Picard’s theorems and Schottky’s theorem
within second order arithmetic
Yoshihiro Horihata Tohoku university
Joint work with Keita Yokoyama
27 May, 2011
Tohoku logic seminar
Today’s topics
1. Picard’s theorem via Riemann mapping theorem. 2. Picard’s theorem via Schottky’s theorem.
bababababababababababababababab
Picard theorem via RMT
Definition (RCA0)
✓ ✏
f : [0, 1] → R: continuous.
We say that f is effectively integrable on [0, 1] when we can find a modulus of integrability for f .
Here, modulus of integrability for f is a function h : N → N such that for all n ∈ N and for all partitions
∆1, ∆2 of [0, 1],
|∆1|, |∆2| < 2−h(n) → |S ∆1( f ) − S ∆2( f )| < 2−n+1.
✒ ✑
Theorem
✓ ✏
The following assertions are equivalent over RCA0. 1. WKL0.
2. Heine/Borel theorem.
3. Every continuous function [0, 1] → R is effectively integrable.
✒ ✑
Theorem
✓ ✏
The following assertions are equivalent over RCA0. 1. WWKL0.
2. Weak-Heine/Borel theorem.
Here, weak-Heine/Borel theorem is following:
If [0, 1] ⊆ Sn∈N B(an; rn) then there exists h(bi j, ci j)j≤li | i ∈ N, li ∈ Ni such that
[0, 1] ⊆ [
n<i
B(an, rn) ∪ [
j≤li
(bi j, ci j)∧
i→∞lim
X
j<li
|ci j − bi j| = 0.
✒ ✑
Theorem
✓ ✏
The following assertions are equivalent over RCA0. 1. WWKL0.
2. Weak-Heine/Borel theorem.
3. Every bounded continuous function is effectively integrable.
✒ ✑
Next, we see some basic theorems of complex analysis in SOA.
Definition (holomorphic function)
✓ ✏
The following definition is made in RCA0. D ⊆ C: open,
f , f ′ : D → C: continuous.
Then a pair ( f , f ′) is said to be holomorphic if
∀z ∈ D lim
w→z
f (w) − f (z)
w − z = f
′(z).
✒ ✑
Theorem
✓ ✏
The following assertions are equivalent over RCA0. 1. WKL0.
2. Cauchy’s integral theorem.
✒ ✑
Cauchy’s integral theorem
✓ ✏
D ⊆ C: open,
f : D → C: holomorphic. Then, for all △abc ⊆ D, R
∂△abc f (z) dz exists and Z
∂△abc
f (z) dz = 0.
✒ ✑
RCA0 version of Cauchy’s integral theorem
✓ ✏
D ⊆ C: open,
f : D → C: holomorphic and effectively integrable. Then, for all △abc ⊆ D, R∂△abc f (z) dz exists and
Z
∂△abc
f (z) dz = 0.
✒ ✑
Theorem
✓ ✏
RCA0 proves the RCA0 version of Cauchy’s integral theorem.
✒ ✑
Definition
✓ ✏
f : D = {z | 0 ≤ R1 < |z − a| < R2} → C: holomorphic.
Then, a is said to be an isolated essential singularity if there exists {an}n∈Z such that f (z) = Pn∈Z an(z − a)n for
all z ∈ D and ∀m ∈ N∃k ≥ m (a−k , 0).
✒ ✑
Theorem
✓ ✏
WWKL0 proves Riemann’s theorem on removable sin- gularities:
D := {z | 0 < |z − a| < r},
f : D → C : holomorphic.
If there exists r′ > 0 such that r′ < r and f is bounded on {z | 0 < |z − a| < r′}, then there exists a holomorphic function ˜f : D ∪ {a} → C such that ˜f (z) = f (z) for all z ∈ D.
✒ ✑
Theorem
✓ ✏
WWKL0 proves Casorati/Weierstraß theorem: D := {z | 0 < |z − a| < r},
f : D → C : holomorphic,
a is isolated essential singularity. Then f (D) is dense in C.
✒ ✑
Theorem
✓ ✏
WWKL0 proves Casorati/Weierstraß theorem: D := {z | 0 < |z − a| < r},
f : D → C : holomorphic,
a is isolated essential singularity. Then f (D) is dense in C.
✒ ✑
Picard’s little theorem
✓ ✏
f : C → C : holomorphic.
If f is non-constant, then f attains all values of C with at most one exception.
✒ ✑
bababababababababababababababab
(1) Liouville theorem.
bababababababababababababababab
(1) Liouville theorem. (2) Lifting lemma.
bababababababababababababababab
(1) Liouville theorem. (2) Lifting lemma.
(3) We can take ∆(1) as a covering space of C \ {0, 1}.
bababababababababababababababab
(1) Liouville theorem. ⇐ provable in WWKL0. (2) Lifting lemma.
(3) We can take ∆(1) as a covering space of C \ {0, 1}.
