Corollary 2
to Thereom 13, the equation f A f = 0 has f = 0 for its only solution. Hence Corollary to
Thereom 14 implies that the equation x Ax = y has a solution for all y.
Thereoms 13 and 16 show that for the equation
x Ax = y (122)
8.2 Fredholm theorems for linear integral operators and the Fred-
holm resolvent
One can formulate the Fredholm theorems for linear integral operators
Z b
A(x) = K(x, y)(y)dy (123)
a
will be called regular values. The numbers for which the Fredholm resolvent (x, y, ) is not
defined (they belong to the point spectrum) will be characteristic values of the IE (124), and
they coincide with the poles of the resolvent. The inverse of characteristic values are eigenvalues
of the IE.
Formulate, for example, the first Fredholm theorem in terms of regular values of the resol-
vent.
If is a regular value, then the Fredholm IE (124) has one and only one solution for arbitrary
f (x). This solution is given by formula (99),
Z b
(x) = f (x) + (x, y, )f (y)dy. (125)
a
26
In particular, if is a regular value, then the homogeneous Fredholm IE (124)
Z b
(x) K(x, y)(y)dy = 0 (126)
a
can be represented, by changing the order of summation and integration, in the form
n
X Z b
(x) ai (x) bi (y)(y)dy = f (x). (129)
i=1 a
Here one may assume that functions ai (x) (and bi (y) ) are linearly independent (otherwise, the
numer of terms in (128)) can be reduced).
It is easy to solve such IEs with degenerate and separable kernels. Denote
Z b
ci = bi (y)(y)dy, (130)
a
and the problem reduces to the determination of unknowns ci . To this end, substitute (131)
into equation (129) to obtain, after simple algebra,
n
( Z b " n
# )
X X
ai (x) ci bi (y) f (y) + ck ak (y) dy = 0. (132)
i=1 a k=1
27
Denoting
Z b Z b
fi = bi (y)f (y)dy, aik = bi (y)ak (y)dy (134)
a a
D() is a polynomial in powers of of an order not higher than n. Note that D() is not
identically zero because D(0) = 1. Thus there exist not more than n different numbers = k
such that D(k ) = 0. When = k , then system (135) together with IE (129) are either
not solvable or have infinitely many solutions. If 6= k , then system (135) and IE (129) are
uniquely solvable.
Example 10 Consider a Fredholm IE of the 2nd kind
Z 1
(x) (x + y)(y)dy = f (x). (137)
0
Here
2
X
K(x, y) = x + y = ai (x)bi (y) = a1 (x)b1 (y) + a2 (x)b2 (y),
i=1
where
a1 (x) = x, b1 (y) = 1, a2 (x) = 1, b2 (y) = y,
is a degenerate kernel, f (x) is a given function, and is a parameter.
Look for the solution to (137) in the form (131) (note that here n = 2)
2
X
(x) = f (x) + ci ai (x) = f (x) + (c1 a1 (x) + c2 a2 (x)) = f (x) + (c1 x + c2 ). (138)
i=1
Denoting
Z 1 Z 1
fi = bi (y)f (y)dy, aik = bi (y)ak (y)dy, (139)
0 0
28
so that
Z 1 Z 1
1
a11 = b1 (y)a1 (y)dy = ydy = ,
0 0 2
Z 1 Z 1
a12 = b1 (y)a2 (y)dy = dy = 1, (140)
0 0
Z 1 Z 1
1
a21 = b2 (y)a1 (y)dy = y 2 dy = ,
0 0 3
Z 1 Z 1
1
a22 = b2 (y)a2 (y)dy = ydy = ,
0 0 2
and
Z 1 Z 1
f1 = b1 (y)f (y)dy = f (y)dy, (141)
0 0
Z 1 Z 1
f2 = b2 (y)f (y)dy = yf (y)dy,
0 0
and changing the order of summation and integration, we rewrite the corresponding equation
(133) as
2
X
ci aik ck = fi , i = 1, 2, (142)
k=1
or, explicitly,
(1 /2)c1 c2 = f1 ,
(/3)c1 + (1 /2)c2 = f2 ,
29
f
D1 = 1
= f1 (1 /2) + f2 =
f2 1 /2
Z 1 Z 1 Z 1
= (1 /2) f (y)dy + yf (y)dy = [(1 /2) + y]f (y)dy, (146)
0 0 0
(1 /2) f
1
D2 = = f2 (1 /2) + f1 /3 =
/3 f2
Z 1 Z 1 Z 1
= (1 /2) yf (y)dy + (/3) f (y)dy = [y(1 /2) + /3]f (y)dy, (147)
0 0 0
Here
K(x, y) = sin x cos y = a(x)b(y)
is a degenerate kernel, f (x) is a given function, and is a parameter.
Rewrite IE (151) in the form (131) (here n = 1)
Z 2
(x) = f (x) + c sin x, c= cos y(y)dy. (152)
0
Multiplying the first equality in (152) by b(x) = cos x and integrating from 0 to 2 we obtain
Z 2
c= f (x) cos xdx (153)
0
30
Therefore the solution is
Z 2
(x) = f (x) + c sin x = f (x) + sin x cos yf (y)dy. (154)
0
where K(x, y) = x2 + y 2 is a degenerate kernel f (x) = x2 is a given function, and the parameter
= 1.
the kernel K(x, y and f (x) are continuous functions in the square
= {(x, y) : a x b, a y b}.
Then (x) will be also continuous. Approximate IE (156) with an IE having a degenerate kernel.
To this end, replace the integral in (156) by a finite sum using, e.g., the rectangle rule
Z b n
X
K(x, y)(y)dy h K(x, yk )(yk ), (157)
a k=1
where
ba
h= , yk = a + kh. (158)
n
The resulting approximation takes the form
n
X
(x) h K(x, yk )(yk ) = f (x) (159)
k=1
31
calculating the determinants
1 hK(x1 , y1 ) hK(x1 , y2 ) . . . hK(x1 , yn )
hK(x , y ) 1 hK(x2 , y2 ) . . . hK(x2 , yn )
2 1
D(n) () = . (162)
. . ... .
hK(x , y ) hK(x , y ) . . . 1 hK(xn , yn )
n 1 n 2
and so on.
To reconstruct the approximate solution (n) using the determined {i } one can use the IE
(159) itself:
n
X
(x) (n) = h K(x, yk )k + f (x). (163)
k=1
One can prove that (n) (x) as n where (x) is the exact solution to IE (159).
Solving this IE (164) by the described approximate method and passing to the limit n
we obtain the resolvent as a ratio of two power series
D(x, y, )
(x, y, ) = , (165)
D()
where
X (1)n
D(x, y, ) = Bn (x, y)n , (166)
n=0 n!
X (1)n
D() = cn n , (167)
n=0 n!
K(x, y) K(x, y1 ) . . . K(x, yn )
Z b Z b
K(y1 , y) K(y1 , y1 ) . . . K(y1 , yn )
Bn (x, y) = ... dy1 . . . dyn ; (168)
a a . . ... .
K(y , y) K(y , y ) . . . K(yn , yn )
n n 1
c0 = 1,
32
K(y1 , y1 ) K(y1 , y2 ) . . . K(y1 , yn )
Z b Z b
K(y2 , y1 ) K(y2 , y2 ) . . . K(y2 , yn )
cn = ... dy1 . . . dyn . (169)
a a . . ... .
K(y , y ) K(y , y ) . . . K(yn , yn )
n 1 n 2
D() is called the Fredholm determinant, and D(x, y, ), the first Fredholm minor.
If the integral Z Z b b
|K(x, y)|2 dxdy < , (170)
a a
then seires (167) for the Fredholm determinant converges for all complex , so that D() is an
entire function of . Thus (164)(167) define the resolvent on the whole complex plane , with
an exception of zeros of D() which are poles of the resolvent.
