It is proved that if a smooth function
, such that
where UN is the normal derivative of u on S, has a closed smooth surface S of zeros, then the function
has also a closed smooth surface Se of zeros. Here u is a smooth function and
is a sufficiently small number.
zero surfaces, perturbation theory
Let
be a bounded domain containing inside a connected closed C3-smooth surface S, which is the set of zeros of a function
, so that Consider the scattering problem:
(1)
Let N = Ns be the unit normal to S, such that
where UN is the normal derivative of u on S. Let
, where
is sufficiently small. Assume that
(2)
The purpose of this paper is to prove Theorem 1.
Theorem 1. Under the above assumptions there exists a smooth closed surface Se such that Ue = 0 on Se.
In Section 2 Theorem 1 is proved.
Although there are many various results on perturbation theory, see [2], [3], the result formulated in Theorem 1 is new.
Proof of Theorem 1
Consider the following equation for t:
(3)
where N = N(s) is the normal to S at the point s and t is a parameter. Using the Taylor's formula and relation (1), one gets from (3)

where
is the Lagrange remainder in the Taylor's formula and

Since the functions u and
belong to C3(D), the function
has a bounded derivative with respect to t uniformly with respect to 
Consider equation (4) as an equation for t = t(s) in the space C(S). Rewrite (4) as

Let us check that the operator B satisfies the contraction mapping theorem in the set

where
is a small number, and 
First, one should check that B maps M into itself. One has

We have chosen N so that
This is possible because equation (1) implies that
is orthogonal to S at the point
Assumption (2) implies that for sufficiently small ∈ one has

Since
is continuously differentiable, one has


Thus, if (13) holds then B maps M into itself.
Let us check that B is a contraction mapping on M. One has
is sufficiently small. Indeed,

if δ is suciently small. Here C4 is a constant.
Thus, B is a contraction on M. By the contraction mapping principle, equation (6) is uniquely solvable for t. Its solution t = t(s) allows one to construct the zero surface Se of the function Ue by the equation r = s + t (s) N , where r = r (s) is the radius vector of the points on Se.
Theorem 1 is proved.
Remark 1. Condition (2) is a sufficient condition for the validity of Theorem 1. Although this condition is not necessary, if it does not hold one can construct counterexamples to the conclusion of Theorem 1. For example, assume that
Then the function
does not have zeros in i 3.
Remark 2. In scattering theory the following question is of interest: assume that u (x) is an entire function of exponential type,
Assume that u = 0 on S, where S is a closed smooth connected surface in i 3.
Is there another closed smooth connected surface of zeros of an entire function Ue of exponential type, 
We will not use Theorem 1 since assumption (2) may not hold, but sketch an argument, based on the fact that S in the above question is the intersection of an analytic set with i 3, see, for example, [1] for the definition and properties of analytic sets. The functions u and ue in Remark 2 solve the differential equation
The function uN may vanish on S at most on the closed set
which is of the surface measure zero (by the uniqueness of the solution to the Cauchy problem for equation (16)). For every point
the argument given in the proof of Theorem 1 yields the existence of t(s), the unique solution to (6). Since S is real analytic the set Se , defined in the proof of Theorem 1, is analytic and is a part of the analytic set defined by the equation ue = 0. In our problem S is a bounded closed real analytic surface. The set Se can be continued analytically to an analytic set which intersects the real space i 3 over a real analytic surface Se. It is still an open problem to prove (or disprove) that the analytic continuation of the set Se intersects i 3 over a bounded closed real analytic surface 
- Fuks B (1963) Theory of analytic functions of several complex variables, AMS, Providence RI.
- Kato T (1984) Perturbation theory for linear operators, Springer Verlag, New York.
- Ramm AG (2005) Inverse problems, Springer, New York.