ROCZNIKI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO Seria I: PRACE MATEMATYCZNE VII (1962)
R.
Ta ber sk i(Poznań)
Singular integrals depending on two parameters
1. Introduction. Let L2n be the class of all functions 27r-periodic and Lebesgue-integrable in the interval <—7c,7c>. Write L(_a,b} for the class of functions Lebesgue-integrable in <a, by. Denote, once for all, by K( t , £) a function defined for all t and £eE (where E is a given set of numbers), 2Tt-periodic, even, bounded and measurable with respect to t for every fixed £eE. Suppose that | 0 is an accumulation point of E.
It is easily observed that Fejer’s theorem concerning the convergence of singular integrals
7Г
'(1) U ( x , £ , f ) = j f ( t ) K ( t - x , ( ) d t ( f t L 2K)
----7T
can be extended to the convergence (x, S) -> (x0, | 0) on an arbitrary set of points of the plane. Namely,
1.1. If
7Г
(2) lim f K ( t , i ) d t = 1 (ŚeE),
— 7T
П
(3) J IK( t , 1)1 dt < 0 on E (C = const),
—П
(4) lim sup \K[t, f)| = 0 for д > 0 (д < те, £еЕ), ł->ło
then we have
(5) lim U(x, i , f ) = f(x0) ЦеЩ
(ж, i)— > (Xq, Iq)
at every x0 at which f is continuous (cf. [4], I, p. 89, [1], p. 418).
X
Delation (5) holds at the points of differentiability of f f(t)dt, under 5
the assumptions of Bomanovski’s theorem, if the convergence (x , I) ->
-> (x0, £0) is restricted to some set of points of the plane. Also a theorem of the Faddeev type and a result concerning the convergence of deriva
tives drU(x, i , f ) j dxr can be stated. The present note is devoted to these
problems.
1 7 4 R. T a b e r s k i
2. A generalization of Natanson’s lemma. We shall prove the following fundamental lemma (cf. [3], p. 243).
2.1. Let <p{t) be a function of bounded, variation in every interval
b
by (0 < у < b — a), such that f v(s)ds < oo, where
v(s) = var cp{t) {a < s < b) , v(b) — 0.
S t < b
Then, if
a + h
M = sup — f f(t)dt
0 < h t ą b —a h J
< o o (f e L ( a , b y),
the improper Lebesgue integral I = ff(t)<p(t)dt exists and a+
\i\ < m J [ * ( * ) + И & ) [ ] * > -
Proof. Write
F(t) = Jf( u)du (a <&),
a
P
Ia, n= f f(t)<p{t)dt (a < a < ft ^.b).
a
Integrating by parts, we obtain
fi /3
/ <p{l)dF(t) =<p(t)F(t)\f - f F(t)d?(t).
a a
Since
fi P
pj J
F(t)dq>{t)\ < M
J(t — a ) d [ —v(t)] = M^[a — t)v(t) j v{t)dt\,
a o . a
a
(a —a)\(p(a) — (p{b)\ ^ (a — a)v(a) < J v(t)dt,
a b
the integral fF(t)dy(t) exists and
ь ь
I f F{t)dcp{t)] < M f v(t)dt.
a+
a +
Our conclusion is now evident.
Singular integrals depending on two parameters 17 5
ь
R em ark. If ip(t) is such that / w(s)ds < oo, where
a
then
with
w{s) = var ip(t) (a < s < b), w{a) = 0,
a < < < s
b - b
I / f ( t ) y { t ) d t \ ^ Ж f [w(s)-\-\ip(a)\']ds,
a a
N = sup
0 < h < b—a
< O O .
3. Theorems of the Romanovski and the Faddeev type. Some results on the convergence of integrals (1) will now be given.
3.1. Suppose that the function K ( t, £) is non-negative and non-inereasing in t on <0, тс) for every £eE, that it satisfies condition (2) and that
(6)
Let
lim K( d, £) = 0 for 0 < д < тг (£ eF?).