We may conclude that for all complex that do not coincide with the poles of the resolvent,
the IE Z b
(x) K(x, y)(y)dy = f (x) (171)
a
is uniquely solvable, and its solution is given by formula (99)
Z b
(x) = f (x) + (x, y, )f (y)dy. (172)
a
For practical computation of the resolvent, one can use the following recurrent relationships:
B0 (x, y) = K(x, y), c0 = 1,
Z b
cn = Bn1 (y, y)dy, n = 1, 2, . . . , (173)
a
Z b
Bn (x, y) = cn K(x, y) n K(x, t)Bn1 (t, y)dt, n = 1, 2, . . . . (174)
a
Example 13 Find the resolvent of the Fredholm IE (137) of the 2nd kind
Z 1
(x) (x + y)(y)dy = f (x). (175)
0
We will use recurrent relationships (173) and (174). Here
K(x, y) = x + y
is a degenerate kernel. We have c0 = 1, Bi = Bi (x, y) (i = 0, 1, 2), and
B0 = K(x, y) = x + y,
Z 1 Z 1
c1 = B0 (y, y)dy = 2ydy = 1,
0 0
Z 1
B1 = c1 K(x, y) K(x, t)B0 (t, y)dt =
0
Z 1
1 1
= x+y (x + t)(t + y)dt = (x + y) xy ,
0 2 3
Z 1 Z 1
1 1 1
c2 = B1 (y, y)dy = (y + y) y 2 dy = ,
0 0 2 3 6
Z 1
B2 = c2 K(x, y) K(x, t)B1 (t, y)dt =
0
Z 1
1 1 1
= (x + y) 2 (x + t) (t + y) ty dt = 0.
6 0 2 3
33
Since B2 (x, y) 0, all subsequent c3 , c4 , . . ., and B3 , B4 , . . ., vanish (see (173) and (174)),
and we obtain
D(x, y, ) =
1
X (1)n
= Bn (x, y)n = B0 (x, y) B1 (x, y) =
n=0 n!
1 1
= x + y (x + y) xy , (176)
2 3
D() =
2
X (1)n 1
= cn n = c0 c1 + c2 2 =
n=0 n! 2
1 1
= 1 2 = (2 + 12 12)
12 12
and
D(x, y, ) x + y 12 (x + y) xy 1
3
)
(x, y, ) = = 12 . (177)
D() 2 + 12 12
The solution to (175) is given by formula (99)
Z 1
(x) = f (x) + (x, y, )f (y)dy,
0
(x, y) = (y, x) x, y R,
(x1 + x2 , y) = (x1 , y) + (x2 , y), x1 , x2 , y R,
(x, y) = (x, y), x, y R,
(x, x) 0, x R, (x, x) = 0 if and only if x = 0.
34
is called the Euclidian space.
In a complex Euclidian space the inner product satisfies
(x, y) = (y, x),
(x1 + x2 , y) = (x1 , y) + (x2 , y),
(x, y) = (x, y),
(x, x) 0, (x, x) = 0 if and only if x = 0.
Note that in a complex linear space
y).
(x, y) = (x,
The norm in the Euclidian space is introduced by
||x|| = x, x.
The Hilbert space H is a complete (in the metric (x, y) = ||x y||) infinite-dimensional
Euclidian space.
In the Hilbert space H, the operator A adjoint to an operator A is defined by
(Ax, y) = (x, A y), x, y H.
The selfadjoint operator A = A is defined from
(Ax, y) = (x, Ay), x, y H.
Assume that K(x, y) is a HilbertSchmidt kernel, i.e., a square-integrable function in the square
= {(x, y) : a x b, a y b}, so that
Z bZ b
|K(x, y)|2 dxdy , (180)
a a
35
Theorem 18 Equality (182) and condition (180) define a completely continuous linear opera-
tor in the space L2 [a, b]. The norm of this operator is estimated as
s
Z bZ b
||A|| |K(x, y)|2 dxdy. (183)
a a
Proof. Note first of all that we have already mentioned (see (75)) that if condition (180) holds,
then there exists a constant C1 such that
Z b
|K(x, y)|2 dy C1 almost everywhere in [a, b], (184)
a
i.e., K(x, y) L2 [a, b] as a function of y almost for all x [a, b]. Therefore the function
Z b
(x) = K(x, y)(y)dy is defined almost everywhere in [a, b] (185)
a
We will now show that (x) L2 [a, b]. Applying the Schwartz inequality we obtain
Z 2 Z b Z b
b
2
|(x)| = K(x, y)(y)dy |K(x, y)|2 dy |(y)|2 dy = (186)
a a a
Z b
= ||||2 |K(x, y)|2 dy.
a
Integrating with respect to x and replacing an iterated integral of |K(x, y)|2 by a double integral,
we obtain Z b Z bZ b
2 2 2
||A|| = |(y)| dy |||| |K(x, y)|2 dxdy, (187)
a a a
which proves (1) that (x) is a square-integrable function, i.e.,
Z b
|(x)|2 dx ,
a
By setting
N
X
KN (x, y) = amn m (x)n (y). (189)
m,n=1
define an operator AN which is finite-dimensional (because it has a degenerate kernel and maps
therefore L2 [a, b] into a finite-dimensional space generated by a finite system {m }N
m=1 ) and
therefore completely continuous.
36
Now note that KN (x, y) in (189) is a partial sum of the Fourier series for K(x, y), so that
Z bZ b
|K(x, y) KN (x, y)|2 dxdy 0, N . (190)
a a
||A AN || 0, N . (191)
Proof. Indeed, let and be two different eigenvalues of a selfadjoint operator A, i.e.,
Ax = x, x H, x 6= 0; Ay = y, y H, y 6= 0.
Then we have
(x, y) = (x, y) = (Ax, y) = (x, Ay) = (x, y) = (x, y), (193)
which gives
( )(x, y) = 0, (194)
i.e., (x, y) = 0.
Below we will prove all the Fredholm theorems for integral operators in the Hilbert space.
37
in the operator form (182),
= A + f, (197)
or
T = f, T = I A. (198)
T 0 = 0, (199)
T = g, T = I A , (200)
and
T 0 = 0. (201)
We will consider equations (198)(201) in the Hilbert space H = L2 [a, b]. Note that (198)
(201) may be considered as operator equations in H with an arbitrary completely continuous
operator A.
The four Fredholm theorems that are formulated below establish relations between the
solutions to these four equations.
Theorem 21 The inhomogeneous equation T = f is solvable if and only if f is orthogonal to
every solution of the adjoint homogeneous equation T 0 = 0 .
To prove the Fredholm theorems, we will use the definiton of a subspace in the Hilbert space
H:
M is a linear subspace of H if M closed and f, g M yields f + g M where and are
arbitrary numbers,
and the following two properties of the Hilbert space H:
(i) if h is an arbitrary element in H, then the set horth of all elements f H orthogonal to h,
horth = {f : f H, (f, h) = 0} form a linear subspace of H which is called the orthogonal
complement,
and
(ii) if M is a (closed) linear subspace of H, then arbitrary element f H is uniquely represented
as f = h + h0 , where h M , and h0 M orth and M orth is the orthogonal complement to M .
38
Proof. Assume that yn Im T and yn y. To prove the lemma, one must show that y Im T ,
i.e., y = T x. We will now prove the latter statement.
According to the definition of Im T , there exist vectors xn H such that
yn = T xn = xn Axn . (202)
Lemma 2 The space H is the direct orthogonal sum of closed subspaces Ker T and Im T ,
Ker T + Im T = H, (203)
and also
Ker T + Im T = H. (204)
Proof. The closed subspaces Ker T and Im T are orthogonal because if h Ker T then (h, T x) =
(T h, x) = (0, x) = 0 for all x H. It is left to prove there is no nonzero vector orthogonal to
Ker T and Im T . Indeed, if a vector z is orthogonal to Im T , then (T z, x) = (z, T x) = 0 for
all x H, i.e., z Ker T . The lemma is proved.
Lemma 2 yields the first Fredholm theorem 21. Indeed, f is orthogonal to Ker T if and
only if f Im T , i.e., there exists a vector such that T = f .
Next, set H k = Im T k for every k, so that, in particular, H 1 = Im T . It is clear that
. . . H k . . . H 2 H 1 H, (205)
39
Lemma 3 There exists a number j such that H k+1 = H k for all k j.
Proof. If such a j does not exist, then all H k are different, and one can construct an orthonormal
sequence {xk } such that xk H k and xk is orthogonal to H k+1 . Let l > k. Then
The totality of statements of Lemmas 4 and 5 constitutes the proof of Fredholms alternative
22.
