Journal Pre-proof Diffeomorphism on path connected components and applications
Hong Ding
PII:
S0022-247X(20)30091-3
DOI:
https://doi.org/10.1016/j.jmaa.2020.123929
Reference:
YJMAA 123929
To appear in:
Journal of Mathematical Analysis and Applications
Received date:
7 July 2019
Please cite this article as: H. Ding, Diffeomorphism on path connected components and applications, J. Math. Anal. Appl. (2020), 123929, doi: https://doi.org/10.1016/j.jmaa.2020.123929.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2020 Published by Elsevier.
Diffeomorphism on path connected components and applications Hong Ding College of Applied Sciences, Beijing University of Technology, Beijing, P.R. China E-mail:
[email protected] Abstract In this paper, we prove the following: Let F = (F1 , F2 ) ∈ C ∞ (R2 , R2 ). Let R > 0. And suppose det(DF (x)) > 0, ∀x ∈ B(0, R). Suppose there exist K > 0, r ∈ (0, K] and a ∈ R2 satisfying F −1 (B(a, r)) ∩ B(0, R) = ∅ and F (∂B(0, R)) ⊂ B(a, K) \ B(a, r). Let H be any path connected component of F −1 (B(a, r)) ∩ B(0, R). Then F |H : H → B(a, r) is an onto diffeomorphism. We give some applications of this result.
Keywords: diffeomorphism, Hausdorff measure, escape lemma. AMS Subject Classification: 58C25, 57R50, 26B10. 1. Introduction A classical theorem in the area of global inversion states that a local diffeomorphism f : X → Rn is a diffeomorphism ( here X ⊂ Rn a compact set) if f |∂X is one-to-one ( see [9]). Naturally, we want to know what we can get if we don’t have the condition that f |∂X is one-to-one. In this paper, we get a conclusion in R2 without using the injective condition on the boundary. We ask the boundary of the disk ∂B(0, R) to be mapped into an annular region.(Here the closed disk B(0, R) is the domain of our local diffeomorphism.) And we can still get a result of diffeomorphism over what could be a quite large region. Actually, our Theorem 1 states the following: ” Let F = (F1 , F2 ) ∈ C ∞ (R2 , R2 ). Let R > 0. And suppose det(DF (x)) > 0, ∀x ∈ B(0, R). Suppose there exist K > 0, r ∈ (0, K] and a ∈ R2 satisfying F −1 (B(a, r)) ∩ B(0, R) = ∅ and F (∂B(0, R)) ⊂ B(a, K) \ B(a, r). Let H be any path connected component of F −1 (B(a, r)) ∩ B(0, R). Then F |H : H → B(a, r) is an onto C ∞ -diffeomorphism.” And after we get Theorem 1, we have two applications for Theorem 1. One of the applications is to give a proof of a version of the Hadamard’s global inverse function theorem in the plane for functions that map a closed disk centered at origin into another closed disk centered at origin. The other application is Theorem 2. 2. The lemmas Throughout this paper we use the definition: B(a, r) = {x ∈ R2 | x − a < r}. Here a ∈ R2 . The norm is the Euclidean norm. 1