(7) lim —
л.—> о h
J f ( t ) d t = f ( x 0) ( f €^2n) at some x0. Then integrals (1) tend to f(x0) as (x, £) ->
set Z in which the function
(x0, £0) on any plane
Ц х, I) = ( x - x 0) K ( 0, £) (£eE) is bounded (cf. [3], p. 245, [4], I, p. 101, (7.9)).
Proof. Of course, it is sufficient to show that T C
I(x, £) = f [f(t)—f(x0)]K(t — x, £ ) d t ^ 0
— 7T
as (x, £) -> (x0, £0) on Z.
We consider only the case — т с < x0 < 0 . By (7), given e > 0 there is a <5 > 0 such that
1 h
x 0 ± h
Xq
£ when 0 < h < 3.
Suppose that 6 <
tc+ ^
o, 0 < x0 — x < and write
Xq— d X 0 + 6 n
Ц о и , £ ) = ( ) ' + ] ' + j ) u m - f ( 4 № ( t - x , £ ) Ш = Ц + Ц + Ц . -•к ж0-<5 a'o+ó
176 E . T a b e r s k i
It is easy to see that
Xq— 0 7Г
|Za| J |/(«)-№ о)|< й < % № , ( ) / \ № - f ( x 0)\dt
— I t — 7t
and
I t I t
u,\ } \ m - n x 0)\m.
Xq -{- <5 — Tt
Hence, by (6),
lim l!
=0
=lim l3 as
(x, £) -> (x0, £0), £eE.(x, o (x, $>
In view of 2.1,
Xq
\I2\ < £ f [ var К
(t — x, £ )+ K (x 0 — x — d,£)]ds +
Xq- 8 x o ~ d < ł < s
x 0 +d
+ e f [ var
K ( t —x ,£ )-\-K (x 0—x -\-d ,£ ))d sXq S < < < Xq + 6
Tt
< « [ /X (* ,f )< fo + 2 (^ -® ) X (0 ,f)] .
— 7t
Therefore, if the points (я?,
£) eZare sufficiently near to
(x0, £0),we have
\IZ\ < 2 г (ф + 1 ), where
Q =sup
\(x — ж0)1Г(0, |) | ((x, £)eZ).Thus, the proof is finished. Similarly, the following result can be obtained.
3.2.
Let K* (t, £) be. non-negative and non-increasing in t on<0,
tt) for£eE, and'satisfy the same assumptions as K( t , £) in §
1,
above1.1,
and let\ K(t , 1)1 < K * ( t , I) (fe <0,7u>, U B ) .
Suppose that conditions
(2), (4)
for K( t , £) and condition(3)
for K * (t, £) hold. Then the relationlimT Г \fit)—f{x0)\dt = 0 (feL2n)
h-> 0 h J
x0
implies(5),
i.e.7 t
lim
f f ( t ) K { t — x, £)dt = f ( x 0),(*,f)
- nas (x, I) -> (x0, £q) on a plane set Z* in which the function A*(a?, f) = ( x - x o)K*(0, £) (£eE)
is bounded
(cf. [3], p. 246).
Singular integrals depending on two parameters 1 7 7
4. Differentiation of singular integrals. The following theorem hold.
4.1.
Let the function K( t ,£)
and its derivatives dvK ( t , £)/dtv (v == 1, 2, . . . , r) be continuous with respect to t on ( — oo, oo) for every fixed
£eE. Suppose that conditions
(2), (3), (4)
together with(8)
and
(9) lim sup
sup I sinrt
ę e is J
drK( t , £)
drK ( t , I)
dtr
dt < oodtr 0
for all 3 >0 (<5 < тс, £e.Zy)
are satisfied. Denote by S a set of points (x, | ) (£eE) such thatdrK( t , £)
(10)
\x — x0\vJsinr
vtdf
dt < 6Y „
(v — 1, 2, . . . , r ) ,ж0 fixed, where Cv are certain positive constants. Then, if the function f e L 2n possesses at x0 a finite derivative f (r\ x f ) , we have
(11) = f / m
( x , f ) dxr
(*.*)
drK{ t — x, £)
M
dt = f {r)(x0) as (x, £ ) —» (x0, £0) on S(ef. [4], II, pp. 60-61, [1], pp. 419-420).