Now let us prove the third Fredholms Theorem 23.
To this end, assume that Ker T is an infinite-dimensional space: dimKer T = . Then this
space
conyains an infinite orthonormal sequence {xn } such that Axn = xn and ||Axl Axk || =
2 if k 6= l. Therefore, sequence {Axn } does not contain a convergent subsequence, which
contradicts to the fact that A is a completely continuous operator.
Now set = dimKer T and = dimKer T and assume that < . Let 1 , . . . , be an
orthonormal basis in Ker T and 1 , . . . , be an orthonormal basis in Ker T . Set
X
Sx = T x + (x, j )j .
j=1
Operator S is a sum of T and a finite-dimensional operator; therefore all the results proved
above for T remain valid for S.
Let us show that the homogeneous equation Sx = 0 has only a trivial solution. Indeed,
assume that
X
Tx + (x, j )j = 0. (206)
j=1
According to Lemmas 2 vectors psij are orthogonal to all vectors of the form T x (206) yields
Tx = 0
and
(x, j ) = 0, quad1 j .
40
We see that, on the one hand, vector x must be a linear combination of vectors psij , and on
the other hand, vector x must be orthogonal to these vectors. Therefore, x = 0 and equation
Sx = 0 has only a trivial solution. Thus according to the second Fredholms Theorem 22 there
exists a vector y such that
X
Ty + (y, j )j = +1 .
j=1
Calculating the inner product of both sides of this equality and +1 we obtain
X
(T y, +1 ) + (y, j )(j , +1 ) = (+1 , +1 ).
j=1
and
k 0, n , (209)
if {k } is an infinite system.
41
11 IEs with symmetric kernels. HilbertSchmidt theory
Consider first a general a HilbertSchmidt operator (181).
Theorem 25 Let A be a HilbertSchmidt operator with a (HilbertSchmidt) kernel K(x, y).
Then the operator A adjoint to A is an operator with the adjoint kernel K (x, y) = K(y, x).
Proof. Using the Fubini theorem and changing the order of integration, we obtain
Z b (Z b )
(Af, g) = K(x, y)f (y)dy g(x)dx =
a a
Z bZ b
= K(x, y)f (y)g(x)dydx =
a
Z b (aZ )
b
= K(x, y)g(x)dx f (y)dy =
a a
Z b (Z )
b
= f (y) K(x, y)g(x)dx dy,
a a
Consider a symmetric IE
Z b
(x) K(x, y)(y)dy = f (x) (210)
a
for complex-valued or
42
Multiply equality (213) by 0 and integrate with respect to x between a and b to obtain
which yields
||0 ||2
0 = . (215)
(K0 , 0 )
||0 ||2
0 = , = (K0 , 0 ), (216)
and prove that is a real number. For a symmetric kernel, we have
= (K0 , 0 ) = (0 , K0 ). (217)
(0 , K0 ) = (K0 , 0 ).
Thus
= (K0 , 0 ) = = (K0 , 0 ),
Proof. Let and and and be two different eigenvalues and the corresponding eigen-
functions of an integral operator with a symmetric kernel, i.e.,
Z b
(x) = K(x, y) (y)dy = 0, (218)
a
Z b
(x) = K(x, y) (y)dy = 0. (219)
a
Then we have
( , ) = ( , ) = (A , ) = ( , A ) = ( , ) = ( , ), (220)
which yields
( )( , ) = 0, (221)
i.e., ( , ) = 0.
One can also reformulate the HilbertSchmidt theorem 24 for integral operators.
43
Theorem 28 Let 1 , 2 , . . . , and 1 , 2 , . . . , be eigenvalues and eigenfunctions (eigenvectors)
of an integral operator with a symmetric kernel and let h(x) L2 [a, b], i.e.,
Z b
|h(x)|2 dx .
a
The Fourier coefficients fn of the function f (x) are coupled with the Fourier coefficients hn of
the function h(x) by the relationships
hn
fn = , hn = (h, n ),
n
so that
X hn
f (x) = n (x) (fn = (f, n ), hn = (h, n )). (224)
n=1 n
and n (y) are the Fourier coefficients of the kernel K(x, y). Let us calculate n (y). We have,
using the formula for the Fourier coefficients,
Z b
n (y) = K(x, y)n (x)dx, (226)
a
44
Replacing x by y abd vice versa, we obtain
Z b
1
n (y) = K(y, x)n (x)dx, (229)
n a
so that
1
n (y) = n (y).
n
Now we have
X n (x)n (y)
K2 (x, y) = . (230)
n=1 2n
and, generally,
X n (x)n (y)
Km (x, y) = . (231)
n=1 m
n
This series may diverge in the sense of C-norm (i.e., uniformly); however it alwyas converges
in L2 -norm.
According to Theorem 18, A is a completely continuous selfadjoint operator in the space L2 [a, b]
and we can apply the HilbertSchmidt theorem 24 to prove the following statement.
45
Theorem 29 If 1 is not an eigenvalue of the operator A then the IE (210), written in the
operator form as
has one and only one solution for every f . If 1 is an eigenvalue of the operator A, then IE
(210) (or (234)) is solvable if and only if f is orthogonal to all eigenfunctions of the operator A
corresponding to the eigenvalue = 1, and in this case IE (210) has infinitely many solutions.
f 0 = 0 ,
xk (1 k ) = bk (k = 1, 2, . . .)
i.e., when
f 0 = 0 ,
bk
xk = , k 6= 1
1 k
bk = 0, k = 1. (239)
The latter gives the necessary and sufficient condition for the solvability of equation (234). Note
that the coefficients xn that correspond to those n for which n = 1 are arbitrary numbers.
The theorem is proved.
46
12 Harmonic functions and Greens formulas
Denote by D R2 a two-dimensional domain bounded by the closed smooth contour .
A twice continuously differentiable real-valued function u defined on a domain D is called
harmonic if it satisfies Laplaces equation
u = 0 in D, (240)
where
2u 2u
u = + (241)
x2 y 2
is called Laplaces operator (Laplacian), the function u = u(x), and x = (x, y) R2 . We will
also use the notation y = (x0 , y0 ).
The function
1 1
(x, y) = (x y) = ln (242)
2 |x y|
is called the fundamental solution of Laplaces equation. For a fixed y R2 , y 6= x, the function
(x, y) is harmonic, i.e., satisfies Laplaces equation
2 2
+ = 0 in D. (243)
x2 y 2
The proof follows by straightforward differentiation.
47
(with the components A1 . A2 being
to the vector function (vector field) A = (A1 .A2 ) C 1 (D)
once continuously differentiable in the closed domain D = D + , Ai C 1 (D),
i = 1, 2) defined
by " #
v v
A = u grad v = u , u , (247)
x y
the components are
v v
A1 = u , A2 = u . (248)
x y
We have ! !
v v
div A = div (u grad v) = u + u = (249)
x x y y
u v 2 v u v 2v
+u 2 + + u 2 = grad u grad v + uv. (250)
x x x y y y
On the other hand, according to the definition of the normal derivative,
v
ny A = ny (ugrad v) = u(ny grad v) = u , (251)
ny
so that, finally,
Z Z Z Z Z Z
v
div Adx = (uv + grad u grad v)dx, ny Ad ly = u dly , (252)
ny
D D
which proves Greens first formula. By interchanging u and v and subtracting we obtain the
desired Greens second formula (245).
be harmonic in the domain D.
Let a twice continuously differentiable function u C 2 (D)
Then Greens third theorem (Greens third formula) is valid
Z !
u (x, y)
u(x) = (x, y) u(y) dly , x D. (253)
ny ny
of radius r (the boundary of a vicinity of a point x D) such that (x, r) D and direct the
unit normal n to (x, r) into the exterior of (x, r). Then we apply Greens second formula
(245) to the harmonic function u and (x, y) (which is also a harmonic function) in the domain
y D : |x y| > r to obtain
Z Z
0= (u(x, y) (x, y)u)dx = (254)
D\(x,r)
Z !
