• Aucun résultat trouvé

A THEOREM ON ENTIRE FUNCTIONS A.G. RAMM Let

N/A
N/A
Protected

Academic year: 2022

Partager "A THEOREM ON ENTIRE FUNCTIONS A.G. RAMM Let"

Copied!
5
0
0

Texte intégral

(1)

A.G. RAMM

LetG(k) =R1

0 g(x)ekxdx,gL1(0,1). The main result of this paper is

Theorem 1. There existsg6≡0, gC0(0,1), such thatG(kj) = 0, kj< kj+1,

j→∞lim kj = ∞, lim

k→∞|G(k)| does not exist, lim sup

k→+∞

|G(k)| = ∞. This g oscillates infinitely often in any interval[1δ,1], however smallδ >0is.

AMS 2010 Subject Classification: 30D15, 42A38, 42A63.

Key words: Fourier integral, entire function, uniqueness theorem.

1. INTRODUCTION Let g∈ L1(0,1) andG(k) = R1

0 g(x)ekxdx. The function G is an entire function of the complex variable k. It satisfies for allk∈Cthe estimate (1) |G(k)| ≤c0e|k|, c0 :=

Z 1

0

|g(x)|dx, it is bounded for <k≤0

(2) |G(k)| ≤c0, <k≤0, and it is bounded on the imaginary axis

(3) sup

τR

|G(iτ)| ≤c0, where Rdenotes the real axis.

It is not difficult to prove that ifgkeeps sign in any interval [1−δ,1], and R1

1−δ|g(x)|dx >0 then lim

k→∞|G(k)|=∞. By∞ we mean +∞ in this paper.

However, there exists ag 6≡0,g ∈C0(0,1), such that lim

k→∞|G(k)| does not exist, but lim sup

k→∞

|G(k)|= ∞. This g changes sign (oscillates) infinitely often in any interval [1−δ,1], however small δ >0 is, and the corresponding G(k) changes sign infinitely often in any neighborhood of +∞, i.e., it has infinitely many isolated positive zeros kj, which tend to ∞.

From the known results the following claim follows (see [1], [2]).

REV. ROUMAINE MATH. PURES APPL.,55(2010),6, 515–519

(2)

Claim. If c1 := lim sup

k→∞

|G(k)|<∞, then g= 0.

We prove this Claim for convenience of the reader.

Our main result is

Theorem 1. There exists g6≡0, g∈C0(0,1), such that lim sup

k→∞

|G(k)|=∞,

k→∞lim |G(k)| does not exist, G(kj) = 0, where kj < kj+1, lim

j→∞kj =∞, and g oscillates infinitely often in any interval [1−δ,1], however small δ >0 is.

In Section 2 proofs are given. The main tool in the proof of the Claim is a Phragmen-Lindel¨of (PL) theorem (see, e.g., [2], [5]). In the proof of Theorem 1 we use a construction, developed in [3], p. 282, (see also [4]), for a study of resonances in quantum scattering on a half line. The conclusion of the Claim that g= 0, one can also derive from a version of Paley-Wiener theorem given in [1], p. 36.

For convenience of the reader we formulate the Phragmen-Lindel¨of (PL) theorem, used in Section 2. A proof of this theorem can be found, e.g., in [2].

By ∂Qthe boundary of the set Qis denoted, M >0 is a constant, and 0< α <2π.

Theorem (PL). Let G(k) be holomorphic in an angle Q of opening πα and continuous up to its boundary. If sup

k∈∂Q

|G(k)| ≤ M, and the order ρ of G(k) is less than α, i.e.,

ρ:= lim sup

r→∞

ln ln max

|k|=r|G(k)|

lnr < α, then sup

k∈Q

|G(k)| ≤M.

In Section 2 Theorem (PL) will be used forα= π2 and ρ= 1< α.

2. PROOFS

The proof of Theorem 1 is based on two lemmas, and the conclusion of Theorem 1 follows immediately from these lemmas. The first lemma is identical with the Claim.

Lemma 1. If

(4) lim sup

k→∞

|G(k)|<∞,

(3)

then

(5) g= 0.

Proof. The entire function G in the first quadrant Q1 of the complex plane k, that is, in the region 0 ≤ argk ≤ π2,|k| ∈ [0,∞), satisfies estimate (1) and is bounded on the boundary of Q1

(6) sup

k∈∂Q1

|G(k)| ≤max(c0, c1) :=c2.

By a Phragmen-Lindel¨of theorem (see, e.g., [5], p. 276) one concludes that

(7) sup

k∈Q1

|G(k)| ≤c2. A similar argument shows that

(8) sup

k∈Qm

|G(k)| ≤c2, m= 1,2,3,4,

where Qm are quadrants of the complex k-plane. Consequently, |G(k)| ≤ c2 for all k ∈ C, and, by the Liouville theorem, G(k) = const. This const. is equal to zero, because, by the Riemann-Lebesgue lemma, lim

τ→∞G(iτ) = 0.