In order to prove the main theorem, we have a few lemmas. In the proof of Lemma 1 and Lemma 2, we use the area formula in geometric measure theory. Let us first state the theorem. We copy the theorem from [10], which is Theorem 3.7 there. The Area Formula Consider a Lipschitz function f : Rm → Rn for m ≤ n. If A is an L m measurable set, then Jm f (x)dL m x = N (f | A, y)dH m y Rn
A
Some explanation of the notations: Here Jm f (x) is the m-dimensional Jacobian of f at x. Computationally, Jm f (x) equals the square root of the determinant of the m×m matrix (Df (x))t Df (x). Here Df (x) is the Jacobian matrix of f at x. N (f | A, y) = card{x ∈ A : f (x) = y} (card(X) = the number of elements in the set X. ) H m is the m-dimensional Hausdorff measure. L m is the Lebesgue measure. Now we have our first lemma. Lemma 1 Let f = (f1 , f2 ) ∈ C 1 (R2 , R2 ), i ∈ {1, 2}, K ⊂ R2 be a compact set and the Jacobian det(Df (x)) > 0 , x ∈ K. Then for every c ∈ fi (K), the 1-dimensional Hausdorff measure of fi−1 (c) ∩ K is finite. Proof. Fix i = 1 or 2. We can see that fi−1 (c) ∩ K is closed and bounded. ∂fi 2 ) + So it is compact. Let (x, y) ∈ fi−1 (c) ∩ K. Then fi (x, y) − c = 0 and ( ∂x 1 ∂fi 2 ∂fi ( ∂x2 ) = 0 at (x, y). Without loss of generality, suppose ∂x1 (x, y) = 0. By the implicit function theorem (Theorem 17.3 of [4]), there exists a neighborhood W = (y − r, y + r) of y, and a C 1 function g : W → R, so that fi (g(x2 ), x2 ) ≡ c on W . And the conditions: | x1 − x |< r, x2 ∈ W, and fi (x1 , x2 ) = c imply that x1 = g(x2 ). In case that g is unbounded on W , we can find a smaller open interval W = (y − 2r , y + 2r ) and let g¯ = g |W , then g¯ is bounded on W . g (x2 ), x2 ) ≡ c on W and 2) The And the other two conditions, [i.e. : 1) fi (¯ r conditions: | x1 − x |< 2 , x2 ∈ W , and fi (x1 , x2 ) = c imply that x1 = g¯(x2 )] are automatically satisfied. So here without loss of generality, we assume that g is bounded on W . Let V = {(x1 , x2 ) ∈ R2 : | x1 − x |< r, x2 ∈ W }. Then V is an open neighborhood of (x, y) ∈ R2 . And fi−1 (c) ∩ V ⊂ {(g(x2 ), x2 ) ∈ R2 | x2 ∈ W }. Now define j : W → R2 , x2 → (g(x2 ), x2 ). Then j is one-to-one on W . By Theorem 3.7(the area formula) of [10], J1 j(x2 )dx2 = N (j | W, y)dH 1 y = H 1 (j(W )) W
R2
≥ H 1 (fi−1 (c) ∩ V )