(x, y) u(y)
u(y) (x, y) dly . (255)
ny ny
(x,r)
48
Since on (x, r) we have
ny
grad y (x, y) = , (256)
2r
a straightforward calculation using the mean-value theorem and the fact that
Z
v
dly = 0 (257)
ny
Proof follows from the first Greens formula applied to two harmonic functions v and u = 1.
i.e., the value of u in the center of the ball is equal to the integral mean values over both the
ball and its boundary.
Proof. For each 0 < < r we apply (259) and Greens third formula to obtain
Z
1
u(x) = u(y)dly , (261)
2
|xy|=
whence the second mean value formula in (260) follows by passing to the limit r. Multi-
plying (261) by and integrating with respect to from 0 to r we obtain the first mean value
formula in (260).
49
Theorem 32 (MaximumMinimum Principle) A harmonic function on a domain cannot
attain its maximum or its minimum unless it is constant.
Proof. Assume that the interior Dirichlet problem has two solutions, u1 and u2 . Their difference
u = u1 u2 is a harmonic function that is continuous up to the boundary and satisifes the
homogeneous boundary condition u = 0 on the boundary of D. Then from the maximum
minimum principle or its corollary we obtain u = 0 in D, which proves the theorem.
Theorem 34 Two solutions of the interior Neumann problem can differ only by a constant.
The exterior Neumann problem has at most one solution.
Proof. We will prove the first statement of the theorem. Assume that the interior Neumann
problem has two solutions, u1 and u2 . Their difference u = u1 u2 is a harmonic function that
u
is continuous up to the boundary and satisifes the homogeneous boundary condition n y
=0
on the boundary of D. Suppose that u is not constant in D. Then there exists a closed ball
B contained in D such that Z Z
|grad u|2 dx > 0.
B
50
From Greens first formula applied to a pair of (harmonic) functions u, v = u and the interior
Dh of a surface h = {x hnx : x } parallel to the Ds boundary with sufficiently small
h > 0, Z Z Z
2 u
(uu + |grad u| )dx = u dly , (265)
ny
Dh
we have first Z Z Z Z
|grad u|2 dx |grad u|2 dx, (266)
B Dh
are called the potentials. Here, x = (x, y), y = (x0 , y0 ) R2 ; E(x, y) is the fundamental solution
of a second-order elliptic differential operator;
=
ny ny
is the normal derivative at the point y of the closed piecewise smooth boundary C of a domain
in R2 ; and (y) and (y) are sufficiently smooth functions defined on C. In the case of Laplaces
operator u,
1 1
g(x, y) = g(x y) = ln . (269)
2 |x y|
In the case of the Helmholtz operator H(k) = + k 2 , one can take E(x, y) in the form
i (1)
E(x, y) = E(x y) = H0 (k|x y|),
4
(1)
where H0 (z) = 2ig(z) + h(z) is the Hankel function of the first kind and zero order (one
1 1 1
of the so-called cylindrical functions) and g(x y) = ln is the kernel of the two-
2 2 |x y|
dimensional single layer potential; the second derivative of h(z) has a logarithmic singularity.
51
14.1 Properties of potentials
Statement 1. Let D R2 be a domain bounded by the closed smooth contour . Then the
kernel of the double-layer potential
(x, y) 1 1
V (x, y) = , (x, y) = ln , (270)
ny 2 |x y|
where x,y is the angle between the normal vector ny at the integration point y = (x0 , y0 ) and
vector x y. Choose the origin O = (0, 0) of (Cartesian) coordinates at the point y on curve
so that the Ox axis goes along the tangent and the Oy axis, along the normal to at this
point. Then one can write the equation of curve in a sufficiently small vicinity of the point y
in the form
y = y(x).
The asumption concerning the smoothness of means that y0 (x0 ) is a differentiable function
in a vicinity of the origin O = (0, 0), and one can write a segment of the Taylor series
x2 00 1
y = y(0) + xy 0 (0) + y (x) = x2 y 00 (x) (0 < 1)
2 2
because y(0) = y 0 (0) = 0. Denoting r = |x y| and taking into account that x = (x, y) and
the origin of coordinates is placed at the point y = (0, 0), we obtain
s s
q q 1 1
r= (x 0)2 + (y 0)2 = x2 + y 2 = x2 + x4 (y 00 (x))2 = x 1 + x2 (y 00 (x))2 ;
4 4
y 1 xy 00 (x)
cos x,y = = q ,
r 2 1 + x2 (1/4)(y 00 (x))2
and
cos x,y y 1 y 00 (x)
= 2 = .
r r 2 1 + x2 (1/4)(y 00 (x))2
The curvature K of a plane curve is given by
y 00
K= ,
(1 + (y 0 )2 )3/2
cos x,y 1
lim = K(y)
r0 r 2
52
which proves the continuity of the kernel (270).
we
where the (exterior) unit normal vector n of is directed into the exterior domain R2 \ D,
have
v 0 (x) = 1, x D,
1
v 0 (x) = , x , (273)
2
v (x) = 0, x R2 \ D.
0
applied to v(y) = (x, y). For x D it follows from Greens third formula
Z !
u (x, y)
u(x) = (x, y) u(y) dly , x D, (275)
ny ny
applied to u(y) = 1 in D.
Note also that if we set
1
v 0 (x0 ) = , x0 ,
2
we can also write (273) as
0 1
v (x0 ) = lim v(x + hnx0 ) = v 0 (x0 ) , x0 . (276)
h0 2
u0 (x) Z (x, y)
= dly , (278)
nx nx
53
we
where the (exterior) unit normal vector nx of is directed into the exterior domain R2 \ D,
have
u0 (x)
= 1, x D,
nx
u0 (x) 1
= , x , (279)
nx 2
u0 (x)
= 0, x R2 \ D.
nx
Theorem 35 Let D R2 be a domain bounded by the closed smooth contour . The double-
layer potential
Z
(x, y) 1 1
v(x) = (y)dly , (x, y) = ln , (280)
ny 2 |x y|
and from R2 \ D
with a continuous density can be continuously extended from D to D to
R2 \ D with the limiting values on
Z
(x0 , y) 1
v (x0 ) = (y)dly (x0 ), x0 , (281)
ny 2
or
1
v (x0 ) = v(x0 ) (x0 ), x0 , (282)
2
where
v (x0 ) = lim v(x + hnx0 ). (283)
h0
Proof.
1. Introduce the function
Z
(x, y)
I(x) = v(x) v 0 (x) = ((y) 0 )dly (284)
ny
where 0 = (x0 ) and prove its continuity at the point x0 . For every > 0 and every > 0
there exists a vicinity C1 of the point x0 such that
Set Z Z
I = I1 + I2 = ... + ....
C1 \C1
Then
|I1 | B1 ,
54
where B1 is a constant taken according to
Z
(x, y)
dl B1 x D, (286)
ny y
u(x) Z (x, y)
= (y)dly (291)
nx nx
55
14.2 Generalized potentials
Let S () R2 be a domain bounded by the closed piecewise smooth contour . We assume
that a rectilinear interval 0 is a subset of , so that 0 = {x : y = 0, x [a, b]}.
Let us say that functions Ul (x) are the generalized single layer (SLP) (l = 1) or double layer
(DLP) (l = 2) potentials if
Z
Ul (x) = Kl (x, t)l (t)dt, x = (x, y) S (),
where
Kl (x, t) = gl (x, t) + Fl (x, t) (l = 1, 2),
1 1
g1 (x, t) = g(x, y 0 ) = ln , g2 (x, t) = g(x, y0 ) [y0 = (t, 0)],
|x y0 | y0
F1,2 are smooth functions, and we shall assume that for every closed domain S0 () S (),
the following conditions hold
and Z
(x, y)
K1 (x) = 2 (y)dly , x . (296)
nx
The kernels of operators K0 and K1 are continuous on As seen by interchanging the order
of integration, K0 and K1 are adjoint as integral operators in the space C() of continuous
functions defined on curve .