Lifting lemma
✓ ✏
(X, D, π, Ui j, Vi, πi j) : covering space, D ⊆ C : open,
f : D0 → D : continuous.
Then, if D0 is simply connected, then there exists a continuous function ˆf : D0 → X such that π ◦ ˆf = f .
✒ ✑
Lifting lemma
✓ ✏
(X, D, π, Ui j, Vi, πi j) : covering space, D ⊆ C : open,
f : D0 → D : continuous.
Then, if D0 is simply connected, then there exists a continuous function ˆf : D0 → X such that π ◦ ˆf = f .
✒ ✑
Theorem
✓ ✏
The following assertions are equivalent over RCA0. 1. WKL0.
2. Lifting lemma.
✒ ✑
bababababababababababababababab
(1) Liouville theorem. ⇐ provable in WWKL0. (2) Lifting lemma. ⇐ equivalent to WKL0.
(3) We can take ∆(1) as a covering space of C \ {0, 1}.
(3)We reason within ACA0.
Since Riemann mapping theorem is equivalent to ACA0 over WKL0, we can construct a biholomorphic function:
f : Ω := {z | Im(z) > 0 ∧ −1 < Re(z) < 1 ∧ |z| > 1} → ∆(1).
−1 1
Ω ∆(1)
f
We can prove Schwarz reflection principle within WWKL0.
We can prove Schwarz reflection principle within WWKL0. We can construct a covering map π : ∆(1) → C \ {0, 1} by
applying Schwarz reflection principle to Ω.
bababababababababababababababab
(1) Liouville theorem. ⇐ provable in WWKL0. (2) Lifting lemma. ⇐ equivalent to WKL0.
(3) We can take ∆(1) as a covering space of C \ {0, 1}. ⇐ provable in ACA0.
bababababababababababababababab
(1) Liouville theorem. ⇐ provable in WWKL0. (2) Lifting lemma. ⇐ equivalent to WKL0.
(3) We can take ∆(1) as a covering space of C \ {0, 1}. ⇐ provable in ACA0.
Therefore, we can prove Picard’s little theorem within ACA0.
But, in (3),
−1 1
Ω ∆(1)
f
since Ω is a simple domain, we expected that we can con- struct this biholomorphic function f within RCA0.
Definition (semi-polygon)
✓ ✏
A semi-polygon is a finite sequence of functions γ = hγ1, · · · , γni where γi : [(i − 1)/n, i/n] → C (1 ≤ i ≤ n) is a line or an arc of a circle, γi(i/n) = γi+1(i/n) for all 1 ≤ i ≤ n and γ1(0) = γn(1).
A semi-polygon γ is said to be simple if γ(t) , γ(s) for all 0 ≤ t < s < 1.
✒ ✑
Theorem (weak-Riemann mapping theorem)
✓ ✏
The following is provable in RCA0. γ : simple semi-polygon on C,
ϕ : linear transformation s.t. 0 ∈ ϕ(Int(γ)) ⊆ ∆(1), D := ϕ(Int(γ)).
Then, there exists a biholomorphic function f : D →
∆(1) such that f (0) = 0.
✒ ✑
But, to prove Picard’s theorem, we need to expand the map f into f : D → ∆(1).
This is very hard.
Theorem’ (weak-Riemann mapping theorem)
✓ ✏
The following is provable in RCA0. γ : simple semi-polygon on C,
ϕ : linear transformation s.t. 0 ∈ ϕ(Int(γ)) ⊆ ∆(1), D := ϕ(Int(γ)).
Then, there exists a biholomorphic function f : D →
∆(1) such that f (0) = 0.
Moreover, we can take f such that f is uniformly continuous on D;
✒ ✑
Theorem’’ (weak-Riemann mapping theorem)
✓ ✏
The following is provable in RCA0. γ : simple semi-polygon on C,
ϕ : linear transformation s.t. 0 ∈ ϕ(Int(γ)) ⊆ ∆(1), D := ϕ(Int(γ)).
Then, there exists a biholomorphic function f : D →
∆(1) such that f (0) = 0.
Moreover, we can take f such that f is uniformly continuous on D; f is limit-expandable into ∂D;
✒ ✑
Consequently,
• RMT for simple semi-polygon is provable in RCA0.
• RMT for Jordan curve is equivalent to WKL0.
• RMT is equivalent to ACA0 over WKL0.
bababababababababababababababab
Picard theorem via Schottky
Lemma
✓ ✏
The following assertions are provable in WKL0. Let R > 0 be a constant and f : ∆(R) → C be a holomorphic function such that f (z) , 0 on ∆(R). Then, there exist holomorphic functions h(z) and g(z) such that
f (z) = eh(z), f (z) = (g(z))n
where n is a positive integer. Then, we denote log f (z) and n p f (z) as h(z) and g(z), respectively.