Here H 1 is the 1-dimensional Hausdorff measure. Meanwhile W J1 j(x2 )dx2 = (g (x2 ))2 + 1 dx2 is finite. So we have shown for every (x, y) ∈ fi−1 (c) ∩ K, W 2
∃ an open neighborhood V of (x, y) so that H 1 (fi−1 (c) ∩ V ) is finite. A finite number of such V ’s cover fi−1 (c) ∩ K . Suppose they are V1 , · · · , Vp . Then fi−1 (c) ∩ K ⊂ V1 ∪ V2 ∪ · · · ∪ Vp p ⇒ fi−1 (c) ∩ K ⊂ fi−1 (c) ∩ (V1 ∪ · · · ∪ Vp ) = j=1 (fi−1 (c) ∩ Vj ), ⇒ H 1 (fi−1 (c) ∩ K) ≤
H 1 (fi−1 (c) ∩ Vj ) < +∞
j
Lemma 2 Let f = (f1 , f2 ) : R2 → R2 , f ∈ C ∞ (R2 , R2 ). Let R > 0. And suppose det(Df (x)) > 0, ∀x ∈ B(0, R). Suppose there exist K > 0, r ∈ (0, K] so that f (∂B(0, R)) ⊂ B(0, K) \ B(0, r). Here ∂B(0, R) represents the boundary of the set B(0, R). Suppose B(0, R) ∩ f −1 (B(0, r)) = ∅. Let ξ ∈ B(0, R) ∩ f −1 (B(0, r)), c = f2 (ξ). Let Γ be the connected component of f2−1 (c) ∩ B(0, R) containing ξ. Then f (Γ) ⊃ {(x, c) ∈ R2 | x2 + c2 < r2 }. Similar result exists for f1 . Proof. Let f = (f1 , f2 ). Define f˜ = f |B(0,R) , f˜i = fi |B(0,R) , i = 1, 2. We have rank(f˜i ) ≡ 1 on B(0, R), i = 1, 2. f˜i : B(0, R) → R is C 1 . Let −1 ξ ∈ B(0, R)∩f −1 (B(0, r)), c = f˜2 (ξ). By Theorem 3.2 of Chapter 1 of [7], f˜2 (c) −1 is a regular submanifold of class C 1 of B(0, R). And dim(f˜2 (c)) = 1. Let Γ −1 be the connected component of f˜2 (c) containing ξ. Then Γ is a connected 1-manifold of class C 1 . By Corollary 3.4.6 of [1], we can always have a global parametrization by arc length of Γ. Suppose the parametrization of Γ is: Γ = {h0 (s) = (h01 (s), h02 (s)) ∈ B(0, R) | (h01 (s))2 + (h02 (s))2 ≡ 1, s ∈ (aΓ , bΓ )} There exists s0 ∈ (aΓ , bΓ ) so that ξ = h0 (s0 ). It can be seen that h0 (s), s ∈ (aΓ , bΓ ) is the maximal integral curve of the initial value problem (IVP): ⎧ ∂ f˜2 ⎪ dx1 ⎪ ∂x2 ⎪ = ⎪ ⎪ ⎪ ds ∂ f˜2 2 ∂ f˜2 2 ⎪ ( ∂x ) + ( ∂x ) ⎪ ⎨ 1 2 ˜
∂ f2 − ∂x dx2 1 ⎪ ⎪ = ⎪ ⎪ ds ∂ f˜2 2 ∂ f˜2 2 ⎪ ⎪ ( ∂x1 ) + ( ∂x ) ⎪ 2 ⎪ ⎩ (x1 (s0 ), x2 (s0 )) = ξ
(1)
For the argument see [3]. So df˜1 (h0 (s)) det(Df˜) = (h0 (s)) > 0, ∀s ∈ (aΓ , bΓ ) ds ∂ f˜2 2 ∂ f˜2 2 ( ∂x ) + ( ) ∂x 1 2
3
(2)
If Γ is diffeomorphic to S1 , then f˜1 (h0 (s)) should be a periodic function of s. This contradicts (2). So Γ can only be diffeomorphic to R.(By the classification of connected 1-manifolds.) So there is a one-to-one correspondence between parameters s and points of Γ. By Lemma 1, (3) H 1 (f˜2−1 (c)) < +∞ Then the domain of (h01 , h02 ) must be finite by (3) and Theorem 3.7(the area formula ) of [10]. So both aΓ , bΓ are finite. By escape lemma (Lemma 5.1 of chapter IV of [2]), for an increasing sequence {tn } converging to bΓ , {(h01 (tn ), h02 (tn ))} cannot converge to a limit in B(0, R). But because {(h01 (tn ), h02 (tn ))} is a Cauchy sequence in R2 , it must converge to a point in R2 . So the limit is on the boundary of B(0, R). So lim distance((h01 (tn ), h02 (tn )), ∂B(0, R)) = 0 n→∞