56
Theorem 36 The operators I K0 and I K1 have trivial nullspaces
N (I K0 ) = {0}, N (I K1 ) = {0}, (297)
The nullspaces of operators I + K0 and I + K1 have dimension one and
N (I + K0 ) = span {1}, N (I + K1 ) = span {0 } (298)
with Z
0 dly 6= 0. (299)
which yields
2v (x) = K0 (x) = 0. (302)
From the uniqueness of the interior Dirichlet problem (Theorem 33) it follows that v = 0 in the
whole domain D. Now we will apply the contunuity of the normal derivative of the double-layer
potential (300) over a smooth contour ,
v(y) v(y)
= 0, y , (303)
ny + ny
where the internal and extrenal limiting values with respect to are defined as follows
v(y) v(y)
= lim = lim ny grad v(y hny ) (304)
ny y, yD ny h0
and
v(y) v(y)
= lim 2 = lim ny grad v(y + hny ). (305)
ny + y, yR \D ny h0
57
Theorem 37 Let D R2 be a domain bounded by the closed smooth contour . The double-
layer potential
Z
(x, y) 1 1
v(x) = (y)dly , (x, y) = ln , x D, (308)
ny 2 |x y|
with a continuous density is a solution of the interior Dirichlet problem provided that is a
solution of the integral equation
Z
x, y)
(x) 2 (y)dly = 2f (x), x . (309)
ny
Proof The integral equation K = 2f of the interior Dirichlet problem has a unique solution
by Theorem 36 because N (I K) = {0}.
Theorem 39 Let D R2 be a domain bounded by the closed smooth contour . The double-
layer potential
Z
(x, y)
u(x) = (y)dly , x R2 \ D, (310)
ny
with a continuous density is a solution of the exterior Dirichlet problem provided that is a
solution of the integral equation
Z
(x, y)
(x) + 2 (y)dly = 2f (x), x . (311)
ny
Proof. The integral operator K : C() C() defined by the right-hand side of (310) is
compact in the space C() of functions continuous on because its kernel is a continuous
function on . Let be a solution to the homogeneous integral equation + K = 0 on
and define u by (310). Then 2u = + K = 0 on R, and by the uniqueness for the exterior
and dly = 0. Therefore + K = 0 which
Dirichlet problem it follows that u R2 \ D,
means
R
according to Theorem 36 (because N (I + K) = span {1}), that = const on . Now
dly = 0 implies that = 0 on , and the existence of a unique solution to the integral
equation (311) follows from the Fredholm property of its operator.
58
Theorem 41 Let D R2 be a domain bounded by the closed smooth contour . The single-
layer potential Z
u(x) = (x, y)(y)dly , x D, (312)
with a continuous density is a solution of the interior Neumann problem provided that is
a solution of the integral equation
Z
(x, y)
(x) + 2 (y)dly = 2g(x), x . (313)
nx
59
We see that Tn and Un are polynomials of degree n.
Let us breifly summarize some important properties of the Chebyshev polynomials.
If |x| 1, the following convenient representations are valid
or
X
f (x) = bn Un (x), x [1, 1]. (328)
n=0
60
1
2Z
bn = f (x)Un (x) 1 x2 dx, n = 0, 1, ... . (330)
1
Note that f (x) may be a complex-valued function; then the Fourier coefficients an and bn
are complex numbers.
Consider some examples of decomposing functions in the FourierChebyshev series
2
4X 1
1 x2 = 2
T2n (x), 1 x 1;
n=1 4n 1
4X 1
arcsin x = T2n+1 (x), 1 x 1;
n=0 (2n + 1)2
X (1)n1
ln (1 + x) = ln 2 + 2 Tn (x), 1 < x 1.
n=1 n
Similar decompositions can be obtained using the Chebyshev
polynomials Un of the second
2
kind. In the first example above, an even function f (x) = 1 x is decomposed; therefore its
expansion contains only coefficients multiplying T2n (x) with even indices, while the terms with
odd indices vanish.
Since |Tn (x)| 1 according to (321), the first and second series converge for all x [1, 1]
uniformly in this segment. The third series represents a different type of convergence: it con-
verges for all x (1, 1] (1 < x 1) and diverges at the point x = 1; indeed, the
decomposed function f (x) = ln(x + 1) is not defined at x = 1.
In order to describe the convergence of the FourierChebyshev series introduce the
functional
spaces of functions square-integrable in the segment (1, 1) with the weights 1/ 1 x2 and
(1)
1 x2 . We shall write f L2 if the integral
Z1
dx
|f (x)|2 (331)
1
1 x2
(2)
converges, and f L2 if the integral
Z1
|f (x)|2 1 x2 dx. (332)
1
converges. The numbers specified by (convergent) integrals (331) and (332) are the squared
(1) (2)
norms of the function f (x) in the spaces L2 and L2 ; we will denote them by kf k21 and kf k22 ,
respectively. Write the symbolic definitions of these spaces:
Z1
(1) dx
L2 := {f (x) : kf k21 := |f (x)|2 < },
1
1 x2
Z1
(2)
L2 := {f (x) : kf k22 := |f (x)|2 1 x2 dx < }.
1
61
(1) (2)
Spaces L2 and L2 are (infinite-dimensional) Hilbert spaces with the inner products
Z1
dx
(f, g)1 = f (x)g(x) ,
1
1 x2
Z1
(f, g)2 = f (x)g(x) 1 x2 dx.
1
Let us define two more functional spaces of differentiable functions defined in the segment
(1,
1) whose first derivatives are square-integrable in (1, 1) with the weights 1/ 1 x2 and
1x : 2
and
c 1 := {f (x) : f L(1) , f 0 (x) 1 x2 L(2) };
W 2 2 2
1
1 x2 1 1 x2 1
Z1 Z 1 Z 1
dx dx
(f, g)W
b1 = f (x) g(x) + f 0 (x)g 0 (x) (1 x2 )3/2 dx
2
1
1 x2 1 1 x2 1
and norms 1 2
Z Z1
2
dx
0 2
kf kW 1 = f (x) + |f (x)| 1 x2 dx,
2
1
1 x2 1
1 2
Z Z1
2
dx
kf kW
b 21 = f (x)
2 + |f 0 (x)|2 (1 x2 )3/2 dx.
1
1x 1
kf SN k1 0 and kf SN k2 0 as N ,
where SN are partial sums of the FourierChebyshev series. The smoother function f (x) in
the segment [1, 1] the faster the convergence of its FourierChebyshev series. Let us explain
this statement: if f (x) has p continuous derivatives in the segment [1, 1], then
a0 XN C ln N
f (x) T 0 (x) a n Tn (x) , x [1, 1], (333)
2 Np
n=1
62
A similar statement is valid for the FourierChebyshev series in polynomials Un : if f (x) has
p continuous derivatives in the segment [1, 1] and p 2, then
XN C
f (x) bn Un (x) p3/2 , x [1, 1]. (334)
N
n=0
We will often use the following three relationships involving the Chebyshev polynomials:
Z1
1 dx
Tn (x) = Un1 (y), 1 < y < 1, n 1; (335)
xy 1 x2
1
Z1
1
Un1 (x) 1 x2 dx = Tn (y), 1 < y < 1, n 1; (336)
xy
1
Z1 (
1 dx ln 2, n = 0;
ln Tn (x) = (337)
|x y| 1 x2 Tn (y)/n, n 1.
1
The integrals in (335) and (336) are improper integrals in the sense of Cauchy (a detailed
analysis of such integrals is performed in [8, 9]). The integral in (337) is a standard convergent
improper integral.
and
Z1
1 dx
(L + K) := (ln + K(x, y)) (x) = f (y), 1 < y < 1. (339)
|x y| 1 x2
1
In equations (338) and (339) (x) is an unknown function and the right-hand side f (y) is a
given (continuous) function. The functions ln (1/|x y|) in (338) and ln (1/|x y|) + K(x, y)
in (339) are the kernels of the integral equations; K(x, y) is a given function which is assumed
to be continuous (has no singularity) at x = y. The kernels go to infinity at x = y and
therefore referred to as functions with a (logarithmic) singulatity, and equations (338) and
(339) are called integral equations with a logarithmic singulatity of the kernel. In equations
(338) and (339) the weight function 1/ 1 x2 is separated explicitly. This function describes
63
the behaviour (singularty) of the solution in the vicinity of the endpoints 1 and +1. Linear
integral operators L and L + K are defined by the left-hand sides of (338) and (339).
Equation (338) is a particular case of (339) for K(x, y) 0 and can be solved explicitly.