P roof. We construct a function Г
д(г) = ^ ап
sin” (t — a?0)
{ - u < x0
< n )n= 0
such that
9т Ы = / <P,(*o) (P = 0 ,1 ,
In this case, the coefficients an are certain linear combinations of the derivatives f {m)(x0) (m <
n),e.g. a0 = f ( x 0), a1= / /(a?0), a2 = ^ f " ( x 0),
а, = 1[/Ч * „ ) + / ' " « ] > «4 o ) + / , r (*o)], «6 = i [ 9 / ( * o ) + + 1 0 //"(^0) + / F(^o)] etc. II is easy to observe that
oi(t)= [/(£) —
~#0O]/sinr(2 — #0) -*■ 0 as t -> x0 ([2], pp. 25-27).
Obviously,
dr U(x, £, g)dxr
u.
- i ) r f g ( t ) drK ( t — x , £)
—r
dt —7Г
■ 7Г
R oczn ik i PTM — P race M atem atyczn e VII 12
1 7 8 R . T a b e r s k i
Hence, by 1.1,
lim - _ g{r) ^ = j(r) ag (a?, f ) -> (a?0, £0) .
( X , S ) 0X
Since
dr U ( x , ę , f ) dr U ( x , ę , g ) dr U ( x , ę , f - g )
dxr dxr + dxr ~ ’
it is sufficient to show that the second term on the right-hand side of the last identity tends to zero as (x , |) -> (x0, £0) on /S'. But
H( x, £)
eter -1»' /co(£)sinr(£ —a?9) drK ( t - x , £)
dt
and lim
o){t)
=0.
Thus, for any e> 0
there is a <5 >0
( ó < min(7r —x0,
<->x0
7г + а70)) such that
\H(x, |)|
-■0 r
/
sinr( t— x0) drK ( t — x, Ś)d f
\ x0 — ® **
+ !( J +
J
Jco(0sinr(f — я?0) е1г (х, l ) + \I2{%, £) + I 9{x, £)|.dt-\-
drK { t —x, £)
d f
dt
Evidently,
Ii(x, f) < / S in r ( / + £I7— Ж0) arir(<, f)
d f
dt
f sin rt d rK ( t , f)
d f dt +
7 Z
j
|sinr(ź +
x — x0) — s h ft I d rK ( t , f)d f
dt
In view of (8), there is a constant C such that J x(£) < 0 ( $ e E ) .
Applying the formulas
an
— bn =(a— b)(an~l -\-an~2bĄ- ...
-|-6n_1), sin a — &m(3 = 2 cos [(a-j-/?)/2] sin [(a — /?)/2], and taking into account (10), we can show that J 2{x, £) < P on 8, where P is a constant; whence I x( x, £) < (7 + P on 8.Singular integrals depending on two parameters 1 7 9
Let \x — %0\ < \ d . Since y(t) = co(t)smr (t — xQ) is Lebesgue-integrable
m < — 7Г , 7T>,
Hence, by (9), lim l3(a?, £) = 0 as (%, |) (a?0, £„). Similarly, limZ2(#, £) =
= 0. The proof is completed.
Finally, we remark that from 1.1, 3.1 and 4.1 theorems 1, 3, 2 of [1], pp. 417-23, follow, respectively. An analogical result can be obtained for Cesaro’s integrals ([4], II, pp. 60-61).
[1] Gr. M. G olusin, Geometric theory of functions of complex variable (in Russian), Moscow-Leningrad 1952.
[2] Y. M atsuoka, Asymptotic formula for Vallee Poussins singular integrals, Sci. Reports of Kagoshima Univ., No. 9, 1960, pp. 25-34.
[3] I. P. N a ta n so n , Theory of functions of real variable (in Russian), Moscow- Leningrad 1950.
[4] A. Z ygm und, Trigonometric series, I, I I , Cambridge 1959.
*o+<5
7T
drK( t — x, I) d f
References