⇒ lim distance( f (h01 (tn ), h02 (tn )), f (∂B(0, R)) = 0 n→∞
Define Γ+ = {(h01 (s), h02 (s)) ∈ B(0, R) | s0 ≤ s < bΓ }. Then we have distance( f (Γ+ ), ∂B(0, K) ) ≤ K − r
(4)
for f (∂B(0, R)) ⊂ B(0, K) \ B(0, r). If we let Γ− = {h0 (s) | aΓ < s ≤ s0 }, then distance(f (Γ− ), ∂B(0, K)) ≤ K − r. And we know by the intermediate value theorem of continuous function f (Γ+ ) = {(x, c) ∈ R2 | x ∈ [f1 (ξ), lim f1 (h0 (s)) )} sbΓ
If [ lim f1 (h0 (s))]2 + c2 < r2 , then we have distance(f (Γ+ ), ∂B(0, K)) > K − r. sbΓ
That contradicts (4). Therefore [ lim f1 (h0 (s))]2 + c2 ≥ r2 sbΓ
Similar result holds for f (Γ− ). So we have f (Γ) = f (Γ+ ) ∪ f (Γ− ) ⊃ {(x, c) ∈ R2 | x2 + c2 < r2 }. Lemma 3 Let f = (f1 , f2 ) ∈ C ∞ (R2 , R2 ). Let R > 0. And suppose det(Df (x)) > 0, ∀x ∈ B(0, R) . Suppose there exist K > 0, r ∈ (0, K], and f (∂B(0, R)) ⊂ B(0, K) \ B(0, r). Let λ ∈ f (B(0, R)) ∩ B(0, r) = ∅, μ ∈ f −1 (λ) ∩ B(0, R). Then there exists a continuous function L : B(0, r) → B(0, R), so that f (L(u)) = u, ∀u ∈ B(0, r) and L(λ) = μ. Proof. Fix λ ∈ f (B(0, R)) ∩ B(0, r) and μ ∈ f −1 (λ) ∩ B(0, R). Let Γ1 be the connected component of f1−1 (f1 (μ))∩B(0, R) containing μ. Then Γ1 is a regular submanifold of B(0, R) of dimension 1 (Theorem 5.8 of chapter III of [2]). And 4
it is connected and of class C 1 . By Corollary 3.4.6 of [1], there exists a global parametrization by arc length of Γ1 : Γ1 = {h1 (s) = (h11 (s), h12 (s)) ∈ R2 : s ∈ (aΓ1 , bΓ1 )} Suppose h1 (0) = μ. Then we can see s → h1 (s) is the maximal integral curve of the IVP of ODE: ⎧ ∂f1 dx1 2 ⎪ = ∂f1 ∂x ⎪ ∂f ⎪ ds ⎪ ( ∂x )2 +( ∂x1 )2 ⎨ 1 2 ∂f
⎪ ⎪ ⎪ ⎪ ⎩
So
dx2 ds
=
− ∂x1 1 ∂f1 2 ∂f ( ∂x ) +( ∂x1 )2 1
here,(x1 , x2 ) ∈ B(0, R)
2
(x1 (0), x2 (0)) = μ
− det(Df ) df2 (h1 (s)) = (h1 (s)) < 0, ∀s ∈ (aΓ1 , bΓ1 ). And by Lemma ∂f1 2 ∂f1 2 ds ( ∂x1 ) + ( ∂x ) 2
2, f (Γ1 ) ⊃ {(f1 (μ), y) ∈ R2 : (f1 (μ))2 + y 2 < r2 }. So, there exists a unique p1 ∈ Γ1 so that f2 (p1 ) = 0. Let u = (u1 , u2 ) ∈ B(0, r). Let Γ2 be the connected component of f2−1 (0) ∩ B(0, R) containing p1 . For Γ2 is a connected C 1 manifold of dimension 1, there exists a global parametrization by arc length of Γ2 (Corollary 3.4.6, [1] ): Γ2 = {h2 (s) = (h21 (s), h22 (s)) ∈ R2 : s ∈ (aΓ2 , bΓ2 )}, (h2 (0) = p1 ). Then h2 (s) is the maximal integral curve of the IVP of ODE: ⎧ dx1 ⎪ ⎪ ⎪ ds = ⎪ ⎨ dx2 ⎪ ⎪ ds