Assume that the right-hand side f (y) W 1 (for simplicity, one may assume that f is a
2
continuously differentiable function in the segment [1, 1]) and look for the solution (x) in
(1)
the space L2 . This implies that integral operators L and L + K are considered as bounded
linear operators
(1)
L : L2 W 1 and L + K : L(1) 1
2 W2 .
2
Note that if K(x, y) is a sufficiently smooth function, then K is a compact (completely contin-
uous) operator in these spaces.
XN
a0
N (x) = T0 (x) + an Tn (x), (340)
2 n=1
where an are unknown coefficients. Expand the given function f (y) in the FourierChebyshev
series
XN
f0
fN (y) = T0 (y) + fn Tn (y), (341)
2 n=1
where fn are known coefficients determined according to (329). Substituting (340) and (341)
into equation (338) and using formula (337), we obtain
XN XN
a0 1 f0
ln 2T0 (y) + an Tn (y) = T0 (y) + fn Tn (y). (342)
2 n=1 n 2 n=1
a0 f0 1
ln 2 = , a n = fn ,
2 2 n
which gives unknown coefficients an
f0
a0 = , an = nfn
ln 2
and the approximation
XN
f0
N (x) = T0 (x) + fn Tn (x). (343)
2 ln 2 n=1
The exact solution to equation (338) is obtained by a transition to the limit N in (343)
X
f0
(x) = T0 (x) + nfn Tn (x). (344)
2 ln 2 n=1
64
In some cases one can determine the sum of series (344) explicitly and obtain the solution
in a closed form. One can show that the approximation N (x) of the solution to equation (338)
(1)
defined by (340) converges to the exact solution (x) as N in the norm of space L2 .
Consider integral equation (339). Expand the kernel K(x, y) in the double FourierChebyshev
series, take its partial sum
N X
X N
K(x, y) KN (x, y) = kij Ti (x)Tj (y), (345)
i=0 j=0
and repeat the above procedure. A more detailed description is given below.
Example 14 Consider the integral equation
Z1 q
1 dx
(ln + xy) (x) = y + 1 y2, 1 < y < 1.
|x y| 1 x2
1
to obtain q
2 4X T2k (y)
y+ 1 y2 = T0 (y) + T1 (y) ,
k=1 4k 2 1
K(x, y) = T1 (x)T1 (y).
Substituting these expressions into the initial integral equation, we obtain
a0 XN
an 2 4 [N/2]
X T2k (y)
ln 2T0 (y) + Tn (y) + a1 T1 (y) = T0 (y) + T1 (y) ,
2 n=1 n 2 k=1 4k 2 1
here [ N2 ] denotes the integer part of a number. Equating the coefficients that multiply Tn (y) with
the same indices n we determine the unknowns
4 2 8 n
a0 = , a1 = , a2n+1 = 0, a2n = 2
, n 1.
ln 2 4n 1
The approximate solution has the form
4 2 8 [N/2]
X n
N (x) = + x 2
T2n (x).
ln 2 n=1 4n 1
65
18 Solution to singular integral equations
18.1 Singular integral equations
Consider two types of the so-called singular integral equations
Z1
1 dx
S1 := (x) = f (y), 1 < y < 1; (346)
xy 1 x2
1
Z1 !
1 dx
(S1 + K1 ) = + K1 (x, y) (x) = f (y), 1 < y < 1; (347)
xy 1 x2
1
Z1
1
S2 := (x) 1 x2 dx = f (y), 1 < y < 1; (348)
xy
1
Z1
!
1
(S2 + K2 ) := + K2 (x, y) (x) 1 x2 dx = f (y), 1 < y < 1. (349)
xy
1
Equations (346) and (348) constitute particular cases of a general integral equation (347)
and can be solved explicitly. Singular integral operators S1 and S1 + K1 are considered as
(1) (2)
bounded linear operators in the weighted functional spaces introduced above, S1 : L2 L2 ,
(1) (2)
S1 + K1 : L2 L2 . Singular integral operators S2 S2 + K2 are considered as bounded linear
(2) (1) (2) (1)
operators in the weighted functional spaces S2 : L2 L2 , S2 + K2 : L2 L2 . If the kernel
functions K1 (x, y) and K2 (x, y) are sufficiently smooth, K1 and K2 will be compact operators
in these spaces.
N
X 1
fN (y) = fk Uk (y), (351)
k=0
where coefficients fk are calculated by formulas (330). Substituting expressions (350) and (351)
into equation (346) and using (335), we obtain
N
X N
X 1
an Un1 (y) = fk Uk (y). (352)
n=1 k=0
66
Coefficient a0 vanishes because (see [4])
Z1
1 dx
= 0, 1 < y < 1. (353)
x y 1 x2
1
Equating, in (352), the coefficients multiplying Un (y) with equal indices, we obtain an =
fn1 /, n 1. The coefficient a0 remains undetermined. Thus the solution to equation (346)
depends on an arbitrary constant C = a0 /2. Write the corresponding formulas for the approx-
imate solution,
N
1X
N (x) = C + fn1 Tn (x); (354)
n=1
and the exact solution,
1X
(x) = C + fn1 Tn (x). (355)
n=1
(1)
The latter series converges in the L2 -norm.
In order to obtain a unique solution to equation (346) (to determine the constant C),
one should specify an additional condition for function (x), for example, in the form of an
orthogonality condition
Z1
dx
(x) 2
= 0. (356)
1
1 x
(1)
(orthogonality of the solution to a constant in the space L2 ). Then using the orthogonality
(325) of the Cebyshev polynomial T0 (x) 1 to all Tn (x), n 1, and calculating the inner
product in (355) (multiplying both sides by 1/ 1 x2 and integrating from -1 to 1) we obtain
1
1Z dx
C= (x) = 0.
1 x2
1
Equation (347) can be solved in a similar manner; however first it is convenient to decompose
the kernel function K1 (x, y) in the FourierChebyshev series and take the partial sum
N N
X X 1
(1) (1)
K1 (x, y) KN (x, y) = kij Ti (x)Uj (y), (357)
i=0 j=0
67
subject to condition (356). Look for an approximate solution in the form
XN
a0
N (x) = T0 (x) + an Tn (x).
2 n1
(a0 + a2 ) 1
a1 = 0, a2 + = , a3 = 0,
16 2
(a0 + a2 )
a4 + = 0; an = 0, n 5.
32
Finally, after some algebra, we determine the unknown coefficients
1 8 1 1
a0 = C, a2 = ( C), a4 = (2C ), an = 0, n 6= 0, 2, 4,
17 68
where C is an arbitrary constant. In order to determine the unique solution that satisfies (356),
calculate
8 1
C = 0, a0 = 0, a2 = , a4 = .
17 68
The final form of the solution is
8 1
(x) = T2 (x) T4 (x),
17 68
or
2 4 18 2 33
(x) = x + x .
17 17 68
Note that the approximate solution N (x) = (x) for N 4. We see that here, one can obtain
the solution explicitly as a polynomial.
Consider equations (348) and (349). Begin with (348) and look for its solution in the form
N
X 1
N (x) = bn Un (x). (358)
n=0
68
Expand the right-hand side in the Chebyshev polynomials of the first kind
XN
f0
fN (y) = T0 (y) + fk Tk (y), (359)
2 k=1
where coefficients fk are calculated by (329). Substitute (358) and (359) into (348) and use
(336) to obtain
N
X 1 XN
f0
bn Tn+1 (y) = T0 (y) + fk Tk (y), (360)
n=0 2 k=1
which is equivalent to the requirement f0 = 0 (according to (329) taken at n = 0). Note that,
when solving equation (348), we impose, unlike equation (356), the condition (361) for equation
(346), on the known function f (y). As a result, we obtain
a unique approximate solution to (348) for every N 2
1 NX
1
N (x) = fn+1 Un (x), (362)
n=0
and
a unique exact solution to (348)
1X
(x) = fn+1 Un (x). (363)
n=0
We will solve equations (348) and (349) under the assumption that the right-hand side
(1) (2) (2)
f L2 and L2 . Series (363) converges in the norm of the space L2 .