IfG= 0, theng= 0 by the injectivity of the Fourier transform. Lemma 1 and, therefore, Claim, are proved.

Lemma 2. There exists g6≡0, g∈C0(0,1), such that lim sup

k→∞

|G(k)|=∞,

k→∞lim |G(k)| does not exist, G(kj) = 0, where kj < kj+1, lim

j→∞kj =∞, and g oscillates infinitely often in any interval [1−δ,1], however small δ >0 is.

Proof. Let ∆j = [xj, xj+1], 0< xj < xj+1 <1, lim

j→∞xj = 1, j= 2,3, . . ., fj(x)≥0,fj ∈C0(∆j),R

jfj(x)dx= 1, andj >0,j+1 < j, be a sequence of numbers such that f(x)∈C0(0,1), where

(9) f(x) :=

X

j=2

jfj(x).

Define

(10) g(x) :=

X

j=2

(−1)jjfj(x), g6≡0.

The functiongoscillates infinitely often in any interval [1−δ,1], however small δ >0 is.

(4)

Let

(11) G(k) :=

X

j=2

(−1)jjGj(k), Gj(k) :=

Z

j

fj(x)ekxdx.

Note thatGj(k)>0 for all j= 2,3, . . . and allk >0, and

(12) lim

k→∞

Gm(k)

Gj(k) =∞, m > j, because intervals ∆j and ∆m do not intersect if m > j.

Fix an arbitrary small number ω > 0. Let us construct a sequence of numbers

bj >0, bj+1> bj, lim

j→∞bj =∞, such that

(13) G(b2m)≥ω, G(b2m+1)≤ −ω, m= 1,2, . . . . This implies the existence of kj ∈(b2m, b2m+1), such that

G(kj) = 0, kj < kj+1, lim

j→∞kj =∞.

Moreover,

(14) lim sup

k→∞

|G(k)|=∞,

because otherwiseG= 0 by Lemma 1, sog= 0, contrary to the construction, see (10).

Let us chooseb2 >0 arbitrary. ThenG2(b2)>0. Making3 >0 smaller, if necessary, one gets

(15) 2G2(b2)−3G3(b2)≥ω.

Choosing b3 > b2 sufficiently large, one gets

(16) 2G2(b3)−3G3(b3)≤ −ω.

This is possible because of (12).

Suppose thatb2, . . . , b2m+1 are constructed, so that (17)

2m

X

j=2

(−1)jjGj(b2p)≥ω, p= 1,2, . . . , m, and

(18)

2m+1

X

j=2

(−1)jjGj(b2p+1)≤ −ω, p= 1, . . . , m.

(5)

Let m→ ∞. Then

(19) lim

m→∞

2m

X

j=2

(−1)jjGj(k) = lim

m→∞

2m+1

X

j=2

(−1)jjGj(k) =G(k),

whereG(k) is defined in (11), and the convergence in (19) is uniform on com- pact subsets of [0,∞). Therefore inequalities (13) hold. Lemma 2 is proved.

The conclusion of Theorem 1 follows from Lemmas 1 and 2.

Theorem 1 is proved.

REFERENCES

[1] P. Koosis,The Logarithmic Integral, Vol. I. Cambridge Univ. Press, Cambridge, 1988.

[2] G. Polya and G. Szeg¨o, Problems and Theorems in Analysis. Springer-Verlag, Berlin, 1983.

[3] A.G. Ramm,Multidimensional Inverse Scattering Problems.Longman/Wiley, New York, 1992.

[4] A.G. Ramm and B.A. Taylor,Example of a potential in one-dimensional scattering prob- lem for which there are infinitely many purely imaginary resonances. Phys. Lett. A124 (1987), 313–319.

[5] W. Rudin,Real and Complex Analysis. McGraw Hill, New York, 1974.

Received 10 September 2010 Kansas State University

Department of Mathematics Manhattan, KS 66506-2602, USA

ramm@math.ksu.edu

Références

Documents relatifs

Give a necessary and sufficient condition on n for the ring Z/nZ to be

If a real number x satisfies Conjecture 1.1 for all g and a, then it follows that for any g, each given sequence of digits occurs infinitely often in the g-ary expansion of x.. This

[r]

Repeat the exer- cise in homology, for cor G×K H×K and tr G×K H×K (in homology, the hypothesis that G be finite can be supressed, but of course we need [G : H ] &lt; ∞ for the

The vector Boolean functions are used in cryptography to construct block ciphers and an important criterion on these functions is high resistance to differential cryptanalysis.. Let q

Rodier, A Few More Functions That Are Not APN Infinitely Often, Finite Fields: Theory and applications, Ninth International conference Finite Fields and Applications, McGuire et

A means to prove this conjecture is to remark that the APN property is equivalent to the fact that the rational points of a certain algebraic surface X in a 3-dimensional space

Montrer que cette équation possède une unique solution dans ]a, b[.. Dessiner le graphe d'une fonction f vériant toutes