⎪ ⎪ ⎩
=
∂f2 ∂x2 ∂f ∂f ( ∂x2 )2 +( ∂x2 )2 1
∂f
− ∂x2
2
here,(x1 , x2 ) ∈ B(0, R)
∂f ∂f ( ∂x2 )2 +( ∂x2 )2 1
1
2
(x1 (0), x2 (0)) = p1
Define θ2 (p1 , s) = h2 (s). By Theorem 17.19 of [8], θ2 ∈ C ∞ . Because det(Df ) df1 (h2 (s)) = (h2 (s)) > 0, ∀s ∈ (aΓ2 , bΓ2 ). ∂f2 2 ∂f2 2 ds ( ∂x ) + ( ) ∂x2 1 And by Lemma 2, f (Γ2 ) ⊃ {(x, 0) ∈ R2 : x2 + 02 < r2 }. So there exists a unique s1 ∈ (aΓ2 , bΓ2 ) so that f1 (h2 (s1 )) = u1 . So s1 = s1 (u1 ) is a function of u1 . By the implicit function theorem (Theorem 17.3 of [4] ), s1 ∈ C 1 . Let p2 = h2 (s1 ) = θ2 (p1 , s1 (u1 )). Now let Γ3 be the connected component of f1−1 (f1 (p2 )) ∩ B(0, R) containing p2 . ⇒ Γ3 is a connected C 1 1-manifold. Let the global parametrization by arc length of Γ3 be: Γ3 = {h3 (s) = (h31 (s), h32 (s)) ∈ R2 : s ∈ (aΓ3 , bΓ3 )}, (h3 (0) = p2 )
5
Then h3 (s) is the maximal integral curve of the IVP: ⎧ dx1 ⎪ ⎪ ⎪ ds = ⎪ ⎨ dx2 ⎪ ⎪ ds
⎪ ⎪ ⎩
=
∂f1 ∂x2 ∂f ∂f ( ∂x1 )2 +( ∂x1 )2 1
∂f
2
− ∂x1 1 ∂f1 2 ∂f ( ∂x ) +( ∂x1 )2 1
here,(x1 , x2 ) ∈ B(0, R)
(5)
2
(x1 (0), x2 (0)) = p2
Define θ1 (p2 , s) = h3 (s). By Theorem 17.19 of [8], θ1 ∈ C ∞ . From (5) we get, − det(Df ) df2 (h3 (s)) = (h3 (s)) < 0, ∀s ∈ (aΓ3 , bΓ3 ). ∂f1 2 ∂f1 2 ds ( ∂x ) + ( ) ∂x 1 2 By Lemma 2, f (Γ3 ) ⊃ {(u1 , y) ∈ R2 : u21 + y 2 < r2 }. So u = (u1 , u2 ) ∈ f (Γ3 ). And ∃ a unique s2 ∈ (aΓ3 , bΓ3 ) so that f2 (h3 (s2 )) = u2 . So s2 = s2 (u1 , u2 ) is a function of (u1 , u2 ). And by the implicit function theorem (Theorem 17.6 of [4]), s2 ∈ C 1 . Now define L(u1 , u2 ) = h3 (s2 ) = θ1 (p2 , s2 (u1 , u2 )) = θ1 (θ2 (p1 , s1 (u1 )), s2 (u1 , u2 )) ∈ B(0, R) Clearly we have L ∈ C 1 (B(0, r), B(0, R)). Furthermore, we have f (L(u1 , u2 )) = (f1 (h3 (s2 )), f2 (h3 (s2 ))) = (u1 , u2 ), ∀(u1 , u2 ) ∈ B(0, r). 3. The main theorem and applications. Proposition 1 Let f = (f1 , f2 ) ∈ C ∞ (R2 , R2 ). Let R > 0. And suppose det(Df (x)) > 0, ∀x ∈ B(0, R). Suppose there exist K > 0, r ∈ (0, K] satisfying f −1 (B(0, r)) ∩ B(0, R) = ∅ and f (∂B(0, R)) ⊂ B(0, K) \ B(0, r). Let H be any path connected component of f −1 (B(0, r)) ∩ B(0, R). Then f |H : H → B(0, r) is an onto C ∞ -diffeomorphism. Remark Please note that when using the proposition, there is no need to know that f |∂B(0,R) is one-to-one . Proof. By C ∞ -diffeomorphism, we use the definition from the book [2](on p.41). So the map needs to be from an open set in R2 onto an open set in R2 . B(0, r) is certainly open in R2 . So we need to prove that H is open in R2 . Because f is continuous, we see that f −1 (B(0, r)) ∩ B(0, R) is open in R2 . By Theorem 25.4 of [11], H is open in R2 , since R2 is locally path connected. (Path connected component is called path component in [11].) 6