The solution of equation (349) is similar to that of (348). The difference is that K2 (x, y) is
approximately replaced by a partial sum of the FourierChebyshev series
N
X 1 X
N
(2) (2)
K2 (x, y) KN (x, y) = kij Ui (x)Tj (y). (364)
i=0 j=0
Substituting (358), (359), and (364) into equation (349) we obtain a linear equation system
with respect to unknowns bn . As well as in the case of equation (348), the solution to (349)
exists not for all right-hand sides f (y).
The method of decomposing in FourierChebyshev series allows one not only to solve equa-
tions (348) and (349) but also verify the existence of solution. Consider an example of an
equation of the form (349) that has no solution (not solvable).
69
Example 16 Solve the singular integral equation
Z1 q !
1 (2)
+ 1 y2 (x) 1 x2 dx = 1, 1 < y < 1; L2 .
xy
1
Look for an approximate solution, as well as in the case of (348), in the form of a series
N
X 1
N (x) = bn Un (x).
n=0
Replace K2 (x, y) = 1 y 2 by an approximate decomposition
(2) 2 4 [N/2]
X 1
KN (x, y) = T0 (y) 2
T2k (y).
k=1 4k 1
The right-hand side of this equation is 1 = T0 (y). Substituting this expression into the equation,
we obtain
N 1 [N/2]
X X 1
bn Tn+1 (y) + T0 (y) 2 T (y) b0 = T0 (y).
2 1 2k
n=0 k=1 4k
Equating the coefficients that multiply T0 (y) on both sides of the equation, we obtain b0 = 1;
equating the coefficients that multiply T1 (y), we have b0 = 0,which yields b0 = 0. This
cobtradiction proves that the considered equation has no solution. Note that this circumstance
is not connected with the fact that we determine an approxiate solution N (x) rather than the
exact solution (x). By considering infinite series (setting N = ) in place of finite sums we
can obtain the same result.
Let us prove the conditions for function f (y) that provide the solvability of the equation
Z1 q !
1
+ 1 y2 (x) 1 x2 dx = f (y), 1 < y < 1
xy
1
with the right-hand side f (y) To this end, replace the decomposition of the right-hand side
which was equal to one by the expansion
XN
f0
fN (y) = T0 (y) + fn Tn (y).
2 n=1
70
We see that the equation has the unique solution if and only if
f0 1
= f1 .
2
Rewrite this condition using formula (329)
1 1
1Z dy 1Z dy
f (y) = y f (y) .
2 1y 2 1 y2
1 1
Here, one can uniquely determine coefficient bn , which provides the unique solvability of the
equation.
gk = gk () = (A()f, k ),
i.e., an infinite sequence (vector) of the Fourier coefficients {gn } = (g1 , g2 , ...) corresponds to
X
an element g H and we can write g = (g1 , g2 , ...). Let f = fk k ; then f = (f1 , f2 , ...). Set
k=1
N
X
X
fN = Pf = fk k and An = akn k , n = 1, 2, . . .. Then,
k=1 k=1
N
X
X X
X
lim Af N = lim fn akn k = Af = akn fn k .
N N
n=1 k=1 k=1 n=1
X
X
On the other side, Af = gk k , hence, gk = akn (Af, k ) = (g, k ). Now we set Ak = gk
k=1 n=1
and
(gk , j ) = (Ak , j ) = ajk ,
71
where ajk are the Fourier coefficients in the expansion of element gk over the basis {j }.
Consequently,
X
Ak = gk = ajk j ,
j=1
X
Ck = (Af, k ) = lim (Af N , k ) = lim kN = akj j .
N N
j=1
and the latter equalities define the matrix representation of the operator A in the basis {k }
of the Hilbert space H.
72
or, in other notation,
1
L = {lnj }
n,j=1 , lnj = 0, n 6= j, lnn = , n = 1, 2, . . . ,
n
and
1 0 ... 0 ...
0 2 ... 0 ...
L1 =
... ... ... ... ...
0 ... n 0 ...
... ... ... ... ...
or
L1 = {lnj }
n,j=1 , lnj = 0, n 6= j, lnn = n, n = 1, . . . .
that play an important role in the theory of integral operators and operator-valued functions
with a logarithmic singularity of the kernel.
Theorem 44 L is a linear bounded invertible operator acting from the space hp to the space hp+2 :
L : hp hp+2 (Im L = hp+2 , Ker L = {}). The inverse operator L1 : hp+2 hp , is bounded,
p0.
2 Let us show that L acts from hp to hp+2 and L1 acts from hp+2 to hp . Assume that a hp .
Then, for La we obtain
X an 2 p+2 X
n
= |an |2 np < .
n=1 n n=1
The norm kLakp+2 = kakp . The element a = L1 b, a hp , is the unique solution of the equation
La = b, and ker L = {}. According to the Banach theorem, L1 is a bounded operator. 2
Note that kLk = kL1 k = 1 and
L1 L = LL1 = E,
73
2 Let hp and A = . Then,
X
X X
np+2 |n |2 = np+2 anj j
n=1 n=1 j=1
X
X |anj |2
kk2p np+2 p
C 2 kk2p ,
n=1 j=1 j
hence, hp+2 . Now we will prove that A is completely continuous. In fact, operator A can
be represented as a limit with respect to the operator norm of a sequence of finite-dimensional
operators AN , which are produced by the cut matrix A and defined by the matrices AN =
{anj }N
n,j=1 :
2 2
X
N
X1 X X
2
p+2
p+2
k(A AN )kp+2 = n
anj j + n anj j
n=1 j=N n=N j=1
X
2 N 1 X
X |anj | X |anj |2 p+2
kk2p np+2 p + kk2p n
n=N j=1 j n=1 j=N jp
h
X X
|anj | 2 X
X
|anj |2 i
kk2p np+2 + np+2 ,
n=N j=1 jp j=N n=1 jp
and, consequently, the following estimates are valid with respect to the operator norm:
X
X N
|anj |2 X X
p+2 |anj |
2
kA AN k np+2 + n 0, N . 2
n=N j=1 jp j=N n=1 jp
Definition 9 Assume that condition (365) holds for the matrix elements of summation oper-
ator A. Then, the summation operator F = L + A : hp hp+2 is called the L-operator.
74
where Tn (x) = cos(n arccos x) are the Chebyshev polynomials and = (0 , 1 , ...n , ...) h2 ;
that is,
X
1
kk22 = |0 |2 + |n |2 n2 < .
2 n=1
In this section we add the coordinate 0 to vector and modify the corresponding formulas. It is
easy to see that all results of previous sections are also valid in this case. Recall the definitions
of weighted Hilbert spaces associated with logarithmic integral operators:
Z1
(1) dx
L2 = {f (x) : kf k21 = |f (x)|2 < },
1
1 x2
Z1
(2)
L2 = {f (x) : kf k22 = |f (x)|2 1 x2 dx < },
1
(1)
f 1 = {f (x) : f L , f 0 L }, (2)
W 2 2 2
and
c 1 = {f (x) : f L(2) , f (x) = f (x) 1 x2 , f 0 L(2) , f (1) = f (1) = 0}.
W 2 2 2
Below, we will denote the weights by p1 (x) = 1 x2 and p1 1 (x).
(1) (2)
Note that the Chebyshev polynomials specify orthogonal bases in spaces L2 and L2 . In
order to obtain the matrix representation of the integral operators with a logarithmic singularity
of the kernel, we will prove a similar statement for weighted space W f 1.
2
2 Assume that . We will show that W f 1 . Function (x) is continuous, and (366) is
2
its FourierChebyshev series with the coefficients
1
2Z
n = Tn (x)(x)p1
1 (x)dx.