✒ ✑
Bloch’s theorem
✓ ✏
The following assertions are provable in WKL0. Let f : ∆(1) → C be a holomorphic function.
Then, if f (0) = 0 and f ′(0) = 1, then, ∆(B) ⊆ f (∆(1)). Here, B = 1/32.
✒ ✑
Schottky’s theorem
✓ ✏
The following assertions are provable in WKL0. Let M > 0 and 0 < r < 1.
Then, there exists C, depending only on M and r, which satisfies the following properties:
If f : ∆(1) → C is a holomorphic function such that
| f (0)| ≤ M and f omits 0 and 1, then, | f (z)| ≤ C on
∆(r).
✒ ✑
Picard’s big theorem
✓ ✏
The following assertions are provable in WKL0. Fix R > 0 and z0 ∈ C.
Let f : B(z0; R) \ z0 → C be a holomorphic function. Then, if z0 is an essential singularity of f , then, for each positive r < R, f assumes, on B(z0; r) \ z0, every value of C with at most one exception.
✒ ✑
We see the proof.
We reason within WKL0.
To prove contradiction, assume that f omits two points. By translation, we may assume that z0 = 0, f is holomor- phic on ∆(2π) \ 0 and omits 0, 1. Since 0 is an essentially singularity of f , limz→0 | f (z)| does not converge (also, is not ∞).
Hence, there exists M > 0 such that for each n, there ex- ists x ∈ Q + iQ such that |x| < 2−n and | f (x)| < M.
We may assume M > | f (1/2)|.
Then, we want to find the sequence hzn | n ∈ Ni satisfies the following:
• limn→∞ zn = 0;
• 1 > |z1| > |z2| > · · · > |zn| > · · · ;
• | f (zn)| ≤ M for each n ∈ N.
To find the sequence, we use primitive recursion.
By primitive recursion, we can define h : N → Q + iQ as
h(0) = 1 2;
h(n + 1) = λxψ(x, h(n))
where ψ(x, h(n)) ≡ x ∈ Q+iQ∧| f (x)| < M ∧|x| < min{h(n), 2−n} is Σ01 formula.
Write ψ(x, h(n)) ≡ ∃kθ(k, x, h(n)), where θ is Σ00 formula. Then, λxψ(x, h(n)) := (µl θ[(l)0, (l)1, h(n)])1.
Then, let hzn | n ∈ Ni be a sequence such that zn = h(n) for each n ∈ N.
Then, hzn | n ∈ Ni satisfies the following:
• limn→∞ zn = 0;
• 1 > |z1| > |z2| > · · · > |zn| > · · · ;
• | f (zn)| ≤ M for each n ∈ N.
Fix j ∈ N. Define a function G : ∆(1) → C as G(ζ) := f (zj exp(2πiζ)).
Then, G is holomorphic on ∆(1), omits 0, 1, and satisfies
|G(0)| ≤ M.
Hence, by Schottky’s theorem, there exists K, which de- pends only on M, such that
|G(ζ)| = | f (zj exp(2πiζ))| ≤ K (ζ ∈ ∆(1/2)).
This especially implies that
| f (zj exp(2πit))| ≤ K (−1/2 ≤ t ≤ 1/2). Hence, | f (z)| ≤ K on |z| = |zj|.
This holds for each j ∈ N, since K depends only on M.
|z j|
|zj+1|
Dj
By the maximum modulus principle, f is bounded by K on each Dj := {z ∈ C | |zj| ≤ |z| ≤ |Z j+1|}.
Then, f is bounded by K on ∆(|z1|) \ {0}, since for each point z ∈ ∆(|z1|) belongs in Dj for some j ∈ N.
Therefore, by Riemann’s theorem on removable sin- gularities, 0 is a removable singularity, and hence con- tradiction.
This completes the proof.
Picard’s little theorem
✓ ✏
The following assertions are provable in WKL0. f : C → C : holomorphic.
If f is non-constant, then f attains all values of C with at most one exception.
✒ ✑
Proof of Picard’s little theorem
Assume that f : C → C is non-constant function.
Put g(z) := f (1/z). If z = 0 is an essential singularity of g, then it follows from Big Picard’s theorem.
If z = 0 is a pole or removable singularity of g, then g(z) − P
n≤N anz−n is entire and bounded.
Here, Pn≤N anz−n is the principal part of g.
Then, by Liouville’s theorem, g(z) − Pn≤N anz−n = C for some constant C ∈ C, and hence f is a polynomial. Hence, by the fundamental theorem of algebra, f attains all val- ues of C with at most one exception.
bababababababababababababababab
Future studies
• RCA0 proves the extension version of Weak RMT?? Study the next theorem within second order arithmetic.
Theorem (K. T. Hahn, 1983)
✓ ✏
Let D ⊆ C be a domain.
Then, the following assertions are equivalent: 1. D has at least two boundary points.
2. D has the Picard property. 3. D has the Schottky property. 4. D has the Landau property.
✒ ✑