Let g = f |H : H → B(0, r). Pick any point μ ∈ H. Let λ = g(μ) = f (μ). Then λ ∈ B(0, r) ∩ f (B(0, R)). ⇒ μ ∈ f −1 (λ) ∩ B(0, R). By Lemma 3, there exists a continuous L : B(0, r) → B(0, R), so that f (L(u)) = u, ∀u ∈ B(0, r). And L(λ) = μ. We know that L(B(0, r)) is path connected. For the reason f (L(u)) = u, ∀u ∈ B(0, r), we have f (L(B(0, r))) ⊂ B(0, r). So L(B(0, r)) ⊂ f −1 (B(0, r)). Certainly, L(B(0, r)) ⊂ f −1 (B(0, r)) ∩ B(0, R). Also we have L(λ) = μ ∈ H, (λ ∈ B(0, r).) So we get L(B(0, r)) ⊂ H (which is a path connected component of f −1 (B(0, r)) ∩ B(0, R). ) Also, g(L(B(0, r))) = f (L(B(0, r))) = B(0, r). We get from L(B(0, r)) ⊂ H, the conclusion that g(H) ⊃ B(0, r). So g is surjective. Next we prove that g is injective. The proof is essentially the same as in the proof of Theorem 7 of [3]. Suppose g is not injective. Then ∃z1 , z2 ∈ H, z1 = z2 , so that f (z1 ) = f (z2 ). By Lemma 3, we can have a continuous L : B(0, r) → B(0, R) such that L(f (z1 )) = z1 . There exists a path δ : [0, 1] → H, so that δ(0) = z1 and δ(1) = z2 . Then by property of L, f (L(f (δ(t)))) = f (δ(t)), ∀t ∈ [0, 1]. So, both δ and L ◦ f ◦ δ are liftings of f ◦ δ : t → f (δ(t)) with respect to f . Because f is a local homeomorphism, by the uniqueness of lifting for local homeomorphism (Theorem 4.8 of [6] ), δ(0) = z1 , L ◦ f ◦ δ(0) = L(f (z1 )) = z1 , ⇒ δ(1) = L ◦ f ◦ δ(1). That is z2 = δ(1) = L ◦ f ◦ δ(1) = L(f (z2 )) = L(f (z1 )) = z1 . This contradicts the assumption z1 = z2 . So we have proven that g is injective. So the inverse g −1 exists. And by the inverse function theorem (Theorem 6.4 of Chapter II of [2]), g : H → B(0, r) is a homeomorphism. What is left is to show that the inverse is C ∞ . Again by the inverse function theorem, this is done. So g : H → B(0, r) is an onto C ∞ -diffeomorphism. Now we get our main theorem. Theorem 1 Let F = (F1 , F2 ) ∈ C ∞ (R2 , R2 ). Let R > 0. And suppose det(DF (x)) > 0, ∀x ∈ B(0, R). Suppose there exist K > 0, r ∈ (0, K] and a ∈ R2 satisfying F −1 (B(a, r)) ∩ B(0, R) = ∅ and F (∂B(0, R)) ⊂ B(a, K) \ B(a, r). Let H be any path connected component of F −1 (B(a, r)) ∩ B(0, R). Then F |H : H → B(a, r) is an onto C ∞ -diffeomorphism. Proof. Let f (x) = F (x) − a, ∀x ∈ R2 . So we can apply Proposition 1 to f. Then we get the conclusion. Using Theorem 1, we can prove a version of Hadamard’s global inverse function theorem in the plane for functions that map a closed disk centered at origin into another closed disk centered at origin. In order to prove the theorem, we need a lemma. Lemma 4 Let f = (f1 , f2 ) : R2 → R2 , f ∈ C 1 (R2 , R2 ). And suppose det(Df (x)) > 0, ∀x ∈ B(0, 1). Furthermore suppose f (∂B(0, 1)) ⊂ ∂B(0, 1). Then we have: f (B(0, 1)) ⊂ B(0, 1). 7
Proof. Because f is continuous on B(0, 1), so f assumes a maximum on B(0, 1). Because f is a local diffeomorphism on B(0, 1), the maximum point of f cannot be in B(0, 1). So the maximum point of f must be on the boundary ∂B(0, 1). Because f (∂B(0, 1)) ⊂ ∂B(0, 1), f = 1 on ∂B(0, 1), f ≤ 1 on B(0, 1). And furthermore we have ∀x ∈ B(0, 1), f (x) < 1. Because if f (x) = 1 for some x ∈ B(0, 1), the point x will be a maximum point of f inside B(0, 1). That contradicts what we have above. So f (B(0, 1)) ⊂ B(0, 1). Next we prove the version of the Hadamard’s global inverse function theorem in the plane via Theorem 1. Corollary 1 ( A version of the Hadamard’s global inverse function theorem in the plane ) Let f = (f1 , f2 ) ∈ C ∞ (R2 , R2 ). Let R > 0. And suppose det(Df (x)) > 0, ∀x ∈ B(0, R). Let K > 0 so that f (∂B(0, R)) ⊂ ∂B(0, K). Then f |B(0,R) : B(0, R) → B(0, K) is an onto C ∞ -diffeomorphism. Proof. Let g = f |B(0,R) . By Lemma 4, g(B(0, R)) ⊂ B(0, K). First we prove that g : B(0, R) → B(0, K) is injective. Suppose there exist a, b ∈ B(0, R), a = b, and p ∈ B(0, K), so that p = f (a) = f (b). Let l be the line segment connecting a and b. Let a, b also represent the vectors which are starting from 0 = (0, 0) and ending at points a and b. Then any point on the line segment l can be represented as l(t) = (1 − t)a + tb, t ∈ [0, 1]. We know that no point f (l(t)) can be on ∂B(0, K). So the distance(f (l(t)), (K cos θ, K sin θ)) > 0, ∀t ∈ [0, 1], θ ∈ [0, 2π]. Because the distance function is a continuous function, so there is a positive minimum α on the compact set [0, 1] × [0, 2π]. So there exists an α > 0, so that distance(f (l), ∂B(0, K)) ≥ α So f (l) ⊂ B(0, K − α). Now choose an R1 ∈ (0, R) close to R enough, so that it satisfies (a) a, b ∈ B(0, R1 ), and (b) f (∂B(0, R1 )) ⊂ B(0, K) \ B(0, r1 ), where r1 ∈ (K − α2 , K). Then we get, a, b ∈ f −1 (B(0, r1 )) ∩ B(0, R1 ) and
l ⊂ f −1 (B(0, r1 )) ∩ B(0, R1 ).
So there is a path connected component H of f −1 (B(0, r1 )) ∩ B(0, R1 ) so that a, b ∈ H. By Theorem 1, (Please note that we do not know that f |∂B(0,R1 ) is one-to-one, but we can still use Theorem 1. ) f |H : H → B(0, r1 ) is an onto C ∞ -diffeomorphism. That contradicts the fact f (a) = f (b). So there can’t be two points a, b ∈ B(0, R), so that f (a) = f (b). So g is injective. To prove g is surjective, pick any point q ∈ B(0, K). Pick a point p ∈ B(0, R), then f (p) ∈ B(0, K). Certainly there exists an r2 ∈ (0, K), so that q, f (p) ∈ B(0, r2 ). Choose R2 ∈ (0, R) close to R enough, so that (1) p ∈ B(0, R2 ) (2) f (∂B(0, R2 )) ⊂ B(0, K) \ B(0, r2 ). 8
So, p ∈ f −1 (B(0, r2 )) ∩ B(0, R2 ). Let H1 be the path connected component of f −1 (B(0, r2 ))∩B(0, R2 ) containing p. By Theorem 1, f |H1 : H1 → B(0, r2 ) is an onto C ∞ -diffeomorphism. So there q ) = q. So g is surjective. Because we also exists q˜ ∈ H1 ⊂ B(0, R) so that f (˜ have that g is injective. By the similar argument at the end of the proof of Proposition 1, g : B(0, R) → B(0, K) is an onto C ∞ -diffeomorphism. Another application is the following theorem. Theorem 2 Let f = (f1 , f2 ) ∈ C ∞ (R2 , R2 ). Suppose the Jacobian satisfies 0 < det(Df (x)) ≤ 1 on R2 . Let R > 0, r > 0, a ∈ R2 , and f −1 (B(a, r)) ∩ B(0, R) = ∅. Suppose f (∂B(0, R)) ⊂ R2 \B(a, r). If r > √12 R, then card f −1 (y) ∩ B(0, R) = 1, ∀y ∈ B(a, r). (Here card(X) represents the number of elements in the set X. ) Proof. Let H be any path connected component of f −1 (B(a, r)) ∩ B(0, R). We know from the proof of Proposition 1 that H is open in R2 . Because f is continuous, there exists a K > 0 such that f (∂B(0, R)) ⊂ B(a, K). So f (∂B(0, R)) ⊂ B(a, K) \ B(a, r). Then by Theorem 1, f |H : H → B(a, r) is an onto diffeomorphism. Then by Theorem (2.47) of [5], | det(Df (x)) | dx = m(B(a, r)) H
m is the Lebesgue measure on R2 . So 1dx ≥ | det(Df (x)) | dx = m(B(a, r)) H
If r >
H
√1 R, 2
1 2 πR (6) 2 Now let y ∈ B(a, r). Because f |H : H → B(a, r) is an onto diffeomorphism, and H ⊂ B(0, R), there exists at least one x ∈ B(0, R) such that f (x) = y. If ∃x1 , x2 ∈ B(0, R), x1 = x2 such that f (x1 ) = f (x2 ) = y. Then x1 , x2 cannot be in the same path connected component of f −1 (B(a, r)) ∩ B(0, R), because f |H : H → B(a, r) is one-to-one. So there exist at least two path connected components in f −1 (B(a, r))∩B(0, R). Denote them by H1 , H2 . Then by (6), m(Hi ) > 12 πR2 , i = 1, 2. For H1 ∩ H2 = ∅, we have m(H1 ∪ H2 ) = 2 m(H 1 ) + m(H2 ) > πR
. This contradicts the fact H1 ∪ H2 ⊂ B(0, R). So −1 card f (y) ∩ B(0, R) = 1. ⇒ m(H) ≥ πr2 >
Acknowledgements: The author would like to thank the reviewers and editors for their careful reading of the manuscript and constructive suggestions.
References [1] Berger,M. , Gostiaux,B. , Differential geometry: manifolds, curves and surfaces, Springer, New York, 1988. 9
[2] Boothby,W. M. , An introduction to differentiable manifolds and Riemannian geometry, second edition, Academic Press, San Diego, 1986. [3] Ding, H. , Behavior on level sets and global inversion, Applicable Analysis, vol.94(2015), No.9, 1838-1850. [4] Fitzpatrick, P. M., Advanced calculus, second edition, PWS Publishing Company, 2006. [5] Folland, G. B., Real analysis, John Wiley & Sons, 1984. [6] Forster, O. , Lectures on Riemann surfaces, Springer, New York, 1981. [7] Hirsch, M. W. , Differential topology, Springer, New York, 1976. [8] Lee, John M. , Introduction to smooth manifolds, Springer, New York, 2006. [9] Meisters, G. H., Olech, C., Locally one-to-one mappings and a classical theorem on schlicht functions, Duke Math. J. 30 1963 63-80 [10] Morgan, F. , Geometric measure theory: a beginner’s guide, Academic Press, 1988 [11] Munkres, James R., Topology, second edition, Prentice Hall, 2000
10