1
X
The series |n |2 n2 converges and, due to the RieszFischer theorem, there exists an element
n=1
(1)
f L2 , such that its Fourier coefficients in the Chebyshev polynomial series are equal to nn ;
that is, setting Un (x) = sin(n arccos x), one obtains
1 1
2Z 1 2Z 0
nn = nTn (x)(x)p1 (x)dx = Un (x)(x)dx
1 1
1 1
2Z 0 dx 2Z dx
= Un (x)(p1 (x)(x)) = Un (x)f (x) , n 1.
p1 (x) p1 (x)
1 1
75
Z1
Since f (x) is an integrable function, the integral |f (x)|p1
1 (x)dx converges, and
1
Zx
f (t)dt
F (x) =
p1 (t)
1
is absolutely continuous (as a function in the form of an integral depending on the upper limit
of integration), F 0 (x) = f (x)p1
1 (x), and
1 1 1
2Z dx 2Z 0 2Z 0
Un (x)f (x) = Un (x)F (x)dx = Un (x)F (x)dx.
p1 (x)
1 1 1
Hence, (x) F (x) = C = const. The explicit form of F 0 (x) yields 0 (x) = f (x)p1
1 (x);
0 (1) 0 (2)
therefore, (x)p1 (x) L2 and (x) L2 .
f 1 and be represented by the Fourier
Now, let us prove the inverse statement. Let W 2
Chebyshev series (366). Form the Fourier series
X
p1 (x)0 (x) = n Un (x), |x| 1,
n=1
where
1 1
2Z 0 dx 2Z 0
n = Un (x)(p1 (x) (x)) = Un (x)(x)dx =
p1 (x)
1 1
1
2n Z (x)dx
= Tn (x) .
p1 (x)
1
76
and the integral equation
L + N = f, f 1,
W (367)
2
where
Z1
ds
N (x) = N = N (q)(s) , q = (x, s),
p1 (x)
1
and
f0 X
f (x) = T0 (x) + fn Tn (x),
2 n=1
where the coefficients
1
2Z dx
fn = f (x)Tn (x) , n = 0, 1, . . . .
1 x2
1
Function N (x) is continuous and may be thus also expanded in the FourierChebyshev series
X
1
N (x) = b0 T0 (x) + bn Tn (x), (369)
2 n=1
where Z Z
dxds
bn = 2 N (q)Tn (x)(s) . (370)
2p 1 (x)p1 (s)
1
f 1 to the
According to Lemma 6 and formulas (368) and (369), equation (367) is equivalent in W 2
equation
X X
0 1 b0
ln 2 T0 (x) + n n Tn (x) + T0 (x) + bn Tn (x)
2 n=1 2 n=1
X (371)
f0
= T0 (x) + fn Tn (x), |x| 1.
2 n=1
Note that since N (x) is a differentiable function of x in [1, 1], the series in (371) converge
uniformly, and the equality in (371) is an identity in (1, 1).
Now, we substitute series (368) for (x) into (370). Taking into account that {Tn (x)} is a
basis and one can change the integration and summation and to pass to double integrals (these
statements are verified directly), we obtain an infinite system of linear algebraic equations
(
2 ln 20 + b0 = f0 , (372)
n1 n + bn = fn , n 1.
77
In terms of the summation operators, this systems can be represented as
L + A = f ; = (0 , 1 , . . . , n , . . .) h2 ; f = (f0 , f1 , . . . , fn , . . .) h4 .
where
1,
n = j = 0,
nj = 2, n 1, j 1,
2, for other n, j.
Lemma 7 Let N (q) satisfy the continuity conditions formulated in (367). Then
X
X 1
|anj |2 n4 + |a0j |2 C0 , (374)
n=1,j=0 j=0 ln2 2
where
1 1 2
1 Z Z N (q)dx ds Z Z
2 dxds
C0 = 2 + p21 (x) |(p1 (x)Nx0 )0x | 2 . (375)
ln 2 p1 (x) p1 (s) p1 (x)p1 (s)
1 1 1
where
1 1
2nm Z ds Z
Cnm = Tm (s) Tn (x)(p1 (x)Nx0 (q))0x dx
p1 (s)
1 1
Z1 Z1
2nm ds
= Tm (s) Tn0 (x)p1 (x)Nx0 (q)dx
p1 (s)
1 1
1 Z1
nnm Z
2
Tm (s)
= ds Un (x)Nx0 (q)dx
2 p1 (s)
1 1
n2 2nm Z Z
N (q)Tn (x)Tm (s)
= dxds = n2 nm anm .
2 p1 (x)p1 (s)
1
78
Let us expand the function
1
1 Z N (q)dx
g(s) =
ln 2 2 p1 (x)
1
Definition 10 Let X and Y be the Banach spaces and let A : X Y be an injective operator.
Let Xn X and Yn Y be two sequences of subspaces with dim Xn = dim Yn = n and
Pn : Y Yn be the projection operators. The projection method generated by Xn and Pn
approximates the equation
A = f
by its projection
Pn An = Pn f.
This projection method is called convergent for operator A, if there exists an index N such that
for each f Im A, the approximating equation Pn A = Pn f has a unique solution n Xn for
all n N and these solutions converge, n , n , to the unique solution of A = f .
In terms of operators, the convergence of the projection method means that for all n
N , the finite-dimensional operators An := Pn A : Xn Yn are invertible and the pointwise
convergence A1n Pn A , n , holds for all X. In the general case, one may expect
that the convergence occurs if subspaces Xn form a dense set:
inf k k 0, n (376)
Xn
for all X. This property is called the approximating property (any element X can be
approximated by elements Xn with an arbitrary accuracy in the norm of X). Therefore,
in the subsequent analysis, we will always assume that this condition is fulfilled. Since An =
Pn A is a linear operator acting in finite-dimensional spaces, the projection method reduces to
solving a finite-dimensional linear system. Below, the Galerkin method will be considered as
the projection method under the assumption that the above definition is valid in terms of the
orthogonal projection.
Consider first the general convergence and error analysis.
79
Lemma 8 The projection method converges if and only if there exists an index N and a positive
constant M such that for all n N , the finite-dimensional operators An = Pn A : Xn Yn are
invertible and operators A1
n Pn A : X X are uniformly bounded,
kA1
n Pn Ak M. (377)
kn k (1 + M ) inf k k. (378)
Xn
Relationship (378) is usually called the quasioptimal estimate and corresponds to the error
estimate of a projection method based on the approximation of elements from X by elements
of subspaces Xn . It shows that the error of the projection method is determined by the quality
of approximation of the exact solution by elements of subspace Xn .
Consider the equation
S + K = f (379)
and the projection methods
Pn Sn = Pn f (380)
and
Pn (S + K)n = Pn f (381)
with subspaces Xn and projection operators Pn : Y Yn .
kn k M kPn S Sk
For operator equations in Hilbert spaces, the projection method employing an orthogonal
projection into finite-dimensional subspaces is called the Galerkin method.
Consider a pair of Hilbert spaces X and Y and assume that A : X Y is an injective
operator. Let Xn X and Yn Y be the subspaces with dim Xn = dim Yn = n. Then, n Xn
is a solution of the equation A = f obtained by the projection method generated by Xn and
operators of orthogonal projections Pn : Y Yn if and only if
80
Assume that Xn and Yn are the spaces of linear combinations of the basis and probe func-
tions: Xn = {u1 , . . . , un } and Yn = {v1 , . . . , vn }. Then, we can express n as a linear combination
n
X
n = n un ,
k=1
and show that equation (382) is equivalent to the linear system of order n
n
X
k (Auk , vj ) = (f, vj ), j = 1, . . . , n, (383)
k=1
Here, K(x, y) is a smooth function; accurate conditions concerning the smoothness of K(x, y)
that guarantee the compactness of integral operator K are imposed in Section 2.
One can determine an approximate solution of equation (384) by the Galerkin method
taking the Chebyshev polynomials of the first kind as the basis and probe functions. According
to the scheme of the Galerkin method, we set
N
X 1
a0
N (x) = T0 (x) + an Tn (x),
2 n=1
81
The same technique can be applied to constructing an approximate solution to the hyper-
singular equation
1 !
d Z 1
(H + K) = + K(x, y) (x) 1 x2 dx = f (y), 1 < y < 1. (386)
dy xy
1
The conditions concerning the smoothness of K(x, y) that guarantee the compactness of integral
operator K are imposed in Section 2.
Choose the Chebyshev polynomials of the second kind as the basis and probe functions, i.e.,
set
XN = {U0 , . . . , UN 1 }, YN = {U0 , . . . , UN 1 }. (387)
Unknown approximate solution N (x) is determined in the form
N
X 1
N (x) = bn Un (x). (388)
n